Datasets:

griffin

/

ChemSum

like 13

Intermolecular interaction analyses on SARS-CoV-2 receptor binding domain and human angiotensinconverting enzyme 2 receptorblocking antibody/peptide using fragment molecular orbital calculation

The spike glycoprotein (S-protein) mediates SARS-CoV-2 entry via intermolecular interaction with human angiotensin-converting enzyme 2 (hACE2). The receptor-binding domain (RBD) of the S-protein has been considered critical for this interaction and acts as the target of numerous neutralizing antibodies and antiviral peptides. This study used the fragment molecular orbital (FMO) method to analyze the interactions between RBD and antibodies/peptides and extracted crucial residues that can be used to epitopes. The interactions evaluated as inter-fragment interaction energy (IFIE) values between the RBD and 12 antibodies/peptides showed a fairly good correlation with the experimental activity pIC50 (R 2 = 0.540). Nine residues (T415, K417, Y421, F456, A475, F486, N487, N501, and Y505) were confirmed as crucial. Pair interaction energy decomposition analyses (PIEDA) showed that hydrogen bonds, electrostatic interactions, and π-orbital interactions are important. Our results provide essential information for understanding SARS-CoV-2-antibodies/peptide binding and may play roles in future antibody/antiviral drug design.

intermolecular_interaction_analyses_on_sars-cov-2_receptor_binding_domain_and_human_angiotensinconve

TOC Graphic<!>Supporting information<!>Notes

<p>Coronavirus disease 2019 (COVID-19) caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has become an urgent health concern 1,2 . As of January 21, approximately 110 million infections and over 2.4 million deaths were confirmed worldwide 3 . Recently, the vaccines Tozinameran 4,5 and mRNA-1273 developed by Pfizer-BioNTech and Moderna 6,7 , respectively, have been approved for emergency use by the Food and Drug Administration (FDA), which raises expectations for the convergence of COVID-19. However, herd immunity via a high percentage of the global population being vaccinated requires more time. Recently, highly infectious variants of SARS-CoV-2 were reported in the UK, South Africa, Brazil, and other countries [8][9][10] .</p><p>Although there is no evidence showing that these variants can increase viral pathogenicity, the effectiveness of underused vaccines may become questionable for specific mutations.</p><p>Therefore, combating this virus may continue for a long time, and more candidates for drugs and vaccines are urgently needed. SARS-CoV-2 belongs to betacoronaviruses, which comprise the spike, envelope, membrane, and nucleocapsid as structural proteins 11 . Among the structural proteins, the glycoprotein (S-protein) mediates entry into the host cell via intermolecular interaction with human angiotensin-converting enzyme 2 (hACE2) 12 at its receptor-binding domain (RBD) [13][14][15][16] , making it a promising target for both antivirals and neutralizing antibodies.</p><p>To date, numerous antibodies targeting the RBD of S-proteins have been reported [17][18][19][20][21][22][23][24][25][26] .</p><p>These antibodies can be classified into four categories based on their binding sites (Figure 1a) 22 . Among them, antibodies that target RBD-hACE2 interaction sites as epitopes are considered particularly crucial, owing to the exceptions for the direct interruption of Sprotein-hACE2 binding. Understanding the detailed binding modes of the RBD with currently reported antibodies/peptides 27,28 can provide useful information for the design of more potent neutralizing antibodies and antiviral drug candidates.</p><p>The fragment molecular orbital (FMO) method is an ab initio quantum chemical calculation method [29][30][31] and has recently been applied to quantitatively and accurately evaluate molecular interactions [32][33][34][35][36][37][38][39][40][41] . Our group comprehensively performed FMO calculations on each target of COVID-19-related proteins, registered the results in the FMO database (FMODB) [42][43][44] , and released the data for use by other researchers.</p><p>Interaction analyses between S-protein and hACE2 or B38 Fab antibodies using the FMO method have been reported [45][46][47] . This study focused on 12 antibodies/peptides (LCB1 27 , LCB3 27 , C105 23 , COVA2-04 21 , BD-604 26 , CB6 17 , B38 17 , BD-629 26 , C102 22 , CC12.3 19 , CC12.1 19 , and CV30 17 ) and analyzed the interactions between these antibodies/peptides and the RBD using the FMO method. To obtain more useful information for rational antibody design, we further estimated the importance of amino acid residues and regions in the RBD for antibody/peptide binding.</p><p>The complex structural data of the RBD and antibodies/peptides were obtained from PDB, and preparations were conducted (see Supporting Information). Subsequently, the FMO calculation at the MP2/6-31G* 48,49 level was conducted using the ABINIT-MP program [50][51][52] . The FMO calculation results were registered in FMODB [42][43][44] (Table S1).</p><p>These data can be downloaded by users from the FMODB based on an FMODB ID or PDB ID, IFIE/PIEDA analysis is also easily possible on the web interface. The interfragment interaction energy (IFIE) values of the complexes were obtained. The IFIE values were further decomposed by the pair interaction energy decomposition analysis (PIEDA) calculation into four components: electrostatic (ES), exchange repulsion (EX), charge transfer with mixed terms (CT+mix), and dispersion (DI) 53,54 .</p><p>The IFIE-sum, which was the sum of IFIEs between RBD and antibodies/peptides, indicated that the estimated binding interaction energies between RBD and antibodies/peptides showed a fairly good correlation with the experimental inhibitory activities (pIC50) of the antibodies/peptides (R 2 = 0.540, Figure 1c, Table S1). A stronger IFIE-sum estimated from FMO would indicate higher antiviral activities of antibodies/peptides targeting RBD. We further investigated the importance of the RBD-antibody/peptide binding region, especially the interface residues. First, the antibody/peptide binding regions were selected per the EX, CT+mix, and DI value except ES 53,54 . These values are only non-zero when close interactions can be detected. The residues binding the antibody/peptide were in two regions: Seq# 403-421 and the receptor-binding motif (RBM, Seq# 438-506) 55 . The latter RBM is the major binding site of hACE2. The steric location of these two regions of one structure (PDB ID 7CH5) is shown in Figure 1b. These two regions are located at the interaction interface with the RBD-BD-629 Fab antibody and nearby regions.</p><p>Previous research on the interaction between SARS-CoV-2 S-protein RBD and ACE2 with FMO calculation 46 also showed that the interaction sites of hACE2 on RBD could be concentrated in these two regions.</p><p>The contributions of residues in these two regions to the RBD-antibody/peptide interactions were investigated using the ratio of the IFIE-sum (Table S2). The ratio of IFIE-sum of region 1 (Seq# 403-421) to the total IFIE-sum is about 40-50% (average 47.3%). Although the number of residues in the region is small (19), the contribution is high. Similarly, the ratio of the IFIE-sum of region 2 (RBM) to the total IFIE-sum is approximately 40%-60% (average 53.4%), and it exceeds 50% in 7 of the 12 structures treated in this study, indicating its high contribution. From the above, regions 1 and 2 play an important role in the RBD-antibody/peptide interaction in these complexes.</p><p>Next, to reveal important residues for molecular recognition between the RBD and antibody/peptide, we selected important residues meeting the criteria 46 from these two regions for all complexes. For any ES, CT+mix, and DI components, the residues with interaction energies less than -3 kcal/mol were considered important (Table S3). The frequencies for the selected residues are listed in Table S4 based on energy components, Figure 2 shows the RBD with a color gradation of the frequency on the molecular surface. important RBD residues found in this study and a previous report 46 on the RBD-hACE2 interaction. Residues that are commonly important for hACE2 and antibodies/peptides in molecular recognition of RBD and residues that are important only for antibodies/peptides are shown in blue and red, respectively. For any ES, CT+mix, and DI components, the residues with interaction energy less than -3 kcal/mol were considered important residues (refer to Table S4).</p><p>Fifty-two residues were identified as important residues in at least one of the structures, while nine residues (i.e., T415, K417, Y421, F456, A475, F486, N487, N501, and Y505) were commonly detected as key residues in all complexes. Furthermore, these residues consist of two important interaction regions (Figure 2a, Table S3, S4). To examine how much these key residues contribute to the RBD-antibody/peptide interactions, we calculated the correlation between the pIC50 and IFIE-sums obtained from only these nine residues. The result showed a slightly higher correlation (R 2 = 0.555), indicating that these residues can sufficiently account for the RBD-antibody/peptide interactions (Figure 1d, Table S5).</p><p>Next, the important residues in the RBD that were found in this study and those found in the previous report on the RBD-hACE2 interaction among SARS-CoV-2, SARS-CoV-2 chimera, and SARS-CoV 46 were displayed and compared (Figure 2b). Most of the important RBD residues overlapped for RBD-antibodies/peptides and RBD-hACE2 interactions (blue-colored regions), except for two residues, N439 and L461. N439 and L461 were previously reported to be important for RBD-hACE2 interaction 46 , but they are located at the binding edge. Therefore, it can be said that the antibodies/peptides can inhibit hACE2 spatially via interaction with most of the important interaction points of hACE2 binding. Furthermore, regions outside the hACE2 binding sites are also used by antibodies/peptides as the epitopes (red-colored regions).</p><p>Finally, the interactions of the nine RBD residues with each amino acid residue of antibodies/peptides were analyzed using both IFIE and PIEDA values in detail. An analysis of the RBD and BD-629 Fab antibody complex (PDB ID: 7CH5) was shown because this antibody has the best activity value among the antibodies calculated in this study (Table S1). As the characteristics of the interactions at the interface between RBD and BD-629 Fab, and XH/π interactions 35,44,46,56 with aromatic amino acids, such as tyrosine and phenilalanine, and hydrogen bonds with the oxygen atom of the main chain were frequently observed. Figure 3 shows the structure around the nine key residues in the RBD and BD-629 Fab complex. Tables S6 and S7 list the hydrogen bond and XH/π interaction energies and their characteristic distances between each fragment for nine residues by IFIE and PIEDA. Hydrogen bonds between side chains were also observed.</p><p>The hydrogen bonds formed between RBD and BD-629 Fab could be categorized into four types based on Figure 3 and Table S6. The first was hydrogen bonds between the side chains of the neutral amino acid residues. The IFIEs between residues containing such hydrogen bonds were approximately -13 to -14 kcal/mol, specifically fragment pairs such as T415spike-Y58H (IFIE = -13.0 kcal/mol) and N501spike-S30L (-14.0 kcal/mol) from Figures 3a and 3h, respectively. Second, the IFIEs between residues that contained a charged side chain on one of the amino acid residues that formed hydrogen bonds between the side chains were confirmed to be larger than the IFIEs of the first type because of the electrostatic interaction. Specifically, fragment pairs such as N487spike-E26H (IFIE = -25.8 kcal/mol) and N487spike-R97H (IFIE = -25.1 kcal/mol) were formed, as shown in Figure 3g. Third, when both amino acid residues contained charged side chains, the electrostatic interaction of salt bridges as well as hydrogen bonds were much larger than the former two types, and the corresponding IFIE between K417spike-D101H was -153.1 kcal/mol (although the hydrogen bond between the K417spike side chain and the oxygen atom of the main chain on D101H was also included) as seen in Figure 3b. Finally, when hydrogen bonds were between the side chain and the oxygen atom of the main chain, the IFIEs were less than -20 kcal/mol, specifically, fragment pairs containing hydrogen bonds between the oxygen atom on the F456spike and the Y33H side chain (IFIE = -23.7 kcal/mol) (Figure 3d), between the oxygen atom on the Gly476spike (assigned as the A475spike fragment) and the N32H side chain (IFIE = -20.9 kcal/mol) (Figure 3e), and between the Y505spike side chain and oxygen atom on the V29L (IFIE = -20.9 kcal/mol) (Figure 3i).</p><p>On the other hand, many π-orbital interactions were observed around aromatic amino acid residues, which can be classified into two categories: the OH/π interaction and CH/π interactions. First, ES and DI values were similar between residues where the OH/π interaction was formed (see Table S7 for each energy component), specifically, the fragment pair such as Y421spike-Y33H (Figure 3c). On the other hand, in many cases, DI was the main component between residues where CH/π interactions were formed (Table S7). Specifically, fragment pairs such as Y456spike-Y99H, F486spike-Y106H, and Y505spike-F32H (Figure 3d, 3f, and 3i, respectively).</p><p>In the supporting information, we explained the binding mode around the key residues in detail. Here, the corresponding FMODB data regarding the detailed energy components between individual residues by IFIE/PIEDA can be referred to using the FMODB ID 42 (Table S1).</p><p>As described above, we found that the nine key residues on the RBD strongly interact with the antibody residues regarded as the epitopes via various binding modes such as electrostatic interactions, hydrogen bonding, and π-orbital interactions (Figure 3). Here, the aromatic amino acids Y421, F456, F486, and Y505 were selected as key residues on the RBD, and as mentioned earlier, they interact with the residues on the antibody side by XH/π interactions. In FMO calculations, the XH/π interaction can be evaluated mainly by the DI term. However, it is difficult to accurately evaluate such XH/π interactions by molecular mechanics-based interaction energy analysis or structure-based geometry analysis with the classical force field 44 . This study accurately and quantitatively evaluated the XH/π interaction from the DI terms obtained by FMO calculations. We successfully detected aromatic amino acids that are important for the RBD-antibody/peptide interaction. This finding indicates the importance of specific aromatic amino acids on the RBD in the interaction of existing antibodies/peptides with RBDs and provides new insight into ACE2-blocking drug design.</p><p>In addition, K417, which was evaluated as a key residue in this study (Figure 3b), was mutated as K417N in the South African variant, 501.V2 variant 9 , and as K417T in the Brazilian variant, Lineage B. 1. 1. 248 10 . These mutations are predicted to lead to the loss of hydrogen bonds and XH/π interactions that K417 forms with residues on the antibody (Figure 3b). N501 (Figure 3h) is mutated as N501Y in these two variants as well as the British variant, VOC-202012/01 (also known as 501.V1 variant) 8 . The N501Y mutation is also predicted to lead to the loss of the hydrogen bond (ca -13 kcal/mol) that N501 forms with the residue on the antibody, while N501Y and its surrounding residue may form XH/π interactions 46 (Figure 3h). Mutations in the RBD can cause conformational changes in the protein itself as well as substructural changes. When antibodies/peptides interact with the RBD, small changes in the distance between residues may significantly affect interaction strength. Recently, it has been reported that the South African and Brazilian variants can escape from existing neutralizing antibodies 57,58 . More attention should be paid to these highly infectious mutants. Further FMO calculations on the interaction between mutant RBDs and existing antibodies/peptides are needed. In conclusion, this study analyzed the RBD-antibody/peptide interactions using FMO calculations. A fairly good correlation register between the calculated IFIE-sum and experimental pIC50 was confirmed. By further analyses, we extracted nine residues (T415, K417, Y421, F456, A475, F486, N487, N501, and Y505) in the crucial region of the RBD as critical for the binding of antibodies/peptides. These residues are also considered vital for S-protein-hACE2-binding. Detailed energy decompositions of IFIE by PIEDA around these residues showed that hydrogen bonds and electrostatic interactions, and πorbital interactions are important. Notably, mutations in some of the critical residues extracted here have been reported to be highly infectious. Our results provide essential information for understanding SARS-CoV-2 and antibodies/peptide binding and may play roles in future antibody/antiviral drug design. Table S1. Structural information about each calculation model and interaction energies for RBD-antibodies/peptides. Table S2. IFIE-sum of region 1 (R403-Y421), region 2 (RBM), and All (RBD) and the ratio of the IFIE-sum of region 1 or region 2 to the total IFIE-sum Table S3. Relationship between residues on the RBD and the number of structures counted as important residues. Table S4. PIEDA results of antibody/peptide-residues on the RBD (only region 1 (seq# 403-421) and region 2 (RBM)).</p><!><p>Table S5. Interaction energies between the nine key residues on the RBD and antibodies/peptide. Table S6. XH-Y hydrogen bonds between nine key residues on SARS-CoV-2 RBD and BD-629 Fab Table S7. XH/π interactions between the nine key residues with SARS-CoV-2 RBD and BD-629 Fab.</p><!><p>The authors declare no competing financial interest.</p>

ChemRxiv

The weakening effect of soluble epoxide hydrolase inhibitor AUDA on febrile response to lipopolysaccharide and turpentine in rat

A still growing body of evidence suggests the importance of epoxyeicosatrienoic acids (EETs) in the regulation of inflammatory response; therefore, drugs that stabilize their levels by targeting the soluble epoxide hydrolase (sEH), an enzyme responsible for their metabolism, are currently under investigation. The effect of sEH inhibitors on molecular components of fever mechanism, i.e., on synthesis of pro-inflammatory cytokines or prostaglandins, has been repeatedly proven; however, the hypothesis that sEH inhibitors affect febrile response has never been tested. The aim of this study was to examine if sEH inhibition affects core body temperature (Tb) as well as Tb changes during febrile response to infectious (lipopolysaccharide; LPS) or non-infectious (turpentine; TRP) stimuli. Male Wistar rats were implanted intra-abdominally with miniature biotelemeters to monitor Tb. A potent sEH inhibitor 12-(3-adamantan-1-yl-ureido)-dodecanoic acid (AUDA) was suspended in olive oil and administrated into animals in the intraperitoneal (i.p.) dose of 15 mg/kg, which, as we showed, has no significant influence on normal Tb. We have found that AUDA injected 3 h after LPS (50 μg/kg i.p.) significantly weakened febrile rise of Tb. Moreover, injection of sEH inhibitor 7 h after turpentine (administrated subcutaneously in a dose of 100 μL/rat) markedly reduced the peak period of aseptic fever. Obtained results provide first experimental evidence that sEH inhibitors possess anti-pyretic properties. Therefore, medicines targeting sEH enzymatic activity should be considered as a complement to the arsenal of topical medications used to treat fever especially in clinical situations when non-steroidal anti-inflammatory drugs are ineffective.

the_weakening_effect_of_soluble_epoxide_hydrolase_inhibitor_auda_on_febrile_response_to_lipopolysacc

Introduction<!>Experimental animals<!>Temperature measurements<!>Reagents and injections<!>Anti-TNF-α antibody injection<!>TNF-α assay<!>Statistics<!><!>Dose-dependent effect of AUDA on Tb in rats<!><!>Discussion<!><!>Discussion<!>Conclusion

<p>Fever, also known as pyrexia, is a regulated rise in body temperature (Tb) that is most frequently associated with infection, inflammation, and trauma. Experimental data strongly suggest important role of cytokines, especially interleukin (IL)-1β and IL-6 and tumor necrosis factor-α (TNF-α), as endogenous mediators of this physiological response [4, 16]. Pyrogenic cytokines are involved in stimulation of acute phase proteins, activation of hypothalamic–pituitary–adrenal (HPA) axis, and initiation of the arachidonic acid (AA) liberation from membrane phospholipids [15]. Free AA is a precursor for eicosanoids—signaling molecules that exert complex control over many bodily systems [3]. The cyclooxygenase (COX) pathway of AA metabolism, which can be inhibited by non-steroidal anti-inflammatory drugs (NSAIDs) or recently developed COXIBs (selective inhibitors of inducible form of cyclooxygenase, COX-2), produces prostaglandin E2 (PGE2)—another important downstream mediator of fever and inflammation [2]. In contrast, cytochrome P-450 monooxygenase (epoxygenase)-derived metabolites of AA, namely epoxyeicosatrienoic acids (EETs), possess anti-inflammatory and anti-pyretic properties. Kozak and coworkers proved that various isomers of EETs administered into the lateral ventricle reduce lipopolysaccharide (LPS)-induced fever in rats [17]. However, EETs have a short half-life limiting their therapeutic application, ranging from seconds to minutes [10], and are rapidly conversed to far less biologically active or inactive dihydroxyeicosatrienoic acids (DHETs) by soluble epoxide hydrolase (sEH) [6]. Thus, inhibitors of the sEH (sEHi) that stabilize and increase the EET levels are increasingly studied in rodent models of various diseases [as reviewed in 21]. Moreover, sEH inhibitors have been shown to downregulate the expression of COX-2 protein and synergize well with NSAIDs towards the reduction of inflammation [26] as well as LPS-induced plasma levels of pro-inflammatory cytokines [27]. Thus, given the evidence that inhibition of sEH activity affects molecular components of fever mechanism: cytokines, prostaglandins, and EET bioavailability, the present study aimed to determine the effect of a potent sEH inhibitor 12-(3-adamantan-1-yl-ureido)-dodecanoic acid (AUDA) on Tb changes in a course of febrile response to infectious and non-infectious stimuli in rat. Inhibition of sEH enzyme constitutes innovative approach in treating fever; thereby, drugs targeting this enzyme might complete the arsenal of available medicines, especially in clinical situations where oral anti-pyretics are only marginally effective, for example in lowering elevated Tb occurring after stroke [40].</p><!><p>Male Wistar Crl:WI(Han) rats aged 8–12 weeks and weighing from 250 to 300 g were purchased from the Mossakowski Medical Research Centre Polish Academy of Sciences (Warsaw, Poland) and were acclimatized for 10 days before starting the experiments. Animals were kept individually in a room at constant relative humidity (60 ± 5%) and temperature (24 ± 1 °C), with a 12:12-h light–dark photoperiod, with lights on at 7:00 h. Rodent laboratory food and drinking water were provided ad libitum.</p><p>All procedures were approved by the Local Bioethical Committee for Animal Care (permission no. 9/2015).</p><!><p>To monitor the core Tb, all animals were implanted intra-abdominally with temperature-sensitive miniature biotelemeters PhysioTels model TA10TA-F40 (Data Sciences International, St. Paul, MN, USA) under sterile condition [for details, see 39]. All surgical procedures were done at least 2 weeks before the start of experiments.</p><!><p>Systemic inflammation was provoked by intraperitoneal (i.p.) injection of bacterial lipopolysaccharide (LPS) while local aseptic necrosis of tissues was induced with turpentine (TRP) administrated subcutaneously (s.c.). It is well established that injections of both agents provoke characteristic, reproducible febrile rise of Tb in rats [for, e.g., see 23, 31, 39].</p><p>LPS derived from Escherichia coli 0111:B4 (Sigma-Aldrich, St. Louis, MO, USA) was dissolved in pyrogen-free 0.9% sodium chloride (saline) to obtain the final concentration of 50 μg/mL. LPS was injected i.p. in a dose of 50 μg/kg to provoke endotoxin fever. Intraperitoneal injection of saline (1 mL/kg) was used as a control.</p><p>Aseptic necrosis of tissues was induced with undiluted turpentine oil (Elissa, Warsaw, Poland). Turpentine was injected s.c. into the right hindlimb at a volume of 0.1 mL/rat.</p><p>sEH inhibitor 12-(3-adamantan-1-yl-ureido)-dodecanoic acid (AUDA) was synthetized according to the procedure [13]. Dose of AUDA was suspended in 500 μL of olive oil, then sonicated, and vortexed to obtain homogeneous suspension. Suspensions were made individually for each animal freshly before use and injected i.p. in a dose of 5, 15, or 30 mg/kg according to the experiment. As a control, animals received i.p. injection of olive oil in a volume of 500 μL.</p><p>All rats were restrained and not anesthetized during injections. The animals were weighed before injections to determine the precise doses of LPS and AUDA.</p><!><p>TNF-α antibodies (rabbit polyclonal IgG anti-rat TNF-α; Thermo Scientific, Waltham, MA USA; cat. no. PRTNFAI) were injected i.p. in a dose of 50 μg/rat in a volume of 500 μL of phosphate-buffered saline 1 h prior to the injection of AUDA. Rabbit IgG (Rockland Immunochemicals, Limerick, PA, USA; cat. no. 011-001-297) was used as a control. The dose of TNF-α antibody (50 μg/rat corresponds to the dose of 200–250 μg/kg) was selected according to the results of our previous experiments [12].</p><!><p>Blood was collected from anesthetized rats (mixture of ketamine/xylazine) by cardiac puncture into the solution of ethylenediaminetetraacetic acid (EDTA, Sigma-Aldrich, St. Louis, MO, USA). Plasma was separated by a centrifugation (20 min 1000×g) within 30 min of collection and was kept frozen at −20 °C until assay. Blood for analyses was collected an hour after LPS injection, at the time of the greatest decrease in Tb of rats observed with biotelemetry.</p><p>The levels of TNF-α were determined by a standard sandwich ELISA kit from Invitrogen (Camarillo, CA, USA cat. no. KRC3011; the minimum detectable dose of rat TNF-α is <4 pg/mL) according to the manufacturer's instructions. Colorimetric changes in the assay were detected using the Synergy HT Multi-Mode Microplate Reader (BioTek, Winooski, VT, USA).</p><!><p>Temperature values are reported as means ± standard error mean (SEM). Five-minute temperature recordings were pooled into 30-min averages for presentation. Mean values ± SEM of TNF-α concentrations in plasma were calculated for five plasma samples, each from different animal in the experimental group, that were assayed in duplicate. ANOVA with repeated measures followed by a Tukey multiple comparison post hoc test was used to determine the differences in time-dependent patterns of temperature among groups. ANOVA followed by Tukey pairwise comparison was used to test for statistical differences among groups at individual time points as well as TNF-α contents. Differences were considered significant at p < 0.05.</p><!><p>Changes of body temperature (°C) over time (h) of rats treated intraperitoneally at 8:00 with 500 μL of olive oil (open squares, a) or with AUDA (b) in a dose of 5 mg/kg (closed squares), 15 mg/kg (open circles), or 30 mg/kg (open triangles). Closed circles represent normal circadian rhythm of body temperature in control rats. Sample size is indicated in parentheses. Black arrowheads represent the time of injection. Values are means ± SEM at 30-min averages. Asterisks indicate significant difference (***p < 0.001) between rats treated with AUDA in a dose of 30 mg/kg and non-treated animals</p><!><p>Considering the obtained results, we decided to use AUDA in a dose of 15 mg/kg for further experiments as the highest dose that has not influenced the normal Tb of rats.</p><!><p>Changes of body temperature (°C) over time (h) of rats treated s.c. at 8:00 with turpentine. Seven hours afterwards, animals received intraperitoneally AUDA in a dose of 15 mg/kg (open circles) or 500 μL of olive oil (closed triangles) as a control. Closed circles represent normal circadian rhythm of body temperature in non-treated rats. Sample size is indicated in parentheses. Black arrowheads represent the time of injection. Values are means ± SEM at 30-min averages. Asterisks indicate significant difference (***p < 0.001) between "TRP/oil" and "TRP/AUDA" experimental groups</p><p>Changes of body temperature (°C) over time (h) of rats treated i.p. at 9:00 with LPS (50 μg/kg). Three hours afterwards, animals received intraperitoneally AUDA in a dose of 15 mg/kg (open circles) or 500 μL of olive oil (closed triangles) as a control. Closed circles represent normal circadian rhythm of body temperature in non-treated rats. Sample size is indicated in parentheses. Black arrowheads represent the time of injection. Values are means ± SEM at 30-min averages. Asterisks indicate significant difference (***p < 0.001) between "LPS/oil" and "LPS/AUDA" experimental groups</p><p>Changes of body temperature (°C) over time (h) of rats treated intraperitoneally at 8:00 with AUDA in a dose of 15 mg/kg (open circles) or with 500 μL of olive oil (closed triangles) an hour before LPS (50 μg/kg) administration. Closed circles represent normal circadian rhythm of body temperature in non-treated rats. Sample size is indicated in parentheses. Black arrowheads represent the time of injection. Values are means ± SEM at 30-min averages. Asterisks indicate significant difference (***p < 0.001) between "oil/LPS" and "AUDA/LPS" experimental groups</p><p>Plasma TNF-α concentration (pg/mL) measured 1 h after LPS (50 μg/kg i.p.) or saline (1 mL/kg i.p.) injection to animals pre-treated intraperitoneally with AUDA (15 mg/kg) or with 500 μL of olive oil as a control. All groups correspond to those shown in previous experiment. Values are means ± SEM calculated for plasma samples obtained from six animals for each experimental group</p><p>Changes of body temperature (°C) over time (h) of rats treated intraperitoneally at 7:00 with TNF-α antibodies (50 μg/rat i.p.) and at 8:00 with AUDA in a dose of 15 mg/kg an hour before LPS (50 μg/kg i.p.) (open circles) or NaCl (open squares) administration. Closed triangles represent Tb of rats treated at 7:00 with IgG (50 μg/rat i.p.) and at 8:00 with AUDA an hour before LPS injection (both in same concentration as above). Closed circles represent normal circadian rhythm of body temperature in non-treated rats. Sample size is indicated in parentheses. Black arrowheads represent the time of injection. Values are means ± SEM at 30-min averages. Asterisks indicate significant difference (***p < 0.001) between experimental groups "TNFab/AUDA/LPS" and "IgG/AUDA/LPS"</p><!><p>Inhibition of sEH represents a novel therapeutic strategy to treat hypertension and inflammation and to reduce pain [28]. However, the usefulness of sEH inhibitors as anti-pyretic drugs has not been examined yet. The present study aimed to test whether sEH inhibition affects the normal Tb and Tb changes in the course of fever in rats. In the experiments, we used 12-(3-adamantan-1-yl-ureido)-dodecanoic acid (AUDA)—a urea-based potent sEH inhibitor most extensively described in the literature [for, e.g., see 5, 19, 29]. We found that AUDA injected in a dose of 5 mg/kg as well as 15 mg/kg had no significant influence on Tb of rats (as can be seen in Fig. 1b), whereas 30 mg/kg induced meager but significant rise in Tb during early hours after injection. There are no experimental data that would clearly explain such result. To date, the testing in rodents and cell culture has found the urea-based sEH inhibitors to have extremely low toxicity and no adverse effects have been observed when treating rodents chronically [9]. However, to the best of our knowledge, Tb of animals treated with sEH inhibitors has never been measured and AUDA was never injected in such high intraperitoneal dose in animal studies. Since this compound does not directly affect phospholipase A2, an enzyme that releases AA from membrane phospholipids [21] and without exogenous stimuli that induces liberation of AA, injection even a massive dose of sEH inhibitor should not lead to an increase in EET levels. Nevertheless, it was recently discovered that EETs constitute a substrate for COX leading to formation of corresponding epoxy-prostaglandins [24]. This should be considered especially when major pathway of EET metabolism (by sEH) is inhibited. Verification of the hypothesis that injection of high dose of sEH inhibitor leads to an increase in epoxy-prostaglandin level in the circulation and that it translates into an increase in Tb requires, however, separate studies. Therefore, in all the following experiments, we used AUDA in a dose of 15 mg/kg. As we also established, this particular dose had no significant influence on hematological parameters in rats (data not shown).</p><p>To examine whether sEH inhibition affects fever, AUDA was given an hour before injection of LPS or turpentine as well as in the course of febrile response to these compounds. Not surprisingly, AUDA injected an hour before TRP had no influence on feverish changes in Tb (data not shown). Turpentine-induced fever is characterized with long (lasting at least 5–6 h) latency period, and as we presume, sEH inhibitor was biologically inactivated before it could affect endogenous components of fever mechanism. It is known that with the adamantine sensitive to P450 oxidation and the fatty acid chain sensitive to β-oxidation, AUDA is rapidly metabolized in vivo [21]. However, when inhibitor was administrated 7 h after turpentine injection, it significantly reduced the peak period of fever in rats (Fig. 2). A similar effect was observed when AUDA was injected 3 h after LPS, in experimental model of systemic inflammation (Fig. 3). These results provide the first experimental evidence that sEH inhibitor AUDA possesses anti-pyretic properties.</p><!><p>Effects of EETs increased by sEH inhibition with AUDA on the main components of molecular mechanism of fever. On the diagram, ↓ arrowheads represent activation while ⊥ inhibition. As a result of AUDA administration in the course of febrile response to inflammatory stimuli, DHET formation is inhibited and EETs produced from arachidonic acid by cytochrome P-450 monooxygenase are increased and available for a prolonged period. EETs acting by the mechanisms described in the discussion section lead to downregulation in fever mediators—cytokines and prostaglandins—thereby weakening fever</p><!><p>Interestingly, we found that AUDA injected an hour before LPS caused significant and rapid drop of Tb that almost completely diminished the first phase of fever (as can be seen in Fig. 4). Initially, we assumed that observed effect results from the TNF-α upregulation. TNF-α is the first cytokine that appears after LPS administration, peaks after 1–2 h, and can exert both pyrogenic or anti-pyretic effects [1, 12, 14]. Surprisingly, we found no significant increase in plasma TNF-α concentration measured 1 h after LPS administration to animals pre-treated with AUDA compared to vehicle (Fig. 5). Furthermore, injection of TNF-α antibodies before AUDA did not protect against observed drop in Tb in LPS-challenged rats (Fig. 6), thus indicating that the discussed effect is TNF-α-independent. Therefore, we assume that a drop in Tb after LPS injection into the animals pre-treated with AUDA might result from the sudden decrease in blood pressure (BP). It is generally accepted that LPS in high intraperitoneal doses (greater than 1 mg/kg) provokes a decrease in blood pressure in rats [35], but still, little is known of the effect of low doses corresponding to those used in our experiments. Soszynski and Krajewska showed that LPS (50 μg/kg) injected i.p. into rats causes a significant increase in plasma level of potent vasodilator nitric oxide (NO) within 3 h [32]; however, some reports suggest that this effect does not have to translate into lowering blood pressure [25]. As we mentioned, recent findings showed that AUDA significantly elevates levels of vasorelaxing EETs in LPS-treated mice [34]. In regard to interactions between EETs and NO, both appear to act independent; nevertheless, it was proved that EETs activate endothelial isoform of nitric oxide synthase (eNOS) [8]. In the light of these results, we presume that the observed drop in Tb is a consequence of a sudden increase in vasorelaxing agents as a consequence of synergizing effect of AUDA and LPS.</p><p>Inhibition of sEH is an emerging strategy for treatment of cardiovascular and inflammatory disorders. We showed that sEH inhibitors should also be considered as potential anti-pyretic drugs. Undoubtedly, non-steroidal anti-inflammatory drugs (NSAIDs) are one of the most widely prescribed medications in the world for treating fever. The major problem with the use of these drugs is that their chronic administration is limited by the metabolic and cardiovascular side effects [30, 33]. Moreover, NSAIDs are only marginally effective in lowering elevated Tb in specific clinical cases, i.e., in treating fever occurring after stroke that is often associated with poor outcomes [40]. We have previously proved in the animal model of cerebral hemorrhage that such rise in Tb is a PGE2-dependent response, however, it appeared to be not sensitive to a COX inhibitor, unlike the fever induced by LPS [37]. Drugs targeting sEH not only are safe in use, but also reduce the undesirable side effects and synergize well with NSAIDs and COX-2 blockers (COXIBs) [28]. They might be attractive in drug combinations allowing to reduce the dose of anti-pyretics used in clinics. Increased circulating levels of EETs caused by sEH inhibition reduce not only NF-κB nuclear translocation but also COX-2-dependent synthesis of PGE2 in both peripheral tissues and central nervous systems [36]. Furthermore, recent findings showed that sEH inhibitors could eliminate pain caused by the injection of the PGE2 that cannot be treated with either NSAIDs or steroids [11]. It clearly shows that sEH inhibitors by stabilizing EETs affect the biology of prostaglandins through several mechanisms that are not fully understood and involvement of receptors and signaling pathways cannot be ruled out as well [41]. Further studies should concentrate on investigating the usefulness of newly discovered sEH inhibitors in combination with NSAIDs especially in clinical cases where topical therapy does not work.</p><!><p>The results of the present study clearly demonstrate that a potent sEH inhibitor AUDA injected in a course of endotoxin or aseptic fever significantly weakened the rise in Tb of rats. This effect of AUDA undoubtedly results from the increase in EET bioavailability that leads to an inhibition of NF-κB transcriptional activity and COX-2 enzymatic activity and in consequence to downregulation of fever mediators—cytokines and prostaglandins. Obtained results constitute the first experimental evidence that sEH inhibitors should be considered as potential anti-pyretic drugs and thereby should be further examined for their suitability in clinics. Since inhibitors of sEH synergize well with NSAIDs, combined therapy could allow to reduce the side effects resulting from chronic intake of NSAIDs or to increase the effectiveness of treatment, especially of patients who experience stroke. The present study provides also a new experimental model for studying the biological effects of newly synthesized sEH inhibitors and may as well contribute to the expansion of therapeutic area of interest for sEH inhibitors.</p>

PubMed Open Access

Probing the molecular frame of uracil and thymine with high-harmonic generation spectroscopy †

In this work we present computed high-harmonic generation (HHG) spectra of uracil and thymine molecules, by means of the real-time time-dependent formulation of Gaussian-based configuration interaction with single excitations (RT-TD-CIS). According to the experimental work [Hutchinson et al., Phys. Chem. Chem. Phys. Comparison of high-order harmonic generation in uracil and thymine ablation plumes, 2013, 15, 12308] a pulse wavelength of 780 nm has been used, together with an intensity of 10 14 W/cm 2 and a pulse duration of 23 optical cycles. In order to examine the effect of pulse polarisation, rotationally-averaged (to mimic the gas-phase sample) and single-polarisation have been computed for both molecules. Our results show that the HHG signal for both molecules possibly originates from different ionisation channels, involving HOMO, HOMO-1, HOMO-2 and HOMO-3 orbitals, which lie within 4 eV. We characterize the HHG spectrum of thymine, supporting the idea that the absence of thymine signal in the original work does not depend on the single-molecule behaviour. Present results for uracil are consistent with the experimental data. Moreover, we have observed that states below and above the chosen ionisation threshold provide different contributions to the HHG spectrum in averaged and single-polarisation calculations.

probing_the_molecular_frame_of_uracil_and_thymine_with_high-harmonic_generation_spectroscopy_†

Introduction<!>Theory<!>Computational details<!>Results and Discussion<!>Uracil<!>Thymine<!>Conclusions<!>Conflicts of interest

<p>A growing interest in attosecond atomic and molecular processes is nowadays triggered by the advent of ultrafast laser technology [1][2][3][4][5][6][7][8] . Novel time-resolved spectroscopies are emerging together with the opportunity to study electron dynamics with unprecedented time resolution 9 . Attosecond light pulses are indeed used to investigate ultrafast electron dynamics in atomic and molecular systems. 5 The research in the attosecond domain can answer challenging questions such as: which is the role of electron correlation in nonlinear optical process? Which is the electron interference in a molecular system ionised from different channels? Which is the ultrafast redistribution of electrons in molecular charge migration that could be then responsible for bond breaking? As a consequence, attosecond resolution became the fundamental requirement to design original experimental analysis to understand a large variety of physical and chemical processes. The attosecond laser resolution relies on the nonlinear optical processes, as high-harmonic generation (HHG). [10][11][12][13] . The HHG spectra, made of the emitted harmonics of the infrared pulse frequency, encode information about the electronic structure and the dynamical effects of the systems from which the HHG signal is emitted. [14][15][16][17][18][19][20][21][22][23][24][25][26][27][28][29][30] . For instance, from HHG spectra one can extract Italy; E-mail: [email protected] † Electronic Supplementary Information (ESI) available:</p><p>[details of any supplementary information available should be included here].</p><p>See DOI: 10.1039/cXCP00000x/ information on the electronic wave function for molecular orbital tomography. 15,31,32 HHG can also detect interference phenomena between ionisation channels in molecules 20,21,28,33,34 , ultrafast charge migration 35 , vibrational signatures 36 , and quantum coherence in atoms 37 . HHG spectroscopy is therefore considered a powerful tool to investigate the ultrafast processes occurring in atoms, molecules and nanostructures 38 , and recently it has also been applied to study organic and biological molecules 39 . However, HHG experiments on organic and biological molecules can present technical difficulties. In fact, the main practical issue is due to the fact that molecules at room temperature can be in solid or liquid phase, and producing a high-density gasphase sample of unfragmented molecules is not trivial. 39 Moreover, these molecules have typically low ionisation energies and show a large variety of different conformers. Uracil and thymine have been widely characterised spectroscopically, from experimental and theoretical works, in terms of valence [40][41][42][43][44] and core excitations 45,46 . But only only one experimental attempt to measure HHG in uracil and thymine is present in literature, by Hutchinson et al.. 47 Authors used a laser ablation technique to produce weakly ionized plasma plumes from solid samples of uracil and thymine. Only the uracil HHG signal has been observed. The difference between the HHG spectra of uracil and thymine was attributed to differences in the ion composition of the ablation plumes of the two molecules, determining conditions of phase mismatching in thymine. That work certainly opened the route to other experimental investigations of HHG in these biomolecules, as many questions remains still open: why the HHG spectrum of thymine was not measured? Which is the role of fragments in the measured HHG spectra? Do electron correlation and interference effects play any role in the strong-field dynamics? Theoretical calculations can definitely help to answer these questions. Simulations on these molecules are numerically challenging because of their size and complexity. To calculate HHG spectra in uracil and thymine we used the all-electron real-time dependent configuration interaction with single excitations (RT-TD-CIS), 37,[48][49][50][51][52][53][54][55][56][57][58][59] using the computational strategy we developed in the recent years. 37,53,55,57,58 Standard Dunning's basis sets 60 were augmented with diffuse and continuum-optimal 61 Gaussian functions. This determines a proper description of Rydberg and low-energy continuum states. This approach represents a reasonable compromise between accuracy and computational cost, which permits an affordable treatment of many-electron systems interacting with strong fields 58 . Moreover, this approach allows one to control and systematically improve Gaussian basis sets. We computed HHG spectra for different linear polarisations of the laser field in the molecular plane of uracil and thymine, in the three-dimensional (3D) space around the molecules. In order to simulate randomly oriented molecules we averaged the time-dependent dipoles projected on the different polarisations in the molecular plane or in the 3D space. We also computed the HHG spectra with a pulse polarisation perpendicular to the molecular plane or along the direction of the ground-state permanent dipole. Moreover, we analysed the contribution to the HHG spectrum due to to the states below and above the selected ionisation energy. Results presented here allow us to give a quantum, microscopic description of HHG in uracil and thymine and pose open questions inherent the strong field-electron dynamics in these biomolecules. The article is organised as follows: in Section 2 the theoretical framework for computing HHG spectra is reviewed, computational details of our simulations are given in Section 3, results are presented and discussed in Section 4, while in Conclusions we summarize the main achievements of the present work and give perspectives for future developments.</p><!><p>The time-dependent Schrödinger equation for a molecular system under the influence of an external time-dependent electric field is (atomic units are used) :</p><p>where Ĥ0 is the time-independent field-free electronic Hamiltonian and V (t) is time-dependent potential that we calculated in the length gauge. In this gauge V (t) = − μ µ µ • E(t) where μ µ µ is the molecular dipole and E(t) is the time-dependent electric field. 51 The electric field is linearly polarised E(t) = E 0 n sin(ω 0 t + φ ) f (t) along the direction n, E 0 is the maximum field strength, ω 0 is the carrier pulse frequency, φ is the field phase, and f (t) is a cos 2 envelope function. 55 We solved Eq. (1) in the framework of time-dependent configuration interaction with simple excitations (RT-TD-CIS). 48,49,54,55,[62][63][64][65] The time-dependent wave function |Ψ(t) is expanded in a discrete basis of the eigenstates of the fieldfree Hamiltonian Ĥ0 composed of the Hartree-Fock ground state (k = 0) and all the CIS excited states 66 (k > 0)</p><p>where c k (t) are time-dependent coefficients. By inserting Eq. ( 2) into Eq. ( 1) and projecting on Ψ l |, the timedependent equation for the coefficients is:</p><p>where c(t) is the vector of the coefficients c k (t), H 0 is the diagonal matrix representation of Ĥ0 , with elements</p><p>The initial wave function at t = t i = 0 is the field-free ground state, i.e. c k (t i ) = δ k0 . The Eq. ( 3) is solved by discretizing the time (∆t is the time step) and using the split-propagator scheme 55 : c(t + ∆t) ≈ e −iV(t)∆t e −iH 0 ∆t c(t).</p><p>(</p><p>The matrix H 0 is diagonal, as a consequence e −iH 0 ∆t is also a diagonal matrix of elements e −iE k ∆t δ lk . The exponential of the nondiagonal matrix V(t) is calculated by the following transformation</p><p>where U is the unitary matrix describing the change of basis between the original eigenstates of Ĥ0 and a basis in which 51,55 From the knowledge of the time-dependent wavefunction |Ψ(t) , the time-dependent dipole µ µ µ(t) is then calculated as</p><p>and by taking its Fourier transform the HHG spectrum is obtained as</p><p>with t f is the final propagation time. In Figure 2 we schematically show how we simulated an HHG for a laser linearly polarised in the direction perpendicular to the molecular plane (panel a)) and along the direction of the permanent molecular dipole (panel b)).</p><p>Next, we considered the molecules randomly oriented in the plane and in the 3D space. The randomly oriented molecules were simulated averaging over many different polarisation direc-tions of the laser. In the case of the plane we used a normalised polarisation vector ns = ( ns,x , ns,y , 0) while in case of the three dimensional space we used ns = ( ns,x , ns,y , ns,z ), where ns,x and ns,y (and also ns,z for the 3D case) are uniform random numbers in [-1:1] interval. Then, we calculated an averaged time-dependent dipole μ(t) defined as</p><p>where S is the number of simulations made with a randomly chosen electric field polarisation ns and µ µ µ s (t) is the corresponding time-dependent dipole. The averaged HHG spectrum is then calculated as</p><p>In Figure 2 we schematically show how we simulated an HHG rotationally averaged in the plane (panel c)) and in 3D (panel d)).</p><p>Rationalizing the generation mechanism of high-order harmonics in thymine and uracil implies the inclusion of several factors. Tunneling ionisation and recombination can involve different channels for the electron removal (as discussed below) and centers of recombinations. By defining a threshold energy (E T ), which corresponds to the ionisation energy of a particular channel, we are able to disentangle the role played by electronic states above and below E T , and, as a consequence, to more deeply investigate their possible contribution to HHG and mutual interactions. 52 The time-dependent dipole moments, µ µ µ(t) (Eq. 6) or μ(t) (Eq. 8), are hence rewritten as a (partial) sum of terms depending only on the states below E T , labeled as B, or states above E T , labeled as A.</p><p>The ground state is indicated as G. Following this decomposition, the time dependent dipole moments become:</p><p>and</p><p>The partial contributions to time-dependent dipole moments can therefore be identified as:</p><p>µ µ|Ψ 0 is the ground-excited contribution with the index i running over the states below E T (B) or above E T (A), and µ µ µ i j (t) = c * j (t)c i (t) Ψ j | μ µ µ|Ψ i is the excited-excited contribution with the index i and j running over the states below E T (B) or above E T (A). µ µ µ B,s (t) and µ µ µ A,s (t) are the s-th realization of the partial µ µ µ B (t) and µ µ µ B (t) dipoles. The HHG spectrum from each partial contribution is then computed as in Eqs 7 or 9. States below and above E T are also named in this work as bound and continuum states, respectively.</p><!><p>The geometrical structures of uracil and thymine are shown in Figure 1. These structures have been optimised in the groundstate at DFT/B3LYP level using the 6-31G(d,p) basis set with the software Q-Chem. 67 We performed CIS calculations 67 to obtain a number of electronic excited-state energies and transition dipole moments of the fieldfree Hamiltonian Ĥ0 , which are then be used to propagate the time-dependent wavepacket in RT-TD-CIS, as shown in Eq. 3.</p><p>The RT-TD-CIS simulations were performed by means of the homemade code Light [51][52][53][54][55] , that propagates the electronic wavepacket under the influence of a time-dependent strong laser field. We used the same pulse wavelength of the experiment of Ref. 47, i. e. 780 nm (equal to ω 0 =1.59 eV). The pulse intensity is defined as</p><p>We used a pulse intensity I = 10 14 W/cm 2 and pulse duration of 23 optical cycles. The phase φ was set to zero. Nuclei were kept frozen at their equilibrium position during the time propagation. For RT-TD-CIS we used the cc-pVDZ Dunning's basis set 60 for the H atoms of uracil and thymine. Instead, for all the other atoms (C, O and N) we augmented the cc-pVDZ basis set in order to increase the number of bound and continuum states. We added to the cc-pVDZ kernel three sets of diffuse functions and three sets of optimised Gaussian functions for continuum (K) for each angular momentum. We thus obtained the 3aug-cc-pVDZ+3K basis set. The K functions 61 determine a progression of continuum states, i.e. above the ionisation energy E T , which mimics the true manifold of molecular continuum states. [54][55][56] Ionisation in the simulated strong-field electron dynamics was treated by means of the heuristic lifetime model, originally reported in Ref The experimental first, i.e. lowest, ionisation energy of uracil is in the 9.45-9.68 eV range [70][71][72][73][74] , while for thymine is in the range 9.02-9.20 eV [71][72][73] . Our theoretical estimates, computed at Hartree-Fock level of theory with the 3aug-cc-pVDZ+3K basis set, are reported in Table 1. Our calculations overestimate by around 0.5 eV the lowest ionisation energy, corresponding to minus HOMO energy according to Koopmans' theorem. The order of magnitude of the discrepancy between experimental and theoretical ionisation energy values is the same found for CO 2 , with a similar theoretical approach 28 . In order to also include in the propagation contributions from higher ionisation channels (namely, a larger amount of energy is needed to remove the electron from inner orbitals) we considered that electrons can also be extracted from HOMO-1, HOMO-2 and HOMO-3. This choice corresponds to use minus HOMO-3 energy as E T . With this choice the 3aug-cc-pVDZ+3K basis set shows a reasonable balance between energy states below and above E T .</p><!><p>In the experimental work of Hutchison et al. 47 , harmonics were generated from the uracil plasma plume using driving laser pulses at wavelengths of 780 and 1300 nm. The largest harmonic was found to be H33 and H39 in different conditions (Figure 3 and 4 in Ref. 47) with 780 nm. On the other hand, no thymine signal was experimentally recorded. The analysis of the ablation plumes with mass spectrometry shows that the plumes contain a large number of fragments of the parent molecular ion, with various shape and electronic structure. A higher degree of atomization in the thymine plume was found for thymine, with a corresponding higher density of plasma free electrons. Authors argue that abundance of free electrons in the thymine plasma can induce a strong phase mismatch of the HHG signal, making the recording of it not possible. Furthermore, the observed uracil HHG signal is probably the result of a mixed contribution from the entire molecule and its fragments.</p><p>Our goal is the characterization of the HHG spectrum of the two molecules by an ab initio all-electron dynamics, as explained in Section 2. It is worth comparing the information encoded in the experimental spectra of Ref. 47 with what is computed by means of our theoretical approach. The outcomes of our simulations can be compared with the experimental results, but it is necessary to discuss the advantages, drawbacks and limits of the theoretical comparison we propose. The experimental sample is a plume containing the entire molecules and a number of atomic and molecular fragments, as reported above, while we compute the HHG spectrum of unfragmented uracil and thymine molecules: analysis of the fragmentation pattern [75][76][77] will allow us to also simulate, as a next step in future works, the HHG signal from fragments, and to possibly dissect fragment contribution in the final HHG spectrum. For all the simulations we have employed an intensity of 10 14 W/cm 2 , which is a typical value for getting HHG. In Ref. 47 authors report that the infrared probe pulse acquires an intensity up to 10 15 W/cm 2 in the experiments: even when the energy of probe pulse is reported, extracting the precise intensity value corresponding to the recorded HHG spectrum is not easy, making a quantitative comparison with our spectra impracticable. Moreover, in our calculations phase matching conditions can not be simulated, but signatures of electronic structure and dynamics encoded in the HHG spectrum are mainly due to single-molecule behaviour. 39 With these (current) limitations in mind, we are however able to provide a detailed description of the main features of HHG spectra of uracil and thymine. The present work should be therefore seen as a first and original proposal for the application of a full quantum-dynamics approach to the study of complex molecules of biological interest.</p><!><p>In Figure 3 HHG spectra of uracil have been reported. The calculations were carried out following the simulation schemes presented in Figure 2 and described in Section 2.</p><p>The two averaged (plane and 3D) HHG spectra show the same general shape for the low-energy and plateau regions and a substantial difference around the cutoff. Harmonics up to around H39-H41 are seen for the 3D rotational average, whereas the largest harmonic for the molecular-plan average is H29-H31. The three-step model 68,69 estimation of the cutoff is approximately H22. This comparison suggests that contributions to HHG from out-of-plane polarisations are important for higher harmonics. Moreover, the HHG spectra averaged in the molecular plane show higher intensity harmonics in the H13-H31 region. The background of the molecular-plane spectrum is much higher than that of the 3D-averaged spectrum beyond H17.</p><p>The HHG spectrum with polarisation perpendicular to the molecular plane strongly differs from the averaged spectra starting from H9. From H9 to H11 the harmonics are much higher than in the two averaged HHG spectra. Moreover, the spectrum presents a second plateau from H21 to H35, but the intensity of these harmonics is much lower than those of the first part of the spectrum, up to H19. For the second plateau, the cutoff is around H45. The HHG spectrum with polarisation in the direction of the permanent dipole has similar behaviour as the perpendicularpolarisation HHG spectrum. Low-intensity harmonics in the high energy region of the spectrum are still present. In this case even harmonics are clearly visible, due to the inversion-symmetry breaking. For perpendicular polarisation inversion symmetry is given at much less extent, resulting in nearly negligible even peaks.</p><p>Fig. 3 HHG spectra of uracil with pulse wavelength and intensity of 780 nm and 10 14 W/cm 2 .</p><p>The difference between averaged and perpendicularpolarisation spectra can be rationalised by examining the possible ionisation channels and their symmetry. For this reason, a selection of Hartree-Fock uracil molecular orbitals is reported in Figure 4. In our calculations we included possible ionisation from HOMO, HOMO-1, HOMO-2 and HOMO-3 (see Table 1) which lie within 4 eV. With the energy provided by the laser pulse, the four ionisation channels can be activated. The net effect of the perpendicular pulse polarisation on the HHG spectrum is to produce high-energy harmonics, beyond H25. These harmonics are much more intense than those of the molecular-plane average spectrum, when present. They are also higher than the corresponding ones in the HHG spectrum along the permanent dipole. One could formulate the hypothesis that this feature is exclusively determined by the ionisation from orbitals with the proper π symmetry, say HOMO and HOMO-1, which are more sensitive to the perpendicular polarisation than HOMO-2 and HOMO-3. The interaction between HOMO (and/or HOMO-1) and a perpendicular pulse enhances the HHG signal at high energies. In other words, one could argue that ionization from HOMO and/or HOMO-1 is responsible for the appearance of the second plateau in the perpendicular-polarisation HHG spectrum. A confirmation is indirectly given by the presence of harmonics up to H39 in the 3D-average spectrum, though softened by the average along the perpendicular out-of-plane component ns,z . These harmonics are instead missing in the molecular-plane HHG spectrum. Looking at the molecular-plane spectrum, one can ask: since the out-of-plane component of the pulse is exactly zero, do the harmonics of the spectrum (cutoff at H29-H31) derive only from σ orbitals as HOMO-2 and/or HOMO-3, which have the good symmetry, or, anyway, HOMO and/or HOMO-1 still contribute to the harmonic generation, despite the "unfavorable" symmetry? We can not provide a complete answer using the present computational protocol: a detailed analysis of the role of the various ionisation channels during the strong-field dynamics is mandatory to dissect the possible contributions and interference effects. By summarizing, the HHG spectrum of uracil seems not to originate from HOMO ionisation only. Multiple ionisation channels, close in energy, seem to contribute to the HHG spectra. The cutoff of the 3D-averaged spectrum in Figure 3, H39-H41, is consistent with the experimental findings of H33 and H39.</p><p>In Figure 5 we disentangle the contribution of energy states below and above E T . We remind that E T value used in our simulations corresponds to minus HOMO-3 energy. In the left panel, we show the HHG spectrum averaged in the plane of the molecule (Equation 8), the same type of HHG spectrum limited to states below E T (Equation 11), and to states above E T (Equation 13). Below-and above-E T states contribute similarly to the full HHG spectrum. Above-E T states are seen to give slight higher H17-H21 harmonics, and generally provide a higher background of the below-E T . In the right panel of Figure 5 the same analysis is shown for the HHG spectra with perpendicular polarisation. Spectra were computed from the dipoles given in Equations 10 and 12, obtained by below-E T and above-E T states, respectively. In this case the states above E T play a role only for the harmonics belonging to the first plateau of the HHG spectrum, while the energy states below E T contribute everywhere and have a central role for the second plateau between H20 and H50. Differences between the above-E T and below-E T spectra are in this case dramatic, at variance with what occurs with the rotational average in the plane. The same behaviour is observed for the HHG spectrum polarised in the direction of the molecular permanent (in-plane) dipole (not shown). This could suggest that the behaviour observed for the HHG spectra in the left panel of Figure 5 is mainly due to the spatial average, which reduces differences between above-E T and below-E T states. The HHG spectrum averaged in 3D contains by construction the behaviour of both the spectrum averaged in the plane together and of the spectrum with perpendicular polarisation (not shown).</p><!><p>Results for thymine show the same general behaviour observed for uracil. This finding is not unexpected, since the electronic structure of uracil and thymine are rather similar. The HHG spectrum of thymine with a perpendicular pulse polarisation shows much higher harmonics in the high-energy region with respect to the averaged ones, as collected in Figure 6. As for uracil, a second plateau is observed at around H19-H33. When compared with uracil results, the difference for thymine is even more pronounced. By averaging the dipole moment only with respect to the molecule plane, one gets a HHG spectrum which is very similar to that obtained by a full rotational average (Figure 6) up to H19. Cutoff values for the average molecularplane and 3D spectra are H23 and H39-H41, respectively. The comments applied to uracil can be automatically done for thymine: Hartree-Fock molecular orbitals of thymine (Figure 7) have the same symmetry of the uracil ones. HOMO, HOMO-1, HOMO-2 and HOMO-3 are even closer in energy than the uracil orbitals. Hence, also for thymine different ionisation channels are supposed to play a significant role in the strong-field electron dynamics, with the same caveats reported for uracil. By using our all-electron quantum approach the HHG spectrum of thymine naturally arises, as expected. This result points out that the absence of HHG signal in the experimental work 47 does not originate from intrinsic features of the molecule, confirming that the issue has an extra-molecule explanation 47 . However, comparison between the HHG spectra in Figures 3 and 6 shows the thymine 3D-averaged high-energy signal is lower than that of thymine at high energies. Though all the ionisation energies (minus orbital energies, see Table 1) for thymine are smaller than those of uracil, one observes a larger cutoff for thymine, when the 3D average is considered: this finding could suggest that uracil and thymine are characterised by different couplings of the various ionisation channels, making the cutoff value a complicated function of individual ionisation energies. Contribution analysis for thymine is reported in Figure 8. The trend observed for uracil also characterizes the thymine spectra, with few major exceptions: i) HHG spectra from above-and below-E T are nearly identical for the average case; ii) contribution from below-E T states for the perpendicular pulse polarisation is considerably small in the middle-energy range. For both molecules, the smaller above-E T contribution to the HHG spectrum can be quantitatively explained by the smaller number of above-E T states described by the 3aug-ccpVDZ+3K basis set, with respect of that of below-E T states. It is worth mentioning that a simplified description of the electronic structure, as provided by only the ground and continuum states, is not accurate enough to account for the complex richness of the HHG signal of many-electron systems as uracil and thymine: accounting for bound, below-E T states is mandatory to successfully represent their HHG spectrum. However, a detailed investigation on the interplay between the chosen basis set and the ionisation energy could give a deeper insight on the general quality of the propagated wavepacket.</p><!><p>In this work he presented HHG spectra of uracil and thymine. HHG spectra have been computed using the all-electron RT-TD-CIS computational protocol based on the real-time propagation of the electron wavepacket, obtained using CIS and Gaussian basis sets adapted for the continuum. We simulated randomly oriented molecules by averaging the time-dependent dipoles projected on the different polarisations in the molecular plane or in the 3D space. We also computed the HHG spectra with a pulse polarisation perpendicular to the molecular plane or along the direction of the ground-state permanent dipole. This theoretical work contributed to the understanding and characterisation of the experiments by Hutchinson et al. 47 . We discussed the differences between the sample in the original work (unfragmented molecule and fragments) and in our calculations (unfragmented molecule). In the case of the 3D-averaged HHG spectrum our results can be directly compared with those in Ref. 47, and the computed uracil spectrum is consistent with experimental data. Differences between rotational-averaged and single-orientation HHG spectra indicate that the HHG signal can originate from more than one ionisation channel. We discussed the possible role of π and σ orbitals in modulating the shape of the HHG spectrum. Moreover, we obtain a clear HHG spectrum of thymine for all the different calculations schemes. Though less intense than the uracil HHG spectra at high energy, the thymine HHG spectra are certainly obtained. For this reason our work enforces the hypothe-sis given in Ref. 47 according to which the absence of the thymine HHG spectrum in the experiment is not caused by intrinsic features of the molecular system. Furthermore, for both systems the contributions from energy states below or above the ionisation from the HOMO-3 is generally more important for spectra from single polarisations. In conclusion, our approach has been seen to be accurate and reliable in describing HHG of biomolecules, and opens the way to predictive and interpretative studies of strong-field processes occurring in large molecules.</p><!><p>There are no conflicts to declare.</p>

ChemRxiv

Leveraging peptide substrate libraries to design inhibitors of bacterial Lon protease

Lon is a widely-conserved housekeeping protease found in all domains of life. Bacterial Lon is involved in the recovery from various types of stress, including tolerance to fluoroquinolone antibiotics, and is linked to pathogenesis in a number of organisms. However, detailed functional studies of Lon have been limited by the lack of selective, cell-permeant inhibitors. Here we describe the use of positional scanning libraries of hybrid peptide substrates to profile the primary sequence specificity of bacterial Lon. In addition to identifying optimal natural amino acid binding preferences, we identified several non-natural residues that were leveraged to develop optimal peptide substrates as well as a potent peptidic boronic acid inhibitor of Lon. Treatment of Escherichia coli with this inhibitor promotes UV-induced filamentation and reduces tolerance to ciprofloxacin, phenocopying established lon-deletion phenotypes. It is also non-toxic to mammalian cells due to its selectivity for Lon over the proteasome. Our results provide new insight into the primary substrate specificity of Lon and identify substrates and an inhibitor that will serve as useful tools for dissecting the diverse cellular functions of Lon.

leveraging_peptide_substrate_libraries_to_design_inhibitors_of_bacterial_lon_protease

INTRODUCTION<!>RESULTS AND DISCUSSION<!>HyCoSuL screens<!>Kinetic analysis of substrates<!>Kinetic analysis of inhibitors<!>Proteasome labeling<!>RAW cell viability<!>Bacterial strains and growth conditions<!>UV treatment and microscopy<!>Ciprofloxacin treatment and persister cell quantification<!>Software

<p>Lon is a widely-conserved housekeeping protease, found in bacteria, archaea, and eukaryotic mitochondria and chloroplasts1. All Lon orthologs feature a AAA+ ATPase domain that unfolds protein substrates and a proteolytic domain that catalyzes the hydrolysis of those substrates2. The importance of bacterial Lon has been determined mostly through studies using Escherichia coli lon mutants and via biochemical analyses of recombinant enzyme. Lon has myriad regulatory functions related to stress-response3,4, including roles in the SOS response to DNA damage5, defense against reactive oxygen species6, heat shock7, amino acid starvation8, and phage integration9. Phenotypic consequences of lon deletion include the inability to recover normally from UV-induced DNA damage and the reduced persistence of lon mutants following fluoroquinolone treatment10,11. In the context of pathogenesis, lon mutants of many bacteria are defective for infection. These include Pseudomonas aeruginosa in lung infection models of mice and rats12, Salmonella enterica in macrophages and systemic infection of mice13, and Brucella abortis in macrophages and spleen infections of mice14.</p><p>Due to its roles in stress-response, Lon is an interesting target for small-molecule inhibition. A selective inhibitor would enable dynamic studies of Lon proteolysis in a variety of physiological contexts and, based on the links between Lon and pathogenesis, has the potential to be useful as a therapeutic agent15. Furthermore, specific inhibition of Lon protease activity would allow separation of its proteolytic functions from those involving chaperone activity or binding of DNA and polyphosphate. For example, controlled inhibition of Lon would be useful for clarifying the role the protease plays in persistence. While a defect in fluoroquinolone tolerance in lon mutants has been established for many years16, there has been substantial debate about the mechanism by which Lon contributes to this phenomenon17,18. A recently-proposed model for the role of Lon in persistence which involves the degradation of toxin-antitoxin modules has since been disproven19–21. The current model involves regulation through degradation of the cell-division inhibitor SulA, the same mechanism by which Lon directs recovery from other sources of DNA damage. According to this model, Lon proteolytic activity would be important primarily when SulA is overexpressed as part of the SOS response. The ability to precisely control Lon inactivation (i.e., by addition of a small-molecule inhibitor) would be critical to test this hypothesis.</p><p>While a number of small-molecule Lon inhibitors have been identified, to our knowledge, none have been used to test the consequences of Lon inhibition within live bacterial cells. Lon features a serine-lysine dyad in its active site, notably different from the canonical serine-histidine-aspartic acid found in many serine proteases22,23. A likely consequence of its noncanonical active site is that many broad-spectrum serine protease inhibitors have poor activity against the enzyme. Early studies noted that E. coli Lon could be inhibited by the serine protease inhibitors diisopropyl fluorophosphate24 and dansyl fluoride,25 but only at mM concentrations. Inhibitors with slightly greater potency include 3,4-dichloroisocoumarin, other coumarin derivatives26, and oleanane triterpenoids27. Another class of Lon inhibitors comprises peptidic compounds that couple an amino acid recognition moiety with an electrophilic 'warhead' that covalently reacts with the active-site serine to inactivate the enzyme. Examples of such inhibitors with activity for Lon include Z-Gly-Leu-Phe-chloromethylketone28, as well as the human proteasome inhibitors MG13229,30 (featuring an aldehyde warhead), and MG262 and bortezomib (BZ)31,32 (featuring boronic acid warheads). In addition, a larger, hexapeptide boronic acid inhibitor of Lon was generated from the amino acid sequence of the natural λN Lon substrate33. These peptidic inhibitors take advantage of amino acid sequences that are tolerated by Lon, but none have been optimized for the enzyme, nor counter-screened for potential off-target binding or inhibition.</p><p>One of the most significant issues for current Lon inhibitors is their high level of cross reactivity with the proteasome. This leads to significant toxicity, making them ineffective as tools to study Lon function in cells. It is striking that many proteasome inhibitors have cross-reactivity with Lon, considering the differences in the active-sites: hydrolysis by the proteasome is catalyzed by an N-terminal threonine. Crystal structures of Meiothermus taiwanensis Lon revealed that the boronic acid warheads of MG262 and BZ bind covalently to the active-site serine, like their covalent modification of the threonine hydroxyl in the proteasome32. This strong covalent reactivity of boronates towards active site hydroxyls explains the dual potency for Lon and the proteasome.</p><p>We set out to develop selective inhibitors that could be used to specifically block Lon protease activity in cells. We hypothesized that the identification of highly selective peptide substrates could be leveraged to generate an optimized inhibitor using established electrophilic warheads. This strategy builds on a body of work from our groups and others describing the conversion of peptide substrates to inhibitors and activity-based probes for diverse protease targets34, including caspases35, cathepsins36, human neutrophil elastase37, human neutrophil serine protease 438, both the human39 and Plasmodium40,41 proteasome, and proteases important in Mycobacterium tuberculosis pathogensis42 and Zika virus infection43. By screening a large combinatorial library of peptide substrates, we identified a sequence of amino acids optimized for Lon. Based on this screening data, we designed a peptidic boronic acid inhibitor with potent activity for Lon and reduced potency for the human 20S proteasome. This compound was non-toxic to mammalian macrophages and is able to phenocopy classic lon deletion phenotypes in E. coli. We expect this compound to serve as a tool for studying the role of Lon-mediated proteolysis during stress response and pathogenesis.</p><!><p>The primary sequence specificity of Lon has been examined using individual fluorogenic peptide substrates28 as well as by identifying the cleavage sites for a number of endogenous Lon substrates44–46. However, there has not yet been a comprehensive and unbiased profiling of its amino acid preferences. We therefore performed a screen for fluorogenic peptide substrates using a hybrid combinatorial substrate library (HyCoSuL) that has been successfully applied to other protease targets37, 39, 47. Lon was purified after recombinant expression in E. coli (Figure S1). We chose to use libraries of tetrapeptides in which the P1 residue directly adjacent to the site of hydrolysis was fixed as a phenylalanine in order to ensure recognition by Lon. These libraries are made up of a set of sub-libraries in which 121 natural and non-natural amino acids are scanned through each of the P2 and P3 positions on the substrate (Figure 1a). Cleavage by Lon of each sub-library containing a fixed P2 or P3 residue is used to determine the overall specificity patterns at those positions. Initial analysis of the natural amino acid libraries provides some insight into the potential cleavage sites of native protein substrates. We found that at both the P2 and P3 positions, multiple residues are accepted, suggesting an overall broad specificity of the protease (Figure 1b). At each position, bulky or hydrophobic residues yielded the best substrates. This result is consistent with previous reports of favored peptide substrates that feature Ala, Leu, and Phe residues, and various analyses of the cleavage sites within protein substrates. In addition, these results support the model in which Lon is involved in degrading unfolded hydrophobic domains of endogenous substrates. It should be noted that, while information about peptide substrate preferences may be useful for determining preferred cleavage sites and cleavage rates of natural substrates, data from our peptide library screens cannot be used to determine preferences for protein substrates that are dictated by interactions with other domains of the enzyme (e.g., the substrate recognition domain).</p><p>We next performed substrate cleavage analysis using the libraries containing non-natural amino acids to get a broader perspective on the substrate specificity of Lon (Figure 1c–d; Table S1). Interestingly, Lon accepted a diverse array of non-natural amino acids at the P2 position, with more than 50% of the library exhibiting measurable cleavage. In contrast, it was more stringent at the P3 position and showed a strong preference for a single non-natural amino acid, L-homoarginine (hArg). For both positions, the most-preferred amino acids contained bulky side chains. To verify the results of the combinatorial library screening, we generated a set of fluorogenic tri-peptide substrates that contained the newly identified P3 hArg as well as the fixed P1 Phe and a morpholine acetate N-terminal cap. We then varied the P2 position using amino acids selected from the best substrates identified in the substrate screen (Figure 2a). For comparison, we used Mo-Leu-Leu-Phe-ACC (1), a peptide substrate containing only natural amino acids. This substrate yielded kinetic parameters similar to those previously reported for the Lon substrate, Glt-Ala-Ala-Phe-MNA28,48. In contrast, all of the substrates containing the P3 hArg greatly outperformed 1, with specificity constants (kcat/KM) as much as 12-fold higher for the best substrate, 5, which contains a neopentylglycine (nptGly) at the P2 position (Figure 2b, Figure S2a).</p><p>For many proteases, the P1 position adjacent to the scissile bond is critical for recognition of substrates. To evaluate the importance of this position in combination with the optimized hArg and nptGly residues, we generated a set of substrates featuring P1 amino acids found in endogenous Lon substrates: Ala, Val, Thr, Met, Leu44–46 (6–10, Figure 2a). Substrates with Ala (6), Val (7), and Thr (8) at the P1 position exhibited low cleavage rates while substrates with the bulkier amino acids Met (9) and Leu (10) had catalytic efficiencies similar to 5, with nearly 3-fold lower KM values (Figure 2b, Figure S2a). The decrease in activity observed for some P1 variants highlights the importance of this position for the design of efficient Lon substrates.</p><p>Having determined optimal substrates for Lon, we set out to use these scaffolds to build a potent, covalent inhibitor of Lon. The fact that several classes of covalent inhibitors have been reported suggests that the choice of electrophile is important for the optimal inhibitor design. We therefore screened our existing focused library of electrophilic protease inhibitors49 to identify an appropriate electrophile. This set of compounds includes diverse, reactive moieties that form permanent covalent bonds with active-site serine, threonine, or cysteine residues, including diphenyl phosphonates, vinyl sulfones, epoxy ketones, chloroisocoumarins, vinyl ketones, and triazole ureas. To screen this set of ~1,200 compounds we established an in vitro enzyme assay using our optimized fluorogenic peptide substrate 5. Our initial screen at a high concentration (10 μM) of the compounds identified a small number of hits that abolished Lon activity (Figure S3a–b). While we identified hits within all warhead classes, even the most potent compounds from the screen had IC50 values well above that of the human proteasome inhibitor BZ, which has previously been reported as an inhibitor of Lon (Figure S3c). We therefore decided to focus on using the reversible covalent boronic acid electrophile in BZ to make an optimized Lon inhibitor.</p><p>We suspected that converting any one of the Lon substrates to a boronic acid would yield a potent inhibitor. Because peptide boronic acids have been shown to be highly effective inhibitors of the human proteasome, counter screening for proteasome inhibition is essential to avoid high toxicity due to this cross-reactivity. To identify peptide scaffolds that would likely yield a selective Lon inhibitor, we evaluated cleavage of the substrates by both Lon and the human 20S proteasome (h20S). These results showed that the non-optimized substrate 1 containing the Leu-Leu-Phe sequence was cleaved equally effectively by both Lon and the proteasome while substrates containing the optimized P3 hArg were primarily cleaved by Lon and not the human proteasome. In fact, cleavage of 2–5 by the proteasome was so weak that it did not saturate and as a result, we were unable to determine kinetic parameters for those substrates (Figure S2b–c). In lieu of kinetic constants, we compared normalized cleavage rates for a fixed substrate concentration (Figure 2c). These results confirmed that substrates 2–10 were selective for Lon, with 5 being the most selective. This substrate showed essentially no detectable cleavage by the h20S. This result is consistent with a HyCoSuL screen of h20S that showed that peptides featuring hArg in the P3 position are poor substrates for the β1 and β5 subunits39.</p><p>To leverage the identified substrate specificity of Lon into the design of a selective inhibitor, we synthesized a hybrid compound containing the P1, P2, and P3 positions from substrate 10 combined with the boronic acid warhead and N-terminal pyrazinamide cap of BZ to generate Pyz-hArg-nptGly-Leu-B(OH)2 (11, Figure 3a). Though substrates 5 and 9 exhibited higher selectivity indices, we chose to use Leu at the P1 position (i) for the low KM observed for 10 which indicates tight binding to the Lon active site, (ii) for ease of comparison with BZ which also features a Leu in the P1 position, and (iii) for synthetic simplicity (i.e., Fmoc-Leu-boronate is commercially available). Both 11 and BZ exhibited potent, time-dependent inhibition of recombinant Lon (Figure S4a–b), with IC50 values after 60 min of inhibitor pre-incubation approaching the active-site concentration used in the assay (Figure S4c), suggesting covalent inhibition. Kinetic analyses (Figure 3b–c, Figure S4d–e) showed 11 to be a more potent Lon inhibitor than BZ with a two-fold higher kinact/KI driven primarily by improved potency (i.e., a lower KI value) (Figure 3d). To test for activity toward the human proteasome, we pre-treated purified h20S with each compound and then labeled subunit active sites with the fluorescent, activity-based probe MV151 (Figure 3e)50. Competition for active-site labeling of β1 and β5 subunits of h20S required a 10-fold higher concentration of 11 than BZ (50 vs 5 μM). Inhibition assays using fluorogenic peptides specific for each subunit similarly showed an increase in IC50 values for 11 compared to BZ for the β1 and β5 subunits (Table 1, Figure S4f). Surprisingly, we also saw some inhibition of the β2 subunit by 11, despite the strong preference of this "trypsin-like" subunit for Arg at the P1 position. Together these results confirm that the slight increase in Lon potency of 11 compared to BZ was accompanied by a substantial reduction in binding to the β1 and β5 subunits of the proteasome. More importantly, the Lon inhibitor 11 was not cytotoxic to murine macrophages at doses as high as 10 μM. This is in stark contrast to BZ which kills the same cells with an EC50 of 160 nM (Figure 3f). Thus, the drop in potency of 11 towards the proteasome is sufficient to eliminate the toxicity in mammalian cells and suggests that it should be a valuable new compound for use in cell biological studies of Lon function.</p><p>Although Lon plays important roles in stress response and pathogenesis, lon is a nonessential gene and deletion mutants grow normally in the absence of exogenous stress. We generated a clean-deletion of lon (Figure S5a–b) and found that neither genetic disruption of lon nor treatment with 100 μM 11 or BZ had an effect on exponential growth rates (Figure S5c). One of the first observed consequences of lon mutation in E. coli was the filamentation of cells after UV-induced DNA damage51. DNA damage causes upregulation of the cell-division inhibitor SulA as part of the SOS response. Lon-mediated degradation of SulA allows cells to resume division after recovery from stress. In the absence of Lon, SulA concentrations remain high and cells grow but cannot divide, resulting in extended filaments. We hypothesized that if 11 was a selective inhibitor of Lon then treatment of E. coli should phenocopy the eponymous "long" filamentation phenotype found in lon cells. As expected, outgrowth following UV-stress resulted in long filaments in the lon deletion strain but not in wild-type or sulA mutant cells (Figure 4a). Treatment with 11 during outgrowth following UV-stress led to a dose-dependent increase in filamentation (Figure 4b, Figure S6a). In a sulA mutant strain, 11 had no effect on filamentation, similar to observations of lon sulA double mutants52. Quantification of cell area for more than 500 cells per condition showed increases in the maximum cell area and in the percent of cells that were filamented (i.e., with area greater than 4 μm2, Figure 4c).</p><p>Lon is also implicated in recovery from DNA damage caused by fluoroquinolone antibiotics16, 53. Most cells treated with such antibiotics die, but a small subpopulation (typically 0.01% of the initial population) tolerate antibiotic exposure and can replicate after removal of the antibiotic. So-called persister cells54 are reduced in a lon knockout strain. Like UV-induced filamentation, Lon's role in persistence depends on the presence of SulA, with lon sulA double mutant strains producing a similar number of persisters as wild-type cells10, 17, 55. We therefore predicted that co-treatment of cells with 11 and ciprofloxacin would reduce persister cell formation. Neither lon deletion nor treatment of wild-type cells with 100 μm 11 altered the overall MIC of ciprofloxacin (0.0125 μg/ml). However, compared to wild type, we consistently observed a statistically-significant reduction in the fraction of cells that tolerated ciprofloxacin for both the lon mutant strain and wild-type cells treated with 11 in both rich (LB, Figure 5a) and minimal media (M9, Figure S6b). Importantly, this effect was abrogated in the sulA mutant strain, suggesting it results from inhibition of Lon. Furthermore, the effect was time-dependent, with both lon and 11-treated cells exhibiting faster death than wild type (Figure 5b). The effect was also concentration-dependent, with the extent of effect from 11 treatment matching that of the lon knockout strain at high concentrations (Figure 5c).</p><p>The SulA-dependent effects of 11 on UV-induced filamentation and ciprofloxacin persister formation strongly suggest that the compound inhibits Lon in cells. Incomplete phenocopying and the requirement for a high dose (e.g., 100 μM for cellular effects) are likely due to some combination of active efflux of the compound and permeability barriers. Both of these issues are common challenges for treating gram-negative bacteria with small molecules56. Encouragingly, there is evidence that other boronic acid inhibitors can enter E. coli cells57, 58, so we expect that modifications to increase the permeability of 11 will lead to further improved potency against live cells.</p><p>We leveraged amino acid preferences of Lon to develop both an improved fluorogenic substrate and a boronic acid inhibitor of Lon with increased selectivity over the proteasome. Our substrate screening results build on previous observations that Lon prefers to cleave peptides with bulky, hydrophobic residues, consistent with its role in degrading denatured proteins during stress responses. In our initial screen for inhibitors, Lon was poorly inhibited by electrophiles such as diphenyl phosphonates and chloroisocoumarins, which are potent inhibitors of many proteases with canonical serine-histidine-aspartic acid catalytic triads. This observation, along with the potency of proteasome inhibitors toward Lon, highlight the unusual nature of the serine-lysine dyad in its active site. Structural analyses of Lon inhibition by 11 would confirm the hypothesized covalent interaction with the active-site serine and would help to explain the structural basis for Lon's preferences for bulky amino acids and the role that hArg plays in enhancing substrate and inhibitor binding. In the future, novel Lon inhibitors may be identified by exploring alternative warheads such as β-lactams59, or nitriles60 which have activity toward serine-lysine dyads in signal peptidases and the UmuD family of proteases.</p><p>We expect 11 to be a useful compound for studying the roles that Lon plays in stress-response and pathogenesis. The use of a small molecule inhibitor rather than genetic disruptions (e.g., lon deletion or active-site mutants) introduces a level of dynamic flexibility to studies of Lon. Additionally, it provides a means to disentangle Lon's proteolytic activity from other functions of the multidomain complex, such as ATPase activity and its ability to bind and respond to DNA. In our cellular experiments, we observed 11-mediated effects on cellular physiology both when the compound was added during recovery from (outgrowth after UV exposure) or concurrent with (co-treatment with ciprofloxacin) stress. These observations suggest that Lon inhibition during or after stress has similar effects, at least for the SulA-mediated models of stress response tested here. Our data also show that 11 is not toxic to macrophages, meaning it can be used to test inhibition of bacterial Lon in cell culture models of infection and pathogenesis. Finally, because it is a covalent inhibitor, it can be converted to a fluorescent or otherwise affinity-labeled probe in order to visualize Lon activity within living cells. This compound should therefore greatly expand the scope of future studies of Lon function.</p><!><p>HyCoSuL screens were performed in Corning opaque 96-well plates. Each well contained 99 μl of Lon in assay buffer (250 mM Tris, pH 8.0, 1 M KCl, 100 mM MgCl2, and 1 mM ATP). Lon was added to a final hexamer concentration of 190 nM (P2 library) or 570 nM (P3 library). HyCoSuL substrates were added to a final concentration of 100 μM and kinetic fluorescence measurements (ex. 355 nm, em. 460 nm) were taken at 37 °C for at least 30 min starting immediately after substrate addition (Spectramax Gemini XPS, Molecular Devices). The substrate hydrolysis rate (RFU s−1) was calculated from the linear portion of each progress curve. The amino acid with the highest cleavage rate was set to 100%, and remaining amino acids were adjusted accordingly. Each library was screened twice and results are presented as mean values.</p><!><p>Lon and h20S substrate cleavage assays were performed in black 96- or 384-well plates. For Lon experiments, each well contained 25 μl 2X Lon assay buffer, 0.5 μl of 100 mM ATP (1 mM final concentration), and 40 nM final concentration of Lon hexamer. For ATP regeneration, 0.75 μl of 5 mg ml−1 creatine kinase (75 μg ml−1 final concentration) and 4 μl of 50 mM creatine phosphate (4 mM final concentration) were included. Water was added to a final volume of 40 μl. For h20S experiments, each well contained 25 μl 2X h20S buffer (100 mM Tris, pH 7.5, 200 mM NaCl), 1 mM DTT, 2 nM final concentration of h20S (BostonBiochem), 24 nM final concentration of PA28 (BostonBiochem), and water to a final volume of 40 μl. To begin the reaction, 10 μl of each substrate was added from a 5X stock, and fluorescence (ex. 360 nm, em. 460 nm) was measured every minute for 1 h at 37 °C in a microplate reader (BioTek Cytation 3).</p><!><p>Inhibition assays were performed under the same conditions as for substrate kinetics. Compounds were added from a 100X stock in DMSO (0.5 μl). For pre-incubation, compounds were added to the enzyme mixture in each well and plates were incubated at 37 °C for the indicated time. For experiments without pre-incubation, compounds were added to the working stock of substrate. Substrates (10 μl of 250 μM working stock) were added to the enzyme mixture and fluorescence (ex. 360 nm, em. 460 nm) was measured every minute for 1 h at 37 °C in a microplate reader (BioTek Cytation 3). For Lon, 5 was the substrate. For h20S, Z-LLE-AMC, Boc-LRR-AMC, and Suc-LLVY-AMC were substrates specific for the β1, β2, and β5 subunits, respectively. Inhibition data for the proteasome was determined following 60 min pre-incubation with the enzyme. Proteasome substrates were purchased from BostonBiochem.</p><!><p>For each compound, 1 μl of 20X stock in DMSO was added to a sample of h20S (10 nM) in 19 μl labeling buffer (50 mM Tris, pH 7.5, 5 mM MgCl2, 1 mM DTT) and incubated for 1 h at 37 °C. To label proteasome subunits, 0.5 μl of 80 μM MV151 (final concentration 2 μM) was added and incubated for an additional 2 h at 37 °C. Labeling was quenched by addition of 4X Laemmli sample buffer, samples were incubated for 5 min at 95 °C, and samples were separated by SDS-PAGE. MV151 fluorescence was imaged using a Typhoon 9410 Imager on the Cy3 channel (Amersham Biosciences).</p><!><p>RAW 264.7 murine macrophages were cultured in DMEM with 4.5 g l−1 glucose, 4 mM L- glutamine, and 10% v/v FBS (Invitrogen) at 37 °C with 5% CO2. Cells were split and seeded into a 96 well-plate to 5×103 cells per well with 50 μl of medium. To each well was added 49 μl of medium with 1 μl of 100X compound in DMSO (1% v/v final DMSO concentration). Cells were incubated with compound for 24 h then treated with 20 μl CellTiter-Blue (Promega) for 4 h. Cell viability was quantified by measuring fluorescence in a microplate reader (BioTek Cytation 3). Fluorescence values were normalized to untreated cells. Incubation with 1% v/v DMSO reduced cell viability compared to untreated cells, but the effect was independent of compound or dose.</p><!><p>Bacteria were cultured with shaking in LB (Fisher), 2xYT (Teknova), or M9 at 37°C, unless otherwise indicated. M9 contained 6 g l−1 Na2HPO4, 3 g/l KH2PO4, 1 g l−1 NH4Cl, 0.5 g l−1 NaCl, 0.5% w/v glucose, 1 mM MgSO4, 0.1 mM CaCl2, and 0.34 mg l−1 thiamine HCl. The lon mutant strain was generated by clean deletion of the coding region of lon using homologous recombination with CRISPR-Cas9 selection61. The sulA mutant strain (sulA773(del)::kan) was obtained from the Keio Collection62. Growth rates were determined by measuring OD600 of 100 μl cultures grown at 37°C in a 96-well plate in a microplate reader overnight.</p><!><p>Overnight cultures of wild-type, lon, or sulA strains were grown in LB. Cultures were diluted to OD600 0.1 in LB and grown for 1 h. Cells were pelleted by centrifugation (8,000 rcf for 5 min), resuspended in 0.1 volume of 10 mM MgSO4 and transferred to glass tubes. Cells were irradiated with 900 J cm−2 254 nm light (Stratagene Stratalinker 2400). Control cultures were resuspended in MgSO4 as above, but were not irradiated. Cells were diluted 1:25 into LB with compound added from 100X stock in DMSO (1% v/v final DMSO concentration) and grown for 6 h at 37 °C with shaking in the dark. For imaging, 4 μl of each culture was applied to 2% w/v agarose pads63. Phase contrast microscopy was performed on a Zeiss LSM700 confocal microscope with a Plan- Apochromat 63x/1.4 objective. Twenty-five images were captured via tile scan for each condition. Quantification of cell area was performed with the MicrobeJ64 plugin for ImageJ. Regions of interest containing at least 500 cells were analyzed using default settings for bacterial detection. A minimum cell area of 0.9 μm2 was used to exclude non-cellular debris.</p><!><p>For MIC measurements, overnight cultures of wild-type or lon strains were grown in LB, then diluted 1:50 into Mueller Hinton Broth 2 (Sigma). Diluted cultures (50 μl) were aliquoted into wells in a 96-well plate, each containing 50 μl medium and 2X the final concentration of ciprofloxacin. For persister experiments, overnight cultures of wild-type, lon, or sulA strains were grown in LB or M9. Cultures were diluted to OD600 0.01 in the same medium and incubated for 2 h. Cultures were treated with 10 μg ml−1 ciprofloxacin (Sigma) from a 100X stock in water and compound from a 100X stock in DMSO (1% v/v final DMSO concentration). Aliquots (100 μl) of each culture were removed at the indicated time, pelleted by centrifugation (8,000 rcf for 5 min), washed once with PBS, and resuspended in 100 μl PBS. Cells were serially diluted in PBS, 10 μL spots were spotted onto LB agar plates, and plates were incubated for 16–24 h at 37 °C. Colonies were counted to determine CFU.</p><!><p>Statistical analysis, fitting, and plotting were performed with Python v. 3.6.0, Scipy v. 1.1.0, Numpy v. 1.13.3, Matplotlib v. 3.0.3, and Seaborn v. 0.9.0. Microscopy data were analyzed in ImageJ. DNA sequence analysis was performed in SnapGene 4.3.10. Figures were assembled in Adobe Illustrator CS6.</p>

PubMed Author Manuscript

Preparation of a Sensor Based on Biomass Porous Carbon/Covalent-Organic Frame Composites for Pesticide Residues Detection

In this work, a covalent-organic framework with high carbon and nitrogen content microstructures (named COF-LZU1), assisted by 3D nitrogen-containing kenaf stem composites (represented as COF-LZU1/3D-KSCs), was constructed. Moreover, it was utilized for immobilizing acetylcholinesterase (AChE) for identifying trichlorfon, a commonly applied organophosphorus (OP) pesticide. The development of COF-LZU1/3D-KSC was affirmed by SEM, PXRD, and EDXS. The findings confirmed that COF-LZU1 microstructures were uniformly developed on 3D-KSC holes using a one-step synthesis approach, which can substantially enhance the effective surface area. Also, the COF-LZU1/3D-KSC composite contains not only the nitrogen element in COF-LZU1 but also the nitrogen element in 3D-KSC, which will greatly improve the biocompatibility of the material. The AChE/COF-LZU1/3D-KSC integrated electrode was fabricated by directly fixing a large amount of AChE on the composite. At the same time, the integrated electrode had good detection efficiency for trichlorfon. Improved stabilization, a wide-linear-range (0.2–19 ng/mL), and a lower detection limit (0.067 ng/mL) have been displayed by the sensor. Therefore, this sensor can be used as an important platform for the on-site detection of OP residue.

preparation_of_a_sensor_based_on_biomass_porous_carbon/covalent-organic_frame_composites_for_pestici

Introduction<!>Materials and Reagents<!>Instruments<!>Preparation of COF-LZU1/3D-KSCs Composite<!>Preparation of Integrated AChE/COF-LZU1/3D-KSC Electrodes<!><!>Inhibition Measurement of AChE Biosensor<!>Characterization of AChE/COF-LZU1/3D-KSC Electrodes<!><!>Electrochemical Behaviors of AChE/COF-LZU1/3D-KSC Electrodes<!><!>Influence of pH Value, ATCl, and AChE Concentration<!><!>Influence of pH Value, ATCl, and AChE Concentration<!>Voltammetric Detection of Trichlorfon<!><!>Precision, Stability, and Selectivity of Biosensor<!>Reactivity and Real Sample Analysis<!>Conclusion<!>Data Availability Statement<!>Author Contributions<!>Conflict of Interest<!><!>Supplementary Material<!>

<p>Organophosphorus pesticides (OPs), such as Trichlorfon, have been thoroughly used in agriculture due to their powerful insecticidal ability (Ma et al., 2018). However, due to its inhibition of acetylcholinesterase (AChE), the key enzyme of nerve conduction (Baldissera et al., 2019), it also poses a major threat to overall health (Soreq and Seidman, 2001; Shi et al., 2016b). Consequently, quick and sensitive probes of OPs in food production have turned out to be of considerable importance. Conventional analytical techniques, like HPLC and gas chromatography, mostly combined with mass-selective detectors (Liu et al., 2017; Song et al., 2019a), are slow and costly. These approaches are still carried out in laboratories (Shi et al., 2016a; Liu et al., 2017; Su et al., 2018) butthey are not appropriate for quick field detection (Su et al., 2014; Song et al., 2019b). Therefore, the fabrication of rapid and sensitive OPs detection strategies with fewer limitations are increasingly desired by the food industry and for environmental monitoring.</p><p>An electrochemical AChE biosensor has the potential to replace traditional methods due to its higher sensitivity, fast response, and tiny volume (Zeng et al., 2019). According to the inhibition of OPS on AChE, even smaller concentrations of pesticide can be determined accurately. The sensitiveness and limit of detection of such biosensors is dependent on the amount of enzyme (Zhang et al., 2019), so enzyme immobilization on the electrode surface is a key step for biosensor activity.</p><p>In order to firmly immobilize the enzyme, many smart materials, such as carbon nanotubes (Sotiropoulou and Chaniotakis, 2005) and gold nanoparticles, were employed to fabricate an enzyme-entrapped matrix (Wang et al., 2003; Shi et al., 2019). The incorporation of enzymes in new nanomaterials effectively increased the stability, sensibility, and detection threshold of enzymatic biosensors. Nevertheless, several proteases will stack on the surface of these nanomaterials, which will affect the transmission of electrons and reduce the performance of sensors (Khalilzadeh et al., 2016). Therefore, it is very important to discover an electrode material that can modify a large number of proteases without the stacking effect while maintaining good biocompatibility. Carbon biomass materials have good electrical conductivity and biocompatibility and are very suitable for the preparation of electrode materials for enzyme biosensors (Song et al., 2015; Khalilzadeh et al., 2016; Su et al., 2020). However, due to the large pore size of the biomass carbon material, the transmission of electrons will be affected. Therefore, for bioelectrochemical enzyme sensors, it is crucial to modify micro-materials with good conductivity and biocompatibility in the holes.</p><p>In this study, metal-free frame microstructures utilizing a covalent organic framework (named COF-LZU1) through the (3D N-containing kenaf stem) composites were formed using a one-step method. The COF-LZU1s may spread over the pores of 3D-KSC, and also had good biocompatibility because they contain no metal elements and only carbon, nitrogen, and oxygen elements. The COF-LZU1sshowed pitted surfaces, which, when superimposed with the 3D porous structure of KSC, can be employed to entrap more AChE molecules. Furthermore, AChE molecules were added in the COF-LZU1s by using pits of COFs, that efficiently prevented the agglomeration of enzymes at the electrode surface. Moreover, COF-LZU1s material also has good conductivity (Liu et al., 2016; Song et al., 2016a), which can improve the proton transportability of the whole integrated electrode. Thus, the developed trichlorfon sensor based upon the AChE/COF-LZU1/3D-KSC composites showed a wide-range linearity, lower detection limitations, and good stability.</p><!><p>The kenaf stems (KS) were collected from the Futian farm in Ji'an, Jiangxi Province. Graphite powder (99.95% and 325 mesh) and paraffin were acquired from Aladdin. DMFc, 1,3,5-triformylbenzene, acetylthiocholine chloride (ATCl), 1,4-diaminobenzene, and Acetylcholinesterase (1,000 U/mg), were obtained from Sigma-Aldrich (USA). Trichlorfon was bought from Kanghe Yinong Biotechnology Co., Ltd. Other reagents utilized were of analytical grades and procured from Shanghai Guoyao Group Chemical Reagent (China). Distilled water (18.2 MΩ cm) was employed for making all the solutions and purged by nitrogen prior to experiments. PBS was freshly made using dihydrogen phosphate and sodium disodium hydrogen phosphate.</p><!><p>Cyclic voltammetry (CVS) and differential pulse voltammetry (DPVS) were carried out on the CHI660E electrochemical analyzer. A three-electrodes system with a platinum wire (auxiliary electrode), a saturated calomel electrode (SCE) (reference electrode), and AChE/COF -LZU1/3D-KSCE was adopted as a working electrode. CVs and DPVs were carried out in 10 mL (0.2 M PBS of pH 7.0) under 25°C. SEM was done employing an XL30 ESEM-FEG SEM using accelerating voltage (20 kV) provided with a Phoenix (EDXA). The PXRD data was gathered over a (D/Max 2,500 V/PC) diffractometer via Cu Kα radiation info (λ = 0.154056 nm, 40 kV, and 200 mA).</p><!><p>The carbonization of dried KS synthesized the 3D-KSC in a high-heating furnace following protocol from our former project (Song et al., 2015; Khalilzadeh et al., 2016). The carbonization procedure was executed in a quartz reactor in an N2 environment on heating (5°C min−1) and annealing (2 h at 900°C). 3D-KSC was split up to a cylindrical shape, with the exterior diameter equivalent to the inner diameter of a used pipette tip. Therefore, the cylindrical 3D-KSC can be immobilized firmly in the already treated pipette tip. The prepared 3D-KSC was treated with dilute hydrochloric acid (2 M) for 24 h, and distilled water (24 h) to eliminate the inorganic contaminants, and afterwards was cleaned with ethanol and purified water successively, dried out normally, and placed in a beaker. Subsequently, 1,3,5-Triformylbenzene (0.30 mmol) and 1,4-dia-minobenzene (0.45 mmol) were measured and solubilized in 3 mL of 1,4-dioxane. After that 3D-KSCs were immersed in the solution, shifted in a glass vial (volume 20 mL), and then 0.6 mL of 3 mol L−1 dilute acetic acid was added to the mixture. The glass vial was flash-frozen in liquified nitrogen, subjected to a 19 mbar of internal pressure and flame-sealed, decreasing 10 cm in length. After attaining 25°C, the suspension was kept inside an oven uninterrupted for 3 days at 120°C, resulting in a yellow solid forming across the tube. The modified 3D-KSCs which were obtained after centrifugation were washed with N, N-dimethylformamide (3 × 10 mL) and tetrahydrofuran (3 × 10 mL), and then dried at 80°C in a vacuum for (2 h to produce COF-LZU1, a yellow-colored powder (90% yield), and produce the COF-LZU1-modified 3D-KSCs (COF-LZU1/3D-KSCs). Following our previous work,14,15 3D-KSC were developed by carbonizing dried KS in a higher heating system. The procedure is explained as follows: in a tubular quartz reactor, carbonization is conducted at a rate of 5°C min−1 in an N2 atmosphere, and annealing is carried out at 900°C for 2 h. The 3D-KSC is made into a cylinder such that the outer diameter corresponds to the inner diameter of a processed pipette tip so that the cylindrical 3d-ksc can be firmly fixed on the treated pipette tip. After treatment, the 3D-KSCs were treated with diluted hydrochloric acid (2 M) and distilled water for 24 h to eliminate inorganic impurities, and after that were washed alternately with ethanol and ultrapure water, then dried and placed in a beaker. Then, 1,3,5-trimethyl benzene (0.30 mmol) and 1,4-diaminobenzene (0.45 mmol) were put in vials and dissolved within 3 mL 1,4-dioxane. Then 3D-KSCs was immersed in the solution, the mixture transferred to a glass ampoule (vol 20 ml), and 0.6 ml of 3.0 mol L−1 water acetic acid was added to the mixture. The glass ampoules were quickly frozen in a liquified nitrogen bath, vacuumed to 19 mbar interior pressure, and flame-sealed to reduce the entire length by 10 cm. After the suspension was heated to room temperature, it was put in an oven at 120°C for 3 d. The yellow solid was generated along the test tube. The modified 3D-KSCs were separated by centrifugation, and afterwards washed with N, N-dimethylformamide (3 × 10 mL) and tetrahydrofuran (3 × 10 mL). After vacuum drying at (80°C) for 12 h, the yellow powder COF -LZU1 (90% yield) was obtained. The modified 3D-KSCs of COF -LZU1 (COF -LZU1/3D-KSC s) were obtained.</p><!><p>The COF-LZU1/3D-KSC were incorporated within the processed pippete tip. After that, 0.25 g liquid paraffin was mixed with 1 g powder of graphite and homogenized for 20 min in the agate mortar. Then, the mixture was packed inside the upper portion of the pipette tip to touch the base of COF-LZU1/3D-KSC. Then, a copper wire was inserted into the end of the pipette tip, and connected with the COF-LZU1/3D-KSC at the tip through graphite paste. After the paste was naturally dried at room temperature, as depicted in Figure S1, the copper wire was further fixed with a sealing film or epicote. The AChE/COF-LZU1/3D-KSC electrode was fabricated by dropping a 5 μL AChE solution with various concentrations upon the electrode surface, followed by dessication. The entire preparation process was illustrated by Figure S1, including Scheme 1. Lastly, the modified electrode was washed with purified water to eliminate loosely bounded materials and kept at 4°C, intended for further usage. The obtained AChE/COF-LZU1/3D-KSC electrode was denoted as AChE/COF-LZU1/3D-KSCE.</p><!><p>Schematic of the AChE/COF-LZU1/3D-KSC electrochemical pesticides biosensor.</p><!><p>The trichlorfon assay process was illustrated in detail in Scheme 1. Regarding inhibitory tests, the first (DPV) signal was (IP,control) recorded in 0.1 M PBS of pH 7 alongwith 1 mM ATCl. After that, the electrode was cleaned with distilled water, then placed inside an aqueous solution having a preferred concentration of trichlorfon for about 10 min. Afterwards, residual signal (IP,exp) was also observed in a similar state. The rate of trichlorfon inhibition was computed below:</p><p>In Scheme 1, the analysis process of trichlorfon was described in detail. For the inhibition test, the original (DPV) signal (IP,control) was determined in 0.1 M PBS (pH 7) and 1 mM ATCI. The electrodes were then cleaned by water and stored within an aqueous solution having the required amount of trichlorfon for 10 min. Following an incubation period, residual signals (IP,exp) were recorded under the same conditions. The inhibition rate of trichlorfon was estimated as below:</p><!><p>Figures 1A–C shows the SEM images of 3D-KSC and COF-LZU1/3D-KSC composites. 3D-KSC have a 3D macroporous inner structure (Figure 1A) (Song et al., 2015; Shan et al., 2019). After the growth of COF-LZU1, the procured electrode surface was adequately coated with the COF-LZU1 microstructures (Figures 1B,C). As depicted in Figure 1C, the spherical COF-LZU1 microstructures size is about 150 nm (inset of Figure 1C). The high magnification image in Figure 1C shows a special bumpy morphology, which significantly enhances the surface of electrode and the mass transfer. The EDX spectrums of COF-LZU1 and COF-LZU1/3D-KSC indicate the higher pureness of the composite, containing only O, N, and C (Figure 1D). Simultaneously, it can be seen that not only KSC but also COF-LZU1 contain nitrogen, which may greatly increase the biocompatibility of the composites. Figure 1F displays the XRD pattern of COF-LZU1 and also COF-LZU1/3D-KSC, which shows a microcrystalline solid with a long-range structure. Moreover, diffraction peaks around 4.9, 8.0, 9.4, and 12.1 according to 100, 110, 200, and 210 crystal planes, in accordance with the reported literature (Song et al., 2016b). Also, the XRD diffraction pattern of COF-LZU1/3D-KSC coincides with that of COF-LZU1, which indicates that COF-LZU1/3D-KSC has the same crystal structure as that of a single COF-LZU1 material. When AChE molecules were collected on the COF-LZU1/3D-KSC electrode, the interior wall of pores in the COF-LZU1/3D-KSC electrode became rough and irregular, comprising of an opaque film of a fuzzy-like material which might have resulted from the adsorption of AChE molecules at the inner side of pores (Figure 1E). Furthermore, the structure of AChE molecules could be damaged whenever an electron beam pierced the protein, and consequently, the pore surfaces of the COF-LZU1/3D-KSC composite became fuzzy. The findings clearly established the effective immobilization of AChE molecules on the COF-LZU1/3D-KSC electrode.</p><!><p>(A) SEM image of 3D-KSC and (B,C) SEM images of COF-LZU1/3D-KSC. (D) SEM images of AChE/ COF-LZU1/3D-KSC. (E) EDS curve of COF-LZU1(curve a) and COF-LZU1/3D-KSC (curve b). (F) XRD pattern of COF-LZU1(curve a) and COF-LZU1/3D-KSC (curve b).</p><!><p>For exploring the electrochemical characteristics of the AChE/COF-LZU1/3D-KSCE, the CVs of multiple electrodes, particularly AChE/COF-LZU1/3D-KSCE, AChE/3D-KSCE and AChE/glass carbon electrode (AChE/GCE), were investigated (Figures 2A–C). Figure 2A, curve a, showed the CVs of AChE/COF-LZU1/3D-KSCE in PBS (pH 7) having 1 mM ATCl. The CV of the AChE/3D-EUSE presented an irreversible oxidation peak on 0.68 V (curve a), resulting from thiocholine oxidation, the hydrolyzed material of ATCl, through enzyme catalysis. Contrarily, the maximal current with AChE/glass carbon electrode (AChE/GCE) (Figure 2C, curve a) was much lower. The increased response might be due to the stack effect of the COF-LZU1/3D-KSC composite, possessing a higher surface area that can immobilize additional enzymes. Meanwhile, COF-LZU1/3D-KSC also has the benefit of fast electron transfer owing to its 3D-porous composite structure. Additionally, the good biocompatibility of COF-LZU1/3D-KSC could well-preserve the highest bioaction by immobilized enzymes. Following 10 min placing in 9.0 ng/mL and 18 ng/mL trichlorfon solution, the anodic peak currents (curves b and c, Figure 2A) were significantly reduced compared to the control (curve a Figure 2A), and the reduction in peak current improved with the rising concentration of trichlorfon. It was because trichlorfon, an OP compound, displayed acute toxicity and produced an irreversible inhibitory response upon AChE, which therefore decreased enzymatic action to its substrate. However, the anodic peak currents of AChE/3D-KSCE (curves b and c, Figure 2B) and AChE/GCE (curves b and c, Figure 2C) decreased irregularly. This may be due to the decrease of enzyme modification, resulting in the narrowing of the detection range (Figure 2B) and could also be related to the fact that the glassy surface of the carbon electrode is smooth and the AChE can not be immobilized for a long time (Figure 2C). The trichlorfon concentration can be determined by changes in the voltammetric signal of the AChE/COF-LZU1/3D-KSCE. The principle of detection was pictorially illustrated by Scheme 1.</p><!><p>CVs of the (A) AChE/COF-LZU1/3D-KSC, (B) AChE/3D-KSC, and (C) AChE/GCE in PBS (pH 7.0) containing (1.0 mM ATCl) following 10 min incubation in 0.0 (curve a), 9.0 (curve b), and 18 ng/mL (curve c) trichlorfon solution.</p><!><p>Figure 3A showed the ampere sensitivity of AChE/COF-LZU1/3D-KSCE after adding ATCl. The usual biosensor current-time (response curve) was achieved when adding a substrate continuously in the stirred tank. By increasing ATCl concentration, the current response improved and tended to be stable at 1.0 mM. It might be due to the increase of ATCl concentration which leads to the active-sites' saturation of the enzyme by ATCl, thus reducing the binding sites of new molecules. The growth rate of peak current then shows a downward trend. Therefore, in the next pesticide analysis experiment, 1.0 mM ATCl was selected as the constant concentration.</p><!><p>(A) Correlation among current response to ATCl concentration in 0.1 M PBS of pH 7. (B) Plot of amperometric response vs. AChE concentration of AChE/COF-LZU1/3D-KSCE in 0.1 M PBS of pH 7 with 1 mM ATCl. (C) Effect of pH on current response of AChE/COF-LZU1/3D-KSCE to 1 mM ATCl. (D) Effect of inhibition time on inhibition (percentage) of AChE/COF-LZU1/3D-KSCE in 0.1 M PBS of pH 7 with 1 mM ATCl; inhibition of trichlorfon was 19 ng mL−1.</p><!><p>The immobilization of AChE upon the surface electrode is another essential factor affecting the biosensor efficiency. Figure 3B showed the relationship between AChE concentration and the biosensor (amperometric response). With an increasing AChE concentration, a gradual increase occured in peak current and attained the highest value at around 20 U mL−1. After this point, further addition of AChe will slowly weaken the current response. Its behavior may be ascribed to the presence of lesser AChE amounts, which is not enough to catalyze substrate oxidation, while too thick an AChE-modified layer may hinder mass and electron transfer, thus reducing the catalytic current. Therefore, this turning point may be due to the inhibition of COF-LZU1/3D-KSC to generate thiocholine and electron transfer by a large number of AChE. Therefore, in the next experiment, 20 U mL−1 AChE solution was used to build AChE/ COF-LZU1/3D-KSCE.</p><p>For electrochemical biosensors, pH value is the key factor that affects their stability and sensitiveness. Therefore, the influence of pH value was also studied. As shown in Figure 3C, at pH = 7, the maximum ampere response of AChE/COF-LZU1/3D-KSCE at 1 mM ATCl was obtained, which is consistent with most reported AChE biosensors (Ding et al., 2011; Su et al., 2016). Among the most influential parameters in pesticide assay is the culture time of inhibition. While increasing the incubation time period, there is also an increase in the rate of inhibition. Whereas, the required time of inhibition has been determined on various time-intervals varying from 02 to 60 min (Figure 3D). By prolonging the incubation timeframe, the rate of inhibition elevated to its highest value after incubating with trichlorfon 19 ng/mL for 10 min. Therefore, 10 min is used for the test.</p><!><p>In an optimized state, the inhibition stayed proportionate to varying concentrations of trichlorfon with 0.20–19 ng/mL (Figure 4), having a limit of detection (0.067 ng/mL). The effectiveness of AChE/COF-LZU1/3D-KSCE compared to other stated AChE biosensors is listed in Table 1, which indicated that the current AChE/COF-LZU1/3D-KSCE exhibited an equivalent or lower detection limit, demonstrating that COF-LZU1/3D-KSCE had multifunctions in enzyme immobilization. The high carbon content and nitrogen doted characteristics may sustain the enzymatic activity; moreover, the high specific surface-area with excellent electrical conducting potential of the composite would help a lot improvementn improving the sensitivity.</p><!><p>(A) The biosensor inhibition curve with varying concentrations of trichlorfon (inhibitions corresponded to trichlorfon concentrations of 0.8, 2.5, 5, 7, 9, 14, 19, 22, 35, 60, and 100 ng mL−1, respectively), in 0.1 M PBS with pH 7 consisting of 1 mM ATCl. (B) DPVs and (C) standard curve for trichlorfon assessment in 0.1 M PBS of pH 7 including 1 mM ATCl.</p><p>Comparative evaluation of various AChE biosensors' efficiency used for pesticide detection.</p><!><p>After trichlorfon (10 ng/mL) solution was added after around 10 min, the inter-assay precision of 1.0 mM ATCl was established on five distinct electrodes; the inter-assay precision was 3.9%, which proved that the precision and repeatability were good. The interference of several electro-active phenol derivatives (like nitrophenol, catechol, and hydroquinone) and the detection ofinorganic substances containing oxygen (SO42-, NO3-, sodium citrate) was also studied. As shown in Figure S2, when adding 2 times of nitrophenol, hydroquinone, catechol, SO42-, NO3-, and Na3C6H5O7 in determining trichlorfon (19 ng/mL), the inhibition behavior did not change significantly. The good selectivity of the electrode is confirmed and can be utilized for determining actual amounts of trichlorfon in samples. The enzymatic electrode is placed in 4°C in a dry environment unless used. During the first 5 d of storage, the reaction of ATCl did not decrease significantly. After 30 d of storage, the current response of the sensor was still maintained (94%) at the primary response (Figure S3).</p><!><p>Activation of AChE is a key additional component affecting the effectiveness of biosensors. Irreversible inhibiton of AChE by OPs could be fully activated by the use of nucleophilic agents like praldoximin chloride (PAM-Cl), while, 5 mM PAM-Cl PBS concentration was applied for activation. Then, the biosensor was dipped in the PAM-CL solution to inhibit trichlorfon. After 10 min of regeneration, AChE activity recovered completely. By reactivating the procedure, the biosensor can be reused up to five times with good constancy. The biosensor practicability was further proven by the addition of different amounts of trichlorfon to the Schisandra chinensis samples for the recovery test. Table S1 provides an overview of the results. The recovery was 96.1–105%. The results show that the method has high accuracy, high precision, and good reproducibility. It can be employed for the direct detection of associated samples.</p><!><p>During the present project, a stable and highly sensitive biosensor was fabricated using an AChE-modified COF-LZU1/3D-KSC composite, which makes it possible to detect even trace amounts (0.067 ng/mL) of an organophosphorus compound trichlorfon. The use of COF-LZU1/3D-KSC has significantly enhanced the biosensor efficiency in three ways: (1) COF-LZU1s and porous 3D-KSC provides a synergestic response due to the fully bumpy and hollow surface area that can firmly immobilize additional enzymes; (2) COF-LZU1s significantly improves electrical signaling due to fast electron transfer; and (3) Derived nitrogen elements from COF-LZU1 and 3D-KSC show that the higher bioactivities of the immobilized enzymes are also maintained. Due to these factors, the developed biosensor exhibited tremendously high sensitivity and lower-detection limits, and thus is more reliable to detect trace residues of OP pesticide compared to other AChE biosensors.</p><!><p>The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author/s.</p><!><p>YS conceived and designed the project. YL and MZ analysised experimental date and drafted the manuscript. CJ, JZ, and CH performed research. QS contributed methods and resources.</p><!><p>The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.</p><!><p>Funding. This research was financially supported by the National Natural Science Foundation of China (81860702), the Science and technology project of Jiangxi Provincial Department of Education (GJJ180650), Double First-class Discipline Construction Project Fund of Jiangxi Province (JXSYLXK-ZHYAO067, 069, 070, and 106), Foundation for Doctoral Research Initiation of Jiangxi University of Traditional Chinese Medicine (2018WBZR014), and the Research Fund of Zhangjiagang first people's Hospital (ZKY201852).</p><!><p>The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fchem.2020.00643/full#supplementary-material</p><!><p>Click here for additional data file.</p><p>Click here for additional data file.</p><p>Click here for additional data file.</p><p>Click here for additional data file.</p>

PubMed Open Access

An Expedient Route to 9-arylmethylanthracene Derivatives via Tandem Ni-catalyzed Alkene Dicarbofunctionalization and Acid-promoted Cyclization-aromatization

We report a nickel-catalyzed one pot synthesis of 9-arylmethylanthracene motifs, which find applications in medicinal and material chemistry. In this synthesis, we apply three component alkene dicarbofunctionalization of 2-vinylaldimines with aryl iodides and arylzinc reagent to generate a 1,1,2-diarylethyl scaffold, which then undergoes an acidpromoted cyclization followed by aromatization to furnish 9-arylmethylanthracene cores. With the new method, a number of differently-substituted 9-arylmethylanthracene derivatives can be synthesized in good yields.

an_expedient_route_to_9-arylmethylanthracene_derivatives_via_tandem_ni-catalyzed_alkene_dicarbofunct

<p>Anthracene, a polycyclic aromatic hydrocarbon that exhibits blue fluorescence under ultraviolet light, plays an important role as electron-transporting material.[1] Anthracene derivatives have been extensively studied in multiple fields such as material chemistry, thermochromic or photochromic chemistry and as starting materials in organic light-emitting materials.[2] Literatures have shown that anthracene derivatives are also useful for examining DNA cleavage acting as a potential anticancer drug.[3] Due to their chemical stability and excellent photoluminescence property, they have been widely used in the development of fluorescence sensors, which can be applied to function as selective imaging agents.[3–4] These molecular scaffolds are also known to show antitubercular activity[5] (Figure 1).</p><p>Although several synthetic approaches have been known for the synthesis of anthracenes,[6] methods to prepare 9-substituted anthracene derivatives are still limited. An acid catalyzed synthesis of 9-arylanthracene have been reported by Meier et. al.[7] in which the methoxy group in the aromatic ring plays an important role to perform the intramolecular electrophilic aromatic substitution (Scheme 1). In addition, Ye et. al. have described the synthesis of alkyl and 9-aryl anthracene derivatives through gold-catalyzed cyclization of o-alkynyldiarylmethane.[8] In a separate report by Majumdar et. al.[9] describes the synthesis of 9-benzyl and related 9-allenyl substituted anthracene derivative employing phase transfer catalyst and an anthrone as a starting material. Mou et. al. report the synthesis of anthracene derivatives using Negishi coupling of aryl halides with organozinc chlorides catalyzed by a palladium bipyridyl complex.[10] Herein, we report an efficient one pot synthesis of 9-arylmethylanthracene from vinylarenes, aryl iodides and arylzinc reagents. In this process, we utilize a Ni-catalyzed alkene dicarbofunctionalization reaction[11] in tandem with an acid-promoted cyclization-aromatization process to construct the 9-arylmethylanthracene scaffold.</p><p>Transition-metal catalyzed three-component difunctionalization of alkenes[12] involving the addition of two carbon-based entities across a double bond, known as alkene dicarbofunctionalization, is an efficient technique to construct complex molecular architectures from the readily available feedstock chemicals.[13] A significant progress has been made in this area recently.[14] However, the application of this new method in the construction of synthetically useful complex molecules remains limited.[14i,15] Recently, we reported a Ni-catalyzed diarylation of 2-vinylaldimines in which two carbon–carbon (C–C) bonds were constructed across the styryl double bond (Scheme 2).[11] This method furnished a variety of differently substituted 1,1,2-triarylethane scaffolds.</p><p>In our continued efforts to expand the application of alkene difunctionalization to generate synthetically important compounds, we envisioned that 9-arylmethylanthracene derivatives could be readily accessed from the 1,1,2-triarylethane scaffold (Scheme 3). In this process, we anticipated that the 2-aldimine group in the diarylated product would undergo an acidpromoted cyclization with the ortho-position of the proximal aryl group if this aryl group could be made sufficiently electron-rich followed by the elimination of aniline prompted by the presence of the acid.</p><p>We initially examined the difunctionalization of 2-vinylaldimine 1 with iodobenzene and 3-methoxylphenylzinc iodide in the presence of 2 mol% Ni(cod)2 under our previously reported reaction conditions (Table 1). The crude reaction mixture, without further purification, was then subjected to cyclization in the presence of HCl. The reaction furnished 9-benzyl-2-methoxyanthracene 2 in 67% yield when the crude product was cyclized with 0.50 mL of 6.0 M HCl (3.0 mmol) (entry 1). Cylization with lower concentrations of HCl or in shorter reaction time generated the anthracene product 2 in lower yields (entries 2–4). Organic acids, such as p-toluene-sulfonic acid, trifluoromethanesulfonic acid and acetic acid, furnished the anthracene product 2 in slightly lower yields (entries 5–7).</p><p>After optimizing the reaction conditions for cyclization, we examined the scope of the current reaction with respect to 2-vinylaldimines such as N-phenyl-1-(2-vinylphenyl)methanimine (1), and 1-(4-methoxy-2-vinylphenyl)-N-phenylmethanimine1-(5-fluoro-2-vinylphenyl)-N-phenylmethanimine), along with 3-methoxyphenylzinc and 3,4-dimethoxyphenylzinc iodide, and different aryl iodides (Table 2). The reaction proceeded well with both electron-rich and electron-poor aryl iodides, and tolerated substituents such as alkyl, OMe, Cl, CO2Me, CN, SMe and CF3 on aryl iodides. In addition, the reaction also tolerated ortho-substitution on aryl iodides as demonstrated by 1-naphthyl (7) and ortho-iPr (8). The reaction also tolerated F and OMe groups on vinylaldimine. Examination of reaction with 3,4-dimethoxyphenylzinc iodide (6, 8–11) showed better product yield than the phenylzinc iodide containing one OMe group at meta-position (2–5 and 7) most probably due to the activation of aryl ring for electrophilic cyclization on the imine group.[16] The current reaction produced variously substituted 9-arylmethylanthracene derivatives, the structures of which were further confirmed by an X-ray crystallographic analysis of the anthracene product 6 (Figure 2).</p><p>We propose the following catalytic cycle (Scheme 4), which involves both the Ni-catalyzed alkene dicarbofunctionalization and the acid-catalyzed cyclization-aromatization reactions, to account for the formation of the 9-arylmethylanthracene derivatives with the current method. 2-Vinylaldimine first binds to a Ni(0) species, which oxidatively adds aryl iodide to generate a Ni(II) intermediate 13. The aryl group in species 13 then undergoes migratory insertion to the bound alkene to generate the nickellacycle 15. The intermediate 15 subsequently undergoes transmetalation followed by reductive elimination to furnish the diarylated product 16 and regenerate the Ni(0)-catalyst. The crude product 16 undergoes further cyclization upon the imine group in the presence of HCl to furnish the final 9-arylmethylanthrancene derivative (Scheme 5). The cyclization occurs with the electron-rich aryl group originally derived from arylzinc iodide, which proceeds sequentially through protonation, cyclization and re-aromatization steps to re-aromatize the OMe containing ring as well as create a new central aryl ring of the anthracene core.</p><p>In summary, we have developed a new catalytic protocol for the synthesis of a wide range of 9-arylmethylanthracene derivatives. In this process, we implement a Ni-catalyzed alkene dicarbofunctionalization reaction to assemble all the carbon fragments required to generate the final product in the form of 1,1,2-triarylethyl scaffold. This scaffold is then readily converted to the 9-arylmethylanthracene core upon cyclization of the proximal electron-rich aryl ring onto the imine group. The cyclization process is followed by a re-aromatization event, which constructs the central aryl ring of the anthracene core.</p>

PubMed Author Manuscript

Phytochemical Analysis and Antioxidant Property of Leaf Extracts of Vitex doniana and Mucuna pruriens

Oxidative stress and impaired antioxidant system have been implicated in the pathophysiology of diverse disease states. The phytochemical screening and antioxidant property of fresh leaves of Vitex doniana and Mucuna pruriens, used in the management and treatment of various diseases, were studied. The extracts (ethanol and distilled water) were screened for the presence of phytochemicals, and their inhibition of 2,2-diphenyl-1-picryl-hydrazyl (DPPH) radical was used to evaluate their free radical scavenging activity. Liver levels of malondialdehyde (MDA), superoxide dismutase (SOD), and catalase (CAT) in carbon tetrachloride- (CCl4) treated albino rats were also used to assess the antioxidant activity of the extracts. The animals were treated with 250 mg/kg body weight of the extracts for six consecutive days before a single dose (2.5 mL/kg body weight) of CCl4. Vitamin C was used as the standard antioxidant. Phytochemical screening revealed the presence of saponins, tannins, anthraquinones, terpenoids, and flavonoids in all the extracts, while alkaloids were detected in extracts of Vitex doniana only, and cardiac glycosides occurred in extracts of Mucuna pruriens only. All the extracts inhibited DPPH radical in a concentration-dependent manner, water extract of Vitex doniana producing highest inhibition which was not significantly different (P > .05) from vitamin C. The extracts produced a significant decrease (P < .05) in liver MDA, while the levels of SOD and CAT significantly increased (P < .05) relative to the positive control. These results are an indication of antioxidant potential of the extracts and may be responsible for some of the therapeutic uses of these plants.

phytochemical_analysis_and_antioxidant_property_of_leaf_extracts_of_vitex_doniana_and_mucuna_prurien

1. Introduction<!>2.1. Collection of Plant Leaves<!>2.2. Extraction of Leaves Material<!>2.3. Phytochemical Screening<!>2.4. Measurement of Antioxidant Property<!><!>2.5. Data Analysis<!>3. Results and Discussion<!>4. Conclusion<!>

<p>The use of plants in the management and treatment of diseases started with life. In more recent years, with considerable research, it has been found that many plants do indeed have medicinal values [1]. Some medicinal plants used in Nigeria include Garcina kola, used in the treatment of asthma, Carica papaya, used as a remedy for hypertension, Ocimum basilicum, a cure for typhoid fever, and Cola nitida, for treatment of pile [2]. Vitex doniana (Verbenaceae), commonly called black plum, is widely distributed in the eastern and western parts of Nigeria. Various parts of the plant are used by traditional medicine practitioners in Nigeria in the management and treatment of several disorders which include rheumatism, hypertension, cancer, and inflammatory diseases [1]. Mucuna pruriens (Fabaceae) also called velvet bean is found in Eastern Nigeria, where its seeds are used as soup thickeners. The leaves of Mucuna pruriens are used as remedy for various diseases such as diabetes, arthritis, dysentery, and cardiovascular diseases [3]. Phytochemicals are bioactive compounds found in plants that work with nutrients and dietary fibre to protect against diseases. They are nonnutritive compounds (secondary metabolites) that contribute to flavour colour [4, 5]. Many phytochemicals have antioxidant activity and reduce the risk of many diseases, for example, alkyl sulfide (found in onions and garlic), carotenoids (from carrots), and flavonoids (present in fruits and vegetables) [5]. Reactive oxygen-free radicals (ROS) have been implicated in many diseases and in aging process. These free radicals, which cause tissue damage via oxidative stress, are generated by aerobic respiration, inflammation, and lipid peroxidation. Antioxidant systems minimize or prevent deleterious effects of the ROS [6].</p><p>Lipid peroxidation is an established mechanism of cellular injury and is used as an indicator of oxidative stress. Polyunsaturated fatty acids peroxides generate malondialdehyde (MDA) and 4-hydroxyalkanals upon decomposition [7]. Superoxide dismutase (SOD) decomposes superoxide anion into hydrogen peroxide and oxygen at very high rates. Superoxide radical is involved in diverse physiological and pathophysiological processes [8]. Catalase (CAT) is an antioxidant enzyme ubiquitously present in aerobic cells. It catalyses the decomposition of hydrogen peroxide to water and oxygen. High concentration of hydrogen peroxide is deleterious to cells, and its accumulation causes oxidation of cellular targets such as DNA, proteins, and lipids, leading to mutagenesis and cell death [9].</p><p>The medicinal applications of Vitex doniana and Mucuna pruriens have not been given a scientific base. The present study investigates the phytochemical constituents and antioxidant property of the plants.</p><!><p>Fresh leaves of the plants were collected in June, 2010 from a village in Abakaliki of Ebonyi state, Nigeria. They were identified by Professor S.C Onyekwelu of Biology Department, Ebonyi State University, Abakaliki, Nigeria. The leaves were washed, with distilled water, and used immediately.</p><!><p>The extraction methods described by Agbafor [10] were adopted using distilled water and ethanol separately. The local users make use of water or alcoholic drinks for their extractions. After extraction, the solvents were removed using rotary evaporator, to get gel-like extracts.</p><!><p>The methods of Harbone [11] and Trease and Evans [12] were used to identify the following phytochemicals in the extracts: alkaloids, saponins, tannins, anthraquinones, flavonoids, terpenoids and cardiac glycosides.</p><!><p>The antioxidant activity of the extracts was studied in two ways:</p><p>(i) Slightly modified method of Brand-Williams et al. [13] using Vitamin C (Emzor Pharmaceutical Industries, Nigeria) as a reference antioxidant. Here, the free radical scavenging properties of the extracts against 2,2-diphenyl-1-picryl hydrazyl (DPPH) radical were measured at 517 nm, as an index to their antioxidant activity. The concentrations of the extracts and Vitamin C used were 1.0, 2.0, 4.0, 6.0, 8.0 and 10.0 mg/mL. Free radical scavenging activity was obtained as; (1)%  inhibition=  Ab−AtAb×100. A b: absorbance of blank, and A t: absorbance of test.</p><p>Values were obtained in triplicates.</p><!><p>Animals and Handling. Twenty-eight adult male albino rats, weighing 102–120 g, were brought from the animal house of Biochemistry Department, University of Nigeria, Nsukka, Nigeria. They were placed in seven groups (A–G) of four rats in each group and kept in animals house of Biochemistry Department, Ebonyi State University Abakaliki for seven days to acclimatize. All the rats were allowed free access to feed (rat chaw) and water before and throughout the experiment.</p><p>Animal Groups and Treatments. Solutions of the extracts were made with distilled water. Dose of 250 mg/kg body weight of the extracts and 20 mg/kg body weight of vitamin C (Emzor Pharmaceutical Industries, Nigeria) were given orally to groups A–D and E, respectively, while F and G received distilled water for six consecutive days.</p><p>Inducement of Liver Damage. On the seventh day, groups A–F were treated with a single dose of 2.5 mL/kg body weight of CCl4 and olive oil (1 : 1) intraperitoneally. Group G was given distilled water/olive oil (1 : 1).</p><p>Collection of Samples from the Animals. Blood samples were collected from the animals following an overnight fast through cardiac puncture under mild anaesthesia using diethylether. The samples were put into specimen bottles without anticoagulant. Liver was also quickly excised, perfused with cold normal saline, and homogenized in 0.25 M sucrose in phosphate buffer (0.2 M, pH 7.4).</p><!><p>Statistical analysis was done using analysis of variance (ANOVA). Means were compared for significance using Duncan's multiple range test (P < .05) [17].</p><!><p>Table 1 shows the results of phytochemical analysis of the four extracts. Saponins, tannins, anthraquinones, terpenoids, and flavonoids were found in all the extracts. Alkaloids were detected in extracts of Vitex doniana only, while cardiac glycosides were also present in extracts of Mucuna pruriens only. The medicinal values of the plant leaves may be related to their constituent phytochemicals. According to Varadarajan et al. [18], the secondary metabolites (phytochemicals) and other chemical constituents of medicinal plants account for their medicinal value. For example, saponins are glycosides of both triterpene and steroids having hypotensive and cardiodepressant properties [19], while anthraquinones posses astringent, purgative, anti-inflammatory, moderate antitumor, and bactericidal effects [20]. Cardiac glycosides are naturally cardioactive drugs used in the treatment of congestive heart failure and cardiac arrhythmia [21].</p><p>Percentage inhibition of DPPH is presented in Table 2. All the extracts inhibited DPPH, indicating their antioxidant activity. The percentage inhibition produced by the water extract of Vitex doniana did not show a significant difference (P > .05) from those of vitamin C, the standard antioxidant. On the other hand, the inhibitions shown by the other extracts were significantly lower (P < .05) than their corresponding values for vitamin C. The inhibition produced by water extract of Vitex doniana was higher than that of its ethanol extract while the reverse is the case for Mucuna pruriens. All the extracts showed concentration-dependent inhibition.</p><p>The DPPH test provides information on the reactivity of compounds with a stable free radical DPPH that gives a strong absorption band at 517 nm in visible region. When the odd electron becomes paired off in the presence of a free radical scavenger the absorption reduces and the DPPH solution is decolorized as the colour changes from deep violet to light yellow. The degree of reduction in absorbance is reflective of the radical scavenging (antioxidant) power of the compound(s) [13].</p><p>Results of the effect of the extracts on liver concentrations of MDA, SOD, and CAT are presented in Table 3. There was a significant (P < .05) increase in MDA levels and decrease in SOD and CAT activities of group F, treated with CCl4 only relative to the untreated control group. This reflects hepatotoxicity of CCl4, as observed by Singh et al. [22]. The results were reversed on pretreatment with the leaf extracts or vitamin C. The MDA concentration of the pretreated groups was significantly lower (P < .05) than the untreated. On the hand, the activities of SOD and CAT were significantly higher (P < .05) in the pretreated groups than in the positive control. These observations are indicative of antioxidant property of the extracts.</p><p>Free radical damage and oxidative stress are the major reasons for liver tissue damage. The antioxidant enzymes are the first-line defense against such damage and thus provide protection against the deteriorating outcome [23]. Oxidative injury and lipid peroxidation can be monitored by measuring liver MDA. Lipid peroxidation is regarded as one of the basic mechanisms of tissue damage caused by free radicals [24, 25].</p><p>The antioxidant activity of the extracts may be attributed to the presence of the identified phytochemicals. Flavonoids and tannins are phenolic compounds, and plant phenolics are a major group of compounds that act as primary antioxidants or free radical scavengers [26]. Similarly, terpenoids, as vitamins, act as regulators of metabolism and play a protective role as antioxidants [27].</p><p>The antioxidant property of the extracts may be a strong contributing factor to the applications of the plants in the management and treatment of various diseases. Antioxidants prevent oxidative stress, caused by free radicals, which damage cells and vital biomolecules. They terminate chain reactions triggered by free radicals by removing free radical intermediates and inhibit other oxidation reactions [28].</p><p>These effects of the extracts on liver MDA, SOD and CAT were maximum in the group treated with water extract of Vitex doniana. The effect of water extract of Vitex doniana was comparable with that of vitamin C. These observations are consistent with the pattern of inhibition of DPPH by the extracts.</p><!><p>The presence of the identified phytochemicals makes the leaves pharmacologically active. Their antioxidant activity may be responsible for their usefulness in the management and treatment of various diseases. We are currently studying other possible mechanisms of action of these leaves. Efforts to identify the constituent compounds responsible for this antioxidant activity are also in progress.</p><!><p>Phytochemical composition of water and ethanol leaf extracts of Vitex doniana and Mucuna pruriens.</p><p>+: present, −: absent.</p><p>Percentage inhibition of DPPH by the extracts and vitamins C.</p><p>Values are mean ± SD, n = 3, WVD: water extract of Vitex doniana, EVD: ethanol extract of Vitex doniana, WMP: water extract of Mucuna pruriens, and EMP: ethanol extract of Mucuna pruriens.</p><p>Liver MDA, SOD and CAT levels of the animals after treatment.</p><p>Values are mean ± SD, n = 4.</p>

PubMed Open Access

Investigations into Transition Metal Catalyzed Arene Trifluoromethylation Reactions

Trifluoromethyl-substituted arenes and heteroarenes are widely prevalent in pharmaceuticals and agrochemicals. As a result, the development of practical methods for the formation of aryl\xe2\x80\x93CF3 bonds has become an active field of research. Over the past five years, transition metal catalyzed cross-coupling between aryl\xe2\x80\x93X (X = halide, organometallic, or H) and various \xe2\x80\x9cCF3\xe2\x80\x9d reagents has emerged as a particularly exciting approach for generating aryl\xe2\x80\x93CF3 bonds. Despite many recent advances in this area, current methods generally suffer from limitations such as poor generality, harsh reaction conditions, the requirement for stoichiometric quantities of metals, and/or the use of costly CF3 sources. This Account describes our recent efforts to address some of these challenges by: (1) developing aryl trifluoromethylation reactions involving high oxidation state Pd intermediates, (2) exploiting AgCF3 for C\xe2\x80\x93H trifluoromethylation, and (3) achieving Cu-catalyzed trifluoromethylation with photogenerated CF3\xe2\x80\xa2.

investigations_into_transition_metal_catalyzed_arene_trifluoromethylation_reactions

Introduction<!>Part 1. Aryl trifluoromethylation via high valent palladium<!>Part 2: Aryl trifluoromethylation using AgCF3<!>Part 3. Cu-catalyzed aryl trifluoromethylation with CF3\xe2\x80\xa2<!>Outlook

<p>Trifluoromethyl-arenes and heteroarenes are increasingly important structural features of pharmaceuticals and agrochemicals. The incorporation of a trifluoromethyl group into an organic molecule can dramatically impact a variety of properties, including metabolic stability, lipophilicity, and bioavailability.1 Despite the significance of this functional group in medicinal chemistry, mild, efficient, and functional-group tolerant methods for the formation of aryl/heteroaryl–CF3 linkages have been limited until very recently.1f,2</p><p>On the industrial scale, trifluoromethylated arenes are mainly produced by the Swarts reaction, which was developed in 1892.3 This transformation involves a two-step conversion of toluene derivatives to benzotrifluorides via: (1) radical chlorination followed by (2) treatment with an inorganic fluoride (e.g. SbF5) or anhydrous hydrogen fluoride (eq. 1).3 The requirement for reactive fluorinating reagents and high temperatures render this strategy incompatible with many common functional groups. Thus, the development of mild and flexible alternative methods for the installation of CF3 groups, particularly at late stages in the synthesis of complex molecules, is highly desirable.</p><p>This Account describes our efforts in methods development and mechanistic investigations of transition metal-mediated aromatic trifluoromethylation reactions. When we initiated our work in this area in 2009, three groups had just reported exciting advances in Pd- and Cu-promoted arene trifluoromethylation reactions. For example, in 2006, Grushin demonstrated that (Xantphos)Pd(Ph)(CF3) undergoes stoichiometric Ph–CF3 bond-forming reductive elimination to release trifluorotoluene under mild conditions (80 °C, 3 h, eq. 2).4 This was the first reported example of selective aryl–CF3 coupling from a Pd center. The properties of the Xantphos ligand (particularly its large bite angle) were hypothesized to play an important role in this novel transformation.</p><p>A major advance in the area of Cu-promoted trifluoromethylation was made in 2008, when Vicic reported the first example of an isolable, crystallographically characterized CuI–CF3 complex (eq. 3).5 This complex, which is supported by an N-heterocyclic carbene ligand, was shown to react with aryl iodides under mild conditions (25 °C, 112 h) to liberate trifluoromethylated products (eq. 3). While related Cu-mediated trifluoromethylations were known prior to this report,1f these previous systems generally involved ill-defined "Cu-CF3" intermediates.</p><p>A final significant advance that occurred just prior to our entry into the field was a 2009 report by Amii.6 This work demonstrated the first copper catalyzed trifluoromethylation of aryl iodides. As shown in eq. 4, 1,10-phenanthroline (phen) was used as a supporting ligand for Cu in conjunction with TMSCF3 as the CF3 source. A variety of electron deficient aryl iodides underwent trifluoromethylation under these conditions.</p><p>Our goal was to build on these exciting advances by developing new mechanistic pathways for metal-mediated aryl–CF3 coupling reactions. Over the last 3 years, we have pursued three different strategies to achieve this goal (Scheme 1). Strategy 1 involves aryl trifluoromethylation via reductive elimination from high valent PdIV(aryl)(CF3) intermediates. Strategy 2 involves exploiting AgCF3 intermediates to achieve aryl–CF3 bond formation. Finally, strategy 3 involves the trifluoromethylation of aryl-Cu intermediates with CF3•. All three of these approaches are described in detail below.</p><!><p>Historically, it has proven challenging to achieve aryl–CF3 bond-forming reductive elimination from PdII centers. Only two examples of this transformation have been reported in the literature, and both involve the use of specialized phosphine ligands to induce the desired reactivity. As described above, Grushin reported Ph–CF3 coupling from (Xantphos)PdII(Ph)(CF3) in 2006 (eq. 2).4a More recently, Buchwald has shown that (Brettphos)PdII(aryl)(CF3) (Brettphos = dicyclohexyl(2′-isopropyl-3,6-dimethoxy-4′,6′-dipropyl-[1,1′-biphenyl]-2-yl)phosphine) also undergoes aryl–CF3 bond-forming reductive elimination under mild conditions (80 °C, ~30 min).7</p><p>Our group aimed to achieve aryl–CF3 coupling from Pd using a different, complementary approach. Rather than modifying the ligands at PdII, we sought to achieve the desired reactivity by changing the oxidation state of the Pd center from PdII to PdIV. This idea was predicted on the fact that PdIV complexes are well-known to undergo other reductive elimination reactions (eg, C–F, C–Cl, C–I, C–N, C–O) that have proven challenging at PdII centers.8 To probe the viability of this strategy, we synthesized and studied the reactivity of PdIV(aryl)(CF3) intermediates. Two different synthetic routes were used to access these compounds: the 2e− oxidation of pre-formed PdII(aryl)(CF3) complexes (Scheme 2a) and the oxidation of PdII(aryl) complexes with CF3+ reagents (Scheme 2b).</p><p>We initially pursued the synthesis of PdIV(aryl)(CF3) complexes via the 2e− oxidation of (N~N)PdII(aryl)(CF3) (1).9 4,4′-Di-tert-butyl-2,2′-bipyridine (dtbpy) was selected as the N~N ligand, since its rigid, bidentate structure is known to stabilize PdIV complexes.10 As shown in eq. 5, N-fluoro-2,4,6-trimethyl-pyridinium triflate (NFTPT) proved particularly effective for the oxidation of 1, yielding 2 in 53% isolated yield. This product was fully characterized by NMR spectroscopy and X-ray crystallography.</p><p>The availability of pure samples of 1 and 2 enabled a direct comparison of aryl–CF3 bond formation from dtbpy-ligated PdII versus PdIV centers. As shown in eq. 6, PdII complex 1 was inert towards thermal reductive elimination, affording <5% yield of p-F-Ph–CF3 even after 72 h at 130 °C (mass balance was predominantly recovered starting material). In marked contrast, the analogous PdIV complex underwent high yielding aryl–CF3 bond-forming reductive elimination over 3 h at just 80 °C (eq. 6). Notably, products derived from competing aryl–F or aryl–OTf coupling were not observed from 2, presumably due to the low reactivity of these ligands towards reductive elimination.2a,10a,b,11 Overall, the results in eq. 6 confirmed our original hypothesis that aryl–CF3 coupling can be accelerated by oxidation of a Pd center from PdII to PdIV.</p><p>Experimental and computational mechanistic studies indicate that aryl–CF3 coupling from 2 proceeds via pre-equilibrium triflate dissociation (step i) followed by aryl–CF3 bond-formation from cationic intermediate 3 (step ii, eq. 7). These results led us to propose that replacing the dtbpy ligand with N,N,N′,N′-tetramethylethylenediamine (tmeda) would result in an acceleration of this C–C bond-forming event. Importantly, literature precedent has shown that the more flexible tmeda increases the rate of C–C coupling from the related PdIV complexes (N~N)PdIV(CH3)2(Ph)(I) (N~N = bpy versus tmeda).12 DFT calculations of analogues of 2 were consistent with this hypothesis, predicting that both triflate dissociation and aryl–CF3 coupling would be faster with tmeda. Experimental studies confirmed that the PdIV tmeda complex 5 is significantly more reactive than 2, as substitution of tmeda for dtbpy enables aryl–CF3 coupling to proceed at room temperature rather than 80 °C (eq. 8)!</p><p>This work provides the basis for the development of many different types of PdII/IV-catalyzed aryl–CF3 cross-coupling reactions. A potential catalytic cycle for such transformations is outlined in Figure 1. Step i involves the formation of a PdII(aryl) complex. This could occur, for example, by C–H activation (X = H) or transmetalation (X = B, Sn, Si). Subsequent reaction with TMSCF3 (step ii) would yield PdII(aryl)(CF3) (A). Two-electron oxidation of A (step iii) followed by aryl–CF3 bond-forming reductive elimination (step iv) would then furnish the trifluoromethylated product and regenerate the catalyst.</p><p>This approach is already being adopted to achieve synthetically useful trifluoromethylation reactions. For example, a recent report by Liu and coworkers exploited this strategy in the Pd-catalyzed C–H trifluoromethylation of indoles (eq. 9).13 While detailed mechanistic investigations of this transformation have not yet been conducted, the combination of aryl–H (indole), TMSCF3, and an oxidant [PhI(OAc)2] was proposed to react via a cycle very similar to that in Figure 1. An related pathway has also been proposed for the Pd-catalyzed aryltrifluoromethylation of alkenes.14 Numerous analogous transformations can be envisioned, and we anticipate that this approach could find widespread utility for Pd-catalyzed trifluoromethylation sequences.</p><p>Our second strategy for generating PdIV(aryl)(CF3) intermediates is via the reaction of PdII(aryl) complexes with CF3+ reagents (eq. 10). Here the CF3+ plays two roles. First, it serves to oxidize the PdII to PdIV. Second, it serves as the source of CF3 in the product.</p><p>We initially examined the feasibility of this transformation in the context of the cyclopalladated dimer [(bzq)PdII(OAc)]2 (6). This complex was selected for study for two reasons. First, it contains a rigid cyclometalated σ-aryl ligand, which should stabilize high valent Pd oxidation products.15 Second, it is believed to be a catalytically relevant intermediate in C–H functionalization reactions of benzo[h]quinoline.16 As such, studies of its reactivity could potentially be directly applicable to the development of catalytic ligand-directed C–H trifluoromethylation reactions.</p><p>The reaction of 6 with CF3+ reagents 7–9 in AcOH afforded the PdIV complex 10 (eq. 11).17 This complex was fully characterized by NMR spectroscopy and X-ray crystallography.</p><p>Complex 10 was stable at room temperature, but it decomposed at 60 °C with formation the aryl–CF3 coupled product 11 (eq. 12). However, under all of the conditions examined, the formation of 11 was sluggish, showed an induction period, and proceeded in only modest yield (56% in AcOH), with poor mass balance. While we have not yet been able to completely explain these results, we have identified additives that ameliorate many of these issues. In particular, reactions conducted in the presence of Brønsted acids (e.g., trifluoroacetic acid) or Lewis acids (e.g., Yb(OTf)3) were faster, occurred with minimal induction periods, and proceeded in significantly higher yields than those without these additives (eq. 12).</p><p>The lessons learned from these stoichiometric studies have proven highly relevant to Pd-catalyzed ligand-directed C–H trifluoromethylation reactions. In an elegant recent paper, Yu and coworkers achieved the Pd-catalyzed C–H trifluoromethylation of a variety of aromatic substrates using 9 as the CF3+ source. The optimal conditions for these transformations [using a chlorinated solvent in the presence of a Brønsted acid (TFA) and a Lewis acid (Cu(OAc)2)] are remarkably similar to those identified in our stoichiometric reactions with 6 and 11.18</p><p>This suggested the possibility that PdIV complex 10 could be an intermediate in Yu's catalytic reactions. Consistent with this proposal, the use of 10 as catalyst provided nearly identical yield to Pd(OAc)2. Furthermore, analysis of the initial rates with Pd(OAc)2 vs 10 showed that the PdIV complex is a kinetically competent catalyst for C–H trifluoromethylation.17</p><p>Overall, these studies demonstrate the viability of catalytic cycles like that in Figure 2 for PdII/IV-catalyzed trifluoromethylation. This cycle involves initial generation of PdII(aryl) intermediate B via C–H activation or transmetalation (step i). Oxidation of B with CF3+ (step ii) and subsequent C–CF3 coupling (step iii) then releases the trifluoromethylated product. We anticipate that this pathway could prove broadly useful for a number of different Pd-catalyzed transformations for introducing CF3 groups into organic molecules.</p><!><p>Our second approach to uncovering new pathways for arene trifluoromethylation has been to explore metal catalysts beyond Pd and Cu. Our initial efforts in this area focused on Ag for three reasons. First, AgI has the same electronic configuration as CuI, which suggests the possibility for similar reactivity. Second, AgI salts have recently been used as catalysts for related organometallic reactions, including the fluorination of arylstannanes with F+ reagents.19 These examples suggest the possibility that organometallic Ag complexes can participate in aryl–X bond-forming transformations. Third, AgCF3 is readily synthetically accessible (although its reactivity with organic substrates had not previously been explored extensively).20</p><p>We first examined the reaction between AgCF3 and iodobenzene. PhI was selected as a substrate because it is known to react with CuCF3 complexes to generate trifluorotoluene (eq. 14a).1f Very surprisingly, the treatment of AgCF3 with PhI did not yield the expected cross-coupled product PhCF3. Instead, this reaction afforded a mixture of three isomeric C–H trifluoromethylation products (iodobenzotrifluorides) (eq. 14b).21 This is a particularly exciting result because it shows that moving to a different metal (from Cu to Ag) results in completely complementary reactivity.</p><p>Optimization of this transformation led to a new Ag-promoted trifluoromethylation reaction that is applicable to a variety of substrates.18 As summarized in Scheme 3, reactions of electron rich aromatic and heteroaromatic compounds afforded particularly high yields. Somewhat lower yields were obtained with electron deficient aromatics. In substrates containing more than 1 type of aromatic C–H bond, mixtures of isomeric trifluoromethylated products were generally obtained. This feature could prove valuable to medicinal chemists, as it enables the transformation of a single lead molecule into a variety of different trifluoromethylated analogues in a single operation. This is exemplified in the formation of 12 as a mixture of 4 isomers from the trifluoromethylation of Tricor (a commercial cholesterol-lowering drug).</p><p>While detailed mechanistic studies have not yet been conducted, several pieces of evidence implicate a pathway involving homolysis of the Ag–CF3 bond to generate Ag0 and CF3• followed by C–H functionalization via radical aromatic substitution. First, a Ag mirror is observed at the bottom of the flask at the end of these reactions, indicative of the reduction of AgI to Ag0. Additionally, the high reactivity with electron rich aromatics is consistent with the intermediacy of an electrophilic CF3 radical.22 Finally, the addition of 1 equiv of 2,2,6,6-(tetramethylpiperidin-1-yl)oxyl (TEMPO) to this transformation resulted in a dramatic reduction of the yield, suggesting the possibility that this additive is trapping the key CF3• intermediate.12</p><p>Overall these efforts demonstrate the viability of Ag as a promoter for C–H trifluoromethylation. The AgCF3-mediated transformations proceed under mild conditions and are complementary to analogous reactions of CuCF3. In addition, this work adds to a growing body of evidence suggesting that CF3• is a potent reagent for synthetically useful C–H trifluoromethylation reactions. For example, Baran and MacMillan have recently demonstrated the C–H trifluoromethylation of complex molecules with CF3• generated from either NaSO2CF3/tBuOOH (Baran)23 or CF3SO2Cl/Ru(phen)32+/visible light (MacMillan).24,25 All of these transformations serve as valuable methods for the trifluoromethylation of aromatic/heteroaromatic substrates under mild and functional group tolerant conditions.</p><!><p>A third objective of our efforts in this area has been to identify new pathways for Cu-catalyzed boronic acid trifluoromethylation. Prior work had demonstrated that this transformation can be achieved via transfer of nucleophilic CF3− [derived from, for example, TMSCF3 or K(MeO)3B(CF3)]26,27 or electrophilic CF3+ (derived, for example, from 7–9)28 to the Cu catalyst (eq. 15). These two approaches are limited by the relatively high cost of some CF3−/CF3+ reagents, limited functional group tolerance in the presence of these highly nucleophilic/electrophilic reagents, and the requirement for high temperatures in some systems. We reasoned that these limitations could potentially be addressed by accessing an alternative mechanistic manifold in which CF3 transfer to the metal center occurs via CF3• (eq. 15). A discussed above in Part 2, CF3• can effect C–H trifluoromethylation via radical aromatic substitution. Thus, a key challenge for this approach is to identify a system in which reaction of CF3• with the metal catalyst is faster than competing uncatalyzed C–H trifluoromethylation. We selected Cu-based catalysts based on the fact that they are susceptible to rapid 1e− oxidation reactions.</p><p>Our proposed approach to Cu-catalyzed trifluoromethylation requires a mild and readily available source of CF3•. We were inspired by several recent reports by MacMillan that used CF3I as a precursor to CF3• in the presence of visible light and a photocatalyst.29 As such, our initial studies focused on the Cu-catalyzed trifluoromethylation of boronic acid derivatives with CF3I in the presence of Ru(bpy)32+ (eq 16).</p><p>Our investigations revealed that the reaction of 1,1′-biphenyl-4-ylboronic acid with CF3I in the presence of 20 mol % of Cu(OAc), 1 mol % of Ru(bpy)32+, and visible light (two 26 W household light bulbs) affords the trifluoromethylated product 4-(trifluoromethyl)-1,1′-biphenyl in high yield (eq. 17).30 Importantly, the reaction proceeds in <5% yield when light, Cu, or Ru are excluded from the reaction mixture, indicating that all three of these components are necessary for the major reaction pathway. Furthermore, <2% of competing C–H trifluoromethylation of the substrate or product was observed, indicating that the relative rate of the Cu-catalyzed process is faster than radical aromatic substitution.</p><p>Scheme 4 summarizes the scope of this transformation. As shown, this method is effective for the trifluoromethylation of a wide variety of aromatic and heteroaromatic boronic acid substrates bearing many common functional groups. Importantly, analogous perfluoroalkylation reactions of boronic acids can also be conducted under these conditions using cheap and readily available perfluoroalkyl iodide starting materials.</p><p>This work indicates that Cu-catalyzed trifluoromethylation reactions involving CF3• intermediates can be viable and facile processes. We believe that it is possible (even likely) that many Cu-catalyzed processes that were initially believed to involve CF3− or CF3+ transfer actually involve radical intermediates. For example, several reports have shown that Ag salts serve as promoters for the Cu-catalyzed trifluoromethylation of aryl iodides.20b,31 The role of Ag has been proposed to involve mediating transmetalation of CF3− from Si (in TMSCF3) to Cu.17b However, the current results (along with those in Part 2 of this Account) suggest that the role of Ag may be to generate CF3• in these transformations. In addition, CF3+ reagents can potentially undergo 1e− reduction to form CF3•, suggesting the possibility that Cu-catalyzed reactions of CF3+ reagents may also involve radical intermediates.28c,32</p><p>Our preliminary results also have important implications for the future development of Cu-catalyzed trifluoromethylation sequences. They suggest that combining aryl-Cu species (generated via transmetalation, C–H activation, oxidative addition, etc) with perfluoroalkyl radicals (generated via various possible oxidative or reductive pathways) could prove broadly effective for the construction of new fluorinated molecules.</p><!><p>Over the past 5 years, the field of aromatic trifluoromethylation has experienced an explosion of research activity. In parallel with our own efforts, numerous advances from other research groups around the world1f,2,33 have dramatically expanded the organic chemists' toolbox for installing CF3 groups onto arenes and heteroarenes. As discussed above, our contributions to this area have particularly focused on discovering new oxidation states (e.g., PdIV), new metals (e.g., Ag), and new reaction pathways (e.g., the combination of Cu and photogenerated CF3•) for achieving aryl–CF3 coupling. Moving forward, we anticipate that detailed mechanistic studies of all of these transformations will provide valuable insights for the development of second-generation catalysts. Furthermore, the invention of novel pathways and reagents for these reactions should stimulate advances that further facilitate the assembly of trifluoromethyl-containing molecules.</p>

PubMed Author Manuscript

Survey of Home‐Use UV Disinfection Products †

AbstractThe COVID‐19 pandemic provided a commercial opportunity for traders marketing a range of ultraviolet (UV) radiation products for home‐use disinfection. Due to concerns about the efficacy of such products and the potential for harmful levels of UV exposure to people, a range of products were purchased from on‐line trading platforms. Spectral irradiance measurements were carried out to determine whether the products could be effective against the SARS‐CoV‐2 virus and whether they were likely to exceed internationally agreed exposure limits. It was concluded that many of the devices were not effective and many of those that were potentially effective presented a risk to users.

survey_of_home‐use_uv_disinfection_products_†

<!>INTRODUCTION<!>MATERIALS AND METHODS<!><!>MATERIALS AND METHODS<!>RESULTS AND DISCUSSION<!><!>RESULTS AND DISCUSSION<!><!>RESULTS AND DISCUSSION<!><!>RESULTS AND DISCUSSION<!><!>RESULTS AND DISCUSSION<!>CONCLUSION

<p>This article is part of a Special Issue dedicated to the topics of Germicidal Photobiology and Infection Control.</p><!><p>The acute respiratory syndrome coronavirus 2 (SARS‐CoV‐2) is responsible for more than 116 million confirmed COVID‐19 cases and 2.589 million deaths as recorded on 8 March 2021 (1), as well as a deep drop in global consumption of goods and services and millions of job losses. As the pandemic has progressed, there has been an acceleration in the search for effective controls to limit the spread of the virus.</p><p>The use of ultraviolet (UV) radiation may be an important environmental intervention which can reduce both contact spread and airborne transmission of pathogens (2, 3). Solar UV is the primary natural viricidal agent (4, 5, 6) which may help reduce viral load outdoors. However, seasonal changes in Europe with diminishing contribution of sun exposure to viral control (7, 8) and people moving indoors for occupational, educational and most of their recreational activities, has shifted the emphasis to the deployment of affordable measures to abate COVID‐19 transmission including the use of artificial UV radiation in public places and at home.</p><p>Ultraviolet radiation, UV‐C in particular, has been used for disinfection of air (9, 10, 11), water (12), surfaces (13, 14) and food for decades (15, 16, 17). The World Health Organization (WHO) recognized it as a means for tuberculosis infection prevention and control (3).</p><p>Ultraviolet radiation is generally considered to be carcinogenic, with UV‐B and UV‐A parts of the spectrum being known carcinogens (18). Although there is no evidence that UV‐C alone causes cancer in humans, it can cause erythema, trigger photokeratitis and some UV‐C sources can also emit UV‐B and UV‐A radiation (19, 20).</p><p>The likely routes of transmission of the SARS‐CoV‐2 virus may include transfer from the hands following contact with a contaminated surface to the eyes, nose or mouth and through breathing in virions present in droplets or aerosols. Very early in the pandemic, an increasing range of home‐use products became available on on‐line trading platforms. These products were marketed with claims of reducing the risk of catching COVID‐19 and other diseases. However, the rapid proliferation of UV‐C disinfection technology outside the professional sector raised concerns that some devices may pose a risk to human health and/or produce insufficient inactivation of the virus. As a response, the World Health Organization (WHO) and the Commission Internationale de l´Éclairage (CIE) released position statements concerning the use of UV‐C disinfection products and warned against the use of UV disinfection lamps on hands or any other area of skin (2, 21) unless clinically justified. If UV‐C devices are used for disinfecting surfaces or air, in addition to assessment of germicidal potential, personal safety should be evaluated to ensure that recommendations on the limits of exposure specified in the International Commission on Non‐Ionizing Radiation (ICNIRP) guidelines (22) are not exceeded. For conclusive risk assessment and management, appropriate UV measurements are essential.</p><p>Early in the pandemic, there was a report of three members of a household in Hong Kong who were hospitalized after using a UV‐C source to disinfect their home. The device was purchased over the internet, and the safety information to the user supplied with the device was not actioned (23). It was reasonably foreseeable that similar incidents would occur elsewhere. Photokeratitis was recorded in seven people in Helsinki, and three of which were exposed to lamps at home, three at their workplace and one at a dentist office (24). The patients reported that they did not follow manufacturer instructions and were directly exposed without skin or eye protection for periods from 10 min to 4 h (24).</p><p>In April–July 2020, Public Health England (PHE) together with the Office for Product Safety and Standards (OPSS) carried out a pilot survey of the photobiological safety and potential for viral disinfection of home‐use UV disinfection devices available at that time on the UK on‐line consumer market; 48 devices in total were assessed. The results of this study are reported here.</p><!><p>The spectral irradiance was measured at the distance recommended for use in any user information and, where appropriate, at 200 mm required for the classification in accordance with BS EN 62471: 2008 "Photobiological Safety of Lamps and Lamp Systems" (25). Measurements were carried out under environmentally controlled laboratory conditions using a double‐grating IDR300 spectroradiometer (Bentham Instruments, Reading, UK) calibrated using a 1000 W tungsten‐halogen lamp, calibrated for spectral irradiance to the Physikalisch‐Technische Bundensanstalt (PTB, Germany) traceable reference standards (250–800 nm) and Deuterium lamps, calibrated for spectral irradiance to the National Physical Laboratory (NPL, UK) traceable reference standards (200–400 nm).</p><p>The initial results suggested two areas of concern: that the products may not be effective for inactivating SARS‐CoV‐2 and/or they may present a risk to the eyes or the skin. Spectral irradiance data were used to calculate:</p><!><p>E UV, UV‐effective irradiance weighted with the S(λ) hazard weighting function (actinic hazard) and time tUV to reach the exposure limit of 30 J m−2 of the ICNIRP guidelines;</p><p>E VI, irradiance weighted with the virus inactivation efficacy as a function of wavelength and time to reach 90% inactivation (t VI 90) and 99% inactivation (t VI 99). The viral inactivation action spectrum was determined from Lytle and Sagripanti (4). The weighted radiant exposure required for 90% inactivation of SARS‐CoV‐2 is taken to be 6.9 and 28 J m−2 for 99% inactivation according to Sagripanti and Lytle (5). At the time of writing, there were no internationally agreed weighted exposure values for the inactivation of SARS‐CoV‐2 for real‐world exposure situations, and the values taken from Sagripanti and Lytle (2020) were considered reasonable. However, the data in this paper can be scaled for radiant exposure values from other studies.</p><!><p>To assess the benefit to risk potential, the ratio of times for 90% or 99% viral inactivation with respect to the safe ocular and skin exposure limits was derived as follows:(1)A90=tVI90tIIV, (2)A99=tVI99tIIV.</p><p>These values are unique for the device, depend only on the spectral power distribution of emission and could be considered as modified hazard ratios widely used in the assessment of optical radiation safety (26). If A X < 1, the required level of inactivation X could be achieved without over‐exposure of the eyes or skin. If A X ≥ 1, the required level of inactivation X may result in a risk to the eyes and skin when human access to the radiation is not prevented during use. It should be stressed that A < 1 does not mean, at all, that this device is eye‐ and skin‐safe: only that the time for viral inactivation is shorter than the time of ocular‐safe exposure.</p><!><p>The devices fell into four broad categories: handheld wands (18 units), area exposure units (17 units), enclosures/bags (12 units) and one handheld vacuum cleaner. A total of 24 devices were equipped with mercury lamps and 24 units comprised light‐emitting diodes (LEDs); details are shown in Table 1.</p><!><p>Samples of home‐use UV disinfection products.</p><!><p>The 24 devices incorporating mercury lamps emitting UV‐C radiation, including the 253.7 nm line, were generally capable of inactivating viruses; with A90 in the range of 0.11–0.12 and A99 in the range of 0.46–0.47, i.e., 90% inactivation could be achieved in one‐tenth of the time needed to pose a risk to the eyes or skin. However, all the units incorporating mercury lamps were capable of exposing people to levels of UV‐C that could result in erythema or photokeratitis unless access to radiation was prevented by in‐built safety features.</p><p>The devices incorporating LEDs comprised exposed UV‐C LEDs, UV‐C LEDs covered with a plastic shield and LEDs emitting in the UV‐A or visible spectral regions. Only devices with exposed UV‐C LED emitters could potentially be useful for viral inactivation; 11 (23%) units either had a plastic cover, which blocked short wavelengths UV, or they emitted UV‐A or visible light only, making these devices unsuitable for disinfection despite claims on the packaging and information to the user. This includes two E27 (medium Edison screw) fitting lamps, one white and one blue, marketed for UV‐C sterilization (sic). Note that the term "sterilization" was used on the packaging or the information to the user accompanying many of the products. Sterilization is usually used where the quantity of virus is reduced by at least a factor of one million. This is misleading because none of the 11 products assessed could achieve this level of viral inactivation.</p><p>For the devices with LED UV‐C emitters, A90 varied in the range of 0.25–0.33 and A99 within the range of 1.0–1.35, depending on LED peak wavelength emission. While 90% viral inactivation may be feasible without simultaneous risk to the eyes or skin, 99% eye‐safe inactivation at the same distance is unlikely. Furthermore, the irradiance produced by the LED devices was significantly lower than that emitted by the mercury lamps and the irradiated area was also smaller. Both of the tested area exposure devices with UV‐C LEDs were at least an order of magnitude less effective for inactivation at the same distance compared with devices equipped with mercury lamps. Unlike mercury lamps emitting in 360°, unless fitted with a back reflector, LEDs emit in the forward direction only and may be better suited for irradiating particular surfaces rather than space.</p><p>Shortest time to reach the exposure limit of 30 J m‐2 (t UV) at arm's length, taken as 200 mm, for the area exposure devices and the handheld wands is shown in Fig. 1 on a logarithmic scale. Where there was no actinic risk to the eyes from the product at this distance, the data cell is shown as bold arrow.</p><!><p>Shortest time to reach the exposure limit of 30 J m‐ 2 (t UV) of area exposure (a) and handheld devices (b) assessed at 200 mm.</p><!><p>t UV time at a foreseeable accessible distance of 200 mm is, generally, much shorter for area exposure devices than for handheld wands. The shortest time to reach the exposure limit of 30 J m−2 of the ICNIRP guidelines for the tested area devices (Fig. 1a) is 2 s and 33 s for the handheld ones (Fig. 1b). For seven out of seventeen tested devices, t UV is shorter than 5 s even at 200 mm; four area exposure devices did not emit hazardous actinic radiation, and the emission of one unit was extremely low. Closer to the source, risk of over‐exposure increased, and many area exposure devices presented foreseeable risk to the eyes and skin even at much longer distances. t UV was shorter than 1 min for three out of eighteen handheld devices at 200 mm, within the range of 1–3 min for six units, longer than 10 min for five of them; four handheld wands did not emit hazardous actinic radiation.</p><p>More than half of the enclosures/bags were equipped with safety features preventing direct access to optical radiation (see Table 2), and ocular exposure was prevented in seven out of twelve devices; one unit did not emit hazardous actinic radiation. The shortest ocular‐safe exposures at 200 mm looking inside working devices were 6 s in one unit and an order of minutes in another one; times were much longer with the others and risk to the eyes could be considered insignificant for those devices. However, three enclosures/boxes had a gravity or tilt sensor as a safety feature preventing direct ocular exposure, but it was possible to place them directly on skin and the time to reach the exposure limit of 30 J m−2 of the ICNIRP guidelines could be as short as 10 s.</p><!><p>Safety features of the UV‐C disinfection devices.</p><!><p>Time required for 90% viral inactivation t VI 90 for handheld and area exposure devices is shown in Fig. 2. Where viral inactivation by the product was unlikely, the data cell is shown as bold arrow.</p><!><p>Time required for 90% viral inactivation t VI 90 for area exposure (a) devices assessed at treatment distance of 200 mm and handheld wands (b) assessed at treatment distance of 20 mm.</p><!><p>Seven out of seventeen area exposure devices (Fig. 2a) were capable of 90% viral inactivation in less than 1s within the 200 mm distance, two in less than 10s, and three smaller units in less than 1 min; five devices either did not emit UV‐C or UV‐B radiation or the emission level was negligible.</p><p>Potential for viral inactivation of handheld wands was assessed at recommended treatment distance of 20 mm. Within 20 mm, seven out of eighteen handheld wands (Fig. 2b) had the potential for viral inactivation in less than 1 s, five units in 1–5 s and six required hours of treatment for 90% even at very close distances. However, a main challenge with the use of UV‐C for area disinfection is shadowing, which means that surfaces not in direct line‐of‐sight of the source may not receive sufficient radiant exposure for effective disinfection.</p><p>Accurate quantitative assessment of the potential effectiveness for viral inactivation of enclosures/boxes was not possible due to a combination of fitted safety features, which terminated emission when open or tilted, and geometric factors, for example range of distances from the emitters. Therefore, only an indication of the potential for viral inactivation was made in this case; emission of one unit did not contain UV‐C or UV‐B radiation and viral inactivation was not feasible by this product. It also should be emphasized that items placed inside for disinfection were generally only irradiated on one or two sides, with the other surfaces shielded from the UV by the item placed inside. The effectiveness of all these devices for virus inactivation is critically dependent on line‐of‐sight exposure conditions; the information to the user did not address this.</p><p>For the most of tested devices (but not all), user information, and in some cases—direct labelling on the body of device itself, contained warning to avoid direct exposure of the eyes and skin. In addition, a number of home‐use UV‐C disinfection devices incorporated safety measures including gravity and motion sensors, and interlocks, as listed in Table 2.</p><p>A motion sensor was fitted in a single area exposure device, effectively preventing people's presence in the treatment area; all other area exposure devices did not incorporate any safety features and, for the units equipped with mercury lamps, at close distances of 20–50 cm hazardous exposure levels could be reached in seconds. It should be also noted that 3 of the tested units produced ozone within minutes of operation at sufficient levels to require thorough ventilation of the premises. Ozone is a very strong oxidant that may cause irritation to the airways when breathed in and to the eyes; at higher concentrations (27, 28), it may be toxic or interact with materials.</p><p>The vacuum cleaner was equipped with pressure and gravity (tilt switch) sensors which disabled UV emission if the cleaning head was tilted or lifted off the treated surface normally preventing human access to hazardous emission. However, it would be possible to operate the device if it was placed on a hand.</p><p>Five of the handheld wands incorporated gravity or tilt switch sensors terminating emission if the device was tilted and preventing accidental eye exposure; however, this does not stop accidental or intentional exposure of the skin if, for example, the hands were placed under the unit. Although the risk to the eyes was reduced for such devices, it is reasonably foreseeable that a child could be looking up into a device being used by an adult. Two handheld wands were additionally supplied with protective eyewear.</p><p>Most of the enclosures/bags were effective at minimizing the risk of eye exposure by design restrictions; more than 50% (7 out of 12) were additionally equipped with interlocks terminating emission when the device was open or tilted; for those without interlocks, hands could be intentionally placed inside but this risk is small.</p><p>Data from the OPSS COVID‐19 Consumer Survey (29) demonstrated that, out of 200 people who purchased UV devices, 31% reported purchasing it for use on their skin and 17% for use on their pets. It can only be hypothesized as to whether this was either a replacement for, or complimentary to, hand washing and the use of hand sanitizer. A purchaser of a portable UV‐C device was considered likely to find a way to expose at least their hands to the UV‐C, even if control measures were incorporated into the product. This was particularly relevant for the wand devices. Based on the consumer survey results, it could be hypothesized that some consumers are unlikely to read warnings. In situations where operating instructions are not followed or are missing, the risk of exposure to UV‐C has led to consumers experiencing photokeratitis or requiring hospitalization (23, 24).</p><p>50% of the devices assessed contained low‐pressure mercury lamps. The vulnerability of these lamps to accidental damage varied, but the lamps in several devices appeared very exposed. If the lamp was broken, there would be risks associated with broken glass, plus the chemical hazard associated with mercury.</p><!><p>A total of 48 devices available for UK consumers in April‐July 2020 and marketed for home use‐UV‐C disinfection or sterilization were assessed. Not all devices were marketed specifically for use against SARS‐CoV‐2, but all claimed effectiveness against viruses and bacteria. The surveyed sample set included devices for area exposure, handheld wands, enclosures/boxes and one vacuum cleaner. Devices were equipped with either mercury lamps, UV‐C LEDs or non‐UV‐C LEDs; nine out of 48 devices did not emit radiation effective for inactivation of viruses. OPSS instigated the appropriate corrective actions in relation to the noncompliant and unsafe products identified through testing (30).</p><p>Where UV‐C disinfection devices are used as a mitigation measure for preventing viral spread in indoor environments, it is recommended that their efficacy and safety be demonstrated with relevant data. Effectiveness of disinfection depends on multiple parameters including the underlying technology, design of the device, surface area covered, whether surfaces are in direct line‐of‐sight, exposure time and distance between the UV‐C device and the treated surface.</p><p>In general, for mercury lamp‐based devices, 90% inactivation could be achieved in one‐tenth of the time needed to pose a risk to the eyes or skin. However, all the units incorporating mercury lamps were capable of exposing people to levels of UV‐C that could result in erythema or photokeratitis unless access to radiation was prevented by in‐built safety features.</p><p>Inappropriate use of UV‐C equipment, such as direct exposure of eyes or skin, has been associated with potential health risks. Portable devices should be used with care following manufacturer guidelines, and safety controls should put in place in order to minimize unintended consequences.</p>

PubMed Open Access

Cisplatin Tumor Biodistribution and Efficacy after Intratumoral Injection of a Biodegradable Extended Release Implant

Local delivery of chemotherapeutic drugs has long been recognized as a potential method for reaching high drug doses at the target site while minimizing systemic exposure. Cisplatin is one of the most effective chemotherapeutic agents for the treatment of various tumors; however, its systemic toxicity remains the primary dose-limiting factor. Here we report that incorporation of cisplatin into a fatty acid-based polymer carrier followed by a local injection into the solid tumor resulted in a successful tumor growth inhibition in heterotopic mouse bladder tumor model in mice. Platinum concentration in the tumor tissue surrounding the injected implant remained above the therapeutic level up to 14 days after the injection, while the plasma levels were several orders of magnitude lower comparing to systemic delivery. The reported delivery system increased the maximum tolerated dose of cisplatin 5 times compared to systemic delivery, thus potentially improving antitumor efficacy of cisplatin in solid tumor model.

cisplatin_tumor_biodistribution_and_efficacy_after_intratumoral_injection_of_a_biodegradable_extende

1. Introduction<!>2.1. Materials<!>2.2. Preparation of Formulations and In Vitro Drug Release<!>2.3. In Vitro Cytotoxicity<!>2.4. In Vivo Cisplatin Toxicity<!>2.5.1. Inoculation of MBT Cells<!>2.5.2. Treatment Protocols<!>2.6. Platinum Distribution in Plasma and Tissue<!>2.7. Macroscopic and Histopathological Evaluation<!>2.8. Statistical Analysis<!>3.1. In Vitro Cisplatin Release<!>3.2. In Vitro Cytotoxicity<!>3.3. In Vivo Antitumor Activity<!>3.4. Platinum Tumor Distribution<!>3.5. Platinum Plasma Levels<!>3.6. Macroscopic and Histopathological Evaluation<!>4. Discussion<!>5. Conclusion<!>

<p>Polymer-based gels are potential carriers that may affect a target tumor while reducing the toxic effects of the loaded drug. Specifically, we suggest that it is possible to optimize their delivery and improve the responsiveness of solid tumors to current chemotherapeutic agents [1].</p><p>Cisplatin is widely used for the treatment of testicular, bladder, head and neck, small-cell and non-small-cell lung cancers, but it also possesses substantial side effects, such as nephro-, neuro-, and myelotoxicity [2–4]. Polymer-based cisplatin-loaded drug delivery systems such as liposomes [5], polymeric micelles [6], hydrogels [7], polymeric gels [8], and implants [9] provide an opportunity to deliver high, localized doses of drugs for a prolonged period directly into a tumor or at the site of tumor resection. Implanting a biodegradable device loaded with antineoplastic agent in the cavity created by the tumor removal provides high local concentration of the drug, killing the surviving malignant cells. This may also prevent the systemic side effects of chemotherapy that is normally associated with intravenous administration. Injectable device may also provide sustained, controlled delivery of the drug to the malignant tumor. In addition, clinicians can perform debulking of large tumor prior to the surgery by exposing the tumor to large concentrations of the drug [10].</p><p>Biodegradable polyanhydrides and polyesters are useful materials for controlled drug delivery [11]. In earlier reports we have described the synthesis and various applications of ricinoleic acid-based polyanhydrides as drug carriers [12, 13]. These polymers have hydrophobic backbone with hydrolytically labile anhydride and/or ester that may hydrolyze to dicarboxylic acids and hydroxy acid monomers when placed in aqueous medium. The toxicity, biodegradation, and elimination of polyanhydrides and aliphatic polyesters have been recently reviewed [14, 15]. The fatty acid components of these polymers undergo extensive metabolism in the body and are mainly excreted in the form of carbon dioxide. The in vitro and in vivo toxicity tests indicate that these polymers are well tolerated by the tissues and can be generally considered as biocompatible [14]. Furthermore, Vaisman et al. [16] evaluated the safety and biocompatibility of ricinoleic acid-based polymer in rats in high doses intramuscularly, subcutaneously, and intracranially. No systemic tissue damage, polymer-related lesions, or abnormalities could be detected in animals.</p><p>Herein we report the application of a biodegradable poly (sebacic-co-ricinoleic) acid (P(SA : RA)) polymer for local cisplatin delivery to the solid tumor. Cisplatin can be incorporated by direct mixing with P(SA : RA), which is an injectable fatty acid-based polymer that solidifies in contact with aqueous media [17, 18]. As a result the incorporated drug is released in a sustained manner over a period of days. We hypothesized that cisplatin released from the polymer formulation injected directly in the solid tumor will stop the tumor growth and result in a prolonged survival. Furthermore, we investigated cisplatin local and systemic distribution after single intratumoral injection in heterotopic mouse bladder tumor model.</p><!><p>Poly (sebacic acid-co-ricinoleic acid ester anhydride) 2 : 8 and 3 : 7 were synthesized as previously described [17]. Cisplatin was purchased from AlfaAesar (MA, USA). All solvents were analytical grade from BioLab (Jerusalem, Israel) or Frutarom (Haifa, Israel) and were used without further purification. MBT (mouse bladder tumor) cells were a generous gift from Professor Ofer Gofrit [19] from Hadassah Ein-Karem Hospital (Jerusalem, Israel). Cell culture medium and fetal calf serum (FCS) were obtained from Biological Industries (Beit-Haemek, Israel).</p><!><p>The formulations of P(SA : RA) 2 : 8 and 3 : 7 with 5% w/w of cisplatin were prepared under sterile conditions by direct mixing of the polymer with the drug at room temperature. The composition was mixed until a smooth paste was obtained. All formulations were aseptically loaded in syringes and tested for sterility in TSB medium [20]. The obtained formulations were injectable semisolid pastes at room temperature. In vitro drug release studies were conducted by injecting 20 mg of the pasty formulations sample in a 50 mL phosphate buffer solution (0.1 M, pH 7.4) at 37°C with constant shaking (100 RPM). The paste hardened to a soft solid shortly after addition to the buffer. The release medium was replaced periodically with fresh buffer solution, and platinum concentration in the solution was determined by Inductively Coupled Plasma Mass Spectrometry (ICP-MS, Perkin Elmer SCIEX). The instrument is based on Dynamic Reaction Cell technology (ELAN DRC II) with performance-enhancing Axial Field technology. The validity of the analytical procedure was established through a study of specificity, precision, linearity, and accuracy. The linearity of the analytical procedure was evaluated by plotting the detector response (peak area) against analyte concentration. Linear regression analysis was applied to calculate the slope, intercept, and linear correlation coefficient (R 2). The limit of detection (LOD) was calculated as signal-to-noise ratio of 3 : 1, and the limit of the quantification (LOQ) was determined as signal-to-noise ratio of 10 : 1. The number of points used in each curve was 6. Calibration curves were obtained by programmed injection of different aliquots (10–45 μL) of a standard solution with increments of 5 μL. The concentration of standard solutions was 10 ppb of platinum in double distilled water, while the linear region was observed at concentration between 0.01 and 100 ppb. All experiments were performed in triplicate.</p><!><p>The MBT cells were maintained in monolayer cultures in Dulbecco's modified Eagle's medium containing 10% (v/v) fetal calf serum and supplemented with 200 IU/mL penicillin and 200 μL/mL streptomycin (Beit-Haemek, Israel) in 75 cm2 flasks, in humidified 5% CO2 in air, incubated at 37°C [21]. 2 × 103 MBT cells in 100 μL of culture medium were seeded in 96-well plates and were incubated for 24 h at 37°C. Cisplatin in vitro toxicity was tested by adding serial dilutions of cisplatin in a volume of 10 μL to the cultured MBT cells. Cytotoxicity of the polymer degradation products was tested by adding 10 μL solution from the blank polymer release study (same sample used as a blank control for the in vitro drug release). Cell proliferation was estimated by 3H-thymidine incorporation [22]. All data presented as mean ± STD of triplicate. The data was plotted as a percentage of the data from the control cultures, which were treated identically to the experimental cultures.</p><!><p>In a separate study performed to determine maximal tolerated dose (MTD) for the cisplatin-polymer, mice were injected with increasing doses of cisplatin incorporated in the polymer. The dose of 25 mg/kg was chosen as the treatment dose since it was the maximal cisplatin dose at which mice did not show weight loss throughout the study. The dose of 50 mg/kg was determined as (MTD) for cisplatin-polymer formulation, because all mice survived but showed weight loss. The MTD for intraperitoneal delivery was 5 mg/kg, while LD50 for cisplatin in C3H mice is 10 mg/kg [6].</p><p>The experiments in mice were approved by the Ethical Committee for Animal Experimentation of the Hebrew University (NIH approval no. OPRR A01-5011).</p><!><p>Inbred 8–10-week old female C3H mice, weighing about 20 g (Harlan Laboratories, Israel) were kept under specific pathogen-free (SPF) conditions and given free access to irradiated food and acidified water throughout the experiment. Mice were injected subcutaneously via a 27-gauge needle in the posterolateral flank with 5 × 105 MBT cells suspended in 0.1 mL RPMI medium. Tumors were measured using caliper every other day, and their volumes were calculated by the formula: length × width × height × 0.523.</p><!><p>The treatment was initiated 10 days after tumor cells inoculation, when the tumor was palpable, and the volume range was between 0.12–0.243 cm3. The mice were randomly assigned to one of the three treatment groups or the two control groups (n = 10 in each group). Mice in the control groups received either intratumoral injection of 50 μL of the blank polymer or no treatment at all. The first treatment group was injected with 50 μL of a formulation containing 1% cisplatin in P(SA : RA) 2 : 8 (equivalent to 25 mg/kg dose) intratumorally. The second treatment group was treated with intratumoral injection of 0.1 mL cisplatin solution in saline at a concentration of 1 mg/mL which equals 5 mg/kg, and the third group received intraperitoneal (IP) injection of 0.1 mL of cisplatin solution in saline at a concentration of 1 mg/mL (LD50 for cisplatin in C3H mice is 10 mg/kg). Mice received a single injection during the experiment. The animals were sacrificed when the tumor volume reached 3–3.5 cm3.</p><!><p>Additional group of mice (n = 36) was injected with 5 × 105 MBT cells to induce subcutaneous tumor. However, to determine platinum distribution from the injection site in the tumor mass, the treated tumors should partially escape from the treatment; otherwise, there would be no tumor mass to determine the platinum concentration. Thus, 50 μL of the injectable polymer/cisplatin formulation containing 1% cisplatin was injected 13 days after tumor cells inoculation when the tumors volume was >1.2 cm3. In this case the tumor could not be totally eliminated, and the pattern of platinum distribution in the tumor could be studied. At different time points (1, 2, 3, 5, 7, and 14 days after injection), six mice were sacrificed. The tumor was excised and frozen at ‒20°C, and the blood was collected from cardiac puncture, heparinized, and centrifuged (2500 rpm, 5 min) to obtain plasma. The obtained plasma was separated and kept frozen at ‒20°C. Tissue samples were embedded (O.C.T. Compound, Tissue-Tek, Redding, CA) and sectioned into 50–100-μm-thick sections in a cryostat at ‒20°C. The sections were weighed, decomposed in nitric acid, and diluted in DDW to obtain 100-, 1000-, and 2200-fold dilutions to determine platinum concentrations in solution and plasma in ICP-MS (Perkin Elmer). All sections were inspected for the presence of nondegraded formulation, and the formulation was manually scooped out to avoid biased calculation of the actual drug concentration in the tissue.</p><!><p>For histopathological evaluation animals were sacrificed 5 days after treatment application, and tumors were dissected and fixed in 4% formaldehyde solution. The tissue was processed into paraffin, and 3-μm sections were stained with hematoxylin & eosin for histological evaluation. The examination parameters included necrosis total area, inflammatory cell infiltration, and intact tumor tissue.</p><!><p>All results are expressed as mean ± standard deviation (STD) of the mean and statistically analyzed using GraphPad Instat ANOVA. P values less than  .01 were considered significant for all tests.</p><!><p>The polymer carriers—P(SA : RA) 2 : 8 and 3 : 7 having molecular weight (Mw) ranging from 4000 to 6000 Da were prepared from RA and SA by melt condensation as previously described [17]. The polymer structure is shown in Scheme 1. Incorporation of 5% w/w of cisplatin in the polymer by mixing at room temperature did not affect the molecular weight of the polymer and had no chemical interaction with the polymer, as was confirmed by GPC and 1H-NMR. Cisplatin release from P(SA : RA) had no initial burst effect, and during the first day only 5–7% of the incorporated drug was released (Figure 1), followed by additional 25% on the next day, and reaching 85% in the following ten days.</p><!><p>Growth inhibition of cisplatin in MBT cell culture is shown in Figure 2. The IC50 value of cisplatin was found to be 0.8 μg/mL. Similar results of IC50 for cisplatin were reported earlier: 4.8 μg/mL for MBT-2 cells [6] and 1.5 μg/mL for Meth-AR-1 cells [7].</p><!><p>The efficacy of cisplatin delivered intratumorally was investigated in a heterotopic mouse bladder tumor (MBT) model. The treatment was initiated on the tenth day after tumor cell inoculation. Mice that were not treated and mice injected with the blank polymer were sacrificed 15 and 18 days, respectively, after tumor cells inoculation when the tumor volume exceeded 3.5 cm3. However, the growth rate of the tumor was slower in mice injected with the blank polymer that caused statistically significant (P < .005, ANOVA) delay in tumor progression (Figure 3). The injection of blank polymer into the tumor damaged its structure and delayed its development, but since there was no therapeutic effect on the tumor cells, as was confirmed in in vitro cultured cells, the tumor recovered from the physical injury and continued to grow. The injected dose of cisplatin was the same in intratumoral (IT) and intraperitoneal (IP) injection of cisplatin solution (0.1% w/v in saline, 0.1 mL injection, 5 mg/kg); however, IP delivered drug-inhibited tumor growth more efficiently comparing to IT application of solution. After IT injection of cisplatin solution tumors reached the volume of 3 cm3 13 days after treatment while after IP treatment only after 16 days. Thus, soluble cisplatin delivered in solution either IT or IP showed efficiency compared to nontreatment and blank polymer groups in prolonging mice life. In the cisplatin/polymer group, mice were injected intratumorally with 50 μL of the 1% cisplatin formulation 10 days posttumor inoculation. In 8 mice out of 10 the tumors completely disappeared during the first 10 days after treatment, and mice remained tumorless till the end of the study (40 days posttumor cells inoculation), while, in the other two mice in this treatment group twenty days posttreatment administration, regrowth of a small nodule at the edge of the original tumor appeared, but its dimensions did not increase above 0.3 cm3 till the end of the study (40 days posttumor cells inoculation).</p><!><p>In this study we measured total platinum levels rather than intact cisplatin. Cisplatin is not stable in biological fluids and undergoes ligand-exchange reaction that result in metabolites with different biological activity. Although determining concentrations of both intact cisplatin and its various metabolites would be a more accurate way to predict its activity, measuring total serum platinum is traditionally used for clinical pharmacokinetics assays [8].</p><p>Platinum content in the tumor mass gradually decreased over the time course of 14 days (Figure 4), while 90% of the injected platinum was found in the tumor tissue 24 hours after the injection decreasing to 12% of the injected platinum at 14 days.</p><p>The peak platinum concentrations (C max) in tumor tissue at different time points were defined as the amount of platinum per milligram tumor tissue excluding non-degraded formulation (Figure 5). In the first day after injection, platinum concentration in the tumor tissue was still low, which corresponded with the in vitro release results. However, three days after the intratumoral injection, platinum concentration was the greatest and reached 8.8 μg/mg, while platinum concentration in plasma was only 0.12 μg/ml. C max gradually decreased in the following days, and 14 days after the injection platinum concentration in tumor was 0.3 μg/mg that is still high enough to induce cytotoxic effect [23]. For comparison, systemic injection of 10 mg/kg free cisplatin produced tumor platinum concentration of 0.014 μg/mg [24], which is lower than C max 14 days after intratumoral injection of cisplatin/polymer formulation.</p><p>The cisplatin/polymer formulation was injected into the center of the tumor; therefore, platinum distribution is expected to be radial, while the greater concentrations are found at the injection site with gradual decrease toward the tumor boarder. Each line in Figure 6 represents platinum concentration pattern at one time point from the center of the tumor toward the edge over the distance in millimeters. After 24 hours subsequent to injection, the maximal platinum concentration of 1.1 μg/mg tumor tissue was found close to the injection site, and it decreased gradually to 0.01 μg/mg at a distance of 4.4 mm from the injection site that is still considered above the therapeutic level. The greatest platinum tumor tissue concentration was found at 3 and 4 days after the injection, reaching 8.8 μg/mg and 3.8 μg/mg, respectively. After 14 days, platinum levels at the injection sites were still above the therapeutic level. Interestingly, in the first 4 days after the treatment, greater gradient was observed between the injection site and the distant areas of the tumor. However, at the later period, the platinum content in the tumor tissue was more equally distributed that can be explained by the drug diffusion and the clearance from the injection site.</p><!><p>As reported elsewhere [6], the plasma platinum levels after IV administration of free cisplatin were 11.7 μg/mL at time zero and decreased during the following 12 hours to 1 μg/ml. After intratumoral injection of cisplatin/polymer formulation platinum plasma levels gradually increased from less than 0.1 μg/mL after 24 hours and peaked at 0.15 μg/mL on days 4–6, followed by sharp decrease to 0.06 μg/mL on day 8 (Figure 7). The platinum plasma levels are closely related to the events occurring in the tumor, where the polymer releases the incorporated drug. Importantly, platinum plasma levels after intratumoral polymer injection were several orders of magnitude lower comparing to systemic delivery at all time points.</p><!><p>Figure 8 shows the macroscopic view of the tumors dissected from mice and sectioned in cryostat. Figure 8(a) shows the MBT tumor treated with 50 μL of 1% w/w cisplatin/polymer formulation and removed 3 days after treatment. The formulation's color is yellow because of the cisplatin color and can be easily recognized (designated as (*), and contoured with white line). Three days after formulation injection, large portion of the injected formulation was still found in the center of the tumor tissue, and the formulation did not totally degrade. A region of a reddish tissue surrounded the injected formulation, which was histologically proven to be a necrotic/inflammatory process (designated as (⚪) and contoured with green line), followed by the intact tumor cells (designated as (♦)), as was proved in the histopathological evaluation of the tissue. Seven days after the formulation injection, 20% of platinum was still found in the tumor tissue, but the formulation has already degraded, and the injection site remained free of the polymer and tumor cells (Figure 8(b)). Around the injection site, the necrotic region increased in size, which applies to continuous cytotoxic activity (designated as (⚪) and contoured with green line), followed by the intact tumor cells (designated as (♦)). Blank polymer was injected into the tumors and sectioned similarly to eliminate the possibility of the blank polymer cytotoxic activity. Figure 8(c) shows the tumor with the blank polymer 3 days after injection. The polymer region is surrounded by a mild inflammation region followed by tumor cells, and while partial effect of the polymer on the tumor cells cannot be excluded it was less vigorous than with cisplatin. For comparison, Figure 8(d) shows the tumor without treatment.</p><p>The changes appearing in the tumor tissue at the formulation injection site and in more distant regions are shown in the panoramic view of the histology images of the tumors injected with cisplatin/polymer formulation (Figure 9(a)). Only necrotic cells are present at the formulation-tumor interface (higher magnification, Figure 9(b)). Along with the necrosis progression, inflammation process is evident, and the region of dead tumor cells extends up to 3.5 mm from the cisplatin formulation. Figure 9(c) is a higher magnification of a border region between the necrotic process and the unaffected intact tumor tissue, followed by mainly tumor cells (Figure 9(d)). The region in the tumor where cisplatin did not diffuse and was below therapeutic dose (Figure 9(d)) has similar cells appearance as in the nontreated tumor (Figure 10). Importantly, tumor cells density became noticeably lower in both necrotic regions and the region beyond the border, which results in enhanced drug penetration and the cytotoxic effect [25]. Figure 11 shows the appearance of the MBT tumor without treatment; no necrosis or inflammation process was evident in the tumor tissue.</p><p>Efficacy studies showed that intratumoral injection of a blank polymer caused a delay in tumor growth (Figure 3). Histological evaluation of the tumors injected with the blank polymer revealed a mild inflammatory reaction and no infiltration region around the polymer for 0.3 mm. Beyond the 0.3 mm, intact tumor cells appear again, as shown in Figures 11(a) and 11(b).</p><!><p>The toxicity of conventional systemic cancer chemotherapy has severely limited the safety and effectiveness of such therapy, and its impact on the quality of life of patients hampers its wider clinical application [8]. Plasma concentrations are often used as a marker of cytotoxic exposure; however, drug delivery to the tumor is determined not only by plasma concentrations but also by distribution from plasma into the extracellular fluid (ECF) of the tumor and from the ECF into the cells. Solid tumors have several potential barriers to drug delivery that may limit drug penetration, such as alteration of distribution of blood vessels, blood flow, interstitial pressure, and microcirculation in the tumor [23]. Therefore, high systemic levels of the cytotoxic drug often cause systemic toxicity without reaching effective concentrations in the tumor.</p><p>Various injectable drug delivery systems have been investigated for local delivery of cisplatin and other anticancer agents. Intradose, a collagen gel, loaded with cisplatin and epinephrine, has been shown to be effective for the treatment of head and neck and hepatocellular cancers [26]. Intradose is an injectable gel that releases cisplatin after intratumoral injection. However, the semiliquid phase of the collagen gel released its content in short time period of hours till couple of days that evokes a need for repeated administrations. Another injectable biodegradable PEG, PLGA polymer system for intratumoral chemotherapy, is Atrigel. This system has a similar drug release period; however, the solidification mechanism involves solvent displacement that can result in higher systemic toxicity. Interestingly, platinum serum levels in rats were 10 times higher after Atrigel injection [8] comparing to the mice serum levels reported in this study, hence presenting lower systemic toxicity of the described formulation combined with local efficiency again tumor cells.</p><p>Poly (sebacic co-ricinoleic acid) 2 : 8 used in this study is a hydrophobic polymer, built of natural fatty acids, which can be used for the release of hydrophobic or hydrophilic drugs. The polymeric paste formulation with cisplatin is injectable through a 23-gauge needle and it forms a gel in contact with body fluids by a mechanism that does not involve solvent leaching or temperature change [17, 18]. The purpose of this study was to evaluate the effect of cisplatin-polymer formulation injected intratumorally in heterotopic model in mice compared to immediate release formulation after IP and IT delivery. Interestingly, we noticed that IP cisplatin delivery delayed tumor growth more efficiently compared to the same solution delivered IT. A possible explanation of the superior efficiency in tumor treatment of IP delivery is that the drug was delivered through the blood supply to the tumor and had better chances reaching all the tumor regions, while in the direct injection of cisplatin solution to the tumor the delivery is limited to the injection site, and cisplatin penetration can be inhibited by intratumoral interstitial pressure [27]. Moreover, cisplatin in solution is cleared almost immediately leaving other regions of the tumor untreated. In contrast to IP or IT delivery of cisplatin solution, the described polymeric formulation formed a semisolid implant in situ as was confirmed after tumor excision and sectioning [18]. Because of the hydrophobic nature of the polymer, there was no burst effect upon the formulation injection, and the drug was released at controlled rate, as was determined by plasma platinum levels monitoring. The levels of platinum in the tumors treated with cisplatin/polymer formulation were above the therapeutic threshold during two weeks after the injection, without any signs of systemic toxicity. Importantly, incorporation of cisplatin in the polymer that delivered the drug over prolonged period of time allowed increasing MTD 5 times compared to immediate release formulation, hence injecting 25 mg/kg in polymer versus only 5 mg/kg in solution. Thus, the effectiveness of the polymeric formulation in treating MBT tumor in the described heterotopic model can be the result of higher dose, slow prolonged release and exposure to the drug, and greater surface area of contact between the tumor and the formulation.</p><p>Cisplatin is not stable in biological fluids and undergoes ligand-exchange reaction that results in aquated cisplatin that gradually is transformed to other metabolites through reaction with glutathione, albumin, and nucleotides. Although both intact cisplatin and its metabolites are biologically active, it is the intact cisplatin that is responsible for nephrotoxicity. Thus, it can be assumed that systemic platinum levels found after local delivery of cisplatin in the polymer formulation are mainly in the form of biological metabolites, rather than unmodified cisplatin, because of the longer time cisplatin is exposed to biological fluids [24, 28].</p><p>Moreover, P(SA : RA) has been shown to be effective in treating solid tumors with paclitaxel, which is highly potent hydrophobic chemotherapeutic drug [29, 30]. P (SA : RA)-based polymeric system is unique because it can serve as a vehicle for delivery of both hydrophilic and hydrophobic drugs concomitantly in a single injection, thus increasing the therapeutic potential of the formulation [31–33].</p><!><p>The results of this work indicate that treatment with the polymer formulation of cisplatin had a positive outcome and inhibited the growth of the tumors. Distribution studies of cisplatin after intratumoral injection showed high and effective concentrations in the tumors. Histological studies proved the existence of the necrotic process caused by the cytotoxic drug.</p><!><p>Structure of poly(sebacic acid-co-ricinoleic acid).</p><p>In vitro cumulative release of Pt from P(SA : RA)2 : 8 (triangles, dashed line) and P(SA : RA)3 : 7 (stars, solid line) loaded with 5% w/w cisplatin. Each point represents the mean value ± STD (n = 3). Release was conducted in 0.1 M phosphate buffer, pH 7.4, at 37°C. Pt concentrations were determined by ICP-MS.</p><p>In vitro inhibition effect of cisplatin on MBT cells. Cisplatin at increasing concentrations was added 24 hours after cell incubation in wells.</p><p>Effect of cisplatin on MBT tumor growth in s.c. implanted mice (n = 10). Cisplatin 1% w/w in polymer, 50 μL (∗: solid line); blank polymer (■, dashed line); cisplatin solution (0.1% w/v in saline, 100 μL injection, 5 mg/kg) (∆, dashed line) were injected intratumorally. No treatment group is designated as (♦) with solid line, and the group treated IP with cisplatin solution (1% w/v in saline, 100 μL injection, 5 mg/kg) is designated as (■) with solid line. The tumor volume is expressed as mean ± STD. Statistically significant differences between the groups are signed with a star (P < .05,  *).</p><p>The time profile of platinum remaining in the tumor after intratumoral injection of cisplatin/polymer formulation (1% w/w, 50 μL). Values are expressed as mean ± STD (n = 6).</p><p>The time profile of platinum maximal concentration (C max) in the tumor tissue after intratumoral injection of cisplatin/polymer formulation (1% w/w, 50 μL). Values are expressed as mean (n = 6).</p><p>Cisplatin tumor tissue distribution after intratumoral injection of cisplatin/polymer formulation (1% w/w, 50 μL). Each curve represents a different time point when the mice were sacrificed and their tumors processed. Values are expressed as mean (n = 6).</p><p>Platinum (Pt) levels in plasma. Pt levels in mice plasma were determined by ICP-MS. Each data point represents the average of six mice ± STD.</p><p>Macroscopic view of frozen tumor tissues at cryostat sectioning: (a) tumor treated with cisplatin/polymer formulation and excised 3 days after injection; (b) tumor treated with cisplatin/polymer formulation and excised 7 days after injection; (c) tumor treated with the blank polymer and excised 3 days after injection; (d) untreated tumor. The polymeric formulation is assigned with the star (∗), the necrotic tissue with the white circle (⚪), the infiltration of the inflammation cells with a black circle (•), and the intact tumor cells with a black rhomb (♦).</p><p>Histology of the MBT tumor injected intratumorally with cisplatin/polymer formulation (1% w/w, 50 μL). (a) Magnification ×10, panoramic view of the slice; (b) magnification ×40, enlargement of the cisplatin/polymer region and the surrounding necrotic area; (c) magnification ×40, enlargement of the border between the end of the necrotic area and start of the intact tumor area; (d) magnification ×40, enlargement of the intact tumor area beyond the effect of cisplatin. The polymeric formulation is assigned with the star (∗), the necrotic tissue with the white circle (⚪), the infiltration of the inflammation cells with a black circle (•), and the intact tumor cells with a black rhomb (♦).</p><p>Histology of nontreated MBT tumor (magnification ×40), intact tumor cells.</p><p>Histology of the MBT tumor injected intratumorally with blank polymer. (a) Magnification ×20; (b) magnification ×40, enlargement of the blank polymer and the surrounding mild inflammation area; bar: 0.1 mm; the polymeric formulation is assigned with the star (∗), the necrotic tissue with the white circle (⚪), the infiltration of the inflammation cells with a black circle (•), and the intact tumor cells with a black rhomb (♦).</p>

PubMed Open Access

First Principle Investigations of Long-range Magnetic Exchange Interactions via Polyacene Couplers

The electronic and magnetic properties of polyacenes become quite fascinating as the number of linearly conjugated benzene rings increases. Higher-order conjugated polyacenes develop radicaloid characters due to the transition of electronic structures from closed-shell to the open-shell system. Here we have investigated the role of such polyacenes as the magnetic coupler when placed between the two spin-sources based on nitroxy radicals. To do so, the magnetic exchange interactions (2J) are computed employing electronic structure theories, i.e. broken-symmetry (BS) approach within the density functional theory (DFT) as well as symmetry-adopted wave function based multi-configurational methods. In the former approach, various genre of exchange-correlation (XC) functionals such as generalized gradient approximation (GGA), meta-GGA, hybrid functional, constrained spin density (i.e. CDFT) and on-site Coulomb correlation corrected GGA+U functionals are adopted. All DFT based calculations estimate an exponential increase in 2J values with the length of the couplers, especially for the higher-order acenes. This is indeed an unexpected observation and also there is no experimental report is available in support of the DFT calculations. The complexity in the electronic structure enhances with the increasing number of benzene rings due to an increase 1 in near-degenerate or quasi-degenerate molecular orbitals (MOs) and also the reduction of the energy gap with the low-lying excited states. Consequently, it invokes a sever challenge in the computations of the magnetic exchange interactions in DFT. As an alternative approach, the wave function based multi-reference calculations, e.g. CASSCF/NEVPT2 methods are also adopted. In the later calculations, it has been realized that the π-orbitals of the couplers play a crucial role in the exchange interactions. For larger polyacenes (i.e. hexacene to decacene) such calculations become prohibitively expensive and rigorous as the number of π-orbitals increases, thus expanding the active space enormously. The limited active spaces calculations indicate quite strong exchange interactions, thus in principle, it supports the DFT observations of long-range magnetic exchange interactions, but not the exponential increase of 2J with the length of the couplers. In the current scenario, it is anticipated that a methodology that could account for the entire π-electrons in the active space such as CASSCF-DMRG like approach could resolve the issue.

first_principle_investigations_of_long-range_magnetic_exchange_interactions_via_polyacene_couplers

Introduction<!>Theoretical and Computational Methodology<!>Density functional theory based calculations<!>Spin-Constrained density functional theory (CDFT)<!>Hubbard-U corrected GGA+U functional<!>Multi-configurational calculations<!>Results and Discussions<!>Magnetic exchange interactions vs length of the couplers<!>Multi-reference calculations<!>Conclusions and outlook

<p>Molecular magnets based on the stable organic radicals have attracted enormous attention due to their potential applications in spintronics, 1 magnetic logic devices, 2 and quantum computers. 3 The non-magnetic spacer between the radical centers control the nature and strength of intramolecular magnetic exchange interactions. [4][5][6] The seamless π-conjugations between the radical centers and spacer play a vital role in magnetic couplings. [7][8][9] Strong magnetic exchange interactions are reported mostly for couplers with smaller lengths. 10 However, the fascinating long-range couplers desired 11,12 for magnetic and spintronic applications are relatively less explored. On the top of it accomplishment of long range strong magnetic exchange interactions in molecular system is another challenge. The exchange interactions through the larger molecular spacers were investigated by Matsuda et al. for various π-conjugated couplers and established a correspondence between electrical conductance and the magnetic exchange interaction. 13 Both the molecular conductance and the exchange interactions decay exponentially with an increase in distance between radical centers. They observed that the increased π-conjugations slow down the decay process. 14,15 The strength of the magnetic exchange coupling constants (2J), mediated through π-conjugated molecular spacers was found to decay exponentially with the length of the couplers. 13 The m-phenylene coupler unit is the most commonly used monomer fragment to device into stable ferromagnetic molecular units. Different nitronyl-nitroxide and nitroxide diradicals have been synthesized with the m-phenylene spacer. The influence of different structural parameters on magnetic exchange coupling has also been studied. A small distortion of planarity in the radical units and m-phenylene coupler changes magnetic coupling constant significantly. In a work by Rajca and co-workers, the lower limits for exchange couplings were found to be >139 cm −1 through EPR measurements in the solution phase. While, the same diradical systems exhibited different coupling values of > 208 cm −1 and 278-556 cm −1 in solid-phase depending on structural conformations. 16,17 In the present work, we intended to investigate the long-range magnetic exchange interactions between the two localized radicals centers coupled through linear polyacene spacers. Polyacenes are highly conjugated systems with fused benzene rings and possess a rich and intricate electronic structure. Bendikov et al. discovered that in larger acenes the closed-shell nature of the nonmagnetic acenes switches to open-shell electronic ground state. 18 They also predicted a transition in electronic structure from discrete molecular orbitals to band picture occurs for octacene onwards. 18 Such interconversion in the electronic structure in polyacenes has consequences on its ground state molecular properties such as optical 19 , magnetic 20 , vibrational, 21 and other spectroscopic properties 22 . Here such linear polyacenes are adopted as the spacers in coupling the conformationally restricted nitroxy-diradicals. The conformationally restricted di-radicals are specifically opted here to minimize the influence of torsional angles on magnetic coupling constant values. [23][24][25][26] The nitroxy radicals are connected at m-positions of the polyacene to have a ferromagnetic coupling between these radical centers. The spacer length is uniformly increased by adding a unit of fused benzene ring with 1 corresponding to diradical system with benzene coupler and 10 corresponding to diradical system with decacene coupler.</p><!><p>The complete series of diradicals i.e. 1-10 were optimized in triplet state within the density functional theory (DFT) framework adopting Becke's three-parameter and Lee-Yang-Parr's exchangecorrelation hybrid functional (B3LYP). 27 The atom centred polarized triple-zeta (def2-TZVP) 28,29 basis set was used for all the atoms as employed in quantum chemistry code ORCA 30 with the convergence criteria of 10 −8 Eh for each electronic steps. To speed up the calculations, resolution of the identity (RI) approximation along with the auxiliary basis set def2/J has been used with chain of spheres (COSX) numerical integration. 31 Since triplet was the ground state for all the diradicals, these B3LYP optimized geometries were used to compute the exchange coupling constants by performing single point calculations with different functionals.</p><p>The magnetic exchange interaction between the two magnetic sites A and B could be expressed by the Heisenberg-Dirac-Van Vleck (HDVV) spin Hamiltonian:</p><p>where 2J is the orbital-averaged effective exchange integral between the spin sites A and B, ŜA and ŜB being the respective spin operators of these sites. The positive sign of 2J indicates ferromagnetic and the negative sign indicates antiferromagnetic interactions. In diradicals, the exchange interactions could be extracted from the energy of the pure spin-states as</p><p>where E S=1 and E S=0 correspond to the energies of the triplet and singlet states respectively. In principle, the magnetic exchange interactions could be obtained applying Eq.2 for wave function based multi-reference ab initio calculations of spin multiplet energetics. However, in the spin-unrestricted Kohn-Sham UKS approach, it is not possible to obtain the pure singlet state because of the broken natural spin symmetry for open-shell singlets. This results in a poor description of the spin states energies as well as spin-contaminated states. [32][33][34] Ginsberg and Noodleman proposed a strategy to extract the exchange-interactions by mapping the eigen states of the Heisenberg spin Hamiltonian with the high spin-states and broken-symmetry states as obtained from the UKS solutions. 35 This is popularly known as the broken-symmetry (BS) method or more precisely standard broken-symmetry DFT method. The BS solution is not an eigenfunction of the Hamiltonian, but is an admixture of the singlet and triplet states. In principle, the singlet-state requires multi-determinant representation of the wave function which cannot not be obtained accurately in a single-determinant approach. However, standard broken-symmetry DFT is found to be a successful method in extracting the exchange interactions (2J) for the organic diradicals. 36,37 The</p><p>is generally used to extract the 2J values in the broken-symmetry calculations and popularly known as Ginsberg-Noodleman-Davidson 35 approach. It produces accurate exchange interactions for the systems in which the Heisenberg exchange including the ligand-to-metal spin polarization and direct exchange interactions contribute predominantly in the magnetic exchange interactions. The overlap integral between the magnetic orbitals was neglected in this formulation. In the current work, within DFT regime Eq. 3 were used for the evaluation of the exchange interactions whereas Eq. 2 was used to evaluate 2J from multi-reference based methods.</p><!><p>The standard broken-symmetry DFT calculations were carried out adopting the proposition of Ginsberg and Noodleman et al. [38][39][40] that was also advocated by several groups. [41][42][43] The quantum chemistry code ORCA was used for all the standard broken-symmetry DFT calculations. 30 The computed 2J values for the different magnetic molecules strongly depend on the applied exchangecorrelation functional within the DFT framework. Thus to obtain an in-depth understanding of the exchange interactions obtained from DFT calculations, a number of functionals from Jacob's ladder are selected. These include GGA (BLYP and PBE), meta-GGA (TPSS and M06-L) and hybrid ( B3LYP (20% HF exchange), PBE0 (25% HF exchange) and M06 (27% HF exchange))</p><p>functionals. In addition to the aforementioned functionals, two other popular functionals (CDFT and GGA+U) but relatively less used by the community for organic diradicals are applied and described in the following sections.</p><!><p>In high-symmetric and planner magnetic molecules, the singly-occupied molecular orbitals (SO-MOs) obtained in the DFT calculations are generally over delocalized due to self-interactions error (SIE) in DFT. Hence the spin moments aslo get delocalized from their localized radical centers.</p><p>Such unphysical delocalization leads to the spurious description of the chemical bondings as well as associated molecular properties, especially for open-shell magnetic molecules. Recently, it has been observed that the spin-constrained DFT (CDFT) could rectify this to a certain extent. 44 In CDFT, the ↑↓ state is obtained directly by minimizing the KS energy E KS subject to the constraint that the spin of A i.e., the difference between the number of ↑ N α A and ↓ N β A electrons on A should be S A , while the spin of B should be S B :</p><p>For the ↑↓ state, S A is positive while S B is negative and then Löwdin population is used to define</p><p>To perform the constrained minimization, two Lagrange multipliers were introduced γ A , γ B</p><p>and a new functional of electron density ρ is optimized:</p><p>where density is a sum of orbital densities,</p><p>In the CDFT approach, this new functional W was optimized with respect to the Lagrange multipliers in addition to the spin-densities. CDFT single-point calculations were performed by localizing the spin-densities within the nitroxyl radical using B3LYP/6-31G(d) method as implemented in NWChem quantum chemistry code. 45,46 One has to be careful while performing the CDFT calculations, as the outcome depends strongly on the constrained spin-moments. Thus, to succeed in CDFT calculations, one has to identify the exact amount of localized spin-moments within certain spatial regions in the molecule. This is indeed a cumbersome task, but a systematic procedure was reported in our previous work. 47,48 In this work we have constrained 1.0µ B spin moments within the -NO . regions for all the diradicals (1 to 10).</p><!><p>The GGA+U approach has been introduced as a method to treat excessive delocalization of d and f electrons of transition metal and rare earth element as predicted by standard GGA exchangecorrelation potentials. The "U" treats the strong on-site Coulomb interaction of localized electrons with an additional Hubbard-like term. The GGA+U functional may be formulated as follows:</p><p>where ρ(r) is the electronic density, E GGA [ρ(r)] is the standard GGA energy functional, n lσ mm are generalized atomic occupations (i.e. density matrix elements) with spin σ associated to the I atom, and n lσ is the sum of the occupations corresponding to all m and m orbitals and E U [n lσ mm ] represents the correct and accurate value for the on-site correlation energy between electrons in the m and m orbitals which belong to a given atomic shell (s, p, d, f , ...) centered at the same I atom. However, E GGA [ρ(r)] already contains an approximate estimate of electron correlation effects, including the on-site correlation energy between electrons in these m and m orbitals, hence a term accounting for the GGA estimate of this electron-electron interaction, E DC [n lσ ], must be subtracted to avoid double counting when using a Hubbard scheme to represent this correction as in the following equation:</p><p>Here, the U e f f = U − J has been used to impose Hubbard's parameter on the radicals as proposed by Dudrarev et al. 49 where U is the on-site Coulomb term while J is the site exchange term (not the same as magnetic exchange coupling constant). Most often, the U e f f parameter in these approaches has been considered as an empirical parameter introduced to correct the deficiencies of the GGA exchange-correlation potentials in describing the charge (and spin) distributions of atoms involving d or f electrons. This on-site Coulomb correction can be applied to 2p electrons as well.</p><p>Herein, to take into account the delocalization of spin from radical centers PBE+U [50][51][52] was used to calculate the exchange coupling constant by adding on-site Hubbard's U parameter on the -NO . radical centers. The pseudopotentials with projector augmented-wave (PAW) 53 methods were used with a kinetic energy cutoff of 400 eV. The Hubbard-U parameter was empirically varied in the range of 2 to 5 eV for all the diradicals. The exchange interactions was obtained applying the broken-symmetry methods. 54,55 The GGA+U calculations were performed in VASP 56 by placing the diradical in a periodic large rectangular cuboid box (for e.g. 17.00 x 17.00 x 12.00 Å3 for 1 to 27.00 x 27.00 x 15.00 Å3 for 10) to avoid spurious interaction between the periodic images. The cell size chosen for all the diradicals was such that the images were at least 10 Å apart in all the directions. The exam same amount of on-site Hubbard-U was applied to the 2p-orbitals of both N and O-atoms, that primarily host the unpair electron and behave as spin-centers. The 2J values discussed in the main text are with U e f f =2 eV, the rest are provided in SI (Table S13).</p><!><p>The multi-configurational calculations are essential for computing the magnetic exchange interactions (2J) in the wave-function based methods mainly to account the electron-correlation energies, especially for the highly-correlated magnetic systems. The presence of low-lying excited states, quasi-degenerate or degenerate molecular orbitals (MOs) and the multi-determinant nature of the open-shell singlet (anti-ferromagnetic) states are the other reasons to consider the multiconfigurational methods. Thus, in addition to the DFT calculations, we also computed the magnetic exchange interactions employing the complete active space self-consistent field (CASSCF) method that accounts the static electronic correlations for the chosen active space. [57][58][59] The static correlations are mainly long-range in nature while the dynamical correlations are short-ranged type. The effect of dynamical electronic correlations on the magnetic exchange interactions is accounted through the n-electron valence state perturbation theory (NEVPT2). 60,61 The NEVPT2 methods incorporate the dynamical correlations through the second-order perturbative treatment of the CASSCF optimized wave-functions in the selected CAS spaces. All the CASSCF and CASSCF/NEVPT2 calculations with different active spaces were performed in ORCA software package. 30</p><!><p>The computed exchange interactions for all 1-10 diradicals employing different functionals are tabulated in Table 1. The conformationally restricted nitroxy diradical with m-phenylene coupler similar to diradical 1, was reported by Rajca and co-workers. Employing ESR spectroscopy, the authors could observe the lower limit of the magnetic coupling constants of >202 cm −1 . 16 However, the 2J's obtained with different DFT functionals for 1 are substantially varied between 174 -1356 cm −1 for different functionals. The lowest value for coupling constant is observed with CBS-DFT, but all other functionals produce 2J between 580 to 1356 cm −1 for diradical 62 It was observed that the standard BS-DFT consistently overestimate the exchange interactions, and this is a general trend in DFT based calculations, specially for the radicals in which the spin-centers are coplanar with the conjugated molecular spacers. [63][64][65] The computed 2J values using the DFT functionals as tabulated in Table 1 for all the diradicals 1-10 follow a general trend such as 2J(CBS/CDFT)</p><p>The other interesting observation in Table 1 is the <S 2 >, that increases (for both the spinstates) with the increase in number of benzene rings. The deviations of <S 2 > from its expected values indicate the spin-contamination that generally occurs due to the mixing of low-lying states.</p><p>To extract the 2J values from the strongly mixed states, Yamaguchi's proposition of spin-projection method is also applied 41 . (Table S2, SI)</p><p>The comparisons of <S 2 > values between the BS-DFT and CBS-DFT indicates the resulting spin-sates are closer to pure spin-state for CDFT (till 5) compared to hybrid functionals in the standard DFT formalism. In this context, we would like to draw the attention of the readers to surprisingly better performance for the GGA and meta-GGA functionals compared to the hybridfunctionals which is also reflected in their <S 2 > values as well. The better performance for GGA and meta-GGA functionals in terms of 2J and <S 2 > values are incidental due to less mixing with the excited states and also due to more spin − spilling that will be discussed in the later sections.</p><!><p>In previous reports Ali et al. 66 and Bhattacharya et al. 67 investigated the role of linear acene couplers in magnetic exchange interaction, that indicated the increase in coupling constant with increase in the lengths of the couplers, though the investigations in both of these studies were limited up to the pentacene couplers only. In this study, the density based methods (irrespective of their genre of exchange-correlations functionals) we observed that the magnetic exchange interactions increases exponentially for the higher-order of polyacene couplers. (See Table 1 and Fig. 2) This observation is indeed very unusual and also goes against the general understanding of exchange interactions that usually decay either slowly or strongly depending on the nature of the couplers. A similar observation of increase of 2J values with the length of cumulene couplers was reported by Sarbadhikary et al. using broken-symmetry DFT calculations. 68 However, to best of our knowledge, till date there is no experimental confirmation/indication is available in favor of such theoretical observations. On the other hand Nishizawa and co-workers have reported the exponential decay of the magnetic interaction along the length of the couplers. 14 Thus, the DFT observations of exponential increase in 2J values with the lengths of the couplers seems to be paradoxical, that needs an in depth investigations unravelling the electronic structures of the higher-order polyacene couplers.</p><p>The spin density of broken symmetry states as obtained from standard BS-DFT (B3LYP) formalism is compared with those obtained from CBS-DFT(B3LYP). (see Table 2) Visually quite similar spin-densities are observed for diradicals 1 to 8 and deviations are noted from diradical 9</p><p>onwards. For diradical 10, an abrupt failure of BS-DFT is observed in producing opposite spindensities localized at both the -NO magnetic centers in the antiferromagnetic states. In this respect, CDFT produced better 2J values compared to standard broken-symmetry DFT. However, both the methods produced unphysical high exchange interactions for the diradicals with larger polyacene couplers.</p><p>CDFT(CBS-DFT) turned out to be a successful and convenient method even for 10 especially to obtained a desired magnetic state. Moreover, spin density plots also reveal that for the larger acenes (6/7 onwards) the central parts of the spacer slowly gained spin-polarization with the increased number of benzene rings. This is a direct consequence of the increasing open-shell nature of the couplers and development of radicaloid characters in the spacers. Therefore, a considerable amount of spin-density is found to be in the central region of the spacer in 10. In the previous section, it was established that the GGA and meta-GGA seemed to perform better than hybrid functionals, however, these functionals in the BS-DFT approach even failed to achieve the desired BS state for diradical 9 also. (Table S10) Whereas CDFT overcomes such issues and the desired BS states are achieved for all the diradicals.</p><p>The <S 2 > values of the converged BS and high-spin (HS) states for diradicals 1 to 4 indicate the smooth convergence of the wave functions with expected values of ∼2.00 for triplet HS state and ∼1.00 for BS state. The deviation in <S 2 > generally occurs due to spin contaminations. As the ground spin-state of all the studied diradicals is triplet, we will restrict our discussion only for generacy of the molecular orbitals as n increases. Figure 3 depicts the orbital energy levels in the energy window of -7.5 eV to 0.5 eV. This orbital energy diagram plot indicates that the singly occupied molecular orbitals (SOMOs), the orbitals containing the unpaired electrons responsible for the magnetic properties of the molecules are almost pinned at same energy for all the diradicals 1 to 10. The major change is observed for doubly occupied and un-occupied orbitals wherein the density of MO's increases with n. The other impact with the increased densities of the orbitals is that with unoccupied energy levels rapidly gets stabilized and come in close proximity to the pinned SOMOs. Thus HOMO(SOMO)-LUMO energy gaps also decreases along the increase number of benzene rings from 1 to 10. All these facts indicate the increased degeneracy of the MOs. This is an issue as the single-determinant methods e.g. HF or DFT face dificulties. 69 Employing CDFT, the <S 2 > values improved slightly, but a remarkable improvement of the exchange interactions could be observed for the smaller acene couplers 1-5 (see Table 1). We observed that CDFT also starts to deviate as the conjugation of the couplers increases i.e. diradical 5 onwards. This deviation could be attributed to the fact of natural breakdown of the single-determinant based wave functions as the degeneracy of the molecular orbitals increases as mentioned earlier. Even though CDFT method did successfully converged to the broken symmetry state for 10, the 2J value is still comparable to that obtained from standard broken-symmetry DFT. The other reason for the spin-contamination is the mixing with the low-lying excited spin state especially for the longer polyacene couplers. For example, <S 2 > BS for 10 approximately 2.00 is observed using hybrid functionals, which is almost equivalent to the expected <S 2 > value for a triplet state. For a diradical system the genesis of broken-symmetry occurs due to mixing of pure singlet and triplet multiplets. However, the <S 2 > values of 10 for the BS/HS states are ∼1.95/3.32 indicating that the resulting ground spin state is also not a pure spin state due to mixing with the low lying excited states. The success of the broken-symmetry methods are evident when at least the high-spin state is free from any spin-contaminations i.e. pure-spin state. However, here we observed that the higher-spin states are not pure spin-sate as it get mixed with the quintet (S=2) state. Furthermore, the large spin contamination in both, the triplet as well as the brokensymmetry state is also due to quintet state which can be understood from the plot of quintet state energies relative to triplet state (which is ground state for all the cases) as shown in Figure 4. All the considered functionals reveal qualitatively similar trend of lowering of quintet state energy with respect to the ground state of the molecular systems (Figures S1-S7 and S10-S11, SI). Upon increasing the coupler length, the corresponding energy gap between the quintet-triplet states and quintet-BS states decreases respectively. Therefore, with the increase in coupler length, quintet state also starts to play an important role by being energetically close to the ground and BS state and thus, mixing with them. Thus far, it can be concluded that the orbital degeneracy of individual molecule and mixing from higher excited state are crucial factors that can not be directly taken into account.</p><p>The acenes are well known to develop open-shell character upon increase in its length, with open-shell-singlet (OSS) as its ground state. This increasing radical character of the couplers to some extent could also contribute in the spin-contamination of BS and triplet states of the diradicals 1-10. Cano et al. have shown the effect of increasing open shell character of acenes escalating the spin contamination in both the triplet and BS states with dinuclear copper(II) systems. 70 Applying ab initio calculations here we have quantified the radicaloid characters (y-value) of the pristine polyacene as well as for polyacene in diradicals 1-10 where it acts as a coupler between -NO radical sites and tabulated in Table 3. This has been calculated following the work of Yamagucghi et al. 71 and details of calculations could be found in SI. As illustrated in Table 3 for pristine polyacene, the radicaloid nature increases with the increase of the number of benzene rings from 0.004 for benzene to 0.92 for the decacene. However, on placing between the spin centers, i.e., polyacene as coupler, the radical character decreases in comparison to pristine polyacene such that it significantly reduces to y c = 0.73 in 10 from y p = 0.92 in pristine form of decacene. This decrease in the radicaloid characters of the polyacene in the diradicals clearly reveals the spin spilling from the spin-center to the polyacene couplers. The phenomena of spin spilling can be visually anticipated from spin density plots (Table 2) and also from the value of net µ B on the radical centers (Figure 5). Comparing the magnetic moments on the radical sites, it is observed that the hybrid functionals perform better in localizing the spin moments. Analyzing the spin density plots in Tables 2, S10 and S14, spin density in coupler can be seen increasing with maximum spin moment on carbons near the −NO . spin centers. For higher conjugated systems i.e. 9 and 10, the delocalized spin density in the center of the coupler owns to the OSS nature of the coupler. CBS-DFT method surpasses the phenomena of spin-spilling as the magnetic moment value remains constant for all the diradicals. The ease in spin-spilling is facilitated for studied systems by the fact that the magnetic orbital (p z ) are in plane with the π orbitals of the spacer.</p><p>To summarize the section, the calculations of 2J adopting the density based methods and mapping of the Heisenberg spin Hamiltonian is quite challenging task due to the complicated electronic structures of the higher order polyacene spacers. Thus one has to be careful to conclude about the 2Js in such cases as obtained from the DFT calculations. To circumvent, the observed issues in DFT, we also performed the wave function based multi-reference calculations for these systems but that faces different challenges, which has been described in the following section. : Magnetic moment on one of the radical sites (i.e. the site on which the magnetic moment reverses in BS state) with different functionals used. The positive and negative values correspond to the magnetic moment on -NO in HS (triplet) and BS state respectively. In few cases proper BS state was not achieved for diradicals 9 and 10 providing same sign for spin density on both the radical centers in the structure. The magnetic moments as obtained in BS state for those cases is marked with star by reversing the spin moment sign.</p><!><p>The multi-configurational CASSCF/NEVPT2 methods are quite successful in calculations of the 2J values. However, the choice of the active space is always a matter of concern for such calculations. In our previous work 65,72 and from several report in the literature, it has been realized that the minimal active space i.e. CAS(2,2) are quite successful in computing the magnetic exchange interactions in various organic diradicals when both SOMOs involves both the radical sites and coupler as well. In the minimal CAS approach, only the magnetic orbitals i.e. singly-occupied molecular orbitals (SOMOs) and the two unpaired electrons responsible for the magnetic centers are considered. The inclusion of static and dynamical correlations within the CAS(2,2) space remarkably produces the large part of the exchange interactions. Apart from the estimation of 2J values by CASSCF(2,2)/NEVPT2, we also realized that increasing the active space does not necessarily improves the 2J values remarkably. We trace down the issue as the imbalance accounting of the correlation energies between the ground state and the low lying excited state with the increases of the active space. The imbalance occurs as the ground state becomes more stable compared to the excited states and as a results it produce large 2J values. The computed exchange interactions in the multi-configurational methods such as CASSCF and CASSCF-NEVPT2 with different active spaces for all the ten diradicals are given in Table 4.</p><p>The minimal active space CAS(2,2) calculations were performed using the magnetic orbitals i.e.</p><p>SOMOs localized on the -NO . magnetic subunits. The computed exchange interactions employing CASSCF/NEVPT2 methods produces 2J values that decrease with the length of the couplers. This is in complete contrast to the prediction by DFT based calculations, where an exponential increase in 2J's with respect to the length of the couplers is observed. As mentioned earlier, the CASSCF(2,2)-NEVPT2 was observed to be an acceptable method to compute the magnetic exchange interactions for organic diradicals. However, a close look into the magnetic orbitals reveals that with the increase of the couplers lengths, the overlap of the magnetic orbitals reduces, thus the exchange interactions expectedly decreases, as CAS(2,2) space does not take into account the contribution from other π-orbitals.Thus the outcome of the CAS(2,2) calculations is artifact of the limited active space and could not be considered as physical observations, thus the extended active space are necessary for such polyacene couplers.</p><p>Therefore, further active space (4,4) and (6,6) are chosen by incorporating the orbitals contributing from the center of the acene couplers. For all the diradicals 2J values for the individual diradical increases with the increasing active space (except for 4 and 6 with (6,6) active space), this observation matches with the previous discussion and reports as well. The same holds true for CASSCF-NEVPT2 method as well. On increasing the size of active space to (4,4) and (6,6), the complete series show an even-odd effect. For (4,4) space, in odd-acene series it is the central benzene unit that contributes majorly. While, in even-acene series the central naphthalene unit of the individual systems play a vital role. Along the individual series with (4,4) active space we observe a continuous decrease in the exchange coupling constant values. Similar even-odd effect is observed even when the active space is (6,6), odd-acene series has major role of central anthracene unit in the individual spacers. Therefore, along odd-series with (6,6) space too we observe a continuous decrease in the magnetic exchange. The scenario is somewhat different with even-acene series in (6,6) space. Diradicals 2, 4 and 6 have central naphthalene unit contribution to the active space while for 8 and 10 this contribution is from central tetracene unit. This revelation is important here, as in literature the boundary of the originating open-shell ground state for oligoacene is roughly established at heptacene 18,73,74 and the delocalized radical also has the similar qualitative appearance. 75,76 The exchange coupling constant values in this even-acene series decreases in going from 2 to 4 to 6 but increases in going from 6 to 8. This sudden rise in the coupling constant could possibly the effect of the open shell diradical nature of the coupler that might come into play for such longer polyacene.</p><p>The study so far is incomplete to arrive at a conclusion for such highly correlated systems.</p><p>Even in the limited CAS space it was difficult to go beyond (6,6) as the larger systems became not only more resource extensive but also the defining criteria to choose the active space was no longer simple. This still opens the door towards exciting physics that might be at play and may help in designing organic molecular systems with intrinsic radical character that might act as boosters to enhance the communication between the spin centers over long distances. Still, its a long way to go and other higher-level studies like DMRG along with experimental realization are required to have better understanding for such systems.</p><!><p>The first principle calculations of magnetic exchange interaction for organic diradicals with higherorder linear polyacene couplers turned out to be the most challenging task due to the intriguing electronic structures of the spacers. The radicaloid characters, intrinsic open-shell electronic structure, presence of quasi-degenerate MOs, and the low-lying excited spin-sates add up to the complexity in the polyacene couplers. The order of complexity increases in many-folds with the increase of π-conjugations as the fused benzene rings increases. We further realized that the conformationally restricted planar radical moieties that are conjugated with couplers π-conjugations of the spacers facilitate spin-spilling into the spacers, which further invoke the additional challenges in the density-based calculations. However, CDFT prevents such spin-spilling and produces slightly smaller (or may be better) 2Js in comparison to other BS-DFT methods. Eventually, CDFT for higher-order polyacene couplers also converged as of BS-DFT and produces an exponential increase in 2J values with the increase length of the couplers. The exchange interactions seems to be strongly influenced by radicaloide nature and the spin-spilling into the spacer. This indicates a much richer physics of exchange interaction mechanism that needs to be understood appropriately with further studies. Additional developments in theoretical methodologies to circumvent the observed issues in DFT are also foreseen.</p><p>The multi-configurational methods including dynamical electronic correlations are also severely limited especially for the higher-order acenes due to its requirements of prohibitively large and tedious active spaces. The minimal active space calculations reproduce opposite trends of 2J values to DFT calculations, however, it clarifies the importance of the CAS spaces steaming especially form the central part of the couplers. The extended active space indeed reproduces a much higher 2J values compare to CAS(2,2) calculations. However, the systematic inclusion of different MOs in the CAS space and their resulting 2J indicates a reasonably larger exchange interaction even with the larger decacene coupler. To obtain more reliable or exact numerical values for 2J, the complete pi-electrons should be accounted in the CAS space. Currently, we are aiming for CASSCF-DMRG calculations that is capable to handle quite a large active space as the extension of this work.</p><p>In a nutshell, despite the pragmatic difficulties in the computational methodologies, substantially strong-ferromagnetic exchange interactions (if not the exponential increase) for the higherorder acene couplers have been realized. The magnetic couplings through the OSS polyacene couplers are long-range in nature. Such systems will play a pivotal role in designing the ferromagnetic material with long-range magnetic orders. The participation of the spilled-spins (electrons) into the conduction band enhances such long-range exchange interactions between the spin-sites localized in the distance apart. Thus the synthesis of such organic diradicals with the polyacene couplers, will not only resolve the discussed dilemma, but it will open up a new avenue for the long-range magnetic materials.</p>

ChemRxiv

Molecular basis for the substrate stereoselectivity in Tryptophan Dioxygenase

Tryptophan dioxygenase (TDO) and Indoleamine 2,3 dioxygenase (IDO) are the only two heme-proteins that catalyze the oxidation reaction of tryptophan (Trp) to N-formylkynurenine (NFK). While human IDO (hIDO) is able to oxidize both L and D-Trp, human TDO (hTDO) displays a major specificity towards L-Trp. In this work we aim to interrogate the molecular basis for the substrate stereoselectivity of hTDO. Our previous molecular dynamics simulation studies of Xanthomonas campestris TDO (xcTDO) showed that an H-bond between T254 (T342 in hTDO) and the ammonium group of the substrate is present in the L-Trp-bound enzyme, but not in the D-Trp bound enzyme. The fact that this is the only notable structural alteration induced by the change in the stereo structure of the substrate prompted us to produce and characterize the T342A mutant of hTDO to evaluate the structural role of T342 in controlling the substrate stereoselectivity of the enzyme. The experimental results indicate that the mutation only slightly perturbs the global structural properties of the enzyme, but it totally abolishes the substrate stereoselectivity. Molecular Dynamics simulations of xcTDO show that T254 controls the substrate stereoselectivity of the enzyme by (i) modulating the H-bonding interaction between the NH3+ group and epoxide oxygen of the ferryl/indole 2,3-epoxide intermediate of the enzyme, and (ii) regulating the dynamics of two active site loops, loop250\xe2\x80\x93260 and loop117\xe2\x80\x93130, critical for substrate-binding.

molecular_basis_for_the_substrate_stereoselectivity_in_tryptophan_dioxygenase

<!>Expression and Purification of the wild type hTDO<!>Steady-State Activity Assay<!>Spectroscopic Measurements<!>Starting structures for classical Molecular Dynamics Simulations<!>Classical Molecular Dynamics Simulations<!>Optical absorption and resonance Raman spectra of the T342A mutant of hTDO<!>Activity of the T342A mutant of hTDO towards L and D-Trp<!>Molecular dynamics simulations of the ternary complex of xcTDO<!>Molecular dynamics simulations of the ferryl/indole 2,3-epoxide intermediate of xcTDO<!>Mechanistic implications<!>Conclusions

<p>Tryptophan dioxygenase (TDO) is a heme-containing enzyme that catalyzes the conversion of L-Tryptophan (L-Trp) to N-Formyl kynurenine (NFK), which represents the first and rate-limiting step of the L-Trp catabolism through the kynurenine pathway.1 TDO is a ubiquitous enzyme found in bacteria, insects and mammals. In mammals, it is expressed mainly in the liver, where it is responsible for L-Trp processing that ultimately leads to the biosynthesis of NAD and NADP. 2, 3.</p><p>Indoleamine 2,3-dioxygenase (IDO) is also a heme-containing enzyme that catalyzes the same oxidation reaction of L-Trp. Contrary to TDO, in mammals IDO is expressed in all tissues other than the liver. Recently, significant efforts have been put forth to unravel the reaction and inhibition mechanisms of IDO and TDO, as promoted by the discovery that IDO plays a pivotal role in cancer immune escape.4, 5, 67, 8 Along this line, it has been demonstrated that the combination of a IDO inhibitor, 1-methyl Trp, and cytotoxic chemotherapy leads to significant tumour regression in mouse model systems.4 In this context, it is important to develop potent IDO inhibitors that do not interfere with normal TDO function.</p><p>The crystal structure of mammalian TDO is not known; nonetheless those of two bacterial isoforms of TDO, from Xanthomonas campestris (xcTDO, PDB code: 2NW8)9 and Ralstonia metallidurans (rmTDO, PDB code: 2NOX),10 have been published in 2007. In the case of xcTDO, the enzyme was crystallized in both substrate-free and L-Trp-bound forms. In the structure of the L-Trp bound form, several tight contacts between L-Trp and the enzyme are evident (Fig. 1), including H-bonding interactions between the carboxylate group of L-Trp and Y113/R117, the ammonium group of L-Trp, T254 and the propionate A group of the heme, and the indoleamine group of L-Trp and H55, as well as the hydrophobic interactions between L-Trp, F51, L120 and F116 (it is noted that F116 is not shown in Fig. 1 for clarity). Structure-based sequence alignment data suggest that most of the key interactions between the substrate and the enzyme observed in xcTDO are conserved in human IDO (hIDO).9, 11, 12 Consistently, our hybrid Quantum Mechanics and Molecular Mechanics (QM-MM) studies13, 14 showed that xcTDO and hIDO carry out the Trp dioxygenation reaction with a similar ferryl-based mechanism. Based on this mechanism, the heme iron bound dioxygen is first inserted into the C2 atom of Trp to generate a ferryl/indole 2,3-epoxide intermediate. The subsequent ring opening reaction of the epoxide, catalyzed by a proton transfer from the ammonium group to the epoxide, triggers the insertion of the ferryl oxygen to the C2 atom to generate the NFK product. In this scenario, the proper positioning of the ammonium group of the substrate, as well as the electronic environment surrounding it, are believed to be critical for the catalysis.</p><p>Although TDO and IDO exhibits high structural similarity and follow a similar dioxygenase mechanism, they exhibit several intriguing differences in the substrate stereoselectivity: hTDO shows comparable affinities towards L-Trp and D-Trp, but the kcat for L-Trp is 10-fold higher with respect to D-Trp;12 in contrast, hIDO displays ~170-fold higher affinity towards L-Trp with respect to D-Trp15, but the kcat are similar for the two stereoisomers. The structural factors leading to the differences in the substrate stereoselectivity of the two enzymes are unclear. Nonetheless, our previous molecular dynamics (MD) simulation data16 suggest that most of the interactions between the substrate and the enzyme in xcTDO are conserved in the L-Trp and D-Trp-bound complexes, except that the strong H-bond between the ammonium group of L-Trp and the OH group of T254 (equivalent to T342 in hTDO) is significantly weakened in D-Trp-bound enzyme, implying that T254 plays a critical role in controlling the substrate stereoselectivity of xcTDO.</p><p>To define the role of T342 in the substrate stereoselectivity of hTDO, we have produced and characterized the T342A mutant by using resonance Raman and optical absorption spectroscopies, and determined its activity towards L-Trp and D-Trp. Our data show that the T342A mutation only causes minor changes to the global structure of the enzyme, but it totally abolishes the substrate stereoselectivity. To unveil the structural perturbations to the enzyme underlying its absence of substrate stereoselectivity, the experimental work was complemented with classical MD simulations. As the crystal structure of hTDO is not available, the simulations were carried out with xcTDO. The simulation data of the L- and D-Trp bound O2 complexes, as well as the corresponding ferryl/indole 2,3-epoxide intermediates, indicate that T254 controls the substrate stereoselectivity of xcTDO by modulating the H-bonding interaction between the ammonium group and epoxide oxygen of the ferryl/indole 2,3-epoxide intermediate, and by regulating the interactions between two critical loops, loop250–260 and loop117–130, that sequester substrate in the active site.</p><!><p>The details of protein expression and purification are described elsewhere.12, 17 Briefly, the hTDO protein, with N- and C-terminal tails truncated, plus a 6X-His tag extension at the C-terminus, was over-expressed in E. coli BL21 Star (DE3) cells by using the pET30b vector (Stratagene, La Jolla, CA). The transformant was selected from a single colony on a LB kanamycin agar plate and was used to inoculate a LB medium, supplemented with 50 μg/mL kanamycin. This starter culture was grown overnight at 37 °C, and was subsequently used to inoculate a 1 L of LB medium. The resulting E. coli culture was grown at 37 °C in a shaker at 250 rpm until the optical density at 600 nm reached ~0.8. The expression of hTDO was induced by Isopropyl β-D-1-thiogalactopyranoside (IPTG), with a final concentration of 1 mM. An aliquot of hemin, with a final concentration of 8–10 μM, was added to the culture to ensure the complete incorporation of the heme prosthetic group into the recombinant protein. The culture was grown at 25 °C for additional 6–8 hrs in a shaker at 150 rpm. The cells were harvested by centrifugation and stored at −20 °C until use.</p><p>The recombinant protein was purified by affinity chromatography with a Ni-NTA column (Novagen). The protein was eluted with 250 mM imidazole (Sigma) in 50 mM phosphate buffer (pH 7.8) and 250 mM KCl. To stabilize the protein, 10 mM L-Trp and 5% glycerol was present throughout the purification procedure. To ensure that the protein was in its ferric state, it was oxidized with potassium ferricyanide, and subsequently passed through a sephadex G25 column to remove the ferricyanide, L-Trp, imidazol and glycerol. The purity of the protein was confirmed by SDS PAGE analysis as shown in Fig. S2. The protein thus collected was stored in 100 mM phosphate buffer (pH 7.5) with 150mM KCl at −80°C until use.</p><p>The T342A mutant was created by using the QuikChange II kit (Stratagene) with the following primers: sense primer: 5′-CAGCAAAGCTGGCGCCGGTGGTTCCTC-3′ and antisense primer: 5′-GAGGAACCACCGGCGCCAGCTTTGCTG-3′. The sequence was verified by DNA sequence analysis. The mutant was cloned, expressed, and purified by using the same protocol as that described for the wild type protein.</p><!><p>For the activity assay, the ferric hTDO was rapidly mixed with sodium ascorbate (100-fold excess with respect to the protein) in the presence of a desired amount of L-Trp in pH 7.0 phosphate buffer (100 mM). The temperature was controlled with a water bath at 25 °C. The final concentration of hTDO was 0.5 μM. The reaction rate was followed by monitoring the product formation at 321 nm (ε = 3750 M−1cm−1 for N-formyl kynurenine).12, 18 Two representative kinetic traces obtained with 8 mM L- and D-Trp are shown in Fig. S3. The initial product formation rate was plotted as a function of L-Trp concentration; the data were analyzed by Michealis-Menten curve fitting with Origin 6.1 software (Microcal Software, Inc., MA).</p><!><p>The optical absorption spectra were taken on a spectrophotometer, UV2100, from Shimadzu Scientific Instruments, Inc. (Columbia, MD) with a spectral slit width of 1 nm. The resonance Raman spectra were taken on the instrument described elsewhere.19 Briefly, the 413.1 nm excitation from a Kr ion laser (Spectra Physics, Mountain View, CA) was focused to a 30 μm spot on a spinning quartz cell rotating at 1000 rpm. The scattered light, collected at a right angle to the incident laser beam, was focused on the 100 μm-wide entrance slit of a 1.25 m Spex spectrometer equipped with a 1200 grooves/mm grating (Horiba Jobin Yvon, Edison, NJ), where it was dispersed and then detected by a liquid nitrogen-cooled CCD detector (Princeton Instruments, Trenton, NJ). A holographic notch filter (Kaiser, Ann Arbor, MI) was used to remove the laser line. The Raman shift was calibrated by using indene (Sigma) as a reference. The laser power was kept at ~5 mW for all measurements, except that used for the CO complexes, in which it was kept at <1 mW to avoid photodissociation of the heme-bound CO. The acquisition time was 30 min for all except that of the super high frequency spectra of the CO-bound complexes in which the acquisition time was 180 min. The ferrous derivative was prepared by reducing the ferric protein, prepurged with N2 gas, with sodium dithionite under anaerobic conditions. The CO-bound ferrous complexes were obtained by gentle purging CO gas on the surface of the solution containing the ferrous form of the enzyme under anaerobic conditions. The concentration of the protein samples used for the Raman measurements was 50 μM in pH 7.0 phosphate buffer (100 mM).</p><!><p>Due to the fact that the crystal structure of hTDO is not available, we have performed the simulations on the xcTDO structure. Sequence comparison of hTDO and xcTDO with 7 other TDOs from various organisms (Fig. S1) indicates that most of the residues involved in Trp binding are conserved, including H76 (55 in xcTDO), R144 (117 in xcTDO), F72 (51 in xcTDO), F143 (116 in xcTDO), L147 (120 in xcTDO), with the exception of Y113 of xcTDO that is replaced by a Phe residue in all the analyzed mammalian and insect sequences. Additionally, the loop250–260 and loop117–130 involved in the observed conformational change in the T254A mutant of xcTDO display high sequence homology, with all relevant residues involved in the interactions conserved. In this context, xcTDO structure provides a good model of the human protein justifying its use for the MD simulations.</p><p>The structure of the ternary complex of xcTDO was obtained based on the crystal structure of the ferrous L-Trp-bound xcTDO (PDB code: 2NW8)9, as described eleswhere.13, 16 Briefly, a one-subunit model of xcTDO was constructed by removing residues E19 to S35 from subunit A, and adding residues R21 to S35 of subunit B, which corresponds to a short helix that penetrates into the structure of the A subunit. The structure of D-Trp-bound xcTDO was obtained by a docking protocol described in reference.16 The structure of the ferryl/L or D-indole 2,3-epoxide intermediate of xcTDO was obtained from the QM-MM simulations of the corresponding ternary complexes.13, 20 Starting from the structures of the ternary complexes and the ferryl/indole 2,3-epoxide intermediates of the wild type enzyme, we constructed the corresponding T254A mutant derivatives in-silico. These structures were used for all the MD simulations.</p><!><p>The starting structure for each complex, was immersed in a pre-equilibrated octahedral box of TIP3P water molecules. The standard protonation state at physiological pH was assigned to ionizable residues. Special attention was paid to the protonation states of histidines, which were assigned on the basis of the H-bonding patterns with neighboring residues. All simulations were performed at 300 K and pressure of 1 bar using Berendsen thermostat and barostat21. Periodic boundary conditions and Ewald sums (grid spacing of 1 Å) were used to treat long range electrostatic interactions. The SHAKE algorithm was used to keep bonds involving hydrogen atoms at their equilibrium length. A 2 fs time step was used for the integration of Newton's equations. The Amber ff99SB force field22 was used for all residues but the heme, whose parameters were developed and thoroughly tested by our group in previous works23, 24. The parameters for the ferryl/indole 2,3-epoxide intermediate were obtained using standard Amber procedures, using ab-initio HF calculations and RESP22 method for determining the partial charges. All simulations were performed with the PMEMD module of the AMBER9 package.25 Equilibration consisted of an energy minimization of the initial structures, followed by a slow heating up to 300K. For each structure 20ns MD production runs were performed. Frames were collected at 2ps intervals, which were subsequently used to analyse the trajectories. Finally, in order to estimate the work required to produce the rotation of the Cα-Cβ bond in the ferryl/indole 2,3-epoxide intermediate of the wild type and T254A mutant of xcTDO and make a rough estimation of the free energy profile for the inter-conversion of the two extreme structures, we added a time-dependent potential that moves the system from the initial to the final configuration, resulting E′(r)=E(r)+k[(ξ−ξo)2]26. The constant k was 200 kcal/mol. K in all cases, and the coordinate ξ was chosen as the dihedral angle formed by the N- Cα-Cβ-Cγ atoms of the epoxide adduct (Fig. S9).</p><!><p>Fig. 2 shows the optical absorption spectra of the T342A mutant in the absence and presence of L-Trp. In the substrate-free form, the spectra of the ferric T342A shows a Soret band at 404 nm, and a charge transfer band at 635 nm, characteristic for a six-coordinate (6C) water-bound ferric heme, similarly to that reported for the wild type enzyme.12 The ferrous derivative has a Soret band at 427 nm, and a visible band at 556 nm, consistent with a five-coordinate (5C) high-spin heme. The CO-bound complex, on the other hand, shows a Soret band at 419 nm, and two visible bands at 539 and 565 nm, confirming the presence of a 6C low-spin heme. Like the wild type enzyme, the addition of L-Trp or D-Trp to the mutant does not introduce significant modifications to the spectra, as summarized in Table 1.</p><p>The resonance Raman spectra of the ferric derivatives of the T342A mutant are shown in Fig. 3A. The substrate-free enzyme has a ν4 mode at 1372 cm−1, and ν2/ν3 modes at 1560/1482 and 1581/1507 cm−1, indicating a 6C high-spin/low-spin mixed configuration, in good agreement with a water-bound ferric heme, as indicated by the optical absorption spectral data. The addition of L-Trp or D-Trp to the ferric enzyme does not perturb the spectral feature, except that the low-spin component is slightly increased, similar to that observed in the wild type enzyme (Fig. 3B).</p><p>The ferrous derivative of the T342A mutant exhibits ν3 and ν4 modes at 1470 and 1355 cm−1, respectively, in the absence of substrate, indicating a 5C high-spin heme (data not shown), same as the reported data for the wild type enzyme.12 In the low frequency window of the spectrum, the Fe-His stretching mode (νFe–His) is identified at 227 cm−1 (Fig. 4A), similar to that of the wild type enzyme (Fig. 4B and Table 1),12 indicating that the mutation does not perturb the proximal heme environment. Nonetheless, the relative intensities of several in-plane and out-of-plane heme modes in the 250–450 cm−1 and 700–800 cm−1 region of the spectrum17, 20 are significantly perturbed by the mutation, suggesting conformational changes to the heme prosthetic group due to the mutation. L-Trp or D-Trp binding to the mutant does not introduce noticeable changes to the spectrum, in contrast to the small changes observed in the wild type enzyme (Fig. 4B), manifesting slightly weaker substrate-enzyme interactions in the mutant enzyme.</p><p>CO has been demonstrated to be a useful probe for investigating active site structure of heme-proteins. The spectral features of the CO-complex of T342A mutant (Fig. S4) are similar to those reported for the wild type enzyme,12 and are not affected by the addition of L-Trp or D-Trp, expect that the Fe-CO stretching mode (νFe-CO) at 495 cm−1 found in the substrate-free enzyme shifted to 488 cm−1 in responding to L-Trp or D-Trp binding. The assignments of the νFe-CO modes at 495 and 488 cm−1, along with those of the associated C-O stretching modes (νC-O) at 1963 and 1970 cm−1, were confirmed by 12CO-13CO isotope substitution experiments (insert in Fig. S4). As listed in Table 1, in general, all the spectral features, as well as their responses to substrate binding, of the various derivatives of the T342A mutant are similar to those reported for the wild type enzyme,12 confirming that the mutation does not introduce major structural modifications to the enzyme.</p><!><p>To determine how the T342A mutation affects the enzyme activity, NFK production activities of the wild type and T342A mutant were followed as a function L-Trp or D-Trp concentration. As shown in Fig. 5B, the activity of the wild type enzyme towards either L-Trp (black symbols) or D-Trp (red symbols) follows typical Michaelis-Menten behaviour with kcat/Km of 2.24±0.08 s−1/0.12±0.02 mM and 0.20±0.06 s−1/0.26±0.02 mM, respectively. The kcat/Km values are similar to those reported previously (2.1 s−1/0.19 mM and 0.2 s−1/0.18 mM for L- and D-Trp, respectively);12 the small differences are plausibly a result of the slight difference in the experimental temperatures, as the current data were obtained at 25 °C, while the previously reported data were acquired at ambient temperature. The data indicate that hTDO exhibits similar affinity towards L-Trp and D-Trp, but the kcat towards L-Trp is much higher than D-Trp. The T342 mutation leads to the elevation of the Km to 1.19±0.09 and 1.59±0.13 mM for L-Trp and D-Trp, respectively (Fig. 5A); at the same time the kcat is reduced to 0.101±0.002 and 0.082±0.002 s−1, respectively. As listed in Table 2, the T432A mutation results in similar reduction (~10-fold) in the substrate affinity towards L-Trp and D-Trp, but it leads to much higher reduction in kcat for L-Trp (~20-fold) as compared to D-Trp (~3-fold), indicating the abolishment of the substrate stereoselectivity (as manifested by the similar kcat/Km values of the mutant towards the two stereoisomers). Taken together the data confirm that T342 plays a pivotal role in controlling the substrate stereoselectivity of hTDO.</p><!><p>To understand the molecular basis for the abolished substrate stereoselectivity in the T342A mutant of hTDO, we performed 20 ns MD simulations of the ternary complex of the T254A mutant of xcTDO (equivalent to T342A mutant of hTDO), in which O2 and Trp were bound to the active site. The data show that the structures of both the L- and D-Trp bound complexes remained stable during the 20 ns simulations, with a RMSD<2.5Å with respect to the initial structure. (Figs. S5 and S6). In addition, the final averaged structures of the mutant show RMSD of 1.14 and 1.27 Å with respect to the L and D-Trp-bound wild type enzyme, respectively, indicating that the mutation does not introduce global structural changes to the enzyme, consistent with the conclusion drawn from the aforementioned spectroscopic studies.</p><p>Visual inspection of the obtained MD trajectories, however, revealed important structural differences in the H-bonding network surrounding the NH3+ group of the substrate. In the L-Trp bound wild type enzyme, the substrate forms strong H-bonds with its surrounding environment, including (1) the indoleamine group and H55, (2) the COO− group and the sidechain groups of Y113/R117, and (3) the NH3+ group, the OH group of T254, the propionate A group of the heme and the OH group of S124 (Fig. 6A). The replacement of L-Trp with D-Trp leads to the interruption of the interaction between the NH3+ group and T254, while the rest of the H-binding interactions remain almost the same. It is important to note that T254 and S124 are located in two loop regions that are held close together by an H-bond between the OH group of T254 and the backbone carbonyl group of S123 (Fig. 6A). For clarity, the two loops are denoted as loop250–260 and loop117–130, respectively, hereafter.</p><p>In the L-Trp bound complex of the T254A mutant, the elimination of the H-bond between the NH3+ group of the substrate and T254 leads to only subtle modifications in the H-bonding interactions with H55, Y113 and R117. However, it significantly increases the fluctuations of the loop250–260 (Fig. 6E), as evident in the much wider distance distribution of the A254-S123 pair (Fig. 6D). The opening of the two loops results in a dynamic distal pocket with fluctuating interactions between the NH3+ group of the substrate and its surroundings. In addition, the OH group of S124 rotates to form an H-bond with the NH3+ group, instead of the heme propionate A (Fig. 6B versus 6A). When the L-Trp is replaced by D-Trp in the mutant complex, the opening of the two loops was also observed (Fig. 6D), in addition, the heme propionate A forms an H-bond with S124, instead of the NH3+ group (Fig. 6C versus 6B ).</p><!><p>To gain insight into the role of T254 in controlling catalysis, we performed 20 ns MD simulations of the recently characterized ferryl/indole 2,3-epoxide intermediate of the wild type xcTDO.14, 27, 28 The data show that both the complexes with the L- and D-isomers remained stable during the 20 ns simulation, showing a RMSD<2.5 Å. (Figs. S7 and S8). In the previous studies we showed that, in the intermediate, the NH3+ group forms a strong H-bond with the epoxide oxygen (Fig. 7A). This H-bond was demonstrated to be critical for triggering the ring opening reaction of the epoxide that ultimately leads to product formation.28, 29 As observed in the ternary complex, in the epoxide intermediate of the wild type enzyme the NH3+ group is held in position by H-bonding to the heme propionate A and T254; in addition, the loop250–260 and loop117–130 are held together by an H-bond between T254 and S123 (denoted as "closed" conformation hereafter) (Fig. 7A). When L-indole 2,3-epoxide is replaced with the D-isomer, the H-bond between T254 and S123 is ruptured during the simulation, allowing the separation of the two loops (denoted as "open" conformation hereafter). Moreover, a spontaneous rotation of the Cα-Cβ bond of the epoxide intermediate brings the NH3+ group to a new position, thereby temporarily interrupting its H-bonding interaction with the epoxide.</p><p>In the ferryl/L-indole 2,3-epoxide intermediate of T254A, the structure fluctuates between open and closed conformation (Fig. 7B) due to the absence of the H-bond between T254 and S123. When the L-indole 2,3-epoxide is replaced with the D-isomer only open conformation was observed. In the open conformation of both L- and D-isomers, the H-bond between the NH3+ group and epoxide is broken, again due to the spontaneous rotation of the Cα-Cβ bond. Additional calculations show that the estimated free energy barrier for the rotation of the Cα-Cβ bond in the L-indole 2,3-epoxide intermediate of the wild type enzyme is very high (~39 kcal/mol), as a result of steric hindrance (Fig. S9). The steric hindrance is significantly reduced (~3 kcal/mol) for the D-isomer, and becomes barrierless for both isomer complexes of the T254A mutant (Fig. S9). Finally, we performed additional 6 ns MD simulations of the open conformations of the L- and D-intermediate of the wild type and T254A mutant, without restraints. It was found that, in the L-indole 2,3-epoxide intermediate of the wild type enzyme, the NH3+ group returned back to its initial position spontaneously; while in all other three cases, the NH3+ group remained in the open conformation during the time scale of the simulations, confirming that the conversion of the close to open conformation removes the steric hindrance, thereby allowing the rotation of the Cα-Cβ bond.</p><!><p>The simulation data of the ternary complex of xcTDO (with both L- and D-Trp) showed that the T254A mutation causes the opening of the loop250–260-loop117–130, leading to a more dynamic and open distal pocket, accounting for the much lower substrate affinity (i.e. higher Km) observed in the activity studies (Table 2). On the other hand, the simulation data of the ferryl/L-indole 2,3-epoxide intermediate indicate that the T254A mutation also results in the opening of the loop250–260- loop117–130. It leads to local reorganization of the H-bonding interactions surrounding the NH3+ group of the substrate, resulting in an open conformation, in which the H-bond between the NH3+ and the epoxide is temporarily lost. As the H-bond is critical for catalyzing the ring opening reaction of the epoxide during the dioxygenase chemistry,28, 29 the data account for the ~20-fold reduction of the catalytic activity of the enzyme (see kcat in Table 2). Similar result was observed in the complex of the mutant with the D-isomer, accounting for the similar kcat value observed for the two stereoisomers (Table 2). Intriguingly, similar open conformation is observed for the wild type complex with the D-isomer, accounting for its much lower kcat with respect to the L-isomer, confirming that T254 plays a crucial role in determining the stereoselectivity of xcTDO.</p><!><p>In the previous work, we have revealed that hTDO exhibits similar affinities towards L- and D-Trp, but the kcat for L-Trp is 10-fold higher with respect to D-Trp.12 The mutagenesis and MD simulation data obtained in this work demonstrate the critical role of T342 in hTDO (and the equivalent residue T254 in xcTDO) in controlling substrate binding, as well as substrate stereoselectivity of the enzyme, by modulating the H-bonding interaction between the NH3+ group and epoxide oxygen of the ferryl/indole 2,3-epoxide intermediate of the enzyme, and by regulating the dynamics of the loop250–260 and loop117–130. The crystallographic structural data show that substrate binding to xcTDO induces the closure/organization of the two loops to sequester and stabilize the bound substrate. Sequence alignment along with the structural data of the substrate-free hIDO (the only structure available for hIDO) reveal that the substrate induced loop closure also occurs in hIDO, in addition, the T254 and S124 are fully conserved in hIDO, but S123 is replaced by a Gly.</p><p>It is plausible that, due to the absence of the T254-S123 interaction, hIDO displays ~170-fold higher affinity towards L-Trp with respect to D-Trp, but the kcat are similar for the two stereoisomers,12 opposite to what observed in hTDO. Apparently, T254 (T342) is not the only player in determining the substrate stereoselectivity of TDO; we are in a process of evaluating the role of S123 in the stereoselectivity of xcTDO. In any case, the mutagenesis results presented here agree perfectly with the prediction made in our previous work, in which T254 in xcTDO was revealed as one of the key residues for controlling the stereoselectivity of the enzyme. In this sense, our new experimental work validates the computational methodology, and confirms the predictive power of this tool.</p>

PubMed Author Manuscript

Population Pharmacokinetics of Meropenem and Vaborbactam Based on Data from Noninfected Subjects and Infected Patients

ABSTRACTMeropenem-vaborbactam is a broad-spectrum carbapenem–beta-lactamase inhibitor combination approved in the United States and Europe to treat patients with complicated urinary tract infections and in Europe for other serious bacterial infections, including hospital-acquired and ventilator-associated pneumonia. Population pharmacokinetic (PK) models were developed to characterize the time course of meropenem and vaborbactam using pooled data from two phase 1 and two phase 3 studies. Multicompartment disposition model structures with linear elimination processes were fit to the data using NONMEM 7.2. Since both drugs are cleared primarily by the kidneys, estimated glomerular filtration rate (eGFR) was evaluated as part of the base structural models. For both agents, a two-compartment model with zero-order input and first-order elimination best described the pharmacokinetic PK data, and a sigmoidal Hill-type equation best described the relationship between renal clearance and eGFR. For meropenem, the following significant covariate relationships were identified: clearance (CL) decreased with increasing age, CL was systematically different in subjects with end-stage renal disease, and all PK parameters increased with increasing weight. For vaborbactam, the following significant covariate relationships were identified: CL increased with increasing height, volume of the central compartment (Vc) increased with increasing body surface area, and CL, Vc, and volume of the peripheral compartment were systematically different between phase 1 noninfected subjects and phase 3 infected patients. Visual predictive checks demonstrated minimal bias, supporting the robustness of the final models. These models were useful for generating individual PK exposures for pharmacokinetic-pharmacodynamic (PK-PD) analyses for efficacy and Monte Carlo simulations to evaluate PK-PD target attainment.

population_pharmacokinetics_of_meropenem_and_vaborbactam_based_on_data_from_noninfected_subjects_and

TEXT<!>Pharmacokinetic analysis population.<!><!>Pharmacokinetic data description and outlier analysis.<!>Population pharmacokinetic analyses. (i) Development of the initial population pharmacokinetic models.<!>(ii) Covariate analyses.<!>(iii) Update of the population pharmacokinetic models.<!>(iv) Final population pharmacokinetic models.<!><!>(iv) Final population pharmacokinetic models.<!>Final model evaluation.<!><!>Exposures and secondary pharmacokinetic parameter estimates.<!><!>Exposures and secondary pharmacokinetic parameter estimates.<!><!>Exposures and secondary pharmacokinetic parameter estimates.<!>DISCUSSION<!>Study designs.<!>Drug concentration assay.<!>Demographics.<!>Handling of outliers and samples assayed as having concentrations below the limit of quantitation.<!>Population pharmacokinetic analyses.<!>Development of the initial population pharmacokinetic models.<!>Covariate analyses.<!>Update of the population pharmacokinetic models.<!>Final model evaluation.<!>Calculation of secondary pharmacokinetic parameters and exposure estimates.

<p>Meropenem is a broad-spectrum carbapenem with in vitro activity against Gram-negative bacteria, including Enterobacterales and other important pathogens associated with hospital-acquired infections, such as Pseudomonas aeruginosa and anaerobes (1–3). While meropenem is stable against many beta-lactamases, resistance to meropenem and other carbapenems can be mediated by class A serine carbapenemases, especially Klebsiella pneumoniae carbapenemases (KPC) (4). Vaborbactam is a cyclic boronic acid beta-lactamase inhibitor that has broad inhibitory activity against several clinically important beta-lactamases. These include class A carbapenemases such as KPC-2, KPC-3, KPC-4, BKC-1, FRI-1, and SME-2 and class A extended-spectrum beta-lactamases (ESBLs) such as CTX-M, SHV, and TEM. Vaborbactam also has inhibitory activity against class C cephalosporins (e.g., CMY, P99) (5–8). In vitro and in vivo studies show that meropenem in combination with vaborbactam is highly active against Gram-negative pathogens, including KPC-producing Enterobacterales (9, 10).</p><p>Meropenem-vaborbactam as a fixed-dose combination (2 g–2 g over 3 h every 8 h [q8h] with dose adjustments for renal impairment) was approved by the U.S. Food and Drug Administration for the treatment of patients with complicated urinary tract infections, including pyelonephritis (11). The European Medicines Agency approved the same meropenem-vaborbactam dosing regimen with similar dose adjustments for renal impairment for the treatment of patients with hospital-acquired pneumonia (HAP), including ventilator-associated pneumonia (VAP), and those with complicated intra-abdominal and urinary tract (including acute pyelonephritis) infections (cIAI and cUTI, respectively) (12).</p><p>As part of the development program, data from two phase 1 and two phase 3 studies conducted to evaluate meropenem-vaborbactam were used to develop population pharmacokinetic (PK) models for meropenem and vaborbactam. Results of population PK analyses are critical to enable a better understanding of the drug disposition in subjects and patients. Use of a population PK model, together with individual PK data from phase 3 studies, also allows for the conduct of pharmacokinetic-pharmacodynamic (PK-PD) analyses to further inform efficacy and potential safety events (13, 14). Finally, a population PK model, together with preclinical PK-PD targets, in vitro surveillance data, and Monte Carlo simulation, can be used to confirm late-stage dosing regimens, including for special populations, and supports the decisions for interpretive criteria for in vitro susceptibility testing (15). The objectives of the population PK analyses described herein for meropenem and vaborbactam were the following: (i) to develop population PK models to describe the disposition of meropenem and vaborbactam using data from noninfected subjects enrolled in two phase 1 studies (16, 17) and infected patients enrolled in two phase 3 studies (18, 19) and (ii) to identify individual descriptors associated with the interindividual variability (IIV) in meropenem and vaborbactam PK.</p><!><p>Summary statistics of baseline descriptors for the analysis population consisting of 110 noninfected subjects and 322 infected patients are presented in Table 1. The percentage of males in the model development population was 45%; ages for all subjects or patients ranged from 18 to 92 years. Phase 1 noninfected subjects from study 501 had a moderate range of weight (56.0 to 94.7 kg), and most of these subjects had normal renal function. Phase 1 noninfected subjects from study 504 had a relatively broad range of weight (58.2 to 143 kg) and renal function. Infected patients from two phase 3 studies, Studies 505 and 506, had a broader range of weight (40.0 to 177 kg) and renal function (estimated glomerular filtration rate [eGFR] ranged from 4.50 to 338 ml/min/1.73 m2).</p><!><p>Summary statistics or counts of the subject demographic characteristics of analysis population</p><p>Values for gender are given as number (%) of subjects.</p><p>Initial data from study 506 consisted of 23 patients. An additional 27 patients were available upon study completion.</p><!><p>Data from two phase 1 studies (Studies 501 and 504), one completed phase 3 study (study 505), and one partially completed phase 3 study (study 506) were available for the purpose of developing an initial population PK model. For meropenem, the initial population PK analysis data set contained 4,172 meropenem plasma concentrations from 91 noninfected subjects and 295 infected patients and 834 urine meropenem concentrations from 84 noninfected subjects. For vaborbactam, the initial population PK analysis data set contained 3,988 vaborbactam plasma concentrations from 93 noninfected subjects and 294 infected patients and 746 urine vaborbactam concentrations from 75 noninfected subjects. Samples which were considered outliers and excluded from the analysis were either unreasonably high or low due to potential errors in sampling, data collection, or assay.</p><p>A total of 92 meropenem and 94 vaborbactam concentrations were available from an additional 27 infected patients after the completion of study 506. When combined with the initial data, the final data set contained 4,264 meropenem concentrations from 91 noninfected subjects and 322 infected patients and 4,082 vaborbactam concentrations from 93 noninfected subjects and 321 infected patients.</p><!><p>For both meropenem and vaborbactam, a two-compartment model with zero-order input and first-order elimination best described the plasma and urine concentration-time data. For meropenem, interindividual variability was described for the following parameters using a log-normal parameter distribution: clearance (CL), volume of the central compartment (Vc), and volume of the peripheral compartment (Vp). For vaborbactam, interindividual variability was described for all parameters. Residual variability (RV) for plasma and urine was described using a combined additive plus proportional error model. eGFR was evaluated as a covariate for meropenem and vaborbactam renal clearance (CLR) in the base structural model using either a linear, power, or a sigmoidal Hill-type function, each of which was evaluated with an intercept term to account for nonrenal clearance (CLNR). The sigmoidal Hill-type function with estimation of an intercept term representing CLNR provided a more accurate characterization of CL due to having a larger drop in objective function (68.6 and 144.8 units lower than the power function for meropenem and vaborbactam, respectively) and explaining more of the interindividual variability in CL than did the other functions and was therefore selected to describe the relationship between CL and eGFR for both agents. These models served as the comparator for the subsequent covariate analyses described below.</p><!><p>For meropenem, structural covariate parameters tested included weight on clearance, central volume of distribution, and peripheral volume of distribution, as these were the relationships established in a previously developed population PK model (20). Results of the covariate analysis demonstrated that weight was only significant for Vc and Vp. Examination of the fit of the model to the data from study 504 indicated that the CL in the majority of noninfected subjects with severe renal impairment or end stage renal disease (ESRD) was being substantially overpredicted. Based on the results of the analysis of the data from study 504 (16), various models were attempted in which CLNR was also allowed to vary with changing eGFR. The best fit to the data was obtained when CLNR was allowed to be systematically lower in subjects with an eGFR of ≤30 ml/min/1.73 m2. This relationship was therefore incorporated into the model for meropenem. Delta plots were then checked for any remaining potential covariate relationships. There appeared to be an additional relationship between age and meropenem CL. Age was added to the covariate model for CL and resulted in a significant decrease in the minimum value of the objective function (MVOF). The initial population PK model parameter estimates and standard errors for meropenem are provided in Table S1 in the supplemental material.</p><p>For vaborbactam, the covariate screening plots revealed multiple potential relationships between baseline descriptors and primary PK parameters. During the first step of forward selection, incorporating a shift in total clearance for phase 1 noninfected subjects in which subjects are allowed to have systematically higher clearance provided the largest drop in the objective function (50.617 units). Subsequent rounds of forward selection resulted in the inclusion of five additional parameter-covariate relationships: (i) relationship between height in centimeters (HTCM) and CL (drop of 13.5 units), (ii) relationship between body surface area (BSA) and Vc (drop of 32.9 units), (iii) relationship between BSA and Vp (drop of 18.3 units), (iv) relationship between study phase and Vc (drop of 15.5 units), and (v) relationship between study phase and Vp (drop of 20.2 units).The full model was then subjected to backward elimination with more stringent criteria for retention. Removal of the relationship between BSA and Vp resulted in an improvement of the fit with an 11.5-unit drop in the MVOF. Thus, the relationship between BSA and Vc was removed from the model in the first round. In the second round, all remaining relationships resulted in a significant increase in the MVOF and were retained. Thus, the backward elimination process was considered complete. The initial population PK model parameter estimates and standard errors for vaborbactam are provided in Table S2 in the supplemental material.</p><!><p>The population PK models for both meropenem and vaborbactam were updated using the final data from study 506. Fitting of the models to the full data set was successful and resulted in limited increases in the IIV. For meropenem, the weight relationship was applied to all parameters using allometric scaling functions. Forcing the weight coefficients to the allometric values resulted in a significant increase in the MVOF, but it did not substantially affect the individual fits and also resulted in a modest increase in the IIV associated with CL and was therefore retained moving forward. After fitting the full covariance matrix, which resulted in a significant drop in the MVOF, IIV was placed on distributional clearance (CLd). This updated model resulted in a significant drop in the MVOF and was chosen as the final model for meropenem.</p><p>For vaborbactam, attempts were made to standardize the body size relationships to use weight instead of height or BSA but resulted in model instability. The only modification made was fitting a full covariance matrix, which resulted in a further drop in the MVOF, and was thus chosen as the final model for vaborbactam.</p><!><p>The final population PK model for both agents was a two-compartment model with zero-order infusion and first-order (linear) elimination. The population PK parameter estimates and associated standard errors for the meropenem and vaborbactam models are provided in Tables 2 and 3, respectively. The precision of the PK parameter estimates based on asymptomatic standard error was high throughout with the exception of IIV on CLd for meropenem (283%). In general, the magnitude of the IIV was relatively modest.</p><!><p>Final meropenem population PK modela</p><p>eGFR50, eGFR value at which CLR is half-maximal; WTKG, weight (kg); SEM, standard error of the mean; CV, coefficient of variation; IIV, interindividual variability.</p><p>Final vaborbactam population PK modela</p><p>eGFR50, eGFR value at which CLR is half-maximal; SEM, standard error of the mean; CV, coefficient of variation; IIV, interindividual variability.</p><!><p>Standard goodness-of-fit plots showed excellent fits to the data (Fig. S1 and S2). The overall coefficient of determination (r2) values based on observed versus individual fitted concentrations were 0.777 and 0.849 for meropenem and vaborbactam, respectively. In general, the residual plots showed consistent scatter about zero, indicating that there were no significant biases in the fit of the data across the range of fitted concentrations or over time. These plots demonstrate the adequacy of the model fit across subjects and patients.</p><!><p>The prediction-corrected visual predictive check (PC-VPC) plots for the meropenem and vaborbactam model are provided in Fig. 1. Overall, the models provided a robust and unbiased fit to the data, demonstrating good alignment between observed concentrations and the model-predicted 5th, 50th, and 95th percentiles.</p><!><p>Prediction-corrected visual predictive check plots for the final population PK model for meropenem (top) and vaborbactam (bottom). Circles represent prediction-corrected observed plasma concentrations, while the black lines represent the median (solid line) and 5th and 95th percentiles (dashed lines) of the observed data. The red-shaded region shows the 90% prediction interval for the median simulated values, and the solid red line is the median of the median simulated values. The blue-shaded regions show the 90% prediction intervals for the 5th and 95th percentiles of the simulated values, and the solid blue lines show the median of the 5th and 95th percentiles of the simulated values.</p><!><p>Summary statistics for the key PK exposure parameters (maximum concentration [Cmax], area under the concentration-time curve over 24 h (AUC0-24) on day 1 and at steady-state, the alpha half-life (t1/2, α), and the beta half-life (t1/2, β) are provided in Tables 4 and 5 for meropenem and vaborbactam, respectively.</p><!><p>Summary statistics of key meropenem PK parameters in phase 3 patients receiving meropenem 2 g–vaborbactam 2 g q8h derived from the fit of the updated meropenem population PK model</p><p>Cmax represents the highest concentration observed during the first dose interval.</p><p>Based upon protocol-mandated dose adjustment guidelines, 28 patients with renal impairment in study 505 received a dose of meropenem 1 g–vaborbactam 1 g; similarly, nine patients in study 506 received reduced doses of meropenem-vaborbactam due to renal impairment.</p><p>Steady-state AUC0-24 estimates were not available for four patients from study 506, as these patients received less than three doses of meropenem-vaborbactam.</p><p>Summary statistics of key vaborbactam PK parameters in phase 3 patients receiving meropenem 2 g–vaborbactam 2 g q8h derived from the fit of the updated vaborbactam population PK model</p><p>Cmax represents the highest concentration observed during the first dose interval.</p><p>Based upon protocol-mandated dose adjustment guidelines, 28 patients with renal impairment in study 505 received a dose of meropenem 1 g–vaborbactam 1 g; similarly, nine patients in study 506 received reduced doses of meropenem-vaborbactam due to renal impairment.</p><p>Steady-state AUC0-24 estimates were not available for four patients from study 506, as these patients received less than three doses of meropenem-vaborbactam.</p><p>t1/2, β estimates were excluded for two patients from study 506 due to extremely high values (63.5 and 50.9 h, respectively).</p><!><p>To identify individual descriptors which may have an effect on the exposures of meropenem and vaborbactam, post hoc estimates were assessed relative to various covariates. Statistically significant relationships were identified for both meropenem and vaborbactam between clearance and renal function. The effect of renal function on meropenem and vaborbactam concentration-time profiles for a typical simulated infected patient is shown in Fig. S3. Clearance increased in a sigmoidal fashion with increasing eGFR, as shown in Fig. 2. Of note, the shapes of the two relationships were very similar, suggesting that dose adjustments made based upon eGFR for meropenem would allow for appropriate dosing of vaborbactam.</p><!><p>Relationship between individual post hoc estimates of CL and eGFR incorporated in the final population PK models for meropenem and vaborbactam. The solid blue line represents the sigmoidal Hill-type function based on final population PK model estimates. With the exception of eGFR, all covariates were set to reference values.</p><!><p>Two different measures of body size were identified as significant covariates in the population PK models for meropenem (weight) and vaborbactam (height and BSA). For meropenem, weight relationships were implemented for all parameters using allometric functions. For vaborbactam, BSA was a significant predictor of the IIV in Vc, and height was a significant predictor of the IIV in CL. For both agents, the resultant impact on the therapeutically relevant parameter of day 1 and steady-state AUC0-24 was minimal and indicates that a dose adjustment on the basis of body size is not warranted (Fig. S4).</p><p>Given the correlation between age and renal function, it was important to consider potential changes in exposure across age groups relative to eGFR for both meropenem and vaborbactam. There appeared to be no discernible trend for increased exposure in the oldest patients after taking renal function into account (Fig. S5). Similarly, neither gender nor race was expected to have a clinically significant effect on meropenem or vaborbactam exposures (Fig. S6 and S7).</p><!><p>The objectives of the analyses described herein were 2-fold. The first objective was to develop separate population PK models for meropenem and vaborbactam using PK data from noninfected subjects enrolled in two phase 1 studies and PK data from infected patients enrolled in two phase 3 studies. Using these models, the second objective was to identify any individual descriptors associated with IIV in meropenem and vaborbactam population PK parameters, respectively.</p><p>The data set used to undertake the population PK analyses for meropenem and vaborbactam described herein, which was based on data for 91 phase 1 noninfected subjects and 322 phase 3 infected patients, was robust. This was evidenced by the broad dose range for each agent (1 to 2 g for meropenem and 0.25 to 2 g for vaborbactam) and eGFR range (4.50 to 338 ml/min/1.73 m2) and the large number of plasma and urine concentrations (4,264 and 834 meropenem plasma and urine concentrations, respectively, and 4,082 and 746 vaborbactam plasma and urine concentrations, respectively). The disposition of both meropenem and vaborbactam in noninfected subjects and infected patients was best described by a two-compartment model with linear elimination. The results of the covariate analysis identified several descriptors that were associated with the IIV of meropenem PK. Weight was applied to all PK parameters for meropenem using allometric functions, which is considered standard practice (21). In addition to the impact of renal function on CLR, renal impairment was also incorporated as a covariate on CLNR, as bias in the fit of the model to the meropenem concentration-time data was observed in noninfected subjects with severe renal impairment or ESRD enrolled in study 504. After inclusion of the relationship between eGFR and CLNR, a relationship between age and meropenem CL became evident, which was also included in the final population PK model for meropenem. The current population PK model for meropenem provided a reasonable description of the CL of meropenem across a broad range of eGFR values in both noninfected subjects and infected patients.</p><p>Although the PK of meropenem has been explored extensively, the above-described data set that was used for this analysis is likely the most robust data set evaluated to date. Importantly, this data set contained 322 infected patients, the majority of which contributed at least three PK samples to the analysis. The robustness of the data set is important to consider when comparing the population mean meropenem CL in patients with normal renal function based on this analysis (about 10 liters/h in a 58-year-old patient with an eGFR of 100 ml/min/1.73 m2) with those estimates previously reported. Results of previous population PK analyses for meropenem indicated that the population mean CL of meropenem was closer to 14 liters/h (22–25). In contrast to the current analysis, published population PK analyses for meropenem in infected patients were based on relatively small numbers of patients and somewhat limited sampling schemes. Interestingly, the population mean meropenem CL in patients with an eGFR of 40 ml/min/1.73 m2 based on the current analysis was approximately 7 liters/h, a finding that is consistent with other studies in which the PK of meropenem was quantified in subjects with renal impairment (26–28). In addition, the population mean fraction of CL that is renally cleared in patients with an eGFR of 100 ml/min/1.73 m2 based on the current analysis was approximately 59%, which is within the range of 54.3% to 83% estimated in previous studies (26, 27, 29–37). CLNR decreasing with decreasing renal function was a finding also reported previously in the renal impairment studies (26–28), which may be attributed to renal metabolism, which in turn decreases with decreasing renal function (27, 31). Ultimately, given the robustness of the data from infected patients available for this analysis, it is clear that the CL of meropenem is lower than would have been expected from previous analyses.</p><p>Results of the covariate analysis for vaborbactam demonstrated the following statistically significant covariate relationships: study phase with CL, Vc, and Vp, HTCM with CL, and BSA with Vc. The relationships of HTCM with CL and BSA with Vc were expected given that these two parameters tend to scale to body size. One potential explanation for the significance of the relationships between study phase and CL, Vc, and Vp is the difference in PK sampling (intensive in subjects and informative but relatively sparse in infected patients). Ultimately, the fit of the model to the informative data from phase 3 patients suggests that the estimates of CL are robust in the population of interest.</p><p>Given that the majority of noninfected subjects and infected patients included in the analysis received meropenem and vaborbactam concomitantly, the distribution of demographics evaluated for the two sets of analyses were similar. Covariate analyses for each drug resulted in the identification of relationships between renal function (eGFR) and CL for both meropenem and vaborbactam. Given that meropenem is cleared primarily by the kidneys (3, 16, 17), it was not surprising that eGFR was a strong predictor of total CL of meropenem. Results of noncompartmental analyses of the phase 1 studies, Studies 501 and 504, demonstrated that both meropenem and vaborbactam were cleared primarily by the kidneys (16, 17). Overall, the relationships modeled for CL across the range of eGFR values were similar for both agents. This finding suggests that relative CL of meropenem and vaborbactam is consistent regardless of eGFR and supports harmonized dose reductions for both agents in patients with reduced renal function (11, 12). The effects of the remaining covariates in the final population PK models for meropenem and vaborbactam were not of sufficient magnitude to warrant dose adjustments.</p><p>Population PK models for meropenem and vaborbactam were refined using a pooled data set used to develop the original model and the additional data from infected patients from study 506 after study completion. Updates to the meropenem population PK model included applying allometric scaling, incorporating IIV on CLd, and using a full covariance matrix. The final parameter estimates were comparable to the original parameter estimates, and thus, the impact of the model refinements on meropenem exposures is negligible. Updates to the vaborbactam population PK model only included using a full covariance matrix. The final parameter estimates were comparable to the original parameter estimates, with the exception of the proportional shift with phase on CL (0.517 to 0.264) and on Vp (1.28 to 1.78). Despite these differences, an assessment of the impact demonstrated that the effect on predicted vaborbactam exposures was not impressive. Simulations of typical infected patients demonstrated an increase in AUC of 13% for the final model relative to the initial population PK model following intravenous (i.v.) administration of a 2,000-mg vaborbactam dose on day 1 and at steady-state conditions.</p><p>In conclusion, the excellent individual fits obtained using population PK methods indicated that the primary objective of the analysis was met. A robust description of the plasma PK of both meropenem and vaborbactam in the infected patients studied was achieved, such that the derived measures of meropenem and vaborbactam exposure would be expected to be both accurate and precise. The results of covariate analyses for each drug, which demonstrated the influence of eGFR on CL, provided support for adjustments of dose for renal impairment for both meropenem and vaborbactam. The findings of these analyses were useful for the subsequent execution of clinical PK-PD analyses and Monte Carlo simulations to carry out PK-PD target attainment analyses to support meropenem-vaborbactam dose selection and interpretive criteria for in vitro susceptibility testing (38, 39). Results of such analyses, the foundation of which was the population PK models described herein, served to provide data to support the regulatory approval of meropenem-vaborbactam in the United States (11) and European Union for the indications granted, including in the latter region for indications such as cIAI and HAP/VAP (12), which were not directly studied.</p><!><p>Data for these analyses were obtained from two phase 1 studies, study 501 and study 504, pooled with two phase 3 studies, study 505 and study 506 (16–19). A brief description of each study is provided below. A summary of dosing regimens, sampling strategies, and the number of noninfected subjects or infected patients considered for the population PK analyses by study is provided in Table S3 in the supplemental material.</p><p>Study 501 (ClinicalTrials.gov registration no. NCT01897779) (16) was conducted in noninfected subjects who received various combinations of meropenem (1 or 2 g) and/or vaborbactam (0.25, 1, 1.5, or 2 g) as a single i.v. infusion or multiple i.v. infusions. A total of 80 subjects were enrolled into five dose cohorts with each cohort containing four treatment arms. Subjects received single doses on days 1, 2, and 7 and multiple doses on days 8 through 14 using an infusion duration of 3 h. Plasma PK sampling was performed intensively on each day of single-dose administration. For multiple-dose administration, intensive sampling was performed on day 14. Urine PK samples were collected on days 1, 4, 7, and 14.</p><p>Study 504 (ClinicalTrials.gov registration no. NCT02020434) (16) was conducted in noninfected subjects as well as subjects with renal impairment categorized as having either mild, moderate, or severe renal impairment or end-stage renal disease. A total of 40 subjects were enrolled and received a single dose of 1 g meropenem and 1 g vaborbactam in combination in a 3-h infusion. Plasma and urine PK sampling was performed intensively.</p><p>Study 505 (TANGO I; ClinicalTrials.gov registration no. NCT02166476) (18) was a phase 3 clinical trial conducted in patients with acute pyelonephritis (AP) or complicated urinary tract infections (cUTI). A total of 550 patients were randomly assigned 1:1 to receive either meropenem-vaborbactam (2 g meropenem–2 g vaborbactam) i.v. q8h or piperacillin-tazobactam 4.5 g (piperacillin 4 g–tazobactam 0.5 g) q8h. After a minimum of 15 doses of i.v. therapy, patients could be switched to levofloxacin 500 mg by mouth every 24 h to complete a total treatment course of 10 days. Treatment could be up to 14 days if clinically indicated. Samples were collected on day 1 within 0.5 h and 2 to 3 h after the end of infusion, on day 3, and the day of the end of i.v. therapy within 0.5 h after the end of one of that day's infusions.</p><p>Study 506 (TANGO II; ClinicalTrials.gov registration no. NCT02168946) (19) was a phase 3 multicenter, randomized, open-label study of meropenem-vaborbactam versus the best available therapy (BAT) in the treatment of patients with infections due to confirmed or suspected carbapenem-resistant Enterobacterales. Patients with bacteremia, hospital-acquired/ventilator-associated bacterial pneumonia, and complicated intra-abdominal and urinary tract infections (including acute pyelonephritis) were eligible for enrollment. A total of 77 patients were randomly assigned 2:1 (meropenem-vaborbactam:BAT). Samples were collected for PK analysis on day 1 within 0.5 h and 2 to 3 h after the end of the first infusion and on days 3 and 5 at 0.5 h after the end of one of that day's infusions. The development of the population PK model was initiated before study 506 was completed. Model development was conducted using a data set of 27 patients who had completed therapy with meropenem-vaborbactam. The model was then refined once the final data set became available.</p><!><p>Plasma and urine samples were assayed for meropenem or vaborbactam concentrations using a validated liquid chromatography-tandem mass spectrometry assay at MicroConstants, Inc. (San Diego, CA, USA). The calibration range of the assay for both agents was 0.02 to 100 mg/liter. Samples that were expected to be outside of the validated range were appropriately diluted using blank biological fluid prior to sample analysis.</p><!><p>Demographic and disease characteristics were used to characterize the analysis population and to evaluate their ability to explain a portion of IIV for selected PK parameters. eGFR was calculated from serum creatinine, age, and gender using the Modification of Diet in Renal Function equation (40). eGFR was calculated at the time of each serum creatinine measurement and was treated as a time-varying covariate for the population PK analyses. During the calculation of eGFR, serum creatinine was also capped at a lower bound of 0.5 mg/dl. Demographic information included age, height, weight, BSA, body mass index (BMI), sex, and race. BSA was calculated using the method of DuBois and DuBois (41). BMI was calculated as weight (in kilograms) divided by height (in meters) squared.</p><!><p>An outlier was defined as an aberrant observation that substantially deviated from the rest of the observations within an individual. Outliers were excluded owing to the potential for these observations to negatively impact the convergence and/or parameter estimates as noted in the FDA guidance (42).</p><p>Plasma concentration values that were below the limit of quantitation (BLQ) were flagged in the data set. The population analysis program then applied the Beal M3 method (43) such that the algorithm considered a BLQ value as a normally distributed, random value somewhere between negative infinity and the limit of quantification. The Beal M3 method maximizes the probability that a concentration observed to be BLQ is also predicted to be BLQ.</p><!><p>The population PK analyses for meropenem and vaborbactam were conducted using NONMEM software v7.2 (ICON Development Solutions, Ellicott City, MD, USA), implementing the first-order conditional estimation method with interaction. During various stages of model development, population PK models were minimally assessed using the following criteria: (i) evaluation of individual and population mean PK parameter estimates and their precision as measured by the percent standard error of the population mean estimate, (ii) graphical examination of standard diagnostic and population analysis goodness-of-fit plots with possible stratification by various factors such as patient population or dose group, (iii) graphical examination of the agreement between the observed and individual post hoc predicted concentration-time data, (iv) reduction in both residual variability and IIV, and (v) comparison of MVOF for nested models or Akaike's information criterion for nonnested models.</p><!><p>Since concomitant administration of meropenem and vaborbactam does not affect the plasma or urine PK of either drug (17), separate population PK models were constructed for meropenem and vaborbactam. The population PK model for meropenem was based upon a model that had been previously developed (20). The first step of the PK model development for meropenem involved fitting a two-compartment model (without covariate relationships) to the plasma data for noninfected subjects from study 501 in order to obtain stable priors for the population parameter estimates. For the vaborbactam population PK model, the plasma vaborbactam concentration-time data for noninfected subjects from study 501 was used to determine the initial structure of the model. One-, two-, and three-compartment models with zero-order input and first-order elimination were to be evaluated.</p><p>After establishing the most appropriate structural model for both agents, the urinary excretion and plasma concentration data were comodeled to generate estimates of both CLR and CLNR. The data for noninfected subjects from study 504 were then incorporated into the data set, and the models were fit to the pooled phase 1 data. Given that both agents are cleared almost exclusively by the kidneys (16, 17), eGFR was evaluated for statistical significance using various functional relationships prior to the formal covariate analysis (i.e., as part of the base structural model). After the identification of an appropriate relationship between eGFR and CLR, data for infected patients from study 505 and study 506 were included and the base structural model was fit to the pooled data.</p><!><p>Several baseline demographic and disease characteristics were evaluated for their impact on the primary PK parameters. The variables evaluated included sex, race, age, weight, height, BSA, and BMI. Covariate exploration involved calculating individual deviations for each parameter by subtracting individual post hoc PK parameters from the population mean PK parameter. Plots of the individual deviations for each PK parameter versus each covariate were examined for observable trends and were used to identify an appropriate function to describe the relationship between the PK parameter and the covariate.</p><p>Covariate analyses were conducted separately for each agent using stepwise forward selection and backward elimination. Covariates contributing at least a 3.85-unit reduction in the MVOF (α = 0.05, for 1 df) when added to the model univariately were considered statistically significant; only the most statistically significant covariate was added to the model in each step. This process was repeated until no other subject covariates were statistically significant prior to performing a stepwise univariate backward elimination analysis (α = 0.01, for 1 df) to determine the final population PK model for each agent.</p><!><p>Given that study 506 was completed after the development of the initial population PK models for both meropenem and vaborbactam (which were used to support the FDA new drug application), these models were subsequently refined by including additional data for infected patients from study 506. The models were refined to improve the fit, the process for which also included attempts to apply allometric scaling and use full covariance matrices. The final models were used to support the marketing authorization application submitted to the European Medicines Agency.</p><!><p>To assess the ability of the population PK models to reliably describe meropenem and vaborbactam exposure, a PC-VPC was performed using the parameter estimates from the final population PK models. The 5th, 50th, and 95th percentiles of plasma concentrations from simulated noninfected subjects and infected patients were compared with observed data from the phase 1 noninfected subjects and phase 3 infected patients. Due to the heterogeneity in the dosing times across the phase 3 studies, correcting the observed and simulated values to their respective population predicted values allowed for visualization of data in one plot (44).</p><!><p>Estimates of Cmax after the first dose, day 1 and steady-state AUC0-24, t1/2, α, and t1/2, β were generated for all phase 3 infected patients included in the population PK analyses by using a simulated PK profile for each patient using the individual post hoc PK parameters from the final population PK models and the mrgsolve package in R (45).</p>

PubMed Open Access

Secondary Amino Alcohols: Traceless Cleavable Linkers for Use in Affinity Capture and Release

Capture and release of peptides is often a critical operation in the pathway to discovering materials with novel functions. However, the best methods for efficient capture impede facile release. To overcome this challenge, we report secondary amino alcohol-based linkers for release of peptides after capture. These amino alcohols are based on serine (seramox) or isoserine (isoseramox) and can be incorporated into peptides during solid-phase peptide synthesis via reductive amination. Both linkers are quantitatively cleaved within minutes under NaIO4 treatment. Cleavage of isoseramox produced a native peptide N-terminus. This linker also showed broad substrate compatibility; incorporation into a synthetic peptide library resulted in the identification of all sequences by nanoLC-MS/MS. The linkers are cell compatible; a cell-penetrating peptide that contained this linker was efficiently captured and identified after uptake into cells. These findings suggest secondary amino alcohol-based linkers might be suitable tools for peptide discovery platforms.

secondary_amino_alcohols:_traceless_cleavable_linkers_for_use_in_affinity_capture_and_release

Introduction<!>Results and Discussion<!>Conclusion

<p>Synthetic peptide libraries are used to discover sequences that bind to protein targets, penetrate cells, or facilitate site-selective bioconjugations.1–6 These experiments require the selective capture of hits from chemical or biological samples and subsequent release. After release, the isolated peptides are analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS) to determine their sequences. The best approaches for efficient capture impede quantitative release. Indeed, the widely adopted streptavidin-biotin affinity capture and release approach suffers from this hurdle (Figure 1a). To release biotinylated peptides denaturing conditions [e.g., boiling in sodium dodecyl sulfate (SDS) buffer]7 are required, which can lead to degradation and contamination. Therefore, there is a need for new approaches for efficient retrieval of peptides after capturing with streptavidin.</p><p>Cleavable linkers are attractive tools that enable the release of peptides after affinity capture.8–10 Linkers consist of chemical moieties that can be 1) incorporated into peptides and 2) cleaved under specific conditions. Various cleavable linkers have been described, and their cleavages can be induced with enzymes,8 photoirradiation,11 metals,12–14 nucleophiles,15,16 electrophiles,8 acids,17 reductants,18 and oxidants.19–21</p><p>Vicinal diol-based linkers are compatible with standard solid-phase peptide synthesis (SPPS) (Figure 1b). The 1,2-diol motif is stable toward basic and acidic conditions. Treatment of linker-embedded peptides with sodium periodate facilitates cleavage. These oxidative conditions are compatible with folded proteins20 and with cells.22</p><p>Amino alcohols are an alternative periodate-labile motif and can be cleaved 1000-times faster than diols.21,23 An SPPS compatible linker based on this structure has been reported.19 However, this linker requires a nine-step synthesis, presenting challenges for its widespread applicability.</p><p>Here, we report secondary amino alcohols as a new class of oxidatively cleavable linkers. Secondary amino alcohols exhibit the same reactivity as primary amino alcohols towards NaIO4 oxidation in small molecule studies.23 We designed two cleavable linkers with this motif based on serine (seramox) and isoserine (isoseramox). These structures can be inserted into a peptide backbone during SPPS via reductive amination. Both linkers are quantitatively cleaved by periodate within minutes (Figure 1c). The isoseramox linker releases a native peptide N-terminus upon cleavage; this is the first example of a traceless periodate cleavable linker. Isoseramox was incorporated into a 216-membered peptide library, showing broad substrate tolerance for synthesis and cleavage. Isoseramox was also compatible with cell assays. A cell-penetrating peptide containing isoseramox entered the cell cytosol before it was captured and selectively released by periodate. Secondary amino alcohol linkers are thus enabling tools for experiments requiring the capture and release of synthetic peptides.</p><!><p>Secondary amino alcohols can be incorporated into peptides during standard SPPS. Peptide 1 (SLAGV) was prepared by Fmoc-based SPPS. Gly-derived aldehyde 2 was attached to the N-terminal serine residue via reductive amination on-resin (Scheme 1a). The resulting secondary amine was then Boc-protected. Removal of the Fmoc group enabled the elongation of the peptide chain, generating peptide 3 (Y-seramox-SLAGV). Simultaneous cleavage of the peptide from the resin and side chain deprotection provided seramox-containing peptide 4 in 90% crude yield (Figure S1a).24 The cleavable secondary amino alcohol moiety in the peptide backbone is based on a natural Gly-Ser dipeptide, wherein a single carbonyl is replaced with a methylene group. This Ψ(CH2NH) functionality is considered a bioisostere of an amide.25</p><p>A seramox-containing peptide was quantitively cleaved in 5 min, faster than a diol-containing peptide (Figure 2). The peptides [500 μM in 1X phosphate buffered saline (PBS), pH 7.5] were treated with two equivalents of NaIO4 (1 mM). Under these conditions, less than 30% of diol 4a was cleaved after 60 min. Seramox-containing peptide 4b was cleaved to full conversion after 5 min. Quantitative cleavage of 4b was also accomplished at lower concentrations of peptide and periodate (10 μM and 30 μM respectively) in three hours (Figures S3, S4).</p><p>The cleavage of seramox generates an N-terminal aldehyde product (Scheme 1b). Coordination of seramox to the periodate (A) and cleavage of the Cα–Cβ bond generates an imine (B) in the peptide backbone. This imine is spontaneously hydrolyzed, affording a primary amine on the C-terminal fragment (C), and an aldehyde on the N-terminal fragment (D). Aldehydes could represent a handle for further functionalization.26–28 However, aldehydes suppress MS signals, especially hindering MS/MS sequencing. This issue is not only present with seramox, but also with known diol and amino alcohol linkers (Figure 1b) and was circumvented by transforming the intermediate aldehydes into oximes with H2N-OH•HCl (Figure 2).29</p><p>Secondary amino alcohol linkers based on isoserine (isoseramox) enable traceless release of N-terminal peptides. The generation of amines instead of aldehydes at the N-terminus upon periodate cleavage overcomes the MS sequencing challenges of known linkers. Furthermore, applications requiring the release of peptides for biological purposes may benefit from the regeneration of the native peptide structure.30 In designing the isoseramox linker, we hypothesized that replacing the seramox moiety with an isoserine derivative would reposition the cleavable bond (C1–C2), thus generating a native N-terminal peptide upon periodate treatment (D, Scheme 2a).</p><p>Isoseramox linkers were incorporated into peptides by SPPS. Isoserine-derived aldehyde (8), could be synthesized in three steps. Reductive amination on peptide 9 (LAGV) enabled access to isoseramox-containing peptide 4c (Y-isoseramox-LAGV, Scheme 2b).</p><p>Isoseramox-containing peptide 4c was quantitatively converted to N-terminal peptide 5c (LAGV) in 5 min with NaIO4 (Figure 2). Despite the change from a primary alcohol (seramox) to a more sterically hindered secondary alcohol (isoseramox), reaction times were not affected. To our knowledge, this example is the first reported periodate cleavage that generates native peptides at the N-terminus.</p><p>Both the amine and the alcohol functionality of the (iso)seramox linkers appear to be essential in enhancing cleavage rates. Since changing from a diol to an amino alcohol resulted in faster periodate cleavage, we tested a 1,2-diamine linker. A 2,3-diaminopropionic acid (Dap)-derived linker was incorporated into peptides and its cleavage efficiency was studied (Figure S5). Cleavage of the diamine required 18 hours to proceed to full conversion, slower than (iso)seramox linkers. Protonation of both amines at pH 7.5 may slow their rate of coordination to the periodate. We hypothesize that the (iso)seramox alcohol facilitates coordination to the periodate (Scheme 2a, 4c to A), while the amine accelerates the cleavage step (Scheme 2a, A to B).31</p><p>Secondary amino alcohol linkers have a broad substrate tolerance. We incorporated isoseramox into a combinatorial peptide library to test (1) the efficiency of its addition on several N-terminal residues and (2) the tolerance of its oxidative cleavage in the presence of diverse amino acid side chains. A combinatorial library of 216 members was prepared by split-and-pool SPPS. The library contained 17 canonical amino acids in 8 positions (1234LAYK), four of which were variable positions (Figure 3).32 Position 1 could be Asp, Met, His, Trp, Tyr, Val, Lys and Gln (containing basic, acidic, amidic, aromatic and aliphatic residues), position 2: Ser, Ala and Leu, position 3: Glu, Asn and Gly, and position 4 Phe, Pro and Thr. At this stage, part of the library was cleaved for use as a reference pool (library A). The second part was alkylated with isoseramox and then biotinylated (library B). Library B was immobilized onto magnetic streptavidin beads and were separately desalted by micro solid-phase extraction and analyzed by nanoLC-MS/MS on an Orbitrap Fusion Lumos spectrometer. The MS/MS spectra were analyzed with the sequencing software PEAKS 8.5 and the complete list of sequences obtained was additionally filtered for peptides fitting the library design (correct length and correct monomers in positions 1, 2, 3 and 4).33 Efficient and complete cleavage would result in a product distribution identical to reference library A. Indeed, the sequences identified from the reference library A and the cleaved library B were almost identical: 90% (194 out of 216) and 94% (202 out of 216) of peptides were identified, respectively. Sequences with all eight possible residues in position 1 were identified. This demonstrates that incorporation and cleavage of the isoseramox linker are efficient regardless of the neighboring residues. Oxidized methionine residues were observed in both library A (89%) and library B (96%), though this heterogeneity did not impact sequencing. Oxidized Met is recognized by the PEAKS sequencing software as a post-translational modification. No oxidations or modifications of other residues, such as Trp or His, were observed. Furthermore, sodium periodate cleavage has previously been shown to be compatible in proteomics-based workflows, as the oxidation was found not to interfere with sequencing.10 These results indicate that secondary amino alcohol linkers can be used for capture and release workflows.</p><p>Protein secondary structure and folding is not disrupted by sodium periodate treatment. After studying (iso)seramox in the context of short peptides, we investigated whether periodate-based amino alcohol cleavage could be extended to more complex scaffolds. The miniprotein Z33 was synthesized by SPPS. Z33 adopts a 33-residue helix-helix bundle. The α–helical structure of the synthetic Z33 was confirmed by circular dichroism (CD) spectroscopy. After treating Z33 with 4 mM of sodium periodate (i.e., four times more concentrated than the harshest cleavage condition) for 30 minutes, the CD spectrum showed no alteration (Figure S6). This result indicates that the α–helical structure was not disrupted by periodate treatment. The compatibility of Z33 with NaIO4 is in agreement with several previously published reports using periodate cleavages on folded proteins, antibodies, protein complexes and whole cells.34–36</p><p>(Iso)seramox-based recovery of biotinylated peptides from streptavidin beads is more efficient than SDS denaturation strategies (Figure 4). After capture of biotinylated peptides with streptavidin, it is currently difficult to release the peptides for analysis. A frequently used method to disrupt the strong biotin-streptavidin interaction relies on heating the complex in SDS containing buffer.7 We compared this strategy to our (iso)seramox based recovery workflow. Two sets of peptides were prepared, containing seramox-Lys(Biotin) (10a–c) or Gly-Ser-Lys(Biotin) (12a–c), at their C-termini, respectively. To assess the effect of peptide recovery conditions on sensitive moieties, one peptide in each set contained a phosphoserine (pSer) and one peptide contained a TAMRA fluorophore. All peptides were individually immobilized on magnetic streptavidin beads. For 10a–c, periodate treatment enabled the recovery of cleaved peptides 11a–c (Method A; 0.5 mM, 10 min). Peptides 12a–c were recovered according to described denaturing procedures by heating in SDS buffer with excess biotin (Method B; 0.4% SDS, 20 mM biotin, 95 °C, 15 min). While the recovery solutions of 11a–c were homogeneous, suspensions with precipitated material were observed for 12a–c. LC-MS analysis showed that the seramox-based strategy enabled the recovery of pure material. Alternatively, SDS denaturation caused release of several side products along with the desired peptides (Figure 4). TAMRA and pSer were intact after both seramox and denaturation recovery strategies.</p><p>Isoseramox linkers are also compatible with cell-based assays. When screening for peptide binders to whole cells or for cell-penetrating peptides (CPPs) it is often necessary to recover hit peptides from cells.3 Here we show that the isoseramox linker can be used to recover a CPP from HeLa cells.</p><p>We synthesized a variant of the known CPP d-penetratin, containing both biotin for capture and isoseramox for release (10, Figure 5a). Reactions of 10 (50 μM) with NaIO4 (3 equiv) provided quantitative conversion to cleaved product in 10 min (Figure S7). Peptide 10 contains an oxidizable Met residue; only 30% Met oxidation was observed, indicating that oxidative cleavage of the linker is faster than oxidation of the Met sidechain. However, when performing peptide library assays at the MS detection limit, heterogeneity in the sample could cause loss of sequencing information.</p><p>d-Penetratin-containing isoseramox retains its cell-penetrating activity (Figure 5b). A derivative containing TAMRA (12) was synthesized and incubated with HeLa cells. Confocal imaging of the cells displayed diffuse fluorescence in the cytosol and nucleus, in addition to endosomes. These images suggest that the linker does not disrupt the peptide's cell-penetrating activity and access to the cytosol (Figure S8).</p><p>We then captured and released d-Penetratin derivative 10 after delivery into HeLa cells (Figure 5a). HeLa cells were incubated with 10 μM peptide 10 for 1.5 h, washed extensively to remove membrane-bound peptide, and lysed. The lysate containing internalized biotinylated peptide was then incubated with streptavidin coated magnetic beads. Treatment of the magnetic beads with NaIO4 enabled the recovery of 3 ng of d-penetratin (11),37 enough material for characterization by LC-MS (see Supporting Information). By comparison, results in our laboratory utilizing a protease cleavable linker did not result in the recovery of any peptide.38 These results demonstrate the potential of the isoseramox linker for future applications in cell-based peptide library screening.</p><!><p>We have described secondary amino alcohols as a new class of cleavable linkers. (Iso)seramox exhibits broad applicability for peptide library screening. The linkers can be incorporated into peptides during on-resin synthesis. Cleavage of these secondary amino alcohols was shown to be faster than known diol linkers. Furthermore, isoseramox is the first example of a traceless, periodate-labile linker, where cleavage liberates native N-terminal peptides. The broad substrate compatibility of isoseramox was demonstrated in the synthesis, capture, and release of a peptide library. The recovery of 94% of sequences demonstrates 1) the efficiency of reductive amination of isoseramox onto several N-terminal residues and (2) the tolerance of its oxidative cleavage with various peptides. Additionally, isoseramox facilitated the recovery of peptides after uptake into cells. Taken together, these experiments highlight the applicability of (iso)seramox toward the release of peptides from streptavidin beads after screening assays and affinity capture. The tolerance of cells to NaIO439 could expand the potential of (iso)seramox linkers to controlled release of cargo in living systems.13,40 We are optimistic that these easy-to-use linkers will be valuable tools not only for the discovery of functional peptides and peptidomimetics, but also in small molecule and polymer applications.</p>

PubMed Author Manuscript

Polypharmacology of some medicinal plant metabolites against SARS-CoV-2 and host targets: Molecular dynamics evaluation of NSP9 RNA binding protein

Background: Medicinal plants, as rich sources of bioactive compounds with antiviral properties, are now being explored for the development of drugs against SARS-CoV-2. Aims: Identification of promising compounds for the treatment of COVID-19 from natural products via molecular modelling against NSP9, including some other viral and host targets and evaluation of polypharmacological indications. Main methods: A manually curated library of 521 phytochemicals (from 19 medicinal plants) was virtually screened using Mcule server and binding interactions were studied using DS Visualiser. Docking thresholds were set based on the scores of standard controls and rigorous ADMET properties were used to finally get the potential inhibitors. Free binding energies of the docked complexes were calculated employing MM-GBSA method. MM-GBSA informed our choice for MD simulation studies performed against NSP9 to study the stability of the drug-receptor interaction. NSP9 structure comparison was also performed. Key findings: Extensive screening of the molecules identified 5 leads for NSP9, 23 for Furin, 18 for ORF3a, and 19 for interleukin-6. Ochnaflavone and Licoflavone B, obtained from Lonicera japonica (Japanese Honeysuckle) and Glycyrrhiza glabra (Licorice), respectively, were identified to have the highest potential multi-target inhibition properties for NSP9, furin, ORF3a, and IL-6. Additionally, molecular dynamics simulation supports the robust stability of Ochnaflavone and Licoflavone B against NSP9 at the active sites via hydrophobic interactions, H-bonding, and H-bonding facilitated by water. Significance: These compounds with the highest drug-like ranking against multiple viral and host targets have the potential to be drug candidates for the treatment of SARS-CoV-2 infection that may possibly act on multiple pathways simultaneously to inhibit viral entry and replication as well as disease progression.

polypharmacology_of_some_medicinal_plant_metabolites_against_sars-cov-2_and_host_targets:_molecular_

Introduction<!>Materials and Methods<!>Dataset and ligand preparation<!>Evaluation of the binding site<!>Receptor preparation<!>Virtual screening workflow<!>Post-docking visualization and binding site analysis<!>Molecular dynamics simulation .<!>ADMET properties<!>Rodent acute toxicity studies<!>Comparative analysis of SARS-CoV-2 NSP9, SARS-CoV NSP9, and H-CoV NSP9 at the sequence and structural level<!>Binding pocket identification<!>In silico virtual screening<!>Pharmacokinetics and ADMET screening<!>RAT Acute toxicity assessment<!>Binding interactions with NSP9<!>Interactions of the polypharmacological leads with ORF3a, Furin and IL-6<!>Binding site analysis<!>MM-GBSA Binding Energy of Top Inhibitors<!>Molecular Dynamics Simulation Study<!>Conclusion

<p>COVID-19 was first reported in Wuhan, China in December 2019 and declared a global pandemic by the World Health Organization (WHO) in March 2020 (Cucinotta and Vanelli, 2020;Li et al., 2020). The global outlook reported by ECDC as of 28 August 2020, was 24,473,843 confirmed cases and 832,002 deaths among continents ( www.ecdc.europa.eu ) represented in Figure 1 . The global challenge to health and economy posed by COVID-19 may only be upturned by the development of an effective drug or vaccine for its management. To date, no drug or vaccine has been approved to treat COVID-19. Highly exploited drug targets for COVID-19 include Receptor Binding Domain (RBD) (Dong et al., 2020), spike protein (S), nucleocapsid (N) (Grunewald et al., 2018, Surjit et al., 2006), non-structural proteins (NSPs) including papain-like protease (Plpro), RNA-dependent RNA polymerase (RdRp), 3-chymotrypsin-like protease (3Clpro). The non-structural proteins 9 (NSP9) replicase is among the 16 NSPs of SARS-CoV-2 vital for viral replication and transcription (Ziebuhr et al., 2000). Also, unlike other proteins with a single domain, it comprises a unique single folded beta-barrel. It binds with RNA, interacts further with NSP8, and triggers the essential functions (Sutton et al., 2004). Inhibition of NSP9 further inhibits NSP8 and thereby antagonizes viral replication making it one of the attractive targets for drug discovery. Our interest was NSP9 because of its uncharacterized function in the domain of replication. ORF3a is also an interesting target for drug discovery. It plays an essential role in induced cell apoptosis and activation of NLRP3 inflammasome (Siu et al., 2019), leading to cytokine storms.</p><p>Besides the viral targets, there are host proteins that aid viral replication and promote an excessive inflammatory response. Identification of potential inhibitors of host factors has been less explored via computational studies (Wu et al., 2020;Rahman et al., 2020) when compared with the studies on viral proteins inhibition. Among various host targets (Carmen et al., 2020), we were interested in studying the receptors, furin that plays a major role in activating the S-proteins, implicated in the cleavage and activation of viral coat proteins (Cameron et al. 2000), and interleukin-6, a cytokine with a crucial role in virus-induced immunopathological events which causes fatal pneumonia in severe SARS-Cov-2 infections (Channappanavar and Perlman, 2020).</p><p>The interaction of drug molecules with multiple targets have been described as polypharmacology. It represents a new lead to rational designing of the next cohort of therapeutic agents with improved efficacy and decreased toxicity. The past decade witnessed a paradigm shift in the drug discovery process as regards solely targeting disease with one drug. Lack of efficacy and high toxicity had been the reasons for the rejection of new drug applications, hence the need for this paradigm change. The rational designs of multi-target ligands have been challenging to achieve. However, their benefits over the use of selective compounds include improved efficacy and less complicated pharmacodynamic and pharmacokinetic characteristics when compared to a number of drug combinations (Proschak, 2014).</p><p>Plants are rich sources of bioactive compounds that are reported to be effective against numerous pathogens (Hussain et al., 2017). In this context, we present a virtual screening of small-drug like plant metabolites against four chosen targets: NSP9, ORF3a, furin, and IL-6, and the analysis of their polypharmacological indications. We compared the structure of SARS-2 NSP9 against SARS Cov and H-Cov and examined the nature and properties of the binding pockets of all the chosen receptors. A Molecular Dynamics (MD) simulation study was executed to validate the findings and investigate the stability of the NSP9 protein-ligand complexes.</p><!><p>Determination of NSP9 sequence percentage identity, multiple sequence alignment, and pairwise structural clustering 1. Multiple sequence alignment of the three sequences of NSP9 were performed using ClustalΩ from the MPI Bioinformatics toolkit (Zimmermann et al., 2018, Sievers et al., 2011). EduPyMol version 2.3.4 (The PyMOL Molecular Graphics System, Version 2.3.4, Schrödinger, LLC.) was used to align the proteins and analyze the pairwise structural clustering of SARS-2 NSP9 with SARS NSP9 and H-CoV NSP9.</p><!><p>Phytochemicals from 19 different medicinal plants with potential antiviral and anti-inflammatory activities obtained through literature survey were compiled and manually curated to get a comprehensive database of 521 non-redundant molecules. Table S1 (SI) enlists all the plants with their taxonomic identities. The 2D structures of the compounds were retrieved from PubChem Database (Kim et al., 2019), and the database of molecules was prepared by uploading the retrieved structures. As controls for the screening study, standard inhibitors (Table 3) for all the four receptors were downloaded from PubChem. They were minimized by the conjugate gradient algorithm using the minimize function of Open Babel (Boyle et al., 2011) prior to docking.</p><!><p>Four receptors, including two from SARS-CoV-2: NSP9 and ORF3a and two from the host: furin and IL-6, were chosen for our study. The protein PDB structures of the receptors with PDB IDs 5JXI, 1ALU, 6XDC, and 6W4B with resolutions 2.00 Å, 1.9 Å, 2.9 Å and 2.95 Å respectively were downloaded from Research Collaboratory for Structural Bioinformatics Protein Data Bank (RSCB PDB; ( http://www.rcsb.org /, Berman et al., 2000). The active pockets of the viral targets, NSP9 and ORF3a with PDB IDs 6W4B and 6XDC respectively, were predicted using SiteMap (Schrödinger Release 2018-1: SiteMap, Schrödinger, LLC, New York, USA).</p><p>Potential binding regions and active residues of the host receptors, IL-6 and furin with PDB IDs 1ALU and 5JXI respectively, have been previously identified, experimentally verified, and resolved in high-resolution crystals. The pockets and residues identified in the active site were further evaluated using ConSurf (Ashkenazy et al., 2016;Celniker et al., 2013;Ashkenazy et al., 2010) and PrankWeb (Jendele et al., 2019;Krivak et al., 2018) online servers. . The region marked in shades of purple corresponds to the conserved residues, while shades of cyan correspond to variable amino acid residues. The region in yellow doesn't have sufficient information for conservation analysis. Encircled in red is the active site for the host receptors from reported literature.</p><!><p>Chain A of the polymeric-proteins was chosen for further molecular docking and screening studies. Prior to virtual screening, the protein molecules were cleaned and prepared using AutoDock Tools 1.5.6 (Sanner et al., 1999;Morris et al., 2009). All the non-polar hydrogens, lone-pairs, water molecules, and non-standard residues were removed. Polar hydrogens were added, followed by the addition and merging of Gasteiger charges. The files were exported and saved as pdbqt.</p><p>The Grid configuration parameters were also generated using AutoDock Tools 1.5.6. For each receptor, the GridBox was set near the active site region, and the size of the box was relaxed, ensuring the inclusion of all essential catalytic, conserved structural motifs and substrate binding amino acid residues inside the box that are vital for the functioning of the receptors to reduce the search space for ligand optimization (Trott and Olson, 2010). A grid point spacing of 0.375 Å was used, the grid dimensions and the coordinates of the grid box center for all the targets were preserved for virtual screening represented in Table 2 .</p><!><p>Structure-based virtual screening (SBVS) of the prepared phytochemical library was performed using custom workflows of Mcule Drug Discovery Pipeline ( http://Mcule.com ; Kiss et al., 2012). The Mcule server utilized the 3D structure of the target and fitted the ligands by efficiently using a gradient optimization method in its local optimization procedure. The individual workflow steps were added to filter the hits for this study, namely the basic property filter search based on the number of rule-of-five (RO5) violations and the Docking (Vina) workflow step (Trott and Olson, 2010), were sequentially executed. It has been opined that Lipinski's rule (RO5) has been overstated in its role as a tool, guiding the drug design of small molecules that are orally bioavailable. The reasons include: about half of small molecule drugs that are FDA approved are administered orally and obey the RO5, which excludes essential biologicals. Also, the RO5 excludes natural products that make up about 33% of commercially available small molecule drugs. Therefore, the focus on research applicable to natural products should be encouraged (Zhang and Wilkinson, 2007). It is also pertinent to remember that RO5 doesn't predict pharmacological activity; hence restricting the drug discovery space would not just limit the search but would result in the exclusion of certain compounds that would otherwise possess immense pharmacological potentials. Very recently, for areas with large unmet medical needs like oncology and virology, extensive studies show the focus and demands of beyond-rule-of-five drug candidates. To broaden the search of our study for finding anti-COVID phytochemical candidates, we have allowed a maximum balanced violation of two.</p><p>The molecules were screened by docking using the Vina docking algorithm (Trott and Olson, 2010). The receptor targets were prepared and fed as an input file. The grid dimensions, as well as the coordinates of the center of the binding site for each receptor molecule was fed to the workflow step, and other constraints required were set to default. For the inhibitors used as controls, molecular docking was manually performed using AutoDock Vina.</p><!><p>The structures of the top conformation of the docked Phyto-ligands complexed with the receptor macromolecules having the lowest binding affinity values (Vina docking scores) were retrieved from Mcule, and the poses were visualized using EduPyMol version 2.3.4. The top screened molecules were further analyzed for the receptor-ligand interactions (with the binding site residues), the type of bonds, the bond distances , using the various functionalities available in Discovery Studio Visualiser 2020. Among multiple interactions, we mainly looked for the non-covalent interaction, including the occurrences of hydrogen, Vanderwaal, π-π interaction, π-sigma bond, π-sulfur, and many other hydrophobic interactions among the other existing ones. The characteristics of the binding site pocket were also of interest and hence were analyzed for the nature of aromaticity, H-bond donor/acceptor distribution, hydrophobicity, and solvent accessibility (SAS) using DS Visualiser.</p><!><p>For the challenges with respect to the computational expenses, we chose NSP9 for further evaluation mainly because we were interested to understand the dynamics of the binding site residues of a viral protein that has not been appropriately characterised concerning its biological function. Molecular dynamics simulations were performed using Desmond tool (Bowers et al., 2006; Desmond Molecular Dynamics System, D. E. Shaw Research, New York, NY; Maestro-Desmond Interoperability Tools, Schrödinger, New York, NY, 2018). The ligand-receptor complex was placed at the center of a periodic orthorhombic box, and the addition of TIP3 water molecules performed solvation. Sodium and chloride ions were added to neutralize the system at a concentration of 0.15M to mimic physiological conditions. For long-range electrostatics, the smooth particle mesh Ewald (PME) estimation was used, while for nonbonded interactions at a cut-off of 9 Å MSHAKE algorithm was utilized. The Default Desmond protocol was followed for system relaxation. After energy minimization, the system was gradually heated to a temperature of 300K. It was maintained by Nose-Hoover thermostat for a constant number of particles, volume, and temperature simulation of 500ps with 2fs time step, followed by equilibration of pressure to 1 bar and was maintained via Martyna-Tuckerman-Klein barostat, during constant number of particles, pressure, and temperature (NPT) of 500ps with 2fs time step. MD Simulation was performed for each of the three protein-ligand complexes and the apo-state protein for 100ns with 2fs time step. Computation of the trajectories was executed using multiple time-step RESPA integrators. We analyzed the root mean square deviation (RMSD) and root mean square fluctuation (RMSF) of the main backbone of carbon atoms, RMSD of the ligands and the protein-ligand contacts were also assessed. Simulation results analysis was performed using the Desmond SID in maestro.</p><!><p>The resultant library generated after the virtual screening and the benchmark docking scores for all the four receptors were subjected to in silico predictions of pharmacokinetics and ADMET properties using the admetSAR 2.0 webserver (Yang et al., 2019). The molecules were specifically checked for human intestinal absorption, blood-brain barrier permeance, phosphoglycoprotein substrate binding, cytochrome inhibition (CYP2D6), carcinogenicity, and AMES mutagenicity to give us the final leads for the receptors.</p><!><p>In silico prediction of acute toxicity in rodent models for the molecules with possible polypharmacological prospects was performed using the GUSAR webserver (Languin et al., 2011). Using GUSAR, compounds can be evaluated based on the Quantitative Neighborhoods of Atoms Descriptors and Prediction of Activity Spectra for substances algorithm. The result thus obtained will be correlated using the SYMYX MDL toxicity database. Further, it is classified using the chemical classification manual of the Organisation for Economic Co-operation and Development (OECD) (Salman, et al. 2020).</p><!><p>The comparison at the sequence level for NSP9 protein structures of SARS-2 against SARS and H-CoV is shown in Figure 4 . The sequence of SARS-CoV-2 NSP9 (6W4B) protein structure was run for a protein Basic Local Alignment Search in NCBI (BLASTp) where we got a sequence percentage identity of 97.35% for the SARS CoV ( HKU-39849 ) NSP9 with PDB ID 1UW7 and 44.25% for the NSP9 of H-CoV 229E with PDB ID 2J97 ( Table 1 ). Upon aligning the proteins in EduPyMol program, we got a MatchAlign score of 556 for SARS2-SARS NSP9 with an executive RMSD of 0.707 while a MatchAlign score of 220 was obtained for SARS2-HCoV NSP9 superimposition with an executive RMSD of 1.522. This result indicates that SARS-CoV-2 NSP9 is structurally more close to SARS-CoV NSP9 as compared to H-CoV NSP9 ( Figure 4 ). The analysis can be further utilized for the design of broad-spectrum inhibitors for the prevention of the host cell attack (Gurung et al., 2020). Further, MSA results demonstrate the areas of conservation and disagreement in Figure 5 .</p><!><p>For the proteins, especially NSP9 and ORF3a, which are dimeric and tetrameric with similar chain sequences, chain A was chosen for the study for simplicity. For SARS CoV-2 receptors, NSP9 and ORF3a, the binding pockets were predicted using SiteMap (Schrodinger). Residues in the active site pockets included mainly M13, Y33, G38, G39, R40, F41, V42, L43 , F57, P58, K59, S60, I66, Y67, T68, E69, I92, K93 , G94, L95 and N96, among others. The sequence region 94-97 is found to be rich in polar and hydrophobic residues (Sutton et al., 2004). The active site residues in ORF3a include mainly K75, F79, N82, Q116, S117, N119, F120, R122, I123, L127 and L139, among others. The second or third region is properly responsible for Golgi localization. The residues K75, F79 and N82, lie in the second transmembrane helix domain, while residues Q116, S117, N119, F120, R122 and I123 in the third transmembrane helix region (UniProtKB -P0DTC3; https://www.uniprot.org/ ). For furin binding region, residues D153, H194, A252, S253, W254, P256, E257, N295, D301, E331, and S368 were obtained. Interleukin-6 contained four helices linked with loops (Somers et al., 1997). The C-terminal region (175-181) was found to be the receptor-binding domain with ARG179 as the key residue (Fontaine et al., 1993). A and D helices played vital roles in signal transduction and receptor binding. The degree to which amino acid residues will be conserved is highly dependent on the structural as well as functional significance. The ConSurf server used in this study estimated the same based on the phylogenetic relations between homologous sequences (Ashkenazy et al., 2016;Celniker et al., 2013;Ashkenazy et al., 2010), while PrankWeb uses the Jensen-Divergence method for calculating the conservation scores (Jendele et al., 2 019;Krivak et al., 2018).</p><!><p>Phytochemicals have also been reported to inhibit SARS-CoV-2 via a number of mechanisms (Ghildiyal et al., 2020;Gurung et al., 2020). Most phytochemicals like alkaloids, flavonoids, terpenoids, coumarins, and lignins inhibit the virus in the host due to their antioxidant activities, scavenging abilities, inhibition of DNA and RNA synthesis, or the blocking of viral reproduction (Naithani et al., 2010). Harnessing this potential of plant-based bioactives to speed up the much-desired development of COVID-19 therapeutics is strengthened by advances in computer-aided drug discovery (CADD) that deploys several computational technologies to the drug design process (Ebhohimen et al., 2020). This computational study identified promising phytochemicals from some of the reported antiviral and anti-inflammatory medicinal plants screened against SARS Cov-2 Nonstructural protein 9 (NSP9) RNA binding protein, and further scrutinizing the probable binding interactions it can have with other targets involved in COVID-19, including the open reading frame protein 3a (ORF3a) which is experimentally found to induce cellular apoptosis (Ren et al., 2020), furin convertase that activates the spike glycoprotein of SARS Cov-2 and interleukin-6, one among the many critical inflammatory cytokines that can significantly contribute to, fever, lymphopenia, coagulation, lung injury, and multi-organ failure (MOF) (Abbasifard et al., 2020).</p><p>The compounds collected range from chemical classes, including flavonoids, terpenoids, polyphenolics, saponins, thiophenes, furyl compounds, alkaloids, coumarins, sulfides, to polysaccharides. The custom workflow options of Mcule make the virtual screening process easier for execution, and the cloud computing server facilitates faster processing for quick screening. To identify the potent phytochemicals, virtual hits from the in house generated library of 521 phytochemicals, from the first filter of the workflow, were checked for the total number of RO5 violations; a total of 452 molecules were found to have either two or fewer violations. According to the Lipinski Rule of Five (RO5), if a compound has a molecular weight (Mw) > 500 Da, calculated M log P > 4.15, hydrogen-bond donors > 5, hydrogen-bond acceptors > 10, it is unlikely to be further pursued as a potential drug, because it would perhaps lack properties that are necessary for its absorption, distribution, metabolism, and excretion (Lipinski et al., 1997). Traditionally, a prominent number of drugs that target viral receptors deviate from Lipinski's RO5 (Doak et al., 2017) and have now opened immense potentials for compounds beyond RO5 drug space. Hence, to increase the chemical discovery space, we kept the maximum number of RO5 violations to two for further narrowing down the library of molecules to the most relevant molecules for the receptors of our choice .</p><p>These drug-like compounds were further taken for the next workflow step of Vina docking. The Vina docking algorithm (Trott and Olson, 2010) is the most frequently used tool for docking because of its high precision and robustness. The returned result for each of the receptors after the screening was set to the top 30 molecules with the best docking scores. Docking scores of the resultant phytochemicals ranged from a value of -7.1 to -9.1 kcal/mol for NSP9 (6W4B), -7.0 to -8.5 kcal/mol for ORF3a protein (6XDC), -6.9 to -8.1 kcal/mol for Interleukin-6 (1ALU) cytokine and -8.4 to -9.6 kcal/mol for the furin convertase (5JXI). Based on the binding energies of the virtual hits, the cutoff values of docking scores were selected for all the four receptors for the identification of potential inhibitors. Vina docking was manually performed for known inhibitors of all the four target proteins. Table 3 enlists the docking scores for the inhibitors used against each of the receptors. The docking scores of NSP9 against Verdinexor and Dabrafenib as inhibitors (Gordon et al., 2020) were found to be -7.7 and -6.8 kcal/mol, respectively. Therefore, particularly, based on Autodock Vina docking scores (ΔG) of the known inhibitors for NSP9, the cut off value was set as ≤ -7.8 kcal/mol. Similarly, upon docking ORF3a, with two of its inhibitors Emodin and Tranilast (Schwarz et al., 2011;Adnan, 2020), the docking scores found were -6.9 and -7.0 kcal/mol, respectively. Hence a strict benchmark of binding affinity for the identification of potential Phyto ligands for ORF3a was set as ≤ -7.1 kcal/mol. However, the docking scores of IL-6 and Furin with its inhibitor were far less than the maximum docking scores of the top 30 resultant molecules retrieved after screening. Hence, for Furin and IL-6, all the molecules were chosen for further analysis. While for the SARS CoV2 viral targets, cut-off values ≤ -7.8 kcal/mol for NSP9 RNA binding protein and ≤ -7.1 kcal/mol for ORF protein 3a obtained formed the basis for considering 6 and 23 compounds, respectively for further analysis. The top 6 ligands for NSP9 were further assessed for their MM-GBSA energy scores prior to the Molecular Dynamic simulation study. Table S3 (SI) contains the docking scores for all molecules that passed the docking thresholds for each of the receptors, along with the number of RO5 violations.</p><p>Of interest was how the top six phytochemicals, namely Ochnaflavone, Hispaglabridin B, Corylin, Glabrone, Licoflavone B, and Neoandrographolide, that passed the minimum cut off threshold for binding affinity scores for NSP9, performed for the three other target receptors that were chosen. Among these six potential phytochemicals, we found out that Ochnaflavone, Hispaglabridin B, and Licoflavone B passed the docking threshold for all other receptors. Also, Corylin was found to have passed the docking threshold for Furin and ORF3a, while Glabrone passed the docking threshold for Furin besides NSP9. Neoandrographolide was unique for the Nonstructural protein 9 and was not in the top 30 screened libraries for any of the other three receptors. A few other top NSP9 docked molecules were scrutinized, although they didn't pass the set threshold based on the standard drug chosen, but seemingly have good polypharmacological properties. Hispaglabridin A that had a binding affinity score of -7.7 kcal/mol is possibly a potent inhibitor for the other three receptors since it passed the threshold for ORF3a, Furin, and IL-6 with docking scores -7.3, -9.2, and -7.1 kcal/mol, respectively. Psoralidin, which had a binding affinity value of -7.6 kcal/mol for NSP9, was found to have a docking score of -8.1 kcal/mol against IL-6 and -9.0 kcal/mol against Furin. Licoflavone A, Licoflavonone, Kanzonol U, Hydnocarpin, and Glabrol are among the others that have promising affinity values for multiple targets.</p><!><p>Pharmacokinetic properties of drugs usually refer to the properties defining the movement of the drug into, through, and out of the body. Pharmacokinetics can simply be defined as the variation in drug concentration with time due to several processes involved in the absorption, distribution, metabolism, and excretion. Adverse drug reactions in the body mainly depend on patient-related factors (like renal function, genetic makeup, sex, age), and on the physicochemical properties of the drug, so is pharmacokinetics (Alomar, 2014). Pharmacokinetic properties of the compounds need to be identified before performing in vitro and in vivo studies. Additionally, the behavioral action of compounds inside host organisms needs to be ascertained in terms of its rate of absorption and excretion, metabolism, and distribution. Prediction of the drug ADME properties at the early stages of the drug discovery pipeline not only helps in removing the compounds that have poor ADMET properties, but it also helps in reducing research and development costs. The physicochemical properties of the molecules were obtained from the Mcule server, and admetSAR was used to get the ADMET properties for the drugs that met the threshold docking score. Table 4 displays the results for all the computed molecules. With the exception of Ononin and Calycosin 7-o-glucoside, all the potential ligand hits possessed human gastrointestinal absorption, while not many of them possessed blood-brain permeance capability.</p><p>As the final step to filter potential inhibitors for the targets chosen, among many properties, there was extensive focus on the phosphoglycoprotein (Pgp) substrate binding, cytochrome p450 (CYP2D6) inhibition, carcinogenicity, and AMES mutagenicity to filter out the potential inhibitors for the receptors of our study. In general, cytochrome P4502D6 plays a fundamental role in drug metabolism, involved in the metabolism of a variety of liver substrates (Lynch and Price, 2007). Major cases of drug-drug interactions occur due to its inhibition by any medication. Pgp is an ATP dependent efflux transporter glycoprotein, which reduces the efficacy of drugs that are p-gp substrates (Finch and Pillans, 2014). Interestingly, except for Hispaglabridin B, all the top virtual hits of NSP9 are noncarcinogenic, non AMES mutagenic, non inhibitor of CYP2D6 and are not pgp substrate either, and hence were chosen as the final leads for NSP9. Table 5 depicts the respective plant sources, geographical locations, and compound classes for the 5 final leads of NSP9, and their 2D structures (retrieved from PubChem) are demonstrated in Figure 6. Upon further filtering, the compounds for ORF3a, Furin, and IL-6 targets, 18, 23 and 19 final potential lead compounds were found for each of the receptors, respectively. As a side note, we must not also forget that development in drug discovery research has been successful in manipulating some of the functionalities to modify ADMET properties for efficient lead optimization. Small tweakings in the molecular structures can result in remarkable property manipulations.</p><!><p>Acute toxicity is an adverse effect that occurs after single or multiple exposures to a substance, usually within 24 hours. Thus the knowledge of the rodent acute toxicity of lead compounds is essential in drug design. The prediction of the chosen ligand's acute toxicity was performed using a free web server, GUSAR (Lagunin et al., 2011), which considers various routes of drug administration (subcutaneous, oral, inhalation, intravenous and intraperitoneal). LD50 values correspond to acute toxicity dose that leads to 50% mortality in 24 hours after administration of substance. Acute toxicity determined by inhalation or oral administration is pertinent for assessing overall toxicological risk, whereas that for intravenous is important for drug development (Polish et al. 2019). Rat acute toxicity evaluation was carried out for the five leads of NSP9 (including Hispaglabridin B), which presumably possess polypharmacological properties. Despite Hispaglabridin B failing the ADMET tests, it's predicted LD50 was analysed hence, its inclusion here too. The results of the rodent acute toxicity presented in Table 6 shows that the chosen ligands -Licoflavone B, Ochnaflavone, Corylin, and Glabrone can be considered to be low toxicity drugs (class 4 and 5). Ochnaflavone was non-toxic in intraperitoneal and subcutaneous routes of administration. However, Hispaglabridin B was observed to be of class 3 toxicity in oral administration. This also shows evidence for further experimental preclinical studies for the molecules as lead compounds for the treatment of COVID-19.</p><!><p>We identified various non-covalent interactions of the ligands with the amino acid residues of the binding pocket of NSP9, namely H-bonding, Vanderwaals, pi-pi T shaped, pi-pi stacked, pi-sulfur, pi-sigma, alkyl, pi-alkyl, carbon-hydrogen and pi-donor hydrogen bonds. Table S4 (SI) lists all the existing interactions between ligands and NSP9 active site residues for the ligand-receptor poses. The best ranked lead molecule, Ochnaflavone, interacted with the binding site of the target with a binding affinity of -9.1 kcal/mol. This interaction is strengthened by two hydrogen bonds with GLY94 and ARG40 with bond lengths 2.29 Å and 2.11 Å respectively, Vander waals interaction with the residues THR68, ILE92, GLY 39, GLY38, PHE57, LYS59, MET13, LYS93, LEU95, SER60 of the binding pocket, a pi-pi interaction with PHE41 and alkyl, and pi-alkyl interactions with VAL42, ILE66, and ARG40. Among the other lead molecules that establish hydrogen bonding with pocket residues include Glabrone (with ARG40: bond length 2.42 Å) and Neo-andrographolide (with LYS93 AND PRO58: bond lengths 2.60 Å and 2.31 Å, respectively). Figure 7 shows the NSP9-lead compounds binding interactions (including Hispaglabridin B). Corylin and Licoflavone B develop a pi-sigma interaction with THR68, while Neo-andrographolide interacts with the same via pi-donor hydrogen bond. The interaction of Neo-andrographolide is also interesting because there is a possible interaction with PHE57 and ARG40 via carbon-hydrogen bonds. Glabrone and Licoflavone B are observed to interact by a pi-sulfur bond with MET13. Alkyl interactions with ILE66 and VAL42 are prominent in all the ligands. Among all the pocket residues, it seems LYS93, MET13, ARG40, VAL42, PHE57, ILE66, and THR68 play vital role in strengthening the bonds between the ligands and the receptor.</p><!><p>We evaluated the binding interactions of the target receptors with the two compounds, Ochnaflavone, and Licoflavone B, that took our interest due to their potential multi-target inhibition properties, promising ADMET properties, and RAT acute toxicity results. Figure 8 shows the 2D representation of all the different kinds of interactions that occur between the ligands and IL6, furin, and ORF3a. We found that Ochnaflavone established prominent hydrogen bonds with the residues present in the binding pockets for all the three targets besides NSP9. At the same time, Licoflavone B was seen to form hydrogen bonding interactions just with the residues of Furin. Apart from hydrogen bonds, the major contributions of the interaction of Ochnaflavone comes from pi-alkyl, pi-pi T shaped and stacked, and van der waals interaction with nearby residues. Licoflavone B was seen to establish pi-sigma, pi-anion/cation, alkyl, and pi-alkyl, pi-pi T shaped and stacked, and van der waals interaction with the surrounding residue. Ochnaflavone interacted via hydrogen bonding with ASP301 and via van der waals interaction with GLU331, two key residues of the active site. Licoflavone B interacted with SER176 and ARG179 (one of the most critical active residues of IL-6) via van der waals interaction. Ochnaflavone established pi-alkyl interactions with LEU178 and ARG182, besides having van der waals interaction with ARG179. Both Ochnaflavone and Licoflavone B are observed to develop strong bonding interactions (as is shown in Figure 8 ) with the amino acid residues of the second and third transmembrane helix domain that plays an important role in the Golgi localisation.</p><!><p>Besides analysing the binding interactions of the ligands with the amino acid residues in the pocket site, the physicochemical nature of the binding site was examined, and the spread in the aromaticity, solvent accessibility, hydrophobicity, and distribution of H-bond donor/acceptor sites across the surface of the binding site were evaluated. Figure 9 gives the representation of the distribution of the mentioned properties. It was observed that for ORF3a, which becomes membrane protein in its biochemical pathway, the binding site is highly solvent accessible and is dominated by hydrophobic regions, as is for the case of NSP9. IL-6 binding region does not possess aromaticity. However, furin and interleukin-6 were both dominated by hydrophilic regions ( Figure S5 in SI ). H-bond donor/acceptor sites were approximately equally distributed throughout the binding region for all the four receptors.</p><!><p>The molecular mechanics energies combined with generalized Born Surface Area (MM-GBSA) is one of the attractive approaches that is widely and successfully used to improve the docking scores and results of virtual screening, apart from reproducing experimental outcomes. Post-scoring compounds utilizing MM-GBSA have been shown to have a better correlation to their observed binding affinity when compared to docking (Tripathi, et al., 2013;Greenidge, et al., 2013). Docking scores are not always accurate, and the MM-GBSA method gives a better approach to estimate the free binding energies of the protein-ligand complexes with a notably higher degree of accuracy. The molecular docking scores that we obtained for Ochnaflavone, Hispaglabridin B, Licoflavone B, and Verdinexor against NSP9 were -9.1 kcal/mol, -8.5 kcal/mol, -8.1 kcal/mol, and -7.7 kcal/mol, respectively. We, however, observed differences in the binding energy evaluated using the MM-GBSA approach. Hispaglabridin B returned the highest MM-GBSA score of -42.88 kcal/mol, followed by Licoflavone B with -42.76 kcal/mol; both of them were higher than that of Ochnaflavone (-41.43 kcal/mol) as presented in Table 7 . Compared to the Vina docking scores, the top three inhibitors have been predicted to have better binding energy values when compared with the reference molecule (Verdinexor), suggesting even stronger binding.</p><!><p>In order to assess the protein-ligand complex stability for the top three inhibitors identified for NSP9, MD Simulations of 100 ns each were performed for Licoflavone B, Ochnaflavone, and Hispaglabridin. A 100 ns MD run for the free protein (apo-state) was also carried out. For the investigation of the stability, the root mean square deviation (RMSD) of the Cα atoms of the protein and the ligand bound to the protein, root mean square fluctuations (RMSF) of the Cα atoms of the protein. Also, of interest was understanding the dynamics of the protein-ligand contacts with time. The RMSD values for NSP9 free state were within 1.20 Å to 3.6 Å, while the average RMSD asymmetric carbon was 2.22 Å. The RMSD of the Cα atoms of NSP9 is pretty stable for Ochnaflavone and Hispaglabridin B after 40 ns, while for Licoflavone B, the stability is observed after 60 ns simulation run time. In the case of Licoflavone B ( Figure 10C ), a very stable RMSD of the ligand after 60ns was also observed. Hispaglabridin B ( Figure 10B ) showed a stark increase in its RMSD at around 25 ns and remains stable throughout the rest of the duration. This is perhaps due to a conformational change. On the other hand, Ochnaflavone showed a largely stable RMSD after 50 ns. Figure 11 shows the RMSF values of the apo-NSP9 and protein-ligand complexes. RMSF was employed to follow local changes along the NSP9 amino acid residues for the 100 ns simulation time. It was observed that the alpha helices and beta strands of the apo structure and docked systems oscillated within 0.8 Å to 1. 6 Å. The loop regions of all the simulated systems showed large fluctuations up to 4.8 Å, except for the apo state that showed change over 5.4 Å. Remarkably, there were no noteworthy instabilities in the loop regions when compared across the complexes. Understandably, the large fluctuations of the loops observed were due to their intrinsic flexible nature. Our results showed that the lead molecules complex with NSP9 were found to be stable in the binding site with negligible structural movements and lesser conformational changes to the overall enzyme structure, hence presenting them as potential inhibitors against NSP9. Upon analysis of the protein-ligand contacts, it was found that both Licoflavone B and Ochnaflavone have consistent H-bonding as well as H-bonding mediate by water interaction with VAL42 throughout the entire duration of 100ns of the simulation run. Licoflavone B was found to be in a prominent hydrophobic interaction with PHE41, while Ochnaflavone developed stable interactions with ASN96 and ASN99 via H-bonding mediated by water at 60% of the simulation run time. PRO58 and LYS93 are among the other important residues of contact for Ochnaflavone. Hispaglabridin B shows a variable interaction profile with the residues. It communicated with ARG40 via hydrogen bonding and water bridges at 26% at a time. Consistent interaction with GLN12 was maintained at 52% through H-bonding together with water bridge contacts, and more robust interactions with TYR32 via H-bonding, water bridge, and hydrophobic contacts. Also, compared to Ochnaflavone and Licoflavone B, Figure 11 and Figure 12 Hispaglabridin B shows a much greater number of residues contact with NSP9 throughout the duration; however, most of the contacts were less than 10% the simulation time.</p><!><p>COVID-19 disease has remained a significant menace to the world, and the case to fatality percentage has remained on the rise. In order to tackle this menace, there is a need to develop therapeutic intervention rapidly. Here, extensive screening of our manually curated library via docking thresholds and stringent ADMET filters identified 5 leads for NSP9, and 23, 18, and 19 leads for Furin, ORF3a, and interleukin-6, respectively. Among the top five leads, Ochnaflavone, a biflavonoid extract from Lonicera japonica, and Licoflavone B, a flavonoid extract from Glycyrrhiza glabra showed promising polypharmacological prospects, as identified by their docking scores, ADMET properties, and binding interactions with the receptor site residues. Atomistic simulations of 100ns validated the stability of the Nsp9-Ligand complex. The current study comes up with convincing evidence about the antiviral and anti-inflammatory indications of these phytochemicals with great potential to inhibit simultaneous viral entry, replication, and disease progression, further opening up the quest for pre-clinical experimental studies.</p>

ChemRxiv

Photoactivation of trans platinum diamine complexes in aqueous solution and effect on reactivity towards nucleotides

We show that UVA irradiation (365 nm) of the PtIV complex trans,trans,trans-[PtIVCl2(OH)2(dimethylamine)(isopropylamine)], 1, induces reduction to PtII photoproducts. For the mixed amine PtII complex, trans-[PtIICl2(isopropylamine)(methylamine)] (2), irradiation at 365 nm increases the rate and extent of hydrolysis, triggering the formation of diaqua species. Additionally, irradiation increases the extent of reaction of complex 2 with guanosine-5\xe2\x80\xb2-monophosphate (GMP) and affords mainly the bis-adduct, while reactions with adenosine-5\xe2\x80\xb2-monophosphate (AMP) and cytidine-5\xe2\x80\xb2-monophosphate (CMP) give rise only to mono-nucleotide adducts. Density Functional Theory calculations have been used to obtain insights into the electronic structure of complexes 1 and 2, and their photophysical and photochemical properties. UVA-irradiation can contribute to enhanced cytotoxic effects of diamine platinum drugs with trans geometry.

photoactivation_of_trans_platinum_diamine_complexes_in_aqueous_solution_and_effect_on_reactivity_tow

1. Introduction<!>2. Materials and Methods<!>2.1. Sample preparation<!>2.2. Nuclear magnetic resonance (NMR) Spectroscopy<!>2.3. Mass Spectrometry<!>2.4. Liquid Chromatography-Mass Spectrometry<!>2.5. Irradiation<!>2.6. Computational Details<!>3.1. Irradiation of trans,trans,trans-[PtIVCl2(OH)2(dma)(ipa)] (1) in aqueous solution<!>3.2. Irradiation of trans-[PtIICl2(ipa)(ma)] (2) in aqueous solution<!>3.3. DFT and TD-DFT studies of photoactivation mechanisms for 1 and 2<!>3.4. Photoinduced reactivity of trans-[PtIICl2(ipa)(ma)] (2) with DNA model bases<!>3.4.1. Reactivity of trans-[PtCl2(ipa)(ma)] (2) with GMP<!>3.4.2. Reactivity of trans-[PtCl2(ipa)(ma)] (2) with AMP<!>3.4.3. Reactivity of trans-[PtCl2(ipa)(ma)] (2) with CMP<!>4. Conclusions

<p>In recent years there have been major advances in understanding the factors which control the binding of platinum anticancer drugs (such as cisplatin) to DNA and of the consequences of DNA binding, which ultimately lead to cell death [1, 2]. An important step in the activity of platinum anticancer complexes is activation through hydrolysis [3], which is often a necessary and rate-limiting step in their reactions with DNA and other biomolecules. Not only is there current interest in cis diam(m)ine platinum complexes, but also in trans diam(m)ine complexes which can also exhibit interesting activity [4-11], even though transplatin itself is inactive [12]. Our recent investigations of trans platinum complexes containing aliphatic amines, such as trans-[PtCl2(15N-ipa)(15N-ma)], trans-[PtCl2(15N-dma)(15N-ma)] and trans-[PtCl2(15N-dma)(15N-ipa)] (ipa = isopropylamine, ma = methylamine, and dma = dimethylamine) by 2D [1H,15N] HMQC NMR spectroscopy suggest that the dichlorido (less reactive) species is the predominant form present in cancer cells under physiological conditions [13], despite the high cytotoxic activity exhibited by these complexes.</p><p>Photodynamic therapy (PDT) has attracted considerable attention for the treatment of a variety of cancers. The photochemistry of the platinum complexes is of particular interest since light can induce a wide range of electronic transitions [14]. Photoactivable PtIV-azide complexes represent a promising area for new drug development [15-17].</p><p>Photochemical studies on transplatin (trans-[PtIICl2(NH3)2]), cisplatin (cis-[PtIICl2(NH3)2]) and cisplatin-like complexes have shown that irradiation can induce substitution of chlorido ligands by solvent H2O or DMSO [18-20]. UVA light promotes the loss of the second chlorido of transplatin and also promotes DNA binding and cytotoxicity [21]. Here we investigate the possibility of using irradiation to activate the cytotoxic PtIV mixed diamine trans complex trans-[PtIVCl2(OH)2(dimethylamine)(isopropylamine)], 1, and the PtII analogue, trans-[PtIICl2(isopropylamine)(methylamine)], 2 (Figure 1) through photoreduction [22, 23] and photosubstitution. We report the effect of irradiation on the aqueous behaviour of complexes 1 and 2. In addition, we have studied the influence of irradiation over time on reactions of complex 2 with the nucleotides guanosine-5′-monophosphate (GMP), adenosine-5′-monophosphate (AMP), and cytidine-5′-monophosphate (CMP) (Figure 1). The reactions have been monitored by NMR spectroscopy and the corresponding adducts in each case (Table 1) have been characterized by NMR and mass spectrometry.</p><!><p>Trans,trans,trans-[PtCl2(OH)2(dma)(ipa)], 1, and trans-[PtCl2(ipa)(ma)], 2, were prepared using previously described procedures [24, 25]. Guanosine-5′-monophosphate (GMP), cytidine-5′-monophosphate (CMP) and 9-methyladenine (9MeA) were purchased from Sigma-Aldrich and adenosine-5′-monophosphate (AMP) from Acros Organic.</p><!><p>Stock solutions of 2.5 mM platinum complex were prepared unless otherwise stated. Stock samples were diluted for MS and LC-MS as explained below. Reactions between platinum complexes and mononucleotides were performed at 1:2 molar ratios. The pH of the final solution in each experiment was adjusted in the range 6–7 with 0.1 M NaOH. All pH measurements were made at 298 K on a Martini MI150 pH meter equipped with a chloride-free semi-micro combination electrode (Thermo Fisher Scientific) calibrated with standard buffers (pH 4, 7 and 10, Aldrich).</p><!><p>1D 1H and 2D [1H, 13C] HSQC NMR spectra were recorded on a 500 MHz Bruker spectrometer using 5 mm tubes at 310 K. Solutions were prepared in 100 mM NaClO4 and 90% H2O / 10% D2O with a final concentration [Pt] = 2.5 mM for 1H and [1H,13C] HSQC, and 15 mM for 195Pt NMR. 1H, and 195Pt NMR chemical shifts were internally referenced to TSP via 1,4-dioxane (3.76 ppm), and externally to K2PtCl6 (0 ppm), respectively. Water signals were suppressed using presaturation or Shaka methods [26].</p><!><p>MS characterization studies were carried out on an HCT-plus Bruker ion trap mass spectrometer. The final concentration used in these samples was 15 μM [Pt]. EsquireControl 5.2 software was used to analyse the MS spectra. Isotope modelling of the MS peaks was performed for all the assignments. Pt–9MeA adducts were characterized by ESI-MS on an API Q-START Pulsari Q-TOF mass spectrometer in the positive ion mode diluting the samples 50% with methanol. The final concentration was 50 μM [Pt].</p><!><p>High performance liquid chromatography (HPLC) was carried out on an Agilent ChemStation 1100 Series instrument, with a DAD detector. Part of the outflow was routed to the Bruker HCT-plus ion trap MS. Hystar 3.0 software was used to analyse the LC-MS data. The final concentration used in the LC samples was 1 mM [Pt]. For all analytical separations, a RP C18 column (250 × 4.6 mm, 100 Å, 5 μm, Hichrom) was used, eluting with 5–30% acetonitrile gradients over varying time intervals with 0.1% trifluoroacetic acid as an ion-pairing agent and with UV detection at 275 nm.</p><!><p>The light source was a Luzchem LZC-ICH2 illuminator (photoreactor) oven using both Luzchem LZCUVA (Hitachi) and LZC-VIS (Sylvania cool white) lamps, with no other sources of light filtration. The photoreactor operated at 365 nm (λmax) with temperature controller at 310 K and power level of 1.9 mW/cm2. The power levels were monitored using the appropriate probe window, calibrated with an OAI-306 UV power meter from Optical Associates, Inc.</p><!><p>All calculations were performed with the Gaussian 03 (G03) program [27] employing the DFT method and the PBE1PBE [28] functional. The LanL2DZ basis set [29] and effective core potential were used for the Pt atom and the 6-31G**+ basis set [30] was used for all other atoms. Geometry optimizations of 1 and 2 in the ground state (S0) and lowest-lying triplet state (T1) were performed in the gas phase and the nature of all stationary points was confirmed by normal mode analysis. For the T1 geometries the UKS method with the unrestricted PBE1PBE functional was employed. The conductor-like polarizable continuum model method (CPCM) [31] with water as solvent was used to calculate the electronic structure and the excited states of 1 and 2 in solution. Thirty-two singlet excited states with the corresponding oscillator strengths were determined with a Time-dependent Density Functional Theory (TD-DFT) [32, 33] calculation. The electronic distribution and the localization of the singlet excited states were visualized using the electron density difference maps (EDDMs) [34-36]. GaussSum 1.05 was used for EDDMs calculations and GaussSum 2.2 for the electronic spectrum simulation (νfwhm = 4000 cm−1) [37].</p><!><p>The thermal (dark, 310 K) and photo-induced (365 nm light) behaviour of complex 1 in water (2.5 mM, pH ca. 6) was monitored for 4 h using 1H NMR spectroscopy. Characterization of the products after 4 h was carried out by LC-MS and 195Pt NMR techniques.</p><p>The 1H NMR spectrum of complex 1 in the dark over a period of 4 h (Figure 2A) showed no changes, indicating thermal stability. However, for the sample under irradiation (Figure 2B) several changes were observed. The doublet corresponding to the isopropylamine methyl groups (1.33 ppm) appears to split into two doublets, attributable to either isomerization and/or aquation of the PtIV complex as a result of the irradiation. Such effects have been documented previously for PtIV complexes containing halides and ammonia ligands, and have been used in synthesis [23, 38, 39]. Accordingly, it has been reported that irradiation of PtIV azide complexes gives rise to at least five additional PtIV light-induced photoproducts [16]. Irradiation gave rise to new 1H resonances at 1.27 ppm corresponding to PtII species. This doublet is assignable to the methyl groups of the PtII counterpart [PtCl2(dma)(ipa)]. This doublet appears to be composed of two overlapping doublets, likely to be geometrical isomers cis and trans [23]. Additionally when the sample was irradiated, proton resonances for PtIV–NH and PtIV–NH2 groups (6.28 and 5.47 ppm, respectively) disappeared, and a new peak arose in the region of the PtII–NH2 signals (broad signal at 4.04 ppm, not shown) which provided evidence for photoreduction of complex 1. Several new peaks of low intensity were also observed in the high-field region of the 1H NMR spectrum of the irradiated sample, which can be assigned to photo-isomerisation and photo-substitution products formed upon irradiation.</p><p>In order to identify the nature of the photoproducts, the samples were analyzed by LC-MS. For the sample in the dark, elution produces only one peak at 7 min, and no differences were observed over 4 h (Figure S1A). However, when the sample was irradiated, the chromatogram showed two additional new peaks, eluting at 29 and 30 min (Figure S1B). The positive ion ESI-MS of the peak at 7 min showed a species at 426.96 m/z corresponding to {C5H18Cl2N2O2Pt + Na}+ assignable to complex 1 (Figure S2A). The ESI-MS data obtained for the two new peaks eluting at 29 and 30 min were alike (Figures S2B and S2C, respectively), showing a similar molecular ion at 376.07 and 376.02 m/z, respectively, both assignable to {C5H17ClN2Opt + Na}+, likely to be cis and trans isomers of reduced photoproducts of 1 [23].</p><p>The 195Pt NMR spectrum of a sample of 1 (15 mM, pH ca. 6) was recorded before and after the irradiation. The PtIV complex 1 was the only species present in solution before the irradiation (δ(195Pt) +798 ppm). After irradiation a signal at −2185 ppm appeared in the spectrum (Figure S3). 195Pt NMR chemical shifts are highly dependent on the oxidation state of the metal, with PtII considerably more shielded than PtIV [40, 41]. The signal at −2185 ppm corresponds to a PtII species, supporting the photoreduction reaction [42]. However, the 195Pt NMR spectrum failed to show any other PtII species, possibly because they do not accumulate in high enough concentration to be observed by 195Pt NMR spectroscopy.</p><p>Comparison of the LC-MS and 195Pt NMR results obtained for PtIV and its PtII analogue indicates that 1 undergoes reduction under UVA light, in agreement with previous results obtained by 1H NMR. The reduction mechanism of complex 1 in aqueous solution seems to be triggered, or at least promoted, by irradiation.</p><!><p>We studied the effect of irradiation on the hydrolysis of 2, trans-[PtCl2(ipa)(ma)] by 1H NMR. Figure 3A shows the progress of hydrolysis of complex 2 in the dark. Two sets of 1H NMR signals at 1.27 and 1.31 ppm, assignable to isopropyl groups of the chloride (2) and monoaqua (2A) complexes, respectively. In the presence of light (Figure 3B), the same solution quickly gave rise also to a new set of peaks with a 1H NMR chemical shift of 1.35 ppm, which is assignable to the diaqua species (2AA) [13]. Peaks corresponding to the diaqua species (2AA) are clearly dominant over the monoaqua species after 1.5 h irradiation (Figure 4). The diaqua species was detected only in the non-irradiated sample after 20 h [13].</p><p>Photolytic reactions of square-planar platinum(II) complexes in aqueous solutions include photoaquation and photoisomerizations [18, 20, 22]. For example, the photoaquation of cis- and trans-[Pt(NH3)2Cl2], with the concomitant release of chlorido ligands, has been previously reported and the quantum yield values for the photolysis reaction of both isomers determined [20]. Additionally, irradiation of trans-[Pt(NH3)2Cl2] induced further release of chloride in aqueous solution. The release of the second chloride from transplatin as a result of photoactivation may provide a mechanism for activation of the inactive trans complex [21].</p><!><p>DFT and TD-DFT calculations have been successfully employed to obtain insights in the mechanism of photoactivation of several metal complexes [43-46]. Characterization of singlet and triplet states (in the case of diamagnetic d6 metal complexes) can reveal fundamental information on such process. For this reason, the singlet ground state (S0) and the lowest-lying triplet state (T1) geometries were optimized for both 1 and 2 and thirty-two singlet states were calculated by TD-DFT from the S0 geometries of the two complexes to characterize their UV-Vis properties. A comparison between the calculated and experimental UV-Vis spectra of 1 and 2 is reported in the Supporting Information (Figures S4–S6 and Tables S1–S4), together with the orbital composition of selected singlet transitions. Interestingly, such calculations highlight that all the UV-Vis singlet transitions of 1 and 2 in the 210–500 nm range involve the LUMO orbital (Figure 5), which is σ*-antibonding. The contribution of this orbital gives a strong dissociative character to such transitions, particularly towards the Pt–Cl bonds. This result is consistent with the labilization of the chlorido ligands upon light excitation. Furthermore, the lowest-lying triplet state (generally populated after the intersystem crossing promoted by the metal spin-orbit coupling occurs) is dissociative as well. The highest-SOMO, in fact, corresponds to the LUMO of the ground state (shown in Figure 5), resulting in a dissociative state. The lowest-lying triplet geometry of 1 is distorted and has elongated Pt–Cl bonds, consistent with the dissociation of one of these ligands (Table 2). In the case of 2 the elongation of Pt–Cl bonds upon triplet formation is limited, although there is a strong distortion of the Cl–Pt–Cl angle. As shown previously in the case of other photoactive PtIV derivatives, the dissociation of one or more chlorido ligands and the subsequent formation of hydroxo-species can lead to the reduction of the Pt centre via reductive photoelimination [47].</p><!><p>We studied the effect of irradiation on reactions of complex 2, trans-[PtCl2(ipa)(ma)], with the nucleotides: guanosine-5′-monophosphate (GMP), adenosine-5′-monophosphate (AMP) and cytidine-5′-monophosphate (CMP).</p><!><p>The photoinduced reaction between complex 2 and two mol equivalents of GMP was monitored by 1H NMR by following the changes in the H8 peak of GMP (Figure 6), a resonance which is often useful for monitoring interactions of this nucleotide with metal complexes [48, 49]. The following GMP H8 peaks were assigned: 8.15 ppm, free nucleotide; 8.84 ppm for the mono-nucleotide adduct complex 3, trans-[PtCl(GMP)(ipa)(ma)]+; and 8.97 ppm for the bis-adduct trans-[Pt(GMP)2(ipa)(ma)]2+ complex 4. The intensity of the H8-4 signal (bis-adduct) in the photoinduced reaction was stronger than for the mono-nucleotide adduct after 4.5 h and also increased faster than for the non-irradiated sample (Figure 7). Monitoring the H8 chemical changes of the GMP ligand from the mono- and bis-nucleotide adducts versus pH suggests that the platination occurs at N7 of GMP since chemical shift changes associated with the pKa of the N7 were not observed, i.e. no changes in the H8 chemical shift of platinated GMP in the pH range 2–4 (Figure S7) [50]. Similar results have been reported for complex 2 on reaction with 4 mol equivalents of GMP (10 mM NaClO4, 310 K) under non-controlled lighting conditions [51].</p><p>The presence of mono- and bis-GMP adducts was confirmed by MS. The mass spectrum of the non-irradiated sample (Figure S8A) showed the mono-nucleotide species: [PtCl(GMP)(ipa)(ma)]+ (3), assigned from the peak at m/z 684.12, {C14H28ClN7O8PPt}+, and the sodium adduct at m/z 706.09, {C14H27ClN7O8PPt + Na}+. Minor peaks were detected at m/z 665.16, {C14H29N7O9PPt}+, assignable to the hydrolysed mono-nucleotide adduct [Pt(H2O)(GMP)(ipa)(ma)]2+ (3A) and m/z 1010.23, {C24H41N12O16P2Pt}+, and (paired with sodium ion) 1032.18, {C24H40N12O16P2Pt + Na}+, both assignable to the bis-nucleotide adduct [Pt(GMP)2(ipa)(ma)]2+ (4). In the MS of the irradiated sample, the signals for the bis-nucleotide adduct were significantly more intense (Figure S8B), in agreement with the NMR results. The predominance of the bis-GMP adduct signals over the mono-nucleotide adduct has been reported for complex 2 [51] and other similar trans-platinum(II) complexes, such as trans-[PtCl2(NH3)(c-C6H11NH2)] [52], under thermal conditions. However, in those cases, the reaction requires a significantly longer period of time (14 h for 2, and ca. 24 h for the latter) to achieve a similar percentage of speciation. The use of irradiation appears to activate and enhance the formation of the bis-nucleotide adduct 4. However, unlike the findings for transplatin, where the complex was totally consumed at the end of a 3 h-irradiation [21], most of complex 2 (ca. 75%) stays in the original dichlorido form. On the other hand, total conversion of transplatin into trans-[Pt(NH3)2(GMP)2]2+ under non-irradiating conditions has been reported to occur within five hours at 308 K when the Pt/nucleotide ratio was 0.5 [48].</p><!><p>The effect of irradiation on the reaction of complex 2 with AMP was also investigated and monitored by NMR spectroscopy. The changes in the H8 and H2 1H NMR signals of AMP and its platinated adducts were followed over the first 4.5 hours. Two doublets were detected for both H8 and H2 with chemical shifts of 9.26 ppm (H8-5) and 9.15 ppm (H8-5A) for H8, and 8.40 (H2-5) and 8.38 (H2-5A) for H2, in both non-irradiated and the irradiated samples (Figure 8). These signals in the 1H NMR spectra indicate the presence of two different AMP adducts. The 1H NMR signals of both adducts were of higher intensity after 4.5 h in the dark than under irradiation (Figure S9 shows the species distribution curves over 4.5 h).</p><p>The Pt–AMP adducts were characterized by MS. No bis-adducts were detected in the reaction of 2 with AMP. Two different species were assigned in the MS spectra to the following mono-nucleotide adducts: [PtCl(AMP)(ipa)(ma)]+ (5) to the peak at m/z 668.12 ({C14H28ClN7O7PPt}+), and peak at m/z 690.10, {C14H27ClN7O7PPt + Na}+ and [Pt(OH)(AMP)(ipa)(ma)]+ (5A) to the peak at m/z 649.16, {C14H29N7O8PPt}+ (Figure S10).</p><p>In order to complete the assignment of the peaks observed by NMR, the magnitudes of chemical shift differences for H8C8 and H2C2 between the free base and the metallated base in the 2D [1H,13C] NMR spectra were analysed, since this appears to be a useful diagnostic tool [53, 54] for assignment of the type of binding in metal-nucleotide adducts. The shift changes for the 13C resonances were similar and small (up to ca. 3 ppm) for both mono-nucleotide adducts (Figure 9). The H8-adenine proton resonance, however, was low-field-shifted by up to ca. 0.7 ppm compared to free AMP for both complexes 5 and 5A, while the H2 proton resonance was low-field-shifted by ca. 0.1 ppm compared to free AMP for both adducts. The greater shift changes in both species for proton H8 compared to H2 are attributed to N7 (as opposed to N1) AMP-Pt coordination [54]. A NOESY spectrum of the sample after 4 h of reaction of 2 and 5′-AMP showed cross-peaks only between the H8 of the bound nucleotide and the isopropylamine and methylamine ligands in the platinum complex at 9.26 and 9.15 ppm, also suggesting coordination at N7. Interestingly, the two doublets, assigned to H8 in both mono-nucleotide adducts 5 and 5A ([PtCl(AMP)(ipa)(ma)]+ at 9.26 and [Pt(OH)(AMP)(ipa)(ma)]+ at 9.15 ppm, respectively), coalescenced into one doublet upon lowering the pH with HClO4 (Figure S11). On increasing the pH to give basic solutions, however, the intensity of the H8-signal corresponding to one of the mono-nucleotide adducts decreased with concomitant increase in intensity of the H8-signal of the other mono-nucleotide adduct, hence the assignment of the former signal to 5, and the latter to 5A. Additionally, treatment of the solution mixture with 0.15 M NaCl resulted in an increase in the intensity of the peaks assigned to 5, although the anation was incomplete. Further addition of NaOH solution reversed the equilibrium towards 5A.</p><p>LC-MS studies were carried out (mobile phase 0.1% TFA methanol/water, pH 2). Two different species were separated: AMP was detected as unreacted starting material at 348.06 m/z, {C10H15N5O7P}+, and the mono-nucleotide adduct complex 5, [PtCl(AMP)(ipa)(ma)]+, at 668.09 m/z, {C14H28ClN7O7PPt}+. Under acidic conditions the only mono-nucleotide adduct present was complex 5, [PtCl(AMP)(ipa)(ma)]+.</p><p>The assignation of species 5 and 5A was confirmed by an experiment carried out with another adenine derivative, 9-methyladenine (9MeA). When complex 2 was reacted with 9MeA (Pt/nucleobase ratio 1:2, pH ca. 6.3, 24 h) three adducts were observed in the 1H NMR spectrum after just one hour (Figure S12). The ESI-MS recorded after 24 h showed two peaks at m/z 451.15 ({C10H22N7OPt}+) and 470.12 ({C10H21ClN7Pt}+), assignable to trans-[Pt(OH)(9MeA)(ipa)(ma)]+ and trans-[PtCl(9MeA)(ipa)(ma)]+, respectively. No bis-nucleotide adducts were detected. Analysis of the 2D [1H,13C] HMQC NMR spectra allowed assignment of the three adducts as [PtCl(N7-9MeA)(ipa)(ma)]+ (6), [Pt(OH/OH2)(N7-9MeA)(ipa)(ma)]+/2+ (6A), and [PtCl(N1-9MeA)(ipa)(ma)]+ (7) (Figure 10). The assignment is based on the extent of shift of the 2D [1H,13C] H8C8 and H2C2 signals of the platinated nucleobase when compared to those of the free nucleobase. N7-coordination produced more pronounced low-field shifts (by up to ca. 0.8 ppm) of H8 in N7-metallated adenine, while as little as ca. 0.2 ppm for H2. Conversely, N1-coordination produced low-field shifts by up to ca. 0.6 ppm for H2 in N1-metallated adenine, versus 0.1 ppm for H8. The ESI-MS under acidic conditions (0.5% TFA) showed only the monochlorido/mono-9MeA adduct, [PtCl(9MeA)(ipa)(ma)]+, in agreement with the observations on the reaction of complex 2, trans-[PtCl2(ipa)(ma)], with AMP.</p><p>Similarly, Liu et al. have reported an analogous behaviour for the thermal reaction of trans-[PtCl2(E-iminoether)2] with one mol equivalent of GMP or AMP. Trans-[PtCl2(E-iminoether)2] appears to form two different mono-nucleotide adducts in aqueous solution at equilibrium (298 K), trans-[PtCl(E-iminoether)2(nucleotide)]+ and trans-[Pt(OH2)(E-iminoether)2(nucleotide)]2+. The latter reversibly converts into the chlorido species at low pH and at high chloride concentration [55].</p><p>Interestingly, the proton resonances corresponding to H8 and H2 for the Pt–AMP adducts show doublet splitting whilst the Pt–GMP and Pt–9MeA complexes did not. This splitting effect may arise from hindered rotation about the Pt–N bond [56-58]. However, temperature-dependent NMR experiments failed to show coalescence up to 323 K. In agreement with the observations of this work, Marzilli et al. have observed greater steric effects in PtII(AMP)2 adducts in comparison to PtII(GMP)2 adducts, as confirmed by the coalescence temperature [57]. Additionally, Lippert et al. have previously observed an H2 splitting phenomenon for trans-[Pt(MeNH2)2(1-MeT-N3)(9-MeA-N1)]+ in D2O. The authors attributed this to hindered rotation of the nucleobases about the Pt–N bonds. Most importantly, it appears that hindered rotation depends on the presence of methylamine ligands bound to Pt since the NH3 analogue did not exhibit the same behaviour [59, 60].</p><!><p>Brabec and Leng showed that transplatin forms preferably GC interstrand crosslinks over GG adducts under competitive conditions [61]. The reaction of complex 2 with CMP in the dark and under irradiation at 365 nm was followed by 1H NMR spectroscopy and ESI-MS. The 1H NMR doublet corresponding to H6 of the cytidine mono-adducts was monitored over various irradiation times. Two new low-field-shifted doublets appeared and little difference was observed in the 1D 1H NMR spectra of irradiated samples in comparison to those kept in the dark (Figures S13 and S14). Accordingly, the ESI-MS data showed comparable spectra after 4 h reaction under irradiation and thermal conditions (Figure S15), with the main species being complex 8, [PtCl(CMP)(ipa)(ma)]+, as a molecular ion at m/z 644.10 {C13H28ClN5O8PPt}+ and m/z 666.07 {C13H27ClN5O8PPt + Na}+. No bis-adducts were detected by MS. When the NMR sample was acidified to pH 2 only one Pt–CMP adduct remained (Figure S16). Increasing the pH to 11 did not have an effect on the intensity of the signals. Due to the similarities between this reaction and the reaction of complex 2 with AMP, we tentatively assign the two new adducts as the mono-CMP/mono-chlorido adduct and mono-CMP/mono-aqua adduct, complexes 8 and 8A, respectively. Additionally, Figure S16B confirms CMP preferential binding to the metal centre through N3, which represents the more basic site (pKa 4.6) as compared to N4 (pKa 17) [50].</p><!><p>Irradiation with UVA light (365 nm) of complex 1, trans,trans,trans-[PtIVCl2(OH)2(dimethylamine)(isopropylamine)], in aqueous solution promotes reduction of platinum(IV) to platinum(II). When complex 2, trans-[PtCl2(ipa)(ma)], was irradiated in aqueous solution, the formation of bis-aqua species was facilitated. TD-DFT calculations suggested that all the UV-Vis singlet transitions of 1 and 2 in the 210–500 nm range involve the σ*-antibonding LUMO orbital, which has a strongly dissociative character (favouring Pt–Cl dissociation). The lowest-lying triplet state is dissociative as well. When an aqueous solution of 2 was irradiated in the presence of two mol equivalents GMP, formation of bis-adduct [Pt(GMP)2(ipa)(ma)]2+ was favoured; however, this was not the case for AMP nor CMP. When complex 2 reacted with AMP the mono-nucleotide adduct species [PtCl(AMP)(ipa)(ma)]+, 5, and [Pt(OH)(AMP)(ipa)(ma)]+, 5A, were detectable by NMR as the main Pt–AMP adducts in solution. Interestingly, when 2 was reacted with phosphate- and sugar-free nucleobase derivative 9MeA, a new N1-bound species was found. When complex 2 reacted with CMP, only mono-nucleotide adducts [PtCl(CMP)(ipa)(ma)]+ (8) and [Pt(CMP)(ipa)(ma)(OH/OH2)]+/2+ (8A) could be observed and no bis-nucleotide adduct was detected. Our experiments suggest that irradiation at 365 nm can be used to activate trans-PtIV di-amine complexes and aid the formation of bis-nucleotide adducts with guanine derivatives, the preferred binding site for DNA-targeting platinum drugs.</p>

PubMed Author Manuscript

Simeprevir and eltrombopag as potential inhibitors of SARS-CoV2 proteases: a molecular docking and virtual screening approach to combat COVID-19

Pandemic of COVID-19 has disastrously affected human health and wealth. With no approved drugs until today, the global health system is struggling to find an effective and universal treatment measure. At this crucial juncture, repurposing of existing drugs and medication seems to be the only way out to deal with SARS-CoV2. To find a suitable drug, the present study investigates the binding affinity of 61 selected FDA approved drugs with two key virus proteins, 3-chymotrypsin-and papain-like protease of both SARS-CoV and SARS-CoV2. Furthermore, best binding drugs were investigated for ADMET properties. The 3D structures of SARS-CoV and -CoV2 proteases and drugs were downloaded from PDB and DrugBank databases, respectively. Docking was carried out using AutoDock vina and the output file visualized in Discovery Studio. Druglikeness and toxicity were studied using SwissADME and ADMETlab. Simeprevir (-9.4 kcal/mol) and eltrombopag (-9.3 kcal/mol) were found to be the best 3CLpro binding drugs in SARS-CoV2 and SARS-CoV, respectively. Similarly, eltrombopag showed the best affinity to PLpro in both the viruses. The average docking score of 61 drugs with 3CLpro and PLpro was found to be -7.19 and -6.56 kcal/mol in SARS-CoV2 and -7.11 and -7.06 kcal/mol in SARS-CoV, respectively. ADMET study of the top ten drugs revealed that most of the drugs are highly absorbed by the human intestine and pass through the blood-brain barrier and distribute easily. Furthermore, most of the drugs including simeprevir, saquinavir, and vaniprevir were found to act as strong inhibitors of CYP-enzyme complex which therefore cannot metabolize the drugs before reaching the target site. In-silico toxicity study did not show any carcinogenic property of the top ten drugs. The findings of the present study, thus suggests that the use of existing antiviral drugs as well as unknown antiviral drugs such as eltrombopag, digoxin may prove effective against SARS-CoV2.

simeprevir_and_eltrombopag_as_potential_inhibitors_of_sars-cov2_proteases:_a_molecular_docking_and_v

Introduction<!>Ligand Selection and Preparation<!>Collection and Preparation of Proteins<!>Sequence alignment and Superimposition<!>Molecular Docking<!>Validation based on machine learning algorithms<!>Analysis of druglikeness and ADMET properties<!>Sequence alignment and superimposition<!>Molecular Docking Studies<!>Validation based on machine learning algorithms<!>ADMET study<!>Discussion<!>Conclusion

<p>Coronaviruses (CoV) are a group of positive-sense, single-stranded RNA (ssRNA) viruses belonging to the sub-family Coronavirinae, family Coronaviridae, and order Nidovirales. Viruses belonging to this group possess crown-like protein spikes on the outer surface of the virus-coat and hence named as 'coronavirus'. There are four sub-groups of CoV -Alpha-, Beta-, Gamma-, and Delta-coronaviruses, and the size ranges from 65 -125 nm diameters and the genome size 26 -32 kb ssRNA. Till today there are four known human coronaviruses (HCoV) -HCoV-229E, -OC43, -NL63, and -HKU1 that are known to cause mild infections such as common cold in immunocompetent individuals while two highly pathogenic CoVs -Severe Acute Respiratory Syndrome coronavirus (SARS-CoV) and Middle East Respiratory Syndrome coronavirus (MERS-CoV) induces severe respiratory complications in humans (Su et al., 2016;Cui et al., 2019). Based on sequence homology studies it is known that all the human CoVs have originated from bats while HKU1 and HCoV-OC43 are linked to rodents (Forni et al., 2017). SARS-CoV2 is a new coronavirus that emerged in late 2019 from Wuhan city of China in the form of a global pandemic disease called COVID-19 (Benvenuto et al., 2020). Like SARS and MERS, SARS-CoV2 is also a highly contagious and pathogenic coronavirus that causes severe respiratory problems and even death. Genomic studies have found that SARS-CoV2 is phylogenetically close to other SARS-like coronaviruses (Wu et al., 2020). Till today COVID-19 has invaded almost the entire world affecting more than 17.7 million of people and more than 700 thousands death worldwide (WHO). The attack and pathogenicity of SARS-CoV2 start at the lower respiratory system followed by invasion of pulmonary epithelial cells and hijacks the host cell machinery. The most common symptoms of COVID-19 include cough, fever, malaise, gastrointestinal symptoms, loss of smell, and sore throat (Gandhi et al., 2020).</p><p>At present, there is no clinically approved drug or vaccine against COVID-19. However, attempts have been made to reduce the virus load and improve the symptoms by using known antiviral drugs and related clinical practices (McKee et al., 2020). Researchers around the world are working diligently to develop effective drugs based on the pathogenicity and molecular details of the SARS CoV-2 and related coronaviruses. One of the major therapeutic drug targets of human CoVs is to inhibit the functioning of two protease enzymes -3 Chymotrypsin-like proteases (3CLpro), also known as main protease (Mpro), and Papain-like protease (PLpro) (Li et al., 2020;Zhou et al., 2020). Encoded by Open Reading Frame-1 (ORF-1) of viral genome 3CLpro and PLpro enzymes play a major role in the processing of polyproteins (pp1a and pp1ab) into active non-structural proteins (nsp). These nsps are known to be involved in the transcription and replication of viral genome (Hilgenfeld, 2014;Dai et al., 2020). 3CLpro and PLpro are also known to have crucial role in virus infection, assembly, and generation (Ma-Lauer et al., 2016;Zhou et al., 2019). There two protease enzymes, 3CLpro and PLpro, are therefore considered to be a promising candidate for therapeutic drug target (Wu et al., 2020). Recent studies have demonstrated that the two protease enzyme 3CLpro and PLpro of SARS-CoV share about 96% and 83% sequence identity with SARS-CoV2 at the protein level (Morse et al., 2020). Similarly, crystallographic and molecular modeling studies have also revealed that despite its differences of amino acid sequences in the 3CL and PL proteases of SARS-CoV and -CoV-2, there are striking similarities in the ligand-binding sites of the enzymes (Macchiagodena et al., 2020;Shin et al., 2020). In that perspective, candidate drugs that are used for SARS-CoV may also be tested for its effectiveness against novel coronavirus SARS-CoV2. The molecular docking is a common, widely used in-silico technique that assesses the binding affinity between ligand and protein and can be used to design novel inhibitors against disease-causing biological targets (Bhattacharyya et al., 2019). Using this method virtual screening of a large number of chemicals can be made to select the lead compounds of probable drug candidates (Huang and Zou, 2010;Swargiary et al., 2020). At this time of COVID-19 pandemic, it is of utmost importance to look for speedy alternatives so that the raising spread of SARS-CoV2 can be managed. In an attempt to combat COVID-19, the present study investigates the binding affinities of FDA recommended drugs on 3CL protease and PL proteases of SARS-CoV2 and compared the result with SARS-CoV. Furthermore, we also studied the in-silico ADMET properties of the selected drugs.</p><!><p>A total of 61 FDA recommended COVID-19 drugs were taken from DrugBank database (https://www.drugbank.ca/). The PDB files of the ligands were processed and finally converted into .pdbqt file using AutoDock tool (Trott and Olson, 2010).</p><!><p>Three-dimensional structures of 3CL proteases of SARS-CoV (PDB ID: 4TWY) and SARS-CoV2 (PDB ID: 7BUY) and PL Proteases of SARS-CoV (PDB ID: 3MJ5) and SARS-CoV2 (PDB ID: 6XAA) were downloaded from PDB database. The protein structures were cleaned by removing the water and other hetatms. Polar hydrogens and Kollman charges were added to the structure and finally converted into .pdbqt format for docking using AutoDock Tools.</p><!><p>The amino acid sequence of both the proteins for SARS-CoV and -CoV-2 was collected from the PDB database. Multiple sequence alignment was carried out in ClustalW (https://www.genome.jp/tools-bin/clustalw). Superimpositions of the cleaned protein structures were performed in the Chimera tool (Pettersen et al., 2004).</p><!><p>After the ligands and the target enzymes were prepared docking was carried out in AutoDock Vina (Trott and Olson, 2010). The active pocket amino acids selected for 3CLpro were including the catalytic dyad His-41, Cys-145. Similarly, for PLpro the amino acids were including the catalytic triad, Cys-112-His-273-Asp-287. The grid parameters were set as x, y, z size-coordinate and grid box 46,14.854,and 58,52,22.916, for SARS-CoV2 PLpro. Similarly, for SARS-CoV proteases, the grid parameters were set as x, y, z size-coordinate and grid box centrecoordinate: 66,46,52,and 5.092 for 3CLpro and 48,64,46, for PLpro, respectively. The docking algorithm was carried out by keeping the default exhaustiveness at 8. After docking, the pose scoring the lowest binding energy (binding affinity in kcal/mol) was selected and visualize in Discovery Studio.</p><!><p>Bayesian machine learning models from Assay Central platform (https:/assaycentral.github.io/ #) was used to determine the applicability score for potent compounds (based on docking score) that could function against SARS-CoV2. Assay Central is a method for the development of machine learning models (Bayesian, Random Forest, and Deep Neural Networks, etc.) that can be used ultimately to filter and score target-specific lead compounds prior to wet laboratory validation (Clark et al., 2015). It is a collection of predictive Bayesian and Random Forest (RF) models in a self-contained executable for non-experts to evaluate the likelihood of activity against target of interest. The output includes a prediction score (Pm) and applicability score (AS) (percentage of molecular fragments present in the model input). The Pm is the summation of fingerprint i and contributions, Ci. Contribution is based on the number of active compounds with the fingerprint, Ai, out of the total compounds with the fingerprint, Ti, and the total number of active compounds in the dataset, R (Clark et al., 2015;Krems, 2019).</p><!><p>Top ten docking score drugs and the druglikeness properties were studied using SwissADME (Daina et al., 2017) and PubChem database (https://pubchem.ncbi.nlm.nih.gov/). The absorption, distribution, metabolism, excretion, and toxicity (ADMET) properties of drugs were predicted using online database, ADMETlab (Dong et al., 2018).</p><!><p>Multiple sequence alignment and 3D structure superimposition between SARS-CoV and -CoV-2 protease enzymes, 3CLpro, and PLpro is shown in Fig. 1. The protease enzyme 3CLpro consists of 306 amino acid residues in both the viruses. Sequence comparison studies have shown that SARS-CoV and CoV-2 share 95.72% sequence identity in the main protease enzyme 3CLpro</p><p>(Fig 1a). The main protease showed differences in 13 amino acid residues between the two viruses out of which 6 are having similar properties. The superimposition of the 3D structure of the enzymes showed that the two structures have a root mean square deviation (RMSD) 0.641Å (Fig 1c). Most of the deviations were observed in the coil region of the protein structures. On the contrary, PLpro of two viruses has 82.85% sequence identity. A total of 53 amino acids were found to differ between the two structures out of which 15 residues are similar in nature (Fig 1b).</p><p>The RMSD between the two structures is found to be 1.008Å (Fig 1d).</p><!><p>Docking studies were carried out with a total of 61 FDA recommended COVID-19 drugs and are screened for binding affinity to the active sites of two key therapeutic drug target proteins, 3CLpro and PLpro. Among these drugs, 18 compounds are known protease inhibitors while the others include antimalarial, antiviral, antibacterial, antihelmintic, antitumor, etc (Table 1). The docking score of all the drugs against 3CLpro and PLpro of both CoVs are presented in Fig. 2. It is seen from the study that the drugs showed marked variations in the binding affinities. Most of the drugs showed a better binding affinity to SARS-CoV 3CLpro enzyme compared to SARS-CoV2. However, the range of binding energies is almost the same in both the CoVs. The binding energies of drugs ranged from -4.8 to -9.4 kcal/mol in 3CLpro of SARS-CoV2, while in the case of SARS-CoV, the binding energy ranged from -5.9 to -9.3 kcal/mol. Protease inhibitor simeprevir showed best binding affinity (-9.4 kcal/mol) with 3CLpro of SARS-CoV2 while, eltrombopag, a drug used for the treatment of thrombocytopenia is found to have best binding affinity (-9.3 kcal/mol) with 3CLpro of SARS-CoV. The binding affinity of simeprevir is followed by digoxin, vaniprevir, saquinavir, ivermectin, and finally fingolimod in 3CLpro of SARS-CoV2. Similarly, in case of SARS-CoV, eltrombopag is followed by digoxin, ciclesonide, saquinavir, abemaciclib, and finally favipiravir (Fig. 2a). On the contrary, Eltrombopag showed highest binding affinity with the PLpro of both SARS-CoV (-9.1 kcal/mol) and -CoV-2 (kcal/mol) (Fig 2b). Eltrombopag is followed by simeprevir, digoxin, losartan, and finally chloroquine in SARS-CoV2 PLpro while in the case of SARS-CoV, eltrombopag is followed by abemaciclib, penfluridol, bazedoxifene, and finally chloroquine. Fig. 3a showed the ligand-binding site sphere view simeprevir and 3CLpro of SARS-CoV2.</p><p>Binding studies showed that a total of 16 amino acids (Lys-5, ) of 3CLpro were found to have interaction with the simeprevir. Two amino acids, Asn238 and Asp389 made three hydrogen bonds with simeprevir having 2.332Å, 2.068Å, and 2.474Å bond lengths (Fig. 3b). Amino acid Lys-137 is found to have favorable affinity of making H-bond with simeprevir. The other interactions involved Van der Waals interactions, carbon-hydrogen (C-H) bond, and alkyl bonds. Fig 3d & 3e showed the H-bond donor (HBD) and acceptor (HBA) as well as hydrophobicity of ligand-surrounding amino acids. Among the 3H-bonds, simeprevir acts as 1HBD and 2HBA with 3CLpro. The hydrophobicity surface view revealed that out of 16 amino acid residues surrounding the ligand, 14 residues showed hydrophilic property while two amino acids, Val-171 and Leu-287 were hydrophobic in nature (Fig 3e). Fig. 3c & 3f showed the Ramachandran plot of complete 3CLpro structure and ligand surrounding amino acid residues, respectively. Three amino acids, Gly-2, -143, and -275 found in the turn and coil region of 3CLpro were found to lie in disallowed regions. Furthermore, all the 16 amino acids surrounding simeprevir were in the allowed regions that are involved in βsheet structure of the enzyme (Fig. 3f). In comparison to SARS-CoV2, SARS-CoV 3CLpro showed a slightly different ligand binding pattern. Fig 4(a-f) showed the nature of binding interactions between SARS-CoV 3CLpro and eltrombopag. Fig 4a showed a slightly different ligand binding site in 3CLpro of SARS-CoV. A total of 12 amino acid residues were found to make interactions between 3CLpro and eltrombopag (Fig. 4b). Three amino acid residues, Gln-110, Thr-111, and Thr-292 were found to make three conventional H-bondings with the ligand. Other non-covalent bondings such as Van der Waal's interactions (4 amino acids), pi-bonds, etc were also observed between protein-ligand complexes.</p><!><p>The Assay Central Bayesian model for SARS CoV 3CLpro enzyme has a 5-fold cross-validation receiver operating characteristic (ROC) of 0.92, precision 0.85, recall 0.98, specificity 0.78, F1score 0.91, Cohen's kappa (CK) 0.78, and Matthews correlation coefficient (MCC) 0.78 (Fig. 7a). For SARS-CoV 3CLpro, prediction score (Pm) and applicability score (AS) of potent drug eltrombopag was 1.10 and 0.76 respectively (Fig. 8b). In contrast, the SARS CoV PL protease enzyme has a 5-fold cross-validation receiver operating characteristic (ROC) of 0.71, precision 0.40, recall 0.63, specificity 0.75, F1-score 0.49, Cohen's kappa (CK) 0.31, and Matthews correlation coefficient (MCC) 0.33 (Fig. 7b). For SARS-CoV PL protease, prediction score (Pm) and applicability score (AS) of potent drug eltrombopag was 0.49 and 0.76 respectively (Fig. 8b). Furthermore, for coronavirus disease (COVID-19) associated SARS-CoV2 3C like protease enzyme has a 5-fold cross-validation receiver operating characteristic (ROC) of 0.94, precision 0.43, recall 0.93, specificity 0.82, F1 score 0.59, Cohen's kappa (CK) 0.50 and Matthews correlation coefficient (MCC): 0.56 (Fig. 8a). For SARS-CoV2 3CL protease, the prediction score (Pm) and applicability score (AS) of potent drug, simeprevir was 1.03 and 0.65 (Fig. 8b). Unfortunately, for SARS-CoV2 PL-protease; no data was available in the assay central database and hence, could not be validated by the same method.</p><!><p>The heat map of ADMET properties of all the top ten best-binding drugs with 3CLpro and PLpro is presented in Fig 9 . It is observed that all the drugs have high tendency of being absorbed by human intestine except few drugs such as lopinavir, anidulafungin which showed weaker absorption property by the human intestine. Best docking ligand, simeprevir and eltrombopag were predicted to have high absorption property. All the top protease-binding drugs were found to have high permeability property through blood brain barrier (BBB) except, ciclesonide and camostat. Cytochrome-P450 (CYP) enzyme super family is a group of enzymes that metabolize xenobiotic substances that come inside the body. Many drugs are known to increase or decrease the activity of various CYP-isozymes. CYP3A4 (EC 1.14. 13.97), one of the major enzyme of drug metabolism is found to be inhibited by almost all the drugs. The drugs are predicted to act as a strong substrate and thus bind at the active site of the enzyme and inhibit the enzyme function. Inhibition of CYP3A4 enzyme therefore cannot metabolize the drugs. Similarly, the half life and clearance rate of all the drugs are found to be low. The drugs, on the contrary showed potential candidate for hERG (human Ether-a-go-go-Related Gene) inhibitor thus may cause side effect such as electrical activity in heart. The drugs also showed hepatotoxicity and liver injury in host body.</p><!><p>COVID-19 and its causative organism SARS-CoV2 have caused havoc all across the globe and yet the world is far from approved therapeutic agent(s). Millions of lives have lost so far and the future is still cloudy, and therefore, the current situation demands speedy approval of drugs or vaccines to combat this dreadful virus. Laboratory-based in-vitro and in-vivo studies are tedious, costly, and most importantly, a time-consuming process. As an alternative, repurposing of existing drugs and computer-aided drug screening can accelerate the drug development process (Ashburn and Thor 2004). Several recent studies have investigated the existing antiviral and other miscellaneous drugs to see whether these drugs have any role against SARS-CoV2 (Colson et al., 2020;Kandeel and Al-Nazawi, 2020;Swargiary, 2020). Two important therapeutic drug targets of coronaviruses, 3-chymotrypsin-like protease, and papain-like protease have been investigated by many researchers to find out candidate drug(s) that inhibit the functioning of these proteases (Elmezayen et al., 2020;Wang et al., 2020). The study found that simeprevir, a known Hepatitis C virus NS3/4A protease inhibitor, binds to the active site of 3CLpro much stronger than other drugs. Similarly, Hosseini and Amanlou (2020) also found that simeprevir fit well into the active site of main protease enzyme. In vitro and biochemical characterization studies by Lo et al. (2020) revealed that simeprevir is a potential drug for treating COVID-19. In another study, Mahdian et al. (2020) showed the strong binding affinity of simeprevir with different proteins of coronaviruses suggesting potential drug candidates for candidates to treat COVID-19 infections. Furthermore, simeprevir is also found to bind strongly with papain-like protease. Eltrombopag, an FDA approved drug used to treat low blood platelet count showed strongest binding affinity to PLpro enzyme suggesting potential inhibitor of the enzyme. Similar to the present finding, Gul et al. (2020) also reported strong binding affinity of eltrombopag with 3CLpro and RdRp enzyme. Like many other studies, we also found strong binding affinity of ivermectin, an antihelmintic drug, with both the protease enzymes of SARS-CoV and -CoV-2. In a recent in-vitro study, Caly et al. (2020) reported ivermectin as a potent inhibitor of SARS-CoV2 replication. Patri and Fabbrocini (2020) also believe that use of ivermectin in combination with hydroxychloroquine can help in the treatment of COVID-19. In a recent in-silico study ivermectin has been shown as a promising RdRp inhibitor and possible therapeutic drug against COVID-19 (Swargiary, 2020). Our study, however, showed weaker binding interaction of hydroxychloroquine with 3CLpro and PLpro of both the coronaviruses. Remdesivir and hydroxychloroquine showed more or less similar binding affinities to protease enzymes. Preliminary studies have observed that the use of remdesivir shorten the time of recovery, and lowers the respiratory tract infection (Beigel et al., 2020). Assay Central is a stand-alone set of predictive Bayesian and Random Forest models for non-experts to determine the likelihood of action of bioactive compounds against target of interest. After developing the model, each molecule in the selected 'project' receives a relative score, applicability number (fraction of structural features shared with the training set) (Clark et al., 2015). In general, the prediction score ranges from 0-1; a higher score means the compound is more likely to be active at the modeled target. The current threshold for considering a compound as active is a score of 0.5 or higher. In addition, the applicability score refers to the percent of fragments from the predicted molecule that are present in the model. Lesser score suggests that the prediction is not as effective, because the molecules comprising the model do not contain a large portion of the predicted compound. The attribute of the ROC or the receiver operator can be interpreted as a percent probability of correctly predicting compound activity against target. Additional metrics include F1 Score, Cohen's Kappa, and Matthew's Correlation Constant (MCC) and these statistical parameters ranges from 0-1, and the closer to 1 the more predictive the model is considered to be good (Clark et al., 2015;Krems, 2019). Thus, based on the machine learning algorithm it can be suggested that aforesaid drug candidates such as simeprevir and eltrombopag possess strong binding affinity against SARS associated target proteins and hence showing high prediction and applicability scores.</p><p>In-silico druglikeness and ADMET properties are important parameters for the screening of the possible drug candidate. By using in-silico tools it is now possible to predict the druglikeness of a compound based on its structure, physical, and chemical properties (Dimasi et al., 2003;Swargiary et al., 2020). good drug candidate must have high absorption property by the intestinal cell, must distribute easily to the cells without being metabolized en route, must have a high elimination rate, and less toxic effect to the body (Dong et al., 2018;Guan et al., 2019). All the top protease binding ligands including simeprevir and eltrombopag showed high absorption property by the cells. Cytochrome-P450 (CYP) super family of enzymes is an enzyme complex involved in the metabolism of xenobiots inside the body (Nelson et al., 2004). Metabolism of oral drug by CYP-enzyme super family en route before reaching the target site decreases the effectiveness of any drug. Furthermore, some drugs inhibit those metabolizing enzyme and therefore are not metabolized en route reaching the target site showing high efficacy. In many cases, the drugs may become substrate to the CYP-enzyme complex and are used by the enzymes before reaching the target site (Lunch and Price, 2007). Simeprevir is found to possess inhibitory property against several CYP-super family enzymes which suggest the possibility of reaching to the target sites. Eltrombopag, on the other hand showed moderate to high inhibitory property of certain CYP-super family enzymes while in others it acts as a substrate. Similarly, all the top binding ligands of 3CLpro and PLpro showed low to high inhibitory property to certain CYP-super family enzymes suggesting its high possibility to reach the target site. The FDA approved drugs of present study also found to possess moderate to high toxicity effects in the host body. The present study thus suggests that the FDA approved drug simeprevir, along with eltrombopage, digoxin, and tipranavir may act as a strong inhibitor to 3CLpro and PLpro and therefore, may be a promising drug to combat SARS-CoV2 and COVID-19.</p><!><p>The present study highlights the binding affinities of 61 drugs against two key SARS-CoV2 proteases, 3 chymotrypsin-like and Papain-like proteases. The study revealed that out of 61 drugs simeprevir and eltrombopag have the strongest binding affinity to 3CLpro and PLpro active sites and therefore inhibit the function of both the proteases. All the known protease inhibitors including saquinavir, vaniprevir, lopinavir, danoprevir, and nelfinavir also showed high affinity for both the protease enzymes. Furthermore, other non-antiviral drugs such as digoxin, ivermectin, camostat, ebastine, penfluridole, and abemaciclib also showed strong binding affinity to both the enzymes. The Assay Central Bayesian models further corroborate the finding of molecular docking and ranked simeprevir and eltrombopag as possible lead molecules among others for future validation. The present study thus suggest that the along with known antiviral drugs researchers can also focus on unknown antiviral drugs which may prove effective in combating COVID-19.</p>

ChemRxiv

Ethyl diazoacetate synthesis in flow

Ethyl diazoacetate is a versatile compound in organic chemistry and frequently used on lab scale. Its highly explosive nature, however, severely limits its use in industrial processes. The in-line coupling of microreactor synthesis and separation technology enables the synthesis of this compound in an inherently safe manner, thereby making it available on demand in sufficient quantities. Ethyl diazoacetate was prepared in a biphasic mixture comprising an aqueous solution of glycine ethyl ester, sodium nitrite and dichloromethane. Optimization of the reaction was focused on decreasing the residence time with the smallest amount of sodium nitrite possible. With these boundary conditions, a production yield of 20 g EDA day −1 was achieved using a microreactor with an internal volume of 100 μL. Straightforward scale-up or scale-out of microreactor technology renders this method viable for industrial application.

ethyl_diazoacetate_synthesis_in_flow

Introduction<!>Flow synthesis<!>Univariate optimization<!>Multivariate optimization<!>FLLEX module<!>Conclusion<!>Experimental Physical and spectroscopic measurements<!>Chip dimensions<!>Univariate optimization<!>FLLEX experiment

<p>Diazo compounds are frequently used versatile building blocks in organic chemistry [1,2]. From this class of compounds diazomethane and ethyl diazoacetate (1, EDA) are arguably the synthetically most useful ones. Due to the potentially explosive nature of diazomethane and EDA [3][4][5], however, synthetic routes that involve large scale batchwise handling of such diazo compounds is generally avoided in industrial processes. With the advent of continuous processing over the past decade, new approaches have appeared to conceptually change the way chemical synthesis is performed. In particular continuous-flow microreactor technology offers multiple advantages over batch chemistry, including the inherently safe conducting of reactions due to the small reactor dimensions, efficient heat transport and excellent control over the reaction conditions [6][7][8]. While the synthesis of diazomethane has been extensively explored in batch [9] and in continuous-flow reactors [10,11], EDA is synthesized via different routes in batch [12,13], but relatively little is known about continuous-flow approaches [14]. Considering the importance of EDA in a wide variety of reactions e.g. cyclopropanation, X-H insertion, cycloaddition and ylide formation [13,15], and more recently, in the synthesis of valuable compound classes such as β-keto esters [16] and β-hydroxy-α- diazocarbonyl compounds [17], we aimed to develop an inherently safe continuous-flow EDA process using microreactor and separation technology.</p><p>Ethyl diazoacetate (1) can be synthesized in flow via different pathways. Bartrum et al. [18] published a flow synthesis of numerous diazo esters starting from the corresponding arylsulfonylhydrazones, where the diazo moiety was installed through elimination of the sulfone substituent. Additionally, Ley et al. [19] recently prepared a range of α-hydroxy acids in flow starting from the corresponding amino acids, involving diazotization of the amine to the diazonium salt in a biphasic system. Inspired by Ley's approach, which is significantly more atom efficient than the sulfonylhydrazone pathway, we chose to synthesize EDA (1) from glycine ethyl ester (2) using readily available sodium nitrite [20] (Scheme 1). Although the diazotization step itself resembles the first step of Ley's hydroxy acid synthesis, we specifically aimed to produce and isolate the diazo product, which from there can be used for subsequent reactions.</p><p>Scheme 1: Synthesis of ethyl diazoacetate (1).</p><p>We intended to optimize the process focusing on decreasing the residence time in order to reduce solvent use and gain in throughput. Reaction temperature was considered less of an issue since in an industrial setting energy can generally be efficiently regenerated. In-line phase separation was thought to greatly enhance the usefulness of the EDA flow synthesis. Therefore, the outlet of the microreactor was directly connected to membrane-based phase separator to obtain EDA in the organic phase, which in principle can then be immediately used for either batch [13,15] or continuous-flow [16,17] follow-up reactions. Straightforward scale-up or scale-out of microreactor technology renders this method viable for industrial application.</p><!><p>Ethyl diazoacetate (1) was synthesized from glycine ethyl ester (2) and sodium nitrite in a biphasic system of dichloromethane and an aqueous sodium acetate buffer. Dichloromethane was chosen as the organic phase to dissolve the water insoluble EDA, because of its low water uptake and low boiling point and its compatibility with potential follow-up reactions. In principle, however, any other organic solvent immiscible with water could be used. The pH of the buffer was set to 3.5 which had been identified by Clark et al. as the optimal pH for the reaction [12]. A schematic representation of the initial microreactor set-up is shown in Figure 1. The box with the dotted line indicates the single-glass microreactor containing two mixing units M of the folding flow type [21]. The reactor temperature was controlled by a Peltier element and sensed by a Pt1000 temperature sensor. At the outlet of the microreactor, a back-pressure regulator (BPR, 40 psi) was attached to guarantee a liquid phase even above boiling temperatures of the solvents. To ensure welldefined reaction times during optimization experiments, neat N,N-diisopropylethylamine (DIPEA) was added via syringe 4 to efficiently quench the reaction. The collected product (60 μmol) was analyzed by HPLC to establish the conversion of the reaction.</p><!><p>Determination of the optimal conditions for the reaction started off with investigating the important reaction parameters via a univariate optimization. Based on knowledge obtained from EDA synthesis in batch [12] and other flow reactions [22,23], residence time, temperature and NaNO 2 stoichiometry were chosen as relevant parameters. Temperature was expected to have a large influence on the rate of the reaction. Shortening the residence time to a minimum would minimize the risk of side reactions and reduce costs, and the reaction should be performed with the smallest amount of NaNO 2 possible. The results of the univariate optimization are shown in Figure 2. EDA synthesis was shown to be fast, since within 200 seconds complete conversion was obtained at 15 °C. Additionally, the temperature shows a steep increase between 0-30 °C, indicating a large influence of both parameters on the reaction rate. The amount of NaNO 2 shows only a rather small influence. Based on these univariate optimizations the experimental ranges of the three parameters were determined to investigate the interrelationships via a multivariate optimization.</p><!><p>An experimental design based on a D-optimal algorithm was created from the aforementioned three parameters within their respective ranges, namely 5-120 s, 0-60 °C and 0.7-1.5 equiv of NaNO 2 . Using MATLAB (MathWorks, R2007a), fifty data points were selected of which the corresponding experiments were performed in random order. The resulting HPLC yields were normalized and fitted to a third order polynomial model. In-house-developed FlowFit software [24] was used to calculate the best possible model fit. The results are visualized in 2D-contour plots (Figure 3). These plots show a rather broad optimum for the conversion of glycine ethyl ester (2) into EDA (1). The decrease in the upper left corner of the second contour plot can be explained by the high uncertainty of the model at the edge of the plots. As was expected, temperature has a large influence on the reaction rate. The conversion into EDA shows a steep increase with increasing temperature. High temperatures and increasing amounts of NaNO 2 decrease the residence time to a minimum of 20 seconds while still obtaining complete conversion. Not surprisingly, the minimal amount of NaNO 2 required is 1 equivalent. We aimed to reach complete conversion into EDA (1) maintaining a short residence time with a minimum amount of sodium nitrite, possibly using higher temperatures. Based on these boundary conditions, the optimal parameter settings were fixed at 20 seconds residence time, a temperature of 50 °C using 1.5 equivalents of NaNO 2 . A triple-experiment was performed to prove that this set of optimal parameters indeed provided complete conversion into EDA. The experiment was performed in alternation with two other sets of parameters to rule out potential memory effects. HPLC yields of 95, 96 and 95% for the triple-experiment demonstrate the high reproducibility of the system.</p><!><p>Having established a microreactor protocol for the continuousflow synthesis of EDA, the next issue was to separate the product from the biphasic system in which it was collected. In order to increase safety and decrease the hold-up of EDA, the phase separation ideally had to be performed in flow as well. Therefore, a Flow-Liquid-Liquid-Extraction module (FLLEX) [25] was connected to the system [26,27]. The module utilizes a hydrophobic Teflon membrane and two back-pressure regulators (BPRs) to create a pressure difference, which causes the organic layer, in this case dichloromethane, to pass through the membrane resulting in phase separation. A schematic representation of the whole setup is shown in Figure 4.</p><p>As the conversion into EDA was quantitative, quenching with DIPEA was no longer required. Between the microreactor and the FLLEX module some additional tubing was used to ensure complete partitioning of the compounds over the two phases. The back pressure of the FLLEX was set to 40 psi, similar to the BPR used previously, and a pressure difference of 0.14 bar. Direct full separation of phases resulted in a clean organic phase containing 409 mg EDA (11 wt % solution in CH 2 Cl 2 , after 30 min of collection) while all salts remain in the aqueous phase. This corresponds roughly to an EDA production of 20 g day −1 and a space time yield of 100 kg day −1 dm −3 as compared to a reported industrial scale batch process yielding EDA in 48 g day −1 dm −3 [12].</p><!><p>EDA can be safely synthesized utilizing microreactor and separation technology starting from cheap and readily available starting materials. Optimization of the reaction was aimed at reaching complete conversion into EDA within a minimized residence time using the smallest required amount of sodium nitrite, possibly applying higher temperatures. The optimal reaction conditions identified based on these criteria were a residence time of 20 seconds, a temperature of 50 °C and 1.5 equivalents of NaNO 2 . Repeating the EDA synthesis in flow employing the optimal reaction parameters showed complete conversion and high reproducibility of the results. Additionally, we successfully combined a plug-and-play microreactor setup with a commercially available membrane-based phase separation module to perform a direct in-line extraction of the product. Even in our small set-up (internal volume 100 μL), we were able to generate approximately 20 g of pure EDA per day (11 wt % solution in CH 2 Cl 2 ).</p><!><p>NMR spectra were acquired at ambient temperature with a Bruker DMX 300 MHz spectrometer. 1 H NMR spectra were referenced to TMS or to the residual solvent peak. HPLC analysis was performed using an Agilent 1120 Compact LC, C-18 column, 10% acetonitrile in MilliQ, 254 nm. Pyridine (internal standard) has a retention time of 1.75 min, EDA of 9.67 min.</p><!><p>Three different microchips were used during the experiments.</p><p>1. Single borosilicate glass quench microreactor with an internal volume of 92 μL, a channel width of 600 μm and a channel depth of 500 μm. 2. Single borosilicate glass microreactor with an internal volume of 100 μL, a channel width of 600 μm and a channel depth of 500 μm. 3. Single borosilicate glass quench microreactor with an internal volume of 1 μL, a channel width of 120 μm and a channel depth of 50 μm.</p><!><p>Solution A: Glycine ethyl ester hydrochloride (40 mmol, 5.6 g) dissolved in 20 mL buffer 1. Solution B: CH 2 Cl 2 . Solution C: NaNO 2 (60 mmol, 4.1 g) dissolved in 30 mL degassed MilliQ. The flow rates and temperatures were set based on predetermined conditions of residence times and temperatures (Table 2). Experiments with a residence time of 5 s were performed in a glass microreactor with an internal volume of 1 μL. For longer residence times, a microreactor with an internal volume of 92 μL was used. Solution Q was set at a flow rate 1/3 of the flow rate of solution A. Each experiment had a collection time equal to 30 μL of solution A. The product was collected in 1 mL of acetonitrile and analyzed by HPLC. Results are visualized in Figure 3 as 2D-contour plots.</p><!><p>Solution A: Glycine ethyl ester hydrochloride (10 mmol, 1.4 g) dissolved in 5 mL buffer 2. Solution B: CH 2 Cl 2 . Solution C: NaNO 2 (15 mmol, 1.0 g) dissolved in 5 mL degassed MilliQ. Buffer 2: Sodium acetate trihydrate (100 mmol 13.6 g) dissolved in 80 mL MilliQ. Concentrated hydrochloric acid (37%, 12 M) was added until a pH of 3.5 was reached (7 mL). Additional MilliQ was added to obtain a total volume of 100 mL of buffer.</p><p>Solution A (86.25 μL/min) was combined in a stainless steel T-splitter with solution B (172.5 μL/min). The biphasic mixture immediately entered the glass microreactor (internal volume: 100 µL) where it was mixed with solution C (86.25 μL/min).</p><p>The reaction was performed at 50 °C. After the reaction, the mixture was passed through 15 μL of FEP-tubing (ID = 254 μm) before entering the FLLEX module where phases were separated (40 psi, Δp = 0.14 bar). The set-up was stabilized for 2 min before collecting for 30 min. EDA was obtained as a solution in CH 2 Cl 2 (1.52 g). According to 1 H NMR analysis, clean EDA was obtained. Based on the residual solvent peak in the 1 H NMR spectrum it was calculated to be a 27 wt % solution of EDA in CH 2 Cl 2 meaning 409 mg of pure EDA.</p>

Beilstein

Small molecule analysis and imaging of fatty acids in the zebra finch song system using time-of-flight-secondary ion mass spectrometry

Fatty acids are central to brain metabolism and signaling, but their distributions within complex brain circuits have been difficult to study. Here we applied an emerging technique, time-of-flight secondary ion mass spectrometry (ToF-SIMS), to image specific fatty acids in a favorable model system for chemical analyses of brain circuits, the zebra finch (Taeniopygia guttata). The zebra finch, a songbird, produces complex learned vocalizations under the control of an interconnected set of discrete, dedicated brain nuclei \xe2\x80\x9csong nuclei\xe2\x80\x9d. Using ToF-SIMS, the major song nuclei were visualized by virtue of differences in their content of essential and non-essential fatty acids. Essential fatty acids (arachidonic acid and docosahexaenoic acid) showed distinctive distributions across the song nuclei, and the 18-carbon fatty acids stearate and oleate discriminated the different core and shell subregions of song nucleus LMAN. Principle component analysis of the spectral data set provided further evidence of chemical distinctions between the song nuclei. By analyzing song nucleus RA at three different ages during juvenile song learning, we obtain the first direct evidence of changes in lipid content that correlate with progression of song learning. The results demonstrate the value of ToF-SIMS to study lipids in a favorable model system for probing the function of lipids in brain organization, development and function.

small_molecule_analysis_and_imaging_of_fatty_acids_in_the_zebra_finch_song_system_using_time-of-flig

Introduction<!>Tissue preparation and staining<!>ToF-SIMS and image analysis<!>Principle component analysis (PCA)<!>Immunohistochemistry<!>Fatty acids distinguish the major song nuclei<!>Variable content of Essential Fatty Acids in the song control system<!>Subdivisions of LMAN revealed<!>Discrimination of song nuclei by PCA<!>Developmental changes in the RA<!>Discussion<!>EFAs and song nucleus RA<!>Fatty acid changes during the song learning period<!>Prospects

<p>Fatty acids comprise a diverse group of unbranched carboxylic acids, with variable numbers of carbon chains and degrees of saturation. They are central in cell metabolism and also form a primary structural component of all cell membranes (as phospholipids). They also have major roles in cell signaling – and nowhere more so than in the brain. Membrane phospholipids are precursors for the generation of numerous messenger compounds active in the brain including prostaglandins and platelet-activating factor (Bazan, 2005), endocannabinoids (Piomelli, 2005), and diacylglycerol (Kim et al., 2010). Local variations in specific lipid content contribute to the structure of synaptic junctions (Suzuki, 2002), and growing evidence suggests that specific synaptic lipids regulate vesicle exocytosis and membrane recycling (Davletov & Montecucco, 2010). Of particular interest are the essential fatty acids (EFAs) – polyunsaturated fatty acids that are derived from the diet and important in in cognitive function (Moriguchi et al. 2000), inflammatory and immunological responses (Simopoulos, 2002; Bazan, 2005), and brain development (Innis, 2008). Polyunsaturated fatty acids (PUFAs) are notably vulnerable to oxidative damage (and conversely, may also participate in protection against oxidative damage) and have been implicated as promoting or protective factors in a range of neurological diseases, including Alzheimer's (Hartmann et al. 2007), Parkinson's (Israeli & Sharon 2009) schizophrenia (Fenton et al. 2000) and epilepsy (Cole-Edwards & Bazan 2005).</p><p>A better understanding of the diversity of fatty acids and their specific local distributions and functions will therefore be critical for a broad range of issues in neurobiology. However, the tools for studying fatty acid distributions in brain have been limited compared to the tools available for analysis of proteins and nucleic acids. Studies have typically relied on dissection, homogenization, and detection – methods that do not provide information about cellular or subcellular distributions. Specific lipids can be visualized with fluorophore labels but this is limited to cases where labeled lipids can be applied exogenously, and these labels can adversely affect lipid-lipid and lipid-protein interactions (Maier et al. 2002).</p><p>Time-of-flight secondary ion mass spectrometry (ToF-SIMS) is an alternative approach that allows direct surface visualization of intrinsic, unlabeled small molecules (<1,000 m/z) in tissue sections at a lateral resolution of ~1 μm (Ewing 2006, Johansson 2006, Mas et al. 2008). The technology has been shown to be a robust tool for the detection and imaging of lipids in a diverse array of tissues such as muscle (Tahallah et al. 2008), nervous system (Pernber et al. 2007), and kidney (Nygren et al. 2004). Subcellular imaging with ToF-SIMS has also revealed significant localization of other lipophilic molecules such as vitamin E in isolated neurons from Aplysia californica (Monroe et al. 2005) and has been used in several studies on the mammalian spine (Monroe, et al. 2005, Monroe et al. 2008).</p><p>In this report, we describe the application of ToF-SIMS to develop a primary analysis of the distribution of fatty acids within the brain of both adult and developing zebra finches. The zebra finch (Taeniopygia guttata), a songbird, is an important model for neuronal circuitry development, neurogenesis, learning and memory, and its genome was recently sequenced (Warren et al. 2010). A distinctive feature of the songbird brain, which makes it especially conducive to focal molecular analysis, is the presence of several large anatomical nuclei in the forebrain that are interconnected in a circuit (Fig. 1) that controls the learning and production of birdsong (reviewed in (Brenowitz et al. 1997, Clayton et al. 2009, Zeigler & Marler 2004). These nuclei have been studied intensively for their physiological and developmental properties (Clayton 1997) and they form a valuable reference set for studies of molecular composition in the brain (Lovell et al. 2008). In the zebra finch, this "song control circuit" is completed during a critical period in juvenile development (between about 20 and 40 days post-hatch), and only in males (females do not produce learned vocalizations). It is during this critical period that a young male zebra finch learns to sing by copying the song of an adult tutor. With sexual maturity, the ability to further modify the song or learn a new one is essentially lost.</p><p>Several observations suggest important roles for lipids and fatty acids in song system development and function. One of the first genes shown to be developmentally regulated within the song control circuit encodes the ortholog of alpha-synuclein (originally referred to as synelfin, (George et al. 1995)), a lipid-binding protein subsequently implicated in Parkinson's and other neurodegenerative diseases (Clayton & George 1999). Lipophilic molecules, including retinoic acid (Jeong et al. 2005, Denisenko-Nehrbass et al. 2000), cannabinoids (Soderstrom et al., 2004) and sex steroids (London et al. 2006), are synthesized in or near several of the song control nuclei and are necessary for song learning and song circuit development. Yet the lipid and fatty acid composition of the zebra finch brain has remained largely unexplored. Indeed, information about brain lipids is sparse for all birds in comparison to mammals, and mostly limited to issues related to embryogenesis, metabolic rate and egg production (e.g., (Maldjian et al., 1996; Speake & Wood, 2005; Turner, 2005).</p><p>We previously used ToF-SIMS to image the distributions of several inorganic ions and lipids in the songbird brain (Amaya et al. 2007). Here we analyze the distribution of fatty acids and other lipids across the major forebrain components of the song system, both in adults and during the critical period for song learning and circuit development. In order to facilitate analysis of the complex datasets, we incorporate principle component analysis (PCA) into our measurements to determine regional and age-related differences in content. Our results reveal a complex distribution of fatty acids across different song nuclei, and set the foundation for future targeted studies to address the potential functional importance of these fatty acids in this important model of neural circuit development and function.</p><!><p>All animal procedures were conducted according to protocols approved by the University of Illinois Institutional Animal Care and Use Committee. Male zebra finches were sacrificed by rapid decapitation, the brain removed, and rapidly frozen on dry ice. Post-hatch day 20 birds were sexed postmortem by gonadal identification. Parasagittal sections, 16 μm in thickness, were collected beginning at 2.5 mm to the midline (~5.6 mm) and alternatively placed on either a glass slide for histological staining, or a glass cover slip for ToF-SIMS and immunohistochemical analyses. Sections were stored at –80 °C for no more than 3 wks. Histological staining by cresyl violet was carried out as previously described (Amaya et al. 2007). Sections were identified that contained the major components of the forebrain song system: nucleus RA, nucleus HVC (used as a proper name), lateral magnocellular nucleus of the nidopallium (LMAN), dorsolateral thalamic nucleus (DLM), AreaX, and the NCM/CMM complex comprising the AL (Fig. 1). The adjacent sections then used for ToF-SIMS or immunohistochemical analysis.</p><!><p>Sections from adult males (n=3) were removed from –80 °C storage and immediately vacuum dried (~15 min) at room temperature. Dried sections were placed in a Desk II TFC sputter coater (Denton Vacuum, Moorestown, NJ) equipped with a gold target and ~10 Å was deposited on sample surface to improve the ionization of surface molecules (Altelaar et al. 2006). ToF-SIMS analyses were conducted in the Materials Research Laboratory facilities at the University of Illinois at Urbana-Champaign (UIUC); specifically, the instrument used was a TRIFT ToF-SIMS mass spectrometer (Physical Electronics, Chanhassen, MN) equipped with a gold liquid metal ion source. The primary ion beam used Au1 running at 22 keV, 2 nAmp, and the primary ion dose was kept below the static limit of 1 × 1013 primary ions cm-2 and negative secondary ions collected. Mass spectrometric images were collected using WinCadence software (Physical Electronics, Chanhassen, MN) with its built-in mosaic function, except HVC which was collected as two 600 × 600 μm images. Images measuring 1 × 1 mm were composed of 16 steps, each interrogating 250 × 250 μm; the areas of 1.5 × 1.5 mm also were composed of 16 steps, now measuring 375 × 375 μm each. Images containing HVC and AL were stitched as previously described (Amaya et al. 2007). Individual images are composed of 256 × 256 pixels with each pixel representing a unique mass spectrum of the sample surface. All images were convolved once using the default settings in WinCadence and individually scaled to a relative intensity scale of 0 - 100%. Where multiple ToF-SIMS images are shown to compare distributions of different molecules within a single song nucleus, all images are from the same individual tissue section. Qualitative descriptions of histological patterns described in Results are based on observations made in all three adult males, though only a single representative section is shown in each figure. Where additional procedures were performed (i.e., cresyl violet, IHC) these were performed on sections serially adjacent to the section used for ToF-SIMS.</p><p>After determining the likely fatty acids based on mass matching, our putative identifications relied on comparing the spectra from the brain tissue with spectra from standards purchased either from Sigma-Aldrich or Cayman Chemical (> 95% purity). Standards were either adhered to or deposited on carbon tape and subjected to ToF-SIMS analysis and their mass to charge ratio recorded (Table 1). We confirmed our assignments of these fatty acids in the brain using an alternative approach; more specifically, the region of interest was removed, lipids extracted, and fatty acids analyzed by GC-MS (Supplemental S1 and S2). To determine if esterified lipids undergo chemical fragmentation, a standard of phosphatidylethanolamine (PE) purity of ~ 98% (Sigma-Aldrich) was adhered to carbon tape and subjected to ToF-SIMS analysis (Supplemental S3).</p><!><p>For multivariate statistical analysis, mass spectra from the song nuclei from three different males were collected by using the region of interest function (ROI) in WinCadence. The developmental comparison consisted of three different males, for each age group, with ToF-SIMS collected from five different 200 × 200 μm areas within RA. Spectra were box binned and normalized to carbon (m/z 12), a ubiquitous and biologically significant ion. Contaminating peaks and peaks <m/z 50 were excluded from data analysis. PCA was carried out using PLS_Tool box (Eigenvector Research, Inc. Wenatchee, WA), and data were preprocessed by autoscaling and transformed to a logarithmic scale (log10) to ensure high-intensity peaks were not overrepresented (Wagner et al. 2004).</p><!><p>Sections were removed from –80 °C storage and fixed with 4% paraformaldehyde (pH 7.3) for 5 min and washed three times with 0.025 M PBS (pH 7.0) for 5 min each time. Tissue was blocked for 1 h at room temperature in a humidity chamber with Image-iT FX signal enhancer (I36933, Invitrogen, Carlsbad, CA), followed by a TBS-TX (50 mM Tris HCl, pH 7.4, 150 mM NaCl, and 0.1% Triton-X-100) wash (5 min, × 3). A second blocking step with TBS-TX plus 10% normal donkey serum was done, followed by a TBS-TX wash step.</p><p>Sections were then incubated with primary antibodies against the neuronal marker, NeuN 1:500, (MAB377, Millipore, Billerica, MA) and myelin basic protein, MBP 1:75, (AB9348, Millipore) at room temperature in a humidity chamber. After 1 hr tissues were washed with TBS-TX and incubated 1 hr with the secondary antibodies Alexa 488 (A-21202, Invitrogen) and Alexa 546 (A11040, Invitrogen) diluted 1:500, followed by a TBS-TX wash. Cell nuclei were stained with DAPI (12 μg/mL) for 5 min, washed with TBS-TX, and cover slip mounted with ProLong Gold (P36934, Invitrogen). Slides were then analyzed under a Zeiss Axiovert 200M (Carl Zeiss Microimaging, Inc., Thornwood, NY) fluorescence microscope on site at the Institute for Genomic Biology core facilities, UIUC.</p><!><p>From ToF-SIMS data for each brain section examined (n=3 birds), images were constructed for 13 fatty acids and cholesterol. Using the ROI function, relative concentrations were assessed within the five major song nuclei and the two major subdivisions of the AL (Table 2). By ANOVA, there were significant differences among these brain regions for nine of these lipids, with the largest effect (nearly four-fold) observed for the long polyunsaturated EFA, arachidonic acid (ARA, 20:4 n-6).</p><p>Individual song nuclei were readily distinguishable from the surrounding brain tissue by their content of these lipids. Figure 2 presents an overview of the relative content of these lipids in nucleus RA compared to the surrounding arcopallium and nidopallium, in a single tissue section (images of additional song nuclei and at larger magnification are presented in Supplemental S4-S6). As reference and for comparison to conventional histological methods, adjacent sections were stained using cresyl violet and immunohistochemistry. Cresyl violet staining (Fig. 2B) shows this well-defined song nucleus sitting within the arcopallium and just below (ventral to) the nidopallium. Immunohistochemical labeling with the neuronal marker NeuN (Fig. 2C, green) shows robust labeling of the RA and significant labeling in the nidopallium, but little to no staining in the surrounding arcopallium. Myelin labeling (Fig. 2C, red) shows a clear demarcation and a more intense overall labeling of the RA compared to the arcopallium or nidopallium. Punctate areas of intense labeling in the anterior part of the arcopallium likely correspond to fiber tracks connecting RA to fasciculus prosencephali lateralis, and striations in the nidopallium are neuronal projections that connect the HVC and RA. The labeling of cell nuclei with DAPI (blue) shows no discernable difference in cell density across the image. ToF-SIMS analysis of palmitic acid (16:0) produces a striking image as this lipid is sharply enriched within the boundaries of nucleus RA, with intermediate labeling in the surrounding arcopallium and lower labeling in the overlying nidopallium (Fig. 2D). Fig. 2E summarizes the relative signal intensities for all 13 fatty acids and cholesterol in RA compared to surrounding tissue in this same brain section.</p><!><p>The omega-3 and omega-6 unsaturated EFAs are of particular interest given their importance in cell signaling, brain function, and therapeutics. A survey of the local distributions of these in and around the six song system regions is presented in Figure 3. (Selected images are also shown aligned with reference sections in Supplemental S4. A similar series is shown for non-essential fatty acids in Supplemental S6). Eicosapentaenoic acid (EPA, 20:5 n-3) and arachidonic acid (ARA, 20:4 n-6) are notably reduced in the RA, especially in comparison to the nidopallium (Fig. 3C, D). In contrast, docosahexaenoic acid (DHA, 22:6 n-3) shows increased relative levels in the RA compared to surrounding tissue, as do the EFAs linoleic acid (18:2 n-6) and 11,14-eicosadienoic acid (20:2 n-6). The EFAs 8,11,14, eicosatrienoic acid (20:3 n-6) and adrenic acid (22:4 n-6) are not included in the figure as they show no difference in spatial distributions within these nuclei compared to the surrounding tissue, even though they do show differences in levels within different song nuclei (Table 2). Nucleus LMAN, which projects to RA, and its thalamic afferent (DLM) are generally similar to RA for all these EFAs. HVC also projects to RA, but HVC generally shows less differential content of the EFAs compared to surrounding tissue, with little or no enrichment for linoleic acid nor a reduction in ARA. Area X shows little differential content except for a reduction in ARA. The AL shows evidence of internal subdivisions according to the different distributions of EPA, DHA and ARA.</p><!><p>The song control nucleus LMAN is essential for song learning, and has been described as having a central core of magnocellular neurons surrounded by a shell of parvicellular neurons, with discrete functions proposed for these two subregions (Johnson et al. 1995, Bottjer & Altenau 2010). We noted variation in the anatomical pattern of several lipids across this nucleus and analyzed this in further detail by line scan. ToF-SIMS imaging of the LMAN indicates reciprocal internal distributions of the monounsaturated 18-carbon fatty acid, oleic acid, and the fully saturated form, stearic acid (Fig. 4A, B). Linescan analysis, moving across the images from the upper left to the lower right, quantifies these distributions at each pixel (Fig. 6C). Oleic acid (red line) shows a significant increase in relative signal intensity in the area corresponding to the core of the nucleus. Stearic acid (blue line) shows two spots of maximum signal intensity extending over an area of ~150 μm (Fig. 6C). Both areas of increased stearic acid signal are immediately adjacent to the increased oleic acid signal. Comparing the two images it appears the increased stearic acid signal encircles the increased oleic acid signal (Fig. 6A, B). This is consistent with the prior descriptions of core/shell regions (Johnson et al. 1995, Bottjer & Altenau 2010). Direct overlap of IHC and ToF-SIMS images from adjacent sections (not shown) indicates the lipids are concentrated mostly outside of neuronal nuclei but we cannot discriminate whether they are present in glia or neuropil.</p><!><p>The observations above reveal that known chemicals individually show clear differences in distribution across the major song nuclei. We used PCA to analyze at once the entire set of SIMS spectra from each of the three birds, considering all five song nuclei plus NCM and CMM in the AL (Fig. 5). Principle component 1 (PC1, x-axis) captures 55% of the variation in the data. Much of the variation is between the chemical profiles of the NCM/CMM and RA, along with one spectrum from the LMAN (Fig. 5A). Analysis of the corresponding loadings plot (Fig. 5B) reveals that the NCM/CMM region has the greatest association (increased relative amount) with m/z 79 (phosphate), m/z 122 (ethanolamine), and three unknown peaks (m/z 81, 153, 124). The same loadings plot shows that the RA and one LMAN spectrum have a relative increase in peaks corresponding to four fatty acids: m/z 255 (palmitic acid), m/z 227 (myristic acid), m/z 309 (11-eicosenoic acid), and m/z 327 (DHA) as compared to the NCM/CMM. Spectra from the other song nuclei show little variation in relation to the NCM/CMM and RA based on their position near the x-axis. Principle component 2 (PC2, y-axis) accounts for 16.44% of the variation and captures major differences between striatal nucleus AreaX and the thalamic song nucleus DLM, along with some spectra from the RA and LMAN (Fig. 5A). Much of the difference is captured by an unknown peak, m/z 124, showing relative abundance in AreaX. Other peaks highly associated with AreaX include m/z 565, m/z 163, m/z 153, m/z 63, m/z 79 (phosphate) and m/z 122 (representing phosphoethanolamine) (Fig. 5B). Along PC2 of the corresponding loadings plot, peaks corresponding to m/z 327 (DHA), m/z 309 (11-eicosenoic acid), m/z 283 (stearic acid), and m/z 281 (oleic acid) show higher relative levels in the DLM and some spectra from the RA and LMAN as compared to AreaX (Fig. 5). Spectra from the NCM, CMM, HVC, and some RA tissue show little variation compared to the DLM and AreaX based on their proximity to the y-axis (Fig. 5A).</p><!><p>We examined male birds at three different ages, corresponding to just before onset of the critical period for song learning and circuit maturation (post-hatch day 20, p20), the peak of the critical period (p40) and adulthood (>p90). PCA comparing RAs from three male birds at each age shows significant separation between adult and p20 RAs (Fig. 6A); in contrast comparing SIMS data collected from 5 unique areas within each RA shows less variation. Most of the age-related separation is along PC2, accounting for 21.26% of the total variation in the spectral data, while PC1 accounts for 66.25% of the variation and appears to capture inherent biological variation (Fig. 6A). Analysis of the corresponding loadings plot for PC2 reveals that m/z 79 (phosphate), m/z 124 (unknown), and m/z 122 (phosphoethanolamine) are the major peaks associated (increased relative abundance) with RAs from p20 birds compared to adults (Fig. 6B). The loadings plot also shows that RAs from adult birds show greatest association (relative abundance) with high molecular peaks m/z >800, along with some fatty acids including: DHA (m/z 327), 11-eicosenoic acid (m/z 309), and stearic acid (m/z 283), in addition to a peak previously identified as vitamin E (m/z 429) (Fig. 6B).</p><!><p>ToF-SIMS is a sensitive surface analysis method relying on the bombardment of a primary ion to analyze the top few layers of a sample surface (Sodhi 2004). In the past, extensive chemical fragmentation has limited its application for imaging larger molecules from biological tissues; however this limitation has been ameliorated by using cluster ion sources (Brunelle et al. 2005) and higher molecular weight primary ions (Fletcher et al. 2006). The results presented here confirm the use of ToF-SIMS as a valuable tool used to uncover chemical heterogeneity and the local distributions of fatty acids within the complex anatomy of brain tissue sections. It may also complement conventional histological and immunocytochemical staining procedures for analysis of brain anatomy, as we have shown here using the zebra finch song control system, finding new neuroanatomical features not evident with these other techniques.</p><p>A key feature of ToF-SIMS imaging is both a strength and a weakness: it can image the distribution of any peak detected by mass spectroscopy, without requiring any pre or post hoc knowledge of the chemical responsible for the peak. Assignment of chemical identities to specific masses is challenging as there is no comprehensive molecular ion database for ToF-SIMS; moreover, ionization patterns potentially vary within different subregions of the same tissue section based on variation in the local chemical makeup (Ostrowski et al. 2005). Topographical features can also affect sputtering of secondary ions (McDonnell et al. 2005); however, our use of cryosectioning minimizes these problems. Although our mass assignments must be qualified as putative, many of them were supported by direct ToF-SIMS analysis of known standards (Table 1) and by GC-MS analysis (Supplemental S1 and S2).</p><p>We believe the majority of the fatty acids detected in this study are probably derived from esterified lipids that fragmented upon impact of the primary ion. In support of this, we observed that a pure standard of phosphatidylethanolamine (PE) gave rise primarily to free fatty acids upon ToF-SIMS analysis (Supplemental S3). We also considered the possibility that the signals for the unsaturated EFAs ARA and EPA could have been confounded due to fragmentation. However, we believe these signals represent distinct chemical species present in the tissue since both species were detected in forebrain samples by gas chromatography mass spectrometry (Supplemental S2), and standards subjected to ToF-SIMS exhibited minimal and non-identical patterns of fragmentation (data not shown).</p><p>We constructed images showing the simultaneous distributions within individual sections of zebra finch brain for each of 13 well-defined lipids and cholesterol. The anatomical patterns of the various fatty acids generally respected boundaries within the tissue evident by cresyl violet staining or specific immunohistochemistry, though in some cases additional details were evident that were not apparent by these other methods (e.g., Fig. 2D). For example, we noted complementary patterns of several fatty acids associated with song nucleus LMAN, consistent with the presence of distinct "core" and "shell" subregions previously defined by their distinct neuroanatomical projections (Johnson et al. 1995) and different effects of lesion (Bottjer & Altenau 2010). In our data, both of these subregions show enrichment for the saturated 18-carbon fatty acid stearic acid, whereas the shell only is markedly enriched in the monounsaturated form, oleic acid.</p><p>In general our results show that each song nucleus can be distinguished by its relative complement of fatty acids, with respect both to the surrounding tissue (Figs. 2 and 3) and to the other song nuclei (Table 2), an observation that parallels findings using microarrays and in situ hybridization to measure specific mRNA content (Lovell et al. 2008). We replicated our observations in 3 different adult zebra finch brains, and observed a statistically significant main effect of brain region on signal for 8 of the 14 compounds we explicitly quantified (Table 2). We also used PCA on the entire spectral data set, finding that each song nucleus is discriminated fairly effectively in the first two principal components (Fig. 5). PCA is an unsupervised statistical approach used to reduce large complex chemical data sets into a few manageable variables. It can efficiently extract information representing chemical differences and similarities between samples (Lindon et al. 2007) and has been used in SIMS previously (Sjovall et al. 2004, Wu et al. 2007, Kulp et al. 2006). The clustering of our data according to song nucleus is especially interesting given that our zebra finches are not inbred and still retain robust genetic diversity (Forstmeier et al., 2007). PC1 captures 55% of the spectral variation and seems to segregate the brain regions according sensory versus motor function in the song system. The NCM/CMM is a sensory region that shows significant chemical differences when compared with the RA, which is the motor output nucleus of the telencephalic control system. In contrast, the DLM, HVC, LMAN, and AreaX regions are more clearly involved in both sensory and motor function, and cluster between the RA and NCM/CMM areas. The loadings plot shows a decreased association of the RA, compared to the NCM/CMM, with phosphate and ethanolamine, which may suggest a relative increase in sphingomyelin and glycosphingolipids in the RA (Isaac et al. 2006, Sastry 1985). Along PC2, the major effect is the segregation of the striatal nucleus Area X from the other song regions. Interestingly, AreaX also has a rather different profile of gene expression in the combined microarray studies of the SoNG Initiative (Replogle et al. 2008) (and unpublished).</p><!><p>EFAs are of particular interest given their known roles in cellular signaling and plasticity, and the fact that they are not synthesized locally but must be selectively accumulated from circulating stores derived from dietary sources. The EFAs showed especially marked distribution patterns in song nucleus RA. DHA is notably increased in the RA compared to surrounding brain regions. It is believed that the highly unsaturated nature of DHA provides significant membrane fluidity, required for proper brain function (Bourre et al. 1992) and neuronal development (Calderon & Kim 2004), and a protective role against neurodegenerative disease has been suggested (Oster & Pillot 2010). In other systems, DHA has been shown to be enriched in metabolically active brain areas such as the mammalian basal ganglia and gray matter (Diau et al. 2005), in brain regions with increased synaptic vesicles (McGee et al. 1994), and in nerve growth cones (Martin & Bazan 1992). RA itself has a high level of metabolic activity (Adret & Margoliash 2002) with dense synaptic input from two sources (HVC and LMAN) and a mixture of glutamatergic projection neurons and GABAergic interneurons (Pinaud & Mello 2007).</p><p>In contrast to DHA, two other EFAs implicated in brain function, EPA and ARA, are decreased in RA. EPA is a precursor to DHA, suggesting the possibility that RA neurons favor synthesis of DHA at the expense of EPA. ARA is a polyunsaturated fatty acid with multiple roles in cell signaling and synaptic potentiation (Williams et al. 1989), but it also exhibits pro-inflammatory effects by acting as a precursor in eicosanoid synthesis (Simopoulos 2002) and may mediate processes leading to neurodegenerative disease (Sanchez-Mejia & Mucke 2010). Intriguingly, the alpha-synuclein protein (SNCA) is also notably decreased in the RA of the adult zebra finch (George et al. 1995, Jin & Clayton 1997), and it interacts closely with ARA (Darios et al. 2010) and has reciprocal effects on the metabolism of ARA versus DHA (Golovko et al. 2007, Golovko et al. 2009). These observations suggest a concerted regulation within RA of particular lipid pathways that mediate cellular integrity and synaptic plasticity.</p><p>In the remaining song regions we also measured localized decreases in ARA, with two exceptions, Area X and the AL (NCM/CMM). AreaX is located in the striatum proper of the basal ganglia and contains both striatal and pallidal neurons, which are composed of mostly small γ-aminobutyric acid (GABA)ergic spiny neurons with fewer large GABAergic neurons (Farries & Perkel 2002, Reiner et al. 2004, Grisham & Arnold 1994, Carrillo & Doupe 2004). AL has a complex internal organization and is the central auditory area engaged in song perception; it shows high levels of both ARA and EPA, especially in the caudal part of the NCM subregion (Fig. 3, Supplemental S5). The AL also shows unique distributions for palmitic acid, stearic acid, and oleic acid (Supplemental S5). Molecular studies of the AL have shown retinoic acid receptors with a similar expression pattern as the distribution of ARA, EPA, and stearic acid (Jeong et al. 2005). Retinoic acid production in the song control system is necessary for normal song maturation (Denisenko-Nehrbass et al. 2000), and in mice it has a role in neuronal plasticity and learning and memory (Krezel et al. 1998, Cocco et al. 2002). The differential distribution of fatty acids may reflect distinct cellular populations in the auditory caudomedial telencephalon, with distinct molecular, anatomical, and physiological properties.</p><!><p>A male zebra finch learns to sing during a discrete "critical period" in late juvenile life. Importantly, the entire song learning process occurs after development is otherwise largely complete and the bird has fledged from the nest. A young male first begins to produce immature vocalizations at about P30, the age at which axons abruptly project from nucleus HVC and form synaptic connections onto the major output nucleus of the telencephalic control system (Fig. 1), nucleus RA (Konishi & Akutagawa 1985, Holloway & Clayton 2001). The bird's vocal performance gradually becomes more refined and stereotyped over the next 4-8 weeks, reaching its mature adult form by ~P90, after which no further change in the song will occur for the rest of the bird's life.</p><p>We applied ToF-SIMS to ask whether there were significant differences in lipid composition in nucleus RA before (P20), during (P40), and after (adult) the developmental song learning period. Unlike direct analysis of anatomical distributions observed within individual brain sections, comparisons of sections taken from different animals must confront the potential for artifacts arising from subtle variations in sample preparation. To minimize this risk we used a minimal approach to sample preparation, with the sections simply frozen and then vacuum dried prior to sputter coating, and then employed PCA as a way to objectively include all the spectral data in this analysis (Fig. 6).</p><p>Just as our PCA analysis showed a major effect of brain region (Fig. 5), here PCA revealed a major effect of developmental age on the spectral patterns obtained from nucleus RA, especially along the axis of the second principal component (Fig. 6). PC2 separates p20 and adult RAs, with RAs from p40 birds exhibiting an apparent chemical transition from p20 to adults. The P40 samples in particular show variation along PC2, intermediate between the two extremes of the P20 and adult samples. This variation is suggestive of (and may be related to) the variable progress that different individuals are making in their song development at this age (Tchernichovski et al., 2001). The corresponding loadings plot shows a relative increase in high molecular weight ions (m/z ~800–860) in the adult RA compared to the p20 RAs. Although these have not been positively identified, they may represent esterified lipids such as triglycerides and glycosphingolipids (Pernber et al. 2007, Touboul et al. 2005). In addition to higher levels of many fatty acids in adult RAs (i.e., DHA and 11-eicosenic acid), a peak previously identified as vitamin E (m/z 429) (Monroe et al. 2005) also shows relative higher levels in RAs from adult birds; in this case, however, we do not observe several characteristic SIMS fragmentation ions as previously observed. Increased vitamin E levels have been shown during development of mammalian brains (Zhang et al. 1996). Myelination in the RA has previously been identified as taking place around p50 (Clayton 1997). Therefore, we expect to see higher relative levels of phosphate-containing lipids in p20 RAs compared to adults. Other ions (m/z <150) show significant association with p20 RAs and likely represent fragmentation ions of larger molecules, making their identification difficult.</p><!><p>The subcellular imaging of fatty acids and other small molecules by ToF-SIMS in brain sections opens significant avenues in the study of neurochemistry and disease, especially when applied in an animal model like the zebra finch. Future studies to focus on the identification of unknown peaks in the ToF-SIMS spectra will significantly aid our understanding of small molecule heterogeneities across the zebra finch song system and their prospective biological importance in brain development and learning and memory.</p>

PubMed Author Manuscript

Cell-Selective Metabolic Labeling of Proteins

Metabolic labeling of proteins with the methionine (1) surrogate azidonorleucine (2) can be targeted exclusively to specified cells through expression of a mutant methionyl-tRNA synthetase (MetRS). In complex cellular mixtures, proteins made in cells that express the mutant synthetase can be tagged with affinity reagents (for detection or enrichment) or fluorescent dyes (for imaging). Proteins made in cells that do not express the mutant synthetase are neither labeled nor detected.

cell-selective_metabolic_labeling_of_proteins

<p>Time-dependent changes in cellular proteomes can be monitored via a variety of powerful electrophoretic and spectroscopic methods. Traditionally, radiolabeled amino acids have been used to label proteins synthesized during an amino acid 'pulse'; labeled proteins can be distinguished from preexisting (unlabeled) proteins through electrophoretic separation followed by radiographic detection1. More recently, mass spectrometry has enabled the use of stable isotopes in amino acid pulse labeling2.</p><p>In 2006, Dieterich and co-workers introduced the BONCAT (bio-orthogonal non-canonical amino acid tagging) strategy for selective enrichment and identification of newly synthesized proteins in cells3,4. The BONCAT approach reduces sample complexity and permits direct analysis of the primary protein synthesis response to stimuli. Bio-orthogonal functional groups5 are introduced into proteins by pulse-labeling with reactive, non-canonical amino acids. Labeled proteins are selectively modified with affinity tags for enrichment6,7; removal of unlabeled proteins simplifies subsequent analysis and identification by mass spectrometry.</p><p>All of these methods suffer from limitations when experiments are performed in systems that contain multiple cell types. Because incorporation of amino acids is non-specific with respect to cell identity, proteins from all cell types are labeled. In studies of interactions between different cell types in a single organism, the origin of the identified proteins can be difficult to ascertain because the cells share a common genome. When interactions between cells of different genomes are studied, detection of low abundance proteins can be problematic. In infection studies, for example, the protein content of the larger host cells can overwhelm that of the pathogen8 and limit detection and identification of the proteins of primary interest. In complex bacterial communities where hundreds of organisms can occupy a common biological niche9, probing the proteome of a single species in its natural context is an even greater challenge.</p><p>To address these difficulties, we describe here a versatile method for cell-selective protein labeling in mixed cellular environments. To achieve selective labeling, we employ non-canonical amino acids that are excluded by the endogenous protein synthesis machinery (Fig. 1a). These amino acids face discrimination by the quality control mechanisms found at the level of aminoacyl-tRNA synthetases10; they are not charged to tRNA and are not used in protein synthesis. By screening libraries of methionyl-tRNA synthetase (MetRS) mutants from Escherichia coli11,12 we have identified a mutant synthetase (NLL-MetRS) (Supplementary Fig. 1 online) that efficiently appends azidonorleucine (2, Fig. 1b) to cognate tRNA. Cells bearing the mutant MetRS are able to utilize 2 as a surrogate for methionine (1) in protein synthesis. Wild-type cells are inert to 2; proteins made in these cells utilize only methionine and are not labeled (Fig. 1a). In co-culture, protein labeling is restricted to mutant cells.</p><p>To validate this approach, we first confirmed that incorporation of 2 into newly synthesized proteins is dependent on expression of the mutant synthetase. An E. coli strain (DH10B/pJTN1) constitutively expressing a plasmid-borne copy of NLL-MetRS was pulse-labeled with 2 and compared to a control strain (DH10B/pQE-80L) that did not express the enzyme. Separate cultures of the two strains were grown in minimal medium containing the twenty canonical amino acids. When the cell density reached OD600 = 0.5, cells were pulse-labeled with 1 mM 2 for 10 m. Control cells were pulsed in the same fashion with 1, or incubated with the protein synthesis inhibitor chloramphenicol prior to labeling with 2. Cell lysates of each culture were probed for incorporation of 2 via Cu (I) catalyzed ligation13,14 to biotin-FLAG-alkyne (4) followed by Western blotting with protein detection by anti-FLAG antibody. The results of these experiments indicated that only proteins synthesized in cells constitutively expressing the NLL-MetRS (DH10B/pJTN1) were labeled with 2 and susceptible to ligation to 4 (Supplementary Fig. 2 online). A second control strain, in which the wild-type synthetase was over-expressed, was also inert to labeling (Supplementary Fig. 1 online).</p><p>The behavior observed in separate cultures was maintained when cells were incubated in co-culture to simulate a complex, mixed cellular environment. Two different heterologous proteins were employed as markers for the cells of origin to distinguish between NLL-expressing and wild-type E. coli. An E. coli strain (DH10B/pJTN2) expressing the NLL-MetRS was programmed to express green fluorescent protein (GFP) upon induction with isopropyl β-D-1-thiogalactopyranoside (IPTG). The control strain (DH10B/pJTN3) carried an IPTG-inducible gene for the marker protein DHFR. Both marker proteins were His-tagged to enable Ni-affinity purification and detection with Penta-His antibody. Individual cultures of these bacterial strains were grown to OD600 = 1.0 and a third culture was created by mixing cells in a volumetric ratio of 1:2 (DH10B/pJTN2:DH10B/pJTN3). To initiate labeling, 1 mM 2 was added to each of the three cultures and expression of marker proteins was induced with 1 mM IPTG for 3 h. Cell lysates from all three samples were subjected to Cu (I) catalyzed azide-alkyne ligation to 4, and marker proteins were isolated by Ni-affinity purification. Western analysis (Fig. 2a) of isolated proteins with Penta-His antibody revealed the expected expression patterns from the three cultures. In striking contrast, analysis of blots with streptavidin-HRP (for detection of conjugation to 4) revealed exclusive modification of GFP, the marker protein synthesized in cells expressing NLL-MetRS. DHFR isolated from control cells exhibited no such modification. N-terminal protein sequencing indicated 10-20% replacement of 1 by 2 in the GFP marker protein.</p><p>To demonstrate further the utility of this approach, we used fluorescence microscopy to distinguish proteins made in cells expressing the NLL-MetRS from those made in cells that do not express the mutant synthetase. The control strain (DH10B/pJTN4) expressed an IPTG-inducible GFP while the strain (DH10B/pJTN5) constitutively expressing the NLL-MRS carried an IPTG-inducible DHFR. Labeling of individual and mixed cultures were performed as described earlier. After pulse-labeling with 2 and induction of protein synthesis, cells were collected by centrifugation and washed prior to Cu (I) catalyzed labeling of cells with dimethylaminocoumarin-alkyne (5). After washing with PBS to remove excess dye, cells were imaged by fluorescence microscopy. The results (Fig. 2b) were consistent with those of the Western analysis; the coumarin fluorescence was confined to cells that express NLL-MetRS. Control cells expressing GFP were inert with respect to labeling, as indicated by the absence of coumarin emission from these samples (Supplementary Fig. 3 online).</p><p>Cell-selective protein labeling can also be accomplished in systems containing mixtures of bacterial and mammalian cells. Murine alveolar macrophages were infected with E. coli cells that constitutively express the NLL-MetRS (DH10B/pJTN1) or with control bacterial cells that express a GFP marker protein (DH10B/pJTN4). Prior to infection, 2 mM 2 was added to the macrophage medium; to initiate infection, bacteria were added to the culture medium and co-incubated for 35 m at 37 °C. Cells were fixed, permeabilized, and subjected to Cu (I) catalyzed conjugation to TAMRA-alkyne (6, Invitrogen). Bacteria were both bound and internalized by macrophages as confirmed by confocal microscopy and three-dimensional analysis (Supplementary Movie 1 online). Macrophage-associated bacterial cells that express the NLL-MetRS exhibited strong fluorescence emission from 6 (Fig. 3a). The control bacterial strain was bound and internalized by macrophages (as seen by detection of GFP, Supplementary Fig. 4 online) but exhibited no conjugation to 6 above background (Fig. 3a). To confirm protein synthesis by macrophages during the infection period, cells were treated with azidohomoalanine (3) in medium lacking 1 (Fig. 3b). Both 2 and 3 are susceptible to ligation to alkyne-functionalized probes; however, in contrast to 2, 3 is activated by wild-type MetRS and does not discriminate between cell types4. As shown in Fig 3, both bacterial cells and macrophages were labeled with 3, whereas labeling with 2 was observed only in bacterial cells that express the NLL-MetRS.</p><p>Newly synthesized bacterial proteins can be enriched from such cultures by affinity chromatography. Using an E. coli strain that expresses the NLL-MetRS constitutively and GFP under induction with IPTG, we infected macrophages in medium containing 2 mM 2. Immediately upon infection, IPTG was added to initiate bacterial synthesis of GFP. After 35 m at 37 °C, the total cell mixture was collected by centrifugation and lysed. Proteins were subjected to Cu (I) catalyzed azide-alkyne ligation with alkyne-functionalized biotin. Biotinylated proteins were selectively enriched by collection on neutravidin-agarose beads. After five washes, proteins were eluted from the resin with 2 mM free biotin and 2% SDS. To examine the extent of enrichment, the lysate, resin flow-through, washes and eluent were analyzed by immunoblot (Fig. 3c). The mammalian protein β-actin was detected with anti-β-actin and served as a representative macrophage protein. The bacterial marker GFP was detected with anti-Penta-His antibody. No actin was detected in the eluent, indicating at least 50-fold depletion of the mammalian marker. In contrast, comparison of the GFP band intensities in the eluent and lysate confirmed good recovery of the affinity-tagged bacterial protein.</p><p>The results described here illustrate the use of mutant aminoacyl-tRNA synthetases to enable cell-specific protein labeling with non-canonical amino acids. In mixed cellular systems, newly synthesized proteins in selected cells can be labeled with affinity reagents or fluorescent dyes for enrichment, identification, and visualization. This approach will enable unambiguous determination of the cellular origins of proteins made in complex multicellular systems and will provide new insight into intercellular communication. We are expanding the studies described here by engineering new amino acid/synthetase pairs and by using the azidonorleucine/NLL-MetRS pair to examine a variety of intercellular interactions.</p>

PubMed Author Manuscript

High-content and high-throughput identification of macrophage polarization phenotypes

Macrophages are plastic cells of the innate immune system that perform a wide range of immune-and homeostasis-related functions. Due to their plasticity, macrophages can polarize into a spectrum of activated phenotypes. Rapid identification of macrophage polarization states provides valuable information for drug discovery, toxicological screening, and immunotherapy evaluation. The complexity associated with macrophage activation limits the ability of current biomarker-based methods to rapidly identify unique activation states. In this study, we demonstrate the ability of a 2-element sensor array that provides an information-rich 5-channel output to successfully determine macrophage polarization phenotypes in a matter of minutes. The simple and robust sensor generates a high dimensional data array which enables accurate macrophage evaluations in standard cell lines and primary cells after cytokine treatment, as well as following exposure to a model disease environment.

high-content_and_high-throughput_identification_of_macrophage_polarization_phenotypes

Introduction<!>Supramolecular assembly of sensor<!>Discrimination of M1 and M2 subtypes using RAW 264.7 cells<!>Discrimination of M1 and M2 subtypes with primary macrophages<!>Discrimination of macrophages exposed to conditioned media from different cancer cells<!>Discussion<!>Conclusions<!>Materials<!>PONI-C 3 -guanidine polymer synthesis<!>Fluorescent titration<!>Binding affinity calculation<!>Unknown identication<!>Macrophage polarization via activation agents<!>Macrophage polarization via cancer cell conditioned media<!>Sensing studies<!>RT-PCR preparation<!>RNA extraction and cDNA conversion<!>RT-PCR preparation<!>Quantitative RT-PCR

<p>Macrophages are plastic leukocytes that perform a vast range of immune-and homeostasis-related functions, with their function and behavior dictated by environmental stimuli. Macrophages can be characterized as being activated into two major phenotypes, M1 and M2. 1 M1 macrophages are associated with inammation, including secretion of pro-inammatory cytokines, engulfment of foreign entities, generation of reactive oxygen and nitrogen species, and assistance in T-helper type1 (Th1) cell responses to ght infection. Conversely, M2 macrophages perform anti-inammatory and wound repair functions. 2,3 Disturbance of the mechanisms that govern the balance of M1 and M2 states can result in a number of health problems, including infections, cancer, pregnancy complications, and inammatory and autoimmune diseases. 4,5 Given the signicance and complexity of the roles macrophages play in biology and disease, knowledge of their activation and polarization state can provide critical information regarding the disease microenvironment, and be useful in selecting therapeutic approaches. For example, manipulation of tumor-associated macrophages (TAMs) provides a potential means to combat cancer. The tumor microenvironment releases factors that drive macrophages toward an M2-like phenotype, 6 resulting in secretion of anti-inammatory cytokines, promotion of tumor growth and invasion, and facilitation of metastases. Therapies are being developed to "re-educate" these TAMs from this immune-suppressing state to an antitumor M1 phenotype as a more effective, less toxic cancer treatment. 7,8 The development of such entities would be facilitated by a means to evaluate macrophage characteristics in a straightforward and highthroughput manner.</p><p>Efforts to generate therapies based on macrophage phenotypic conversion (to stimulate immune activation or suppression) and evaluate macrophage immunes responses to other agents in drug discovery and toxicology are challenging due to the complexity of the polarization process. An increasing body of research reveals that macrophage polarization is more intricate than a two-state, M1/M2 conversion; rather, a spectrum of states exists. [9][10][11] M2 macrophages can be further subclassied into M2a, M2b, M2c, among others, depending on the activating stimulus and resulting surface markers displayed. 12 In addition, the macrophage polarization/sub-polarization process is dynamic and can evolve based on changes in the microenvironment. [13][14][15] Complicating the matter further, macrophages can have mixed or overlapping M1 and M2-associated indicators. For instance, macrophages isolated from patients with advanced gastric and pancreatic cancers show high levels of both pro-inammatory and anti-inammatory cytokines. Both sets expressed IL-10 (M2-associated), while the former also had high levels of IL-12, and the latter IL-1b and TNF-a (M1 associated). 16,17 These factors make it challenging to identify macrophage polarization states for diagnostic applications and fundamentally, to understand or identify phenotypes that are relevant to disease states.</p><p>Currently, the presence or levels of cellular and/or secreted biomarkers is most commonly used to detect and characterize macrophage polarization. 12,18,19 While providing useful information, this approach is reliant on the specicity of the markers and requires multiple assays to obtain sufficient information for cellular evaluation. Additional limitations include: expression overlap between different polarization states (as mentioned above), poor phenotypic resolution of similar stimuli, nontranslatable markers between mice and humans, 12 and the fact that mRNA levels do not necessarily signal a robust difference in protein expression/at the functional level. 19 In addition, the techniques used to identify the presence of biomarkers, such as RT-PCR, western blot, and ow cytometry, are expensive and not amenable to multiplexing or high-throughput applications. Thus, there is a strong need for a general high-throughput method that can be used to evaluate these cells and their characteristics to facilitate therapeutic design and understand phenotypic responses of macrophages to stimuli.</p><p>As an alternative to marker-specic approaches, chemical nose or array-based sensing employs and discerns selective interactions between analytes and sensor elements to generate unique patterns for each analyte. The resulting pattern can be further analyzed for quantitative classication. Once trained, the sensor can rapidly identify analytes based on pattern recognition. This approach has been successfully used in a wide range of systems including mammalian cells, [20][21][22] bacteria, [23][24][25] and proteins in biouids. [26][27][28] The strategy is ideal for cell phenotyping because changes in cellular responses yield variations in surface composition (e.g., protein, lipids, glycans, etc.) that result in different ngerprints, providing high-content information for each cellular state. [29][30][31] Because macrophage polarization is accompanied by changes in cellular metabolism and surface protein expression, 1,12,32 we hypothesized that an arraybased sensing strategy would provide a general platform for discriminating macrophage phenotypic and sub-phenotypic states. Incorporation of this strategy into a multi-channel format would enable multidimensional, high-content output from a single microwell, rendering this method readily applicable to high-throughput screening. 33 In this paper, we describe the development and application of a polymer-protein supramolecular assembly as a sensor array to gather high-throughput, high-content information on macrophage polarization state. The sensor is composed of only two elements: a guanidine-functionalized cationic poly(oxanorborneneimide) (PONI) polymer, and an anionic green uorescent protein (GFP). The two entities form a complex through electrostatic interactions, resulting in a Förster resonance energy transfer (FRET) pair. When this sensor is applied to macrophages in different polarization/sub-polarization states, it yields uorescent signals in ve channels. The multidimensional output is then quantitatively analyzed using linear discriminant analysis (LDA) to reproducibly classify different macrophage activation states (Fig. 1). To the best of our knowledge, this combination of sensor elements resulting in a 5-channel output has not been reported previously. We validated the sensor with model macrophage RAW 264.7 cells and primary bone marrow-derived macrophages (BMDMs) stimulated with known M1 and M2 polarizing cytokines. The successful discrimination of M1 and M2 macrophages among the ve subtypes demonstrates the ability of the sensor to accurately differentiate subtle phenotypic changes. We further evaluated the efficacy of the sensor system in a model disease environment, where macrophages were cultured in cancer cellconditioned media, generating distinct patterns for macrophages exposed to different cancer types. Taken together, the sensor platform can classify macrophage phenotypes in a matter of minutes. Furthermore, this platform can read out the effects of subtle environmental changes on macrophages, providing a new tool for diagnostics and for fundamental studies of macrophage behavior. The information generated can provide valuable insights on macrophages in diseases, potentially improving efficiency of existing therapies and facilitating the development of new treatments.</p><!><p>The sensor is designed to provide an information-rich, vechannel output with only two sensor elements. The rst element of the sensor is a cationic poly(oxanorbornene) (PONI) random copolymer scaffold that incorporates a guanidine group and a pyrene dye molecule (C 3 -Gu-Py). The positively charged guanidine group ensures that selective interactions occur only when the complex is close to negatively charged cell surface functionalities. The solvatochromic pyrene molecule will alter its spectral properties when local environmental factors, such as polarity and hydrophobicity, change. 34 In this way, both selectivity and sensitivity of the sensor are ensured. Through electrostatic interactions, cationic C 3 -Gu-Py forms a polymeric complex with an anionic GFP. In practice, the pyrene unit provides three signals, two corresponding to the free pyrene and one to the excimer form. The GFP then adds two channels: free GFP uorescence and FRET with the excimer pyrene channel (Fig. 1b).</p><p>Initial studies focused on the optical characterization of the C 3 -Gu-Py/GFP supramolecular assembly. Polymer C 3 -Gu-Py was titrated with increasing concentrations of GFP. Aer 30 min of incubation, a simultaneous decrease in pyrene emission at 470 nm and increase of GFP emission at 510 nm was observed upon irradiation with 344 nm light (Fig. S1 and S2a †). Efficient uorescence quenching of C 3 -Gu-Py was observed at higher concentrations of assembly (Fig. S2b †). The association constant K a of 7.17 Â 10 5 M À1 was derived by tting the uorescent titration curve.</p><p>The overall spectrum featured ve distinguished output peaks that can be recorded from the sensor: pyrene monomers at 344/390 and 344/420, pyrene excimer at 344/470, GFP at 475/ 510, and FRET signal at 344/510. Based on the spectral exibility, a concentration of 0.5 mM C 3 -Gu-Py and 50 nM GFP was selected for the following experiments. Dynamic light scattering data revealed the polymer assembly was $230 nm in diameter and the size slightly increased to $237 nm when GFP was added (Fig. S3 †). Transmission electron microscopy images conrmed these results (Fig. S4 †), indicating that a supramolecular assembly was formed between C 3 -Gu-Py and GFP.</p><!><p>We rst tested the ability of the sensor system to distinguish among macrophage phenotypes using the RAW 264.7 macrophage cell line. Established cytokines were used to stimulate the cells, with each activating macrophage through a different mechanism (Table 1), generating a distinct phenotypic state. RT-PCR results assessing standard M1 and M2 markers conrmed that cells were polarized into corresponding states aer 48 h activation (Fig. 2). LPS and IFN-g treated cells (M1 stimulation) showed signicant increases in TNF-a and iNOS mRNA expression whereas the IL-4 (M2a stimulation) group had an increase in EGR2 and mannose receptor (MR) expression. Similar TNF-a levels observed between the combination treatment and the control group could be explained by the prolonged 48 h activation time, negative regulators such as NFkB and nuclear factor activated T cells, 35,36 greater production of nitric oxide, and the fast intracellular turnover rate of TNF-a. 37 Although the IL-10 group (M2c stimulation) was tested against multiple M2 markers, including EGR2, MR, and TGF-b, as well as the reduction of M1 marker iNOS, no signicant changes in the levels of expression of any associated genes were observed (Fig. S7 †).</p><p>Having conrmed that polarization had occurred, cells from each treatment group were plated on a 96-well microplate for overnight attachment. Equivalent cell numbers (10 000 per sample) were used to ensure that changes in sensor response were due to alterations in cell surface functionalities, not density. For the sensing process, C 3 -Gu-Py and GFP were premixed for 30 min to allow formation of stable FRET complexes. Subsequently, cells were washed once with phosphate buffered saline (PBS) and incubated with the sensor complex in the dark. Fluorescence signals were recorded every 15 min until equilibrium was reached. The 5-channel readout generated a distinct uorescence pattern for each treatment group (Fig. 3a). We further utilized linear discriminate analysis (LDA) to test whether the ve cell phenotypes could be robustly discriminated based upon their uorescent signatures. As shown in Fig. 3b, the LDA plot revealed ve distinct clusters for M1 and M2 subtypes with a correct classication of 100% (Tables S1 and S2 †), demonstrating that each activation pathway resulted in a distinct cellular response. We further validated the reliability of the sensor by performing unknown sample identication and comparing the results against the training set. Among the 45 tested unknowns, 41 samples were predicted correctly into their corresponding group, giving a high percentage of correct unknown identication of 91% (Table S3 †). The accuracy of unknown identication could be further improved by increasing the size and complexity of the training set if needed. Next, we investigated the necessity of having 5 channels of information from the sensor by comparing the performance of classication and unknown identication using either an individual sensor element or different combinations. The highest percentage of accuracy was achieved when all 5 channels were used, demonstrating the importance of multidimensional data in discriminating complex cell phenotypes (Fig. 3c).</p><!><p>Immortalized macrophage cell lines provide a useful tool for assessing sensor response, however, these models differ in multiple aspects from their primary cell analogs. We next tested the sensor using physiologically relevant primary bone marrowderived macrophages (BMDM). Progenitor cells were isolated from C57/B6 mice and induced to differentiate into macrophages using previously reported procedures. 43 Once macrophage cells were obtained, we exposed them to M1 and M2 subtype polarization stimuli for 48 h as used above for RAW 264.7 cells. RT-PCR results conrmed appropriate activations in each case, with increases in TNF-a and iNOS mRNA expression for M1 related stimuli (LPS and/or IFN-g) and EGR2 and MR mRNA levels for IL-4 stimulated M2 cells. Although IL-10 activation did not show substantial enhancement in EGR2 level, a nearly 6-fold increase in MR expression was observed (Fig. S8 and S9 †). Following macrophage polarization, similar sensor procedure was performed. The uorescence patterns observed were distinct from those of the RAW cells, consistent with differences that exist between the two cell models. Complete discrimination among the ve assessed groups of M1 and M2 phenotypes was achieved with 96% correct classication (Fig. 4, Tables S4 and S5 †). 92% of correct unknown iden-tication conrmed the high reliability of our sensor (Table S6 †).</p><p>When the sensor complex interacts with cells, in most cases, all monitored uorescence channels showed an increase in signal intensity. This suggests that upon interacting with macrophages, the sensor complex disaggregates, exposing its individual components to interact with the cell surface. Depending upon the local environment, the uorescence intensities for individual molecules (pyrene, GFP, and FRET) also change. Since distinct uorescence patterns were consistently observed for each stimulus, we believe this disruption process is modulated by cell surface functionalities and composition. Our previous studies have indicated that the sensor complex is highly sensitive to glycosylation patterns on cell surfaces. 30 However, more mechanistic studies are needed in order to elucidate which other cell components are also interacting with the sensor elements.</p><!><p>The above studies demonstrate that our sensor array was able to discriminate macrophages polarized with specic cytokines. However, biological microenvironments are oen far more complex and have multiple stimuli. Hence, we assessed whether the sensor could discern macrophage phenotype in a model disease environment to address this issue. First, conditioned media was generated by culturing different types of cancer cells (HeLa, cervical carcinoma, and MCF7, mammary carcinoma) until $80% conuency was reached. Then, the culture media was extracted and used to stimulate macrophages for 48 h. RT-PCR results revealed different activation patterns for macrophages activated with media conditioned from different cell lines (Fig. S10 †). C 3 -Gu-Py and GFP complexes were added to cells and the 5-channel uorescence readouts were collected.</p><p>Distinct uorescence signals were obtained for macrophages subjected to each of the conditioned media types. An LDA plot showed three well-separated clusters with 100% classication accuracy (Fig. 5, Tables S7 and S8 †). When macrophages were exposed to cultured media conditioned by cervical cancer versus breast cancer cells, the sensing readout was dramatically different, indicating that a unique state of activation was present following each type of stimulation. A high percentage (96%) of correct unknown identication was also achieved (Table S9 †). These results are exciting because it demonstrates that this method not only functions following single cytokine stimulation, but also in more complex environments. This is promising evidence that with careful evaluation, this sensing method could be applied to prole macrophages from individual patients, offering insights for precision medicine.</p><!><p>Macrophage polarization is a complex and dynamic process. With its roles in homeostasis and disease, it is important to be able to discern macrophages characteristics in a rapid and straight-forward manner. Compared with current methods of characterizing macrophage polarization, the sensor reported in this study has advantages of generating a multidimensional and high-content chemical readout regarding the cell surface in a high-throughput matter. Standard methods, such as RT-PCR and ELISA, can only capture a limited number of wellestablished markers for each cell activation state, and are independent (not multiplexed) assays, requiring a separate analysis for each. Considering the heterogeneity of macrophage polarization and the overlapping nature of M1 and M2 markers, 11,12 it is also difficult to elucidate and differentiate activation states with standard methods. For instance, the multiple IL-10 markers used in our RT-PCR studies did not reveal signicant changes. The ambiguity of a less-well characterized sub-phenotype could be because the end-point evaluation missed the dynamic changes on the macrophage marker expression during the 48 h activation.</p><p>In contrast, the array-based sensor utilizes selective interactions between sensor elements and the entire analyte surface to generate high-content ngerprints for each activation state.</p><p>Once trained, the sensor can rapidly identify target analytes through pattern recognition. Although the C 3 -Gu-Py moiety has been utilized for bacterial sensing, 44 its capability in mammalian cell sensing has not been investigated. By coupling the polymer with simple GFP through supramolecular interactions, sensor can discern less-characterized sub-phenotypes, such as IL-10 stimulated macrophages, which are challenging to identify using traditional methods like RT-PCR (Fig. 3 and 4). The 5channel, high-content information gathered from the sensor is crucial in achieving a high level of classication accuracy and it allows us to address challenging biological questions from a chemical perspective. In addition, running assays like RT-PCR and ELISA can be time-consuming and error prone, with relatively high costs for thorough characterizations consisting of multiple markers. In contrast, the sensor material used here is synthetically easy to generate, and all components can be mixed in one microplate well, which not only reduces sensor material but is also compatible for high-throughput screening applications. What is more, accurate phenotyping can be obtained in less than an hour, making this method simple, robust, and rapid.</p><p>Due to the robust and facile nature of the system, there are many potential applications for the array-based sensing strategy. Altered immune states are a major factor in diseases including cancer, atherosclerosis, and auto-immune disorders. [45][46][47] Macrophage polarization states are key in driving forward disease progression. Rapid assessment of their activation states can provide valuable information in selecting appropriate therapeutic strategies. 48,49 Notably, the high-throughput nature of the method would facilitate the rapid screening of immune states for individual patients, enabling personalized medicinal approaches in tackling these immune-driven diseases. Furthermore, this strategy could be applied to other plastic immune cells, such as dendritic cells and T cells. 50,51 By extending this sensor to other cell types, the status of major components of the immune system could be rapidly determined. This strategy can also greatly improve the drug discovery process, by allowing for rapid identication of altered cell states, and/or evaluation of immunogenicity following agent treatment. 52 Potential immune adjuvants or anti-inammatory entities could be screened together by using the sensor on immune cells in a multi-well plate format. With these capabilities, the sensor system not only has utility as a fundamental research tool, but as a highthroughput, high-content means for therapeutic screening against general plastic cell types.</p><!><p>In summary, we demonstrate the use of a simple and robust chemical system that can quickly capture the overall responses of activated macrophages in a high-throughput format, which is challenging with biomolecular tools. Through the supramolecular assembly of only two elements, a 5-channel output is achieved. The high level of information density enables us to accurately prole a spectrum of activation state of macrophages. The ability to use chemical entities to answer biological questions opens the doors for sensing and beyond.</p><!><p>All reagents were purchased from Thermo-Fisher Scientic except where otherwise noted. All DMSO utilized was cell culture grade (Sigma). RT-PCR data was generated using a CFX Connect Real-Time PCR Detection System (Biorad, Hercules, CA). For assays requiring absorbance and uorescence measurements, a SpectraMax M2 plate reader was used (Molecular Devices, San Jose, CA).</p><!><p>Monomers and polymers were synthesized according to previous reports. 53 Detailed synthetic scheme can be found in the ESI. † Green uorescent protein expression GFP was constructed and characterized according to reported protocols. 54 In short, Escherichia coli strain BL21 was transformed with plasmids containing GFP recombinant protein.</p><p>Aer transformation and induction with IPTG, cells were lysed and puried by Co 2+ nitrilotriacetate columns. Fluorescent proteins were further characterized by SDS-PAGE gel, scanning absorbance and emission spectrum. The results are consistent with previously reported work. 30 Transmission electron microscopy (TEM) TEM samples were prepared by either 0.5 mM of C 3 -Gu-Py only or mixing 0.5 mM of C 3 -Gu-Py with 50 nM of GFP in 10 mM HEPES buffer for 30 min in dark at room temperature. 5 mL of the solutions were then placed on 300 mesh copper grids (with formvar lms) obtained from Electron Microscopy Sciences (EMS FF300-Cu) and allowed to dry overnight. The samples were analyzed using a TEM JOEL 2000FX at an acceleration voltage of 200 kV.</p><!><p>0.5 mM C 3 -Gu-Py polymer was titrated with GFP at a concentration range from 0 to 200 mM in a black 96 well-microplate. The solution was mixed in 10 mM HEPEs buffer. Aer 30 min incubation at room temperature in dark, the uorescence spectrum was measured at an excitation wavelength of 344 nm.</p><!><p>Fluorescence titration was utilized to calculate the binding affinity of the C 3 -Gu-Py polymer with GFP. The uorescence decay of the C 3 -Gu-Py excimer as a function of GFP concentration was tted to a one-site binding equation, 55 which is:</p><p>where I is the uorescence intensity of C 3 -Gu-Py excimer at a given concentration of GFP, I 0 is the uorescence intensity of C 3 -Gu-Py in the absence of GFP, I lim is the uorescence intensity when the quenching reaches a plateau, C 0 refers to the concentration of C 3 -Gu-Py, and C is the concentration of GFP.</p><p>Based on the equation, microscopic binding constant K a was determined by using the non-linear least-squares curve tting analysis in OriginPro (OriginLab Co., Northampton, USA).</p><p>Linear discriminant analysis (LDA)</p><p>LDA was applied on normalized uorescence data to statistically classify each group, using SYSTAT soware (version 11.0, SystatSoware, Richmond, CA, U.S.A.). All variables were used in the complete mode and the tolerance was set as 0.001. Input data was transformed to canonical scores to best separate each group where the between-class variance was maximized while the within-class variance was minimized. Aer transformation, LDA reduces the dimensionality of the. The 2D plot pictorially shows where each data point lies in the new dimensional space. Therefore, the positive and negative values on the axis do not have any physiologically meaning.</p><!><p>The identity of unknown samples was predicted by computing the Mahalanobis distance of the unknown data to the training groups using LDA. 56 First, the normalized uorescence responses of the unknown samples were converted to canonical scores in LDA, using the discriminant functions established from the reference set. Next, Mahalanobis distance of that case to the centroid of each training cluster in the LDA space was computed. 56,57 The unknown sample was predicted to belong to the closest group, dened by the shortest Mahalanobis distance.</p><p>Cell culture RAW 264.7 cells, HeLa and MCF7 cell lines were purchased from American Type Culture Collection (ATCC, Manassas, VA). Primary bone marrow derived macrophages (BMDMs) were isolated from freshly euthanized C57/B6 mice, donated generously by Dr Jessie Mager, Department of Veterinary and Animal Science, University of Massachusetts Amherst. The BMDMs were isolated, differentiated and cultured according to previously reported methods. 43 All cells were cultured at 37 C under a humidied atmosphere containing 5% CO 2 using standard growth media consisted of high glucose Dulbecco's Modied Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% antibiotics (100 mg mL À1 penicillin and 100 mg mL À1 streptomycin). Under the above culture conditions, the cells were sub-cultured approximately once every two to ve days.</p><!><p>Both RAW 264.7 cells and BMDMs were treated with the following polarization stimuli for 48 h to induce the desired polarization state. LPS group: 50 ng mL À1 , IFN-g group: 50 ng mL À1 , combo group: 50 ng mL À1 LPS and IFN-g, IL-4 group: 30 ng mL À1 , and IL-10 group: 30 ng mL À1 . Aer 2 day polarization, cells were washed with PBS, trypsinized, and plated as 10 000 cells per well on a 96-well plate overnight before proceeding to sensing studies.</p><!><p>HeLa and MCF7 cell lines were cultured under DMEM medium supplemented with 10% FBS and 1% antibiotics for 2 days to reach above 80% conuency. The supernatant from each cell line was then collected and centrifuged for 5 minutes. Subsequently, 5 mL of the supernatant was transferred into a T25 culture ask containing RAW cells. Aer 48 h of culture, RAW 264.7 cells were washed with PBS, trypsinized and plated as 10 000 cells per well on a 96-well plate for overnight attachment.</p><!><p>The sensor was prepared by mixing 0.5 mM of C 3 -Gu-Py with 50 nM of GFP in 10 mM HEPES buffer for 30 minutes in dark at room temperature. Subsequently, 150 mL of sensor solution was incubated with and without the cell populations (washed once with PBS) in 96-well microplates. The change in uorescence intensity for each channel was recorded every 15 minutes at its respective wavelength (pyrene monomer: 344/390 nm and 344/ 420 nm, pyrene excimer: 344/470 nm, GFP: 475/510 nm, FRET: 344/510 nm) on a Molecular Devices SpectraMax M2 microplate reader using appropriate lters.</p><!><p>Cells were plated in 24-well plates at a density of 50 000 cells/ well. Cells were treated with the appropriate polarization stimulus for 48 h. Following treatments, RNA was extracted following the procedure below.</p><!><p>Approximately 1.5 mg RNA was harvested using the PureLink RNA Mini Kit (Ambion) following the manufacturer's instructions. SuperScript IV Reverse Transcriptase was used for the conversion of approximately 150 ng of RNA to cDNA, along with RNaseOut, 10 mM dNTPs, and 50 mM Random Hexamers (ThermoFisher, Pittsburgh, PA), also following the manufacturer's instructions.</p><!><p>Cells were plated in 24-well plates at a density of 50 000 cells per well. Cells were treated with the appropriate polarization stimulus for 48 h. Following treatments, RNA was extracted following the procedure below.</p><!><p>RT-PCR was performed on cDNA as prepared above using a CFX connect real-time system with iTaq Universal SYBR Green Supermix (Biorad, Hercules, CA). All DNA primers were purchased from Integrated DNA Technologies (Caralville, Iowa).</p><p>The following primer sequences were used:</p><p>b-Actin (forward) GATCAGCAAGCAGGAGTACGA, b-Actin (reverse) AAAACGCAGCGCAGTAACAGT; iNOS (forward) GTTCTCAGCCCAACAATACAAGA, iNOS (reverse) GTGGACGGGTCGATGTCAC; TNF-a (forward) CCTGTAGCCCACGTCGTAG, TNF-a (reverse) GGGAGTCAAGGTACAACCC; EGR2 (forward) TGAGAGAGCAGCGATTGATT, EGR2 (reverse) ATAACAGTCAGTGTGTCCCC; Mannose receptor (forward) GGATGTTGATGGCTACTGGA, Mannose receptor (reverse) AGTAGCAGGGATTTCGTCTG; TGF-b (forward) GCGGACTACTATGCTAAAGA, TGF-b (reverse) TTCTCATAGATGGCGTTGTT.</p><p>Analyses were performed as follows: the samples were rst activated at 50 C for 2 min, then 95 C for 2 min. Then denaturing occurred at 95 C for 30 s followed by annealing at 57 C; the denature/anneal process was repeated over 40 cycles. Relative gene expression was determined by comparing the C t value of the gene of interest to that of the b-actin housekeeping gene, by the 2 DDC t method. 58 Three biological replicates were performed for each control group and three technical replicates were used for each biological replicate.</p>

Royal Society of Chemistry (RSC)

Predicted Melt Curve and Liquid Shear Viscosity of RDX up to 30 GPa

Recent grain scale simulations of HMX and TATB have shown that predictions for hot spot formation in high explosives are particularly sensitive to accurate determinations of the pressure-dependent melt curve and the shear viscosity of the liquid phase. These physics terms are poorly constrained beyond ambient pressure for the explosive RDX. We adopt an all-atom modeling approach using molecular dynamics (MD) simulations to predict the melt curve of RDX near to detonation conditions (30 GPa) and determine the shear viscosity of the liquid as a function of temperature and pressure above the melt curve. Phase-coexistence simulations were used to determine the melt curve, which is predicted to vary by almost 1100 K as the pressure increases from 0 GPa to 30 GPa. Equilibrium MD simulations and the Green-Kubo formalism were used to obtain the pressure-temperature dependent shear viscosity. The shear viscosity of RDX is predicted to be of similar magnitude to the viscosity of TATB at low GPa-range pressures, and to be roughly an order of magnitude lower than the viscosity of HMX. The temperature dependence of the shear viscosity is Arrhenius at a given pressure, and the exponential pre-factor and activation term exhibit a strong, yet complicated, pressure dependence. An empirical pressure-temperature dependent function for RDX shear viscosity is developed that simultaneously captures a wide range of MD predictions while taking an analytic form that extrapolates smoothly beyond the fitted regime. The relative strength of the pressure and temperature dependencies of these two physics terms is found to be of similar magnitude for RDX, HMX, and TATB, which motivates incorporating these results in future RDX grain scale modeling.

predicted_melt_curve_and_liquid_shear_viscosity_of_rdx_up_to_30_gpa

Introduction<!>General Simulation Details<!>Methods<!>Results<!>Shear Viscosity 4.1 Methods<!>Results<!>Conclusions<!>Force Field Validation

<p>Initiation of high explosives is controlled through hot spots, which are localized regions of elevated temperature and energy that form when shock waves interact with microstructural defects such as pores and grain boundaries [1][2][3][4]. The physics of hot spot formation strongly depends on microstructure and elastic/inelastic material response, while hot spot evolution and growth into a self-sustained burn depends on a complex interplay between thermal conduction and reaction kinetics. Grain scale simulations with coarse grained particles [5][6][7][8][9] or continuum-based multiphysics models [10][11][12][13][14] are a primary tool for explicitly resolving the formation and growth of large hot spots thought to govern initiation response. Accuracy of grain scale model predictions depends on accurate determinations of a wide range of thermodynamic, mechanical, thermal, and chemical material properties that serve as direct inputs or calibrants. Parameters for these properties are often difficult to isolate in focused experiments at high pressures and temperatures so placeholder values are often used in grain scale models. All-atom modeling using molecular dynamics (MD) simulations has proven to be a robust tool for obtaining many of these missing material parameters [15][16][17][18][19][20][21][22][23][24]. Scale bridging between atomistic and grain scale simulations at commensurate</p><p>[a] M. P. Kroonblawd length and time scales has recently been used to calibrate and validate grain scale model terms through direct comparison [25][26][27][28][29][30][31]. Despite the promise of a one-and-done approach to calibration through scalebridging simulations, it is often more advantageous to gradually increase grain scale model complexity and independently determine physics terms due to the highly coupled and nonlinear nature of dynamic material response [24,28,29]. This is especially true for material parameters that vary significantly within the applicable temperature and pressure domain such as the melt curve and liquid shear viscosity [15,21,24]. We focus here on obtaining predictions for two key physics terms influencing hot spot formation, namely the melt curve and liquid shear viscosity, for the explosive RDX (hexahydro-1,3,5-trinitro-s-triazine) to near detonation conditions where there is a lack of data to parametrize and constrain grain scale models.</p><p>The melt curve plays a significant role in hot spot formation in high explosives on ultrafast time scales. Melting places an upper temperature limit on plastic work, which serves as a primary source for energy localization at hot spots and in shock heating of bulk crystal. In the case of HMX (octahydro-1,3,5,7tetranitro-1,3,5,7-tetrazocine), estimates of the melt curve based on the Lindemann law [32] were shown to be substantially lower than predictions obtained from MD simulations at GPa-range pressures [24]. Application of MD-derived melt curves for HMX [24] and TATB (1,3,5-triamino-2,4,6-trinitrobenzene) [21] in grain scale simulations of supported shocks were shown to effectively suppress melting during the collapse of micron-and sub-micron scale pores [24,28,29].</p><p>Predicted peak temperatures in these simulations were hundreds of Kelvin greater compared to analogous grain scale predictions with simpler melting models that give lower melting points at elevated pressures. Direct comparisons of grain scale simulations of shockinduced pore collapse to all-atom simulations showed that accurately capturing the melt curve was critical for consistency in predictions for the hot spot formation process [29]. While omitting melting physics altogether may work well in some situations, pressure unloading in unsupported shocks can lead to substantial melting of plastically worked crystal [24]. Complex variations in the pressure field in simulations of polycrystalline microstructures that might lead to melting are difficult to anticipate even for supported shocks. Accurate determinations of the melt curve are therefore necessary, regardless of whether melting is explicitly included or judiciously omitted in multiphysics explosive models.</p><p>Viscous flow can be a significant source for additional heating of melted material. Dynamic shear viscosity determines the transient deviatoric stresses that develop in a liquid in response to a shearing rate and the corresponding dissipative work. In some circumstances, the shear viscosity of a liquid will exhibit exponential dependence on temperature and pressure [16,[33][34][35] and it can also be a function of shearing rate. The pressure-temperature dependent shear viscosity of HMX and TATB at zero rate varies by orders of magnitude over a range of a few GPa and for a range of temperatures that are typical of shock initiation [15,21,24]. Simulations of viscous heating in melt pools formed at hot spots in HMX is predicted to push peak temperatures to values that are hundreds of Kelvin higher than in inviscid treatments [24]. Atomistic modeling predicts that plastic shear localization dynamically forms nanoscale shear bands of amorphous material in many explosives [9,26,[36][37][38][39][40] and this response has been captured phenomenologically as melting in grain scale models [10,13]. Effective rate-dependent viscosities for HMX in shear bands vary linearly in temperature [23] and approach zero-rate predictions [15,24] for rates at or below 5 × 10 9 s −1 . It remains an open question how best to model shear bands in explosives. Crossinspection of shear band temperatures in TATB [40,41] against the predicted melt curve [21] indicates that the material in TATB shear bands falls well within the solid region of the phase diagram.</p><p>We focus here on characterizing the RDX solidliquid phase boundary and liquid phase shear viscosity, for which little is known under shock initiation and detonation conditions. RDX solid exhibits a rich polymorphism with six phases confirmed through diffraction experiments. The α form [42] and β form [43] occur near normal conditions, although the β form is highly metastable and readily converts to α. Reversable transformations to high-pressure polymorphs have been identified near room temperature, including the γ form (P ~ 4 GPa) [44,45], the δ form (p ~ 18 GPa) [46], and the ζ form (P ~ 28 GPa) [46]. A highly reactive ε form has been isolated within a narrow range of elevated temperatures and pressures [47] that can be recovered at ambient pressure [48]. Most of these forms have verified orthorhombic space groups and exhibit discontinuous volume changes across a transition as well as differences in molecular conformations.</p><p>Considerably less is known about the melting point of RDX. The melting point is constrained by multiple experimental studies to a value of 477-478 K at ambient pressure [49][50][51]. Myint et al. [52] note one conference report by Nauflett et al. [53] that obtained somewhat larger values of the ambient-pressure melting point 480-484 K; the equation of state model developed by Myint et al. yielded a melting point of 478.15 K, in agreement with the other experiments. By comparison, there is only limited macroscopic evidence available for melting at elevated pressure. Dreger and Gupta [47] found a melting onset temperature of 488 K for α-RDX at 0.65 GPa that was quickly followed by reactions. Those authors identified spectroscopic and optical signatures for melting in diamond anvil cell experiments only up to ~2 GPa, beyond which α-RDX either decomposes directly or transforms to the reactive ε form. It is important to note that the above measurements of solid-solid and solid-liquid transitions correspond to integrated macroscale observations on timescales of seconds or longer, which does not necessarily constrain the kinetics of transitions under dynamic loads on ultrafast time scales typical of shock initiation or detonation conditions.</p><p>Classical, non-reactive MD simulation provides what is perhaps the only practical route to determine the melt curve and liquid shear viscosity for non-meltcastable explosives such as RDX. While melting and viscous heating under pressure may play significant roles on ultrafast time scales, the melt is highly reactive which prevents direct experimental measurements. Similarly, while reactive MD simulations provide a solid basis for determining material response coupled to chemistry under these conditions [25][26][27], non-reactive simulations with a reasonable choice of force field (FF) allow for isolating a range of mechanical, thermal, and thermodynamic parameters of the reactant state. We use the well-established FF by Smith and Bharadwaj [54], which has found numerous applications to simulations of HMX and RDX under extreme conditions [7, 9, 15, 17-19, 23, 24, 29, 36, 37, 39, 55-57]. The predicted pressure-volume equation of state and Hugoniot are well described by this FF at GPa-range pressures, as is the relative ordering of the Gibbs energy as a function of pressure for the α and γ phases [55]. Predicted peak positions and line widths of α-RDX in the low frequency (<10 THz) vibrational spectrum are close to experimental values [56], indicating that lattice modes are well described. Most of the predicted α-RDX elastic tensor coefficients near normal conditions are essentially the same as experiment within the scatter generated by the different MD methods used to compute these values [18]. The orientationally averaged thermal conductivity of α-RDX predicted by the FF is close to pressed-powder experiments [17] and the predicted ordering of anisotropic conductivity along the three lattice directions matches single-crystals experiments [57]. Sellers et al. [19] found the Smith-Bharadwaj FF predicted the RDX melting point at atmospheric pressure to within 2.2% of experiment, lending confidence in the model for the present application.</p><p>We use MD simulations of phase coexistence [58,59] to obtain the melt curve of RDX up to 30 GPa. Equilibrium MD simulations of the liquid phase and the Green-Kubo formalism [60] are then applied to predict the dynamic shear viscosity as a function of temperature and pressure above the melt curve. The remainder of the article is organized as follows. In section 2, we provide general simulation details and discussion on modifications to the Smith-Bharadwaj FF that improve simulation stability at high pressure conditions. Section 3 outlines the methods used to obtain the melt curve and discusses our predictions for RDX in the context of the limited experimental data and other computational determinations for the related explosives HMX and TATB. Section 4 outlines the methods used to predict the shear viscosity; a complicated temperature-pressure dependence is identified that we reduce to a straightforward analytic empirical form. Conclusions are summarized in Section 5.</p><!><p>Classical, non-reactive MD simulations were performed using the LAMMPS code [61] and a variant of the wellestablished force field (FF) for RDX and HMX by Smith and Bharadwaj [54]. This FF is of a class-I type and includes harmonic potentials for bonds, angles, and improper dihedrals and cosine series for proper dihedrals. Nonbonded interactions are modeled using the Buckingham potential (exp-6) with electrostatics between fixed partial charges on the nuclei. There are two differences between the original parameterization and that used here. These include a tuned harmonic potential for the N-O bonds [62] and an additional 𝑟 −12 repulsive nonbonded interaction for each pair type that compensates for the divergence in the Buckingham potential at short separation [29,63]. The additional repulsive potential was found to be necessary for simulations of the liquid phase at pressures above 10 GPa and leads to no appreciable difference in liquidphase density predictions or pairwise radial distribution functions at 10 GPa. Nonbonded terms were evaluated in real space up to an 11 Å cutoff. Electrostatic terms were evaluated in real space using the Wolf potential [64] with an 11 Å cutoff and a damping factor of 0.2 Å -1 . Additional implementation and validation details can be found in the Supporting Information.</p><p>Isochoric-isothermal (NVT) and isothermalisobaric (NPT) trajectories were integrated at various points in this study. Unless otherwise noted, isothermal simulations were performed using a Nosé-Hoover-style thermostat [65,66] with a time constant of 100 fs. Isobaric simulations were performed using a Nosé-Hoover-style barostat [67] with a time constant of 1000 fs and the optional drag parameter set to 1.0 (unitless). A triclinic barostat was used for simulations involving the crystal phase and an isotropic barostat was used for simulations of the pure liquid. All trajectories were integrated using a 0.5 fs time step. Simulation snapshots were rendered using OVITO [68].</p><!><p>The melt curve 𝑇 m (𝑃) was determined as a function of pressure using MD simulations and the phasecoexistence approach [58,59], which was previously applied to predict 𝑇 m (𝑃) of HMX [24] and TATB [69,21]. In this approach, an initial metastable two-phase configuration is prepared in which a 2D-periodic crystal slab is placed in contact with an amorphous, liquid-like region in an overall 3D-periodic simulation cell. The specific cell used is shown in Figure 1(a) and includes a 5𝐚 × 5𝐛 × 5𝐜 α-RDX crystal supercell (1000 molecules) constructed using the structure of Choi and Prince [42] with an exposed (100) face in contact with 1000 liquid molecules. Procedures for constructing this cell closely follow earlier work on HMX [24]. The twophase simulation cell serves as the starting point for independent NPT simulations at dozens of (T,P) states that bracket the melt curve. States above the melt curve lead to melting of the crystal region and typically involve an increase in the simulation cell volume. States below the melt curve generally show no loss of translational order for molecules in the crystal region, although typical MD simulations are usually not long enough to sample crystallization of the liquid region. All trajectories were 10 ns long, or less if the crystal region fully melted. We take 10 ns as an upper limit both due to the computational expense of these simulations and because 10 ns is a typical timescale for grain scale simulations of shock initiation [10,13].</p><p>Changes in simulation cell volume (𝑉) with time ( 𝑡 ) are typically used as a metric for determining whether a simulation is above or below the melt curve, with 𝑉 ̇> 0 for 𝑇 > 𝑇 m (𝑃) and 𝑉 ̇< 0 for 𝑇 < 𝑇 m (𝑃) . However, we found that this was not a reliable metric at higher pressures for RDX. For instance, all simulations at 30 GPa exhibited 𝑉 ̇< 0 despite inspection of the trajectory showing clear melting of the crystal region. This likely owes to the initial two-phase simulation cell being prepared at a 0 GPa pressure state. All melting point determinations were made based on visual inspection of the crystal region in the MD trajectories. Figure 1(b) shows snapshots of the crystal region at 30 GPa for different temperatures above and below the melting point. In this case, no melting was observed within 10 ns for 𝑇 ≤ 1540 K, some of the crystal layers melt for 1560 K ≤ 𝑇 ≤ 1580 K , and the crystal completely melts for 𝑇 ≥ 1600 K . Based on these results, we take the 30 GPa melting point to be 1560 K.</p><!><p>Solid-liquid coexistence simulations were performed to determine the melting temperature of RDX at 1 atm (hereafter denoted 0 GPa), and elevated pressures of 1, 3, 5, 10, 20, and 30 GPa. These simulations probe melting from the α phase even at high pressures and temperatures for which there are known solid-solid phase transitions. The Smith-Bharadwaj FF does not predict a prompt solid-solid 𝛼 → 𝛾 phase transition under hydrostatic loading up to at least 10 GPa [55]. This contrasts with NPT-MD simulations performed using density functional theory (DFT) in which a homogeneous solid 𝛼 → 𝛾 transition occurs within 5 ps at P = 3.9 GPa, T = 300 K [70]. This transition was observed for pressures as low as 2.75 GPa in DFT simulations, which was attributed to the large stress fluctuations that arise in a small computational cell [70] and that may also influence the transition kinetics. Shock experiments show that the 𝛼 → 𝛾 transition requires an incubation time on the order of 100 ns [71] at 5.5 GPa. The timescale for decomposition chemistry arising from oriented shocks along three different directions is also on the order of 100s of nanoseconds for pressures between 7 GPa and 20 GPa [72,73]. At the same time, kinetics for the 𝛼 → 𝜀 phase transition under hydrostatic conditions are on the order of hours and the transformation is followed promptly by chemical decomposition [47]. Thus, it is not unreasonable to expect the 𝛼 phase to remain metastable and proceed directly to the melt on ultrafast scales typical of shock initiation. Independent NPT simulations were performed at different temperatures for each pressure value. Following a coarse assessment of the melt curve using short O(1 ns) simulations at temperatures chosen in 100 K increments, the melting point was refined through 10 ns long trajectories performed in 20 K increments. Table 1 shows the minimum ( 𝑇 𝑚𝑖𝑛 ) and maximum ( 𝑇 𝑚𝑎𝑥 ) temperature values considered for these long 10 ns simulations as well as the predicted melting points (𝑇 𝑚 ). As discussed above, the melting point was taken to be the temperature value for which at least one crystal layer lost translational order within 10 ns based on visual inspection of the trajectory (see Figure 1).</p><p>Focusing first on the melting point at 0 GPa, there is an anomalous transition to a pseudo-liquid state at some temperatures. Figure 2 shows snapshots of final crystal-region configurations at and below the predicted melting point of 500 K. While there is a complete loss of rotational and translational order at 500 K, the situation is more complicated at 460 K and 480 K. A clear transformation takes place at 480 K to a state that exhibits rotational disorder but with translational order in two dimensions. Net displacement indicative of mass transport is also apparent, with several crystal molecules highlighted in the figure . A similar state is in the process of forming at 460 K, with a distinct "reordered" region similar to the 480 K result seen in the lower right corner of the left-hand panel. No loss of the original crystal packing order is seen at 440 K. This behavior was only found at 0 GPa, and even partially melted layers exhibited loss of both rotational and translational order at higher pressures.</p><p>The Smith-Bharadwaj FF yields a modest 4.6% overestimate of the melting temperature with the phasecoexistence approach at 0 GPa relative to the experimental value of 477-478 K [49][50][51]. Dreger and Gupta [47] found that melting and decomposition of α-RDX were closely linked at a pressure of 0.65 GPa in diamond anvil cell experiments, with a melting onset temperature of 488 K. Distinct changes in Raman spectra and optical images were observed to occur within minutes of melting that clearly indicated decomposition. Their melting onset value is somewhat lower than interpolations based on FF predictions [𝑇 𝑚 (0.65 𝐺𝑃𝑎) ≈ 552 K], but it is also unclear whether melting and decomposition can be fully decoupled in the macroscopic observations. Decomposition of α-RDX [24], and TATB [21]. Data points correspond to MD predictions determined using phase-coexistence simulations and lines to fits of the Simon-Glatzel equation to the MD data. Dashed lines indicate extrapolations beyond the highest pressure considered in MD simulations.</p><p>without any signs of initial melting was observed in those experiments above 2 GPa, as were transitions to the highly reactive 𝜀 phase.</p><p>Several predictions for the melting temperature of α-RDX at 0 GPa have been reported in the literature that used the Smith-Bharadwaj FF. These include the determinations by Zheng and Thompson [74], who predicted melting from a superheated perfect crystal, and more recently be Sellers et al. [19] via more rigorous thermodynamic integration. Zheng and Thompson found that melting of the perfect crystal occurs at 510 K and is preceded by a subtle solid-solid phase transition near 490 K. It is possible that their observed solid-solid transition bears some connection to the formation of a pseudo-liquid state in our simulations, but it was not possible to verify this based on the available information. The thermodynamic melting point determined by Sellers et al. was 488.75 K, overestimating experiment by approximately 2%. Thus, roughly half the error in the predicted melting point obtained via phase coexistence derives from fundamental errors in the FF description at 0 GPa, with the remainder arising from uncertainty due to coarseness in temperature sampling and the kinetics of melting.</p><p>The melting temperature is clearly a strong function of temperature and increases by nearly 1100 K as the pressure varies from 0 to 30 GPa. Similar to HMX and TATB, the predicted melt curve of RDX is well described by the empirical Simon-Glatzel equation [75],</p><p>This form relates the melting point 𝑇 𝑚,𝑟𝑒𝑓 at a reference pressure 𝑃 𝑟𝑒𝑓 with 𝑎 and 𝑐 as fitting parameters.</p><p>Experimental values are taken as the reference point (𝑇 𝑚,𝑟𝑒𝑓 = 478 K, 𝑃 𝑟𝑒𝑓 = 0.0001 GPa) and the parameters were fit to minimize the root-mean-square error relative to the MD predictions for 𝑃 ≥ 1 GPa . The best-fit parameters are 𝑎 = 0.9631 GPa and 𝑐 = 2.8855 (unitless). Figure 3 shows a comparison of the MD predictions and Simon-Glatzel fit for the RDX melt curve to analogous MD-based Simon-Glatzel fits for the related explosives HMX [24] and TATB [21]. The melting temperature of RDX is substantially lower than either HMX or TATB and the differences generally increase with increasing pressure. The relative ordering of the predicted melting points for these explosives is consistent with the limited available experimental data, which gives 𝑇 𝑚 𝑒𝑥𝑝𝑡 values of 478, 551, and 723 K for RDX [49][50][51], HMX [49], and TATB [76] at atmospheric pressure. Predictions for the HMX melt curve were obtained using the same FF used here for RDX, with the only exception being that the C-H bond vibrations were held rigid in the MD simulations of HMX. Thus, differences between those two sets of predictions can be reasonably attributed to the increased molecular weight of HMX and the differing crystal packing structures.</p><!><p>Dynamic shear viscosity of liquid RDX was determined as a function of temperature and pressure using NVT MD simulations and the Green-Kubo formalism [60]. The Green-Kubo formalism relates the autocorrelation of equilibrium shear stress fluctuations to the shear viscosity as</p><p>Here, 𝜎 𝛼𝛽 (𝑡) denotes the off-diagonal components of the stress tensor as a function of time, 〈𝑓〉 denotes a time average of the stress autocorrelation function, 𝑘 𝐵 is the Boltzmann constant, and 𝑉(𝑇, 𝑃) is the (constant) volume of the simulation cell that is determined by the equilibrium density at the specific (𝑇, 𝑃) state being simulated. Equilibrium 𝑉(𝑇, 𝑃) values were determined as the average over the last 250 ps of a 500 ps NPT trajectory. All viscosity calculation simulations were performed using cubic, 3D-periodic simulation cells containing 512 molecules. We closely follow earlier work that determined the shear viscosity of HMX [24] and TATB [21], and so only outline practical considerations and extensions of the approach used to evaluate Equation 2. Three timescales arise including the period for sampling the stresses ( 𝑡 𝑠𝑎𝑚𝑝𝑙𝑒 ), the maximum time lag of the autocorrelation function (𝑡 𝑏𝑙𝑜𝑐𝑘 ) that sets the upper limit of the integral, and the total simulation time ( 𝑡 𝑠𝑖𝑚 ). These must be respectively chosen to adequately sample stress fluctuations, be long enough for the autocorrelation to decay sufficiently close to zero and be long enough to obtain a good time average of the autocorrelation function. The fundamental sampling period was set to 𝑡 𝑠𝑎𝑚𝑝𝑙𝑒 = 1 fs, but down sampling to 2 fs was found to yield essentially identical results and was used for the final analysis to improve the efficiency of computing the stress autocorrelation function. The last two timescales are coupled as the cumulative trajectory is split into 𝑁 contiguous time blocks of length 𝑡 𝑏𝑙𝑜𝑐𝑘 = 𝑡 𝑠𝑖𝑚 /𝑁.</p><p>Convergence of 𝜂(𝑇, 𝑃) is particularly sensitive to the upper limit of the integral, which we take to be 𝑡 𝑏𝑙𝑜𝑐𝑘 /2. An example convergence study to determine an appropriate value of 𝑡 𝑏𝑙𝑜𝑐𝑘 for the case T = 1000 K, P = 0 GPa is given in Figure 4. Panel (a) shows how the viscosity converges as a function of simulation time for selected values of 𝑡 𝑏𝑙𝑜𝑐𝑘 . Panel (b) shows average viscosity values as a function of block size computed over the second half of the trajectory, with the standard deviation taken as the uncertainty. Too short of 𝑡 𝑏𝑙𝑜𝑐𝑘 yields a precise value that underestimates the true viscosity. In contrast, large 𝑡 𝑏𝑙𝑜𝑐𝑘 ≥ 40 ps yield average values that are the same within uncertainty, but that uncertainty increases with block size as the number of samples 𝑁 decreases. We identify the best 𝑡 𝑏𝑙𝑜𝑐𝑘 value for each state point as the value for which doubling the Very long trajectories were needed for adequate sampling that approach hundreds of nanoseconds. We adopted an ensemble approach wherein many (typically 10-20) independent shorter trajectories were integrated for a single state point to obtain a large cumulative 𝑡 𝑠𝑖𝑚 . Each trajectory was broken into contiguous blocks based on the maximum block size 𝑡 𝑏𝑙𝑜𝑐𝑘,𝑚𝑎𝑥 = 4000 ps . An ensemble time history was assembled in which these 4000 ps blocks were intercalated in a round-robin fashion. All smaller 𝑡 𝑏𝑙𝑜𝑐𝑘 values considered were a common divisor of 𝑡 𝑏𝑙𝑜𝑐𝑘,𝑚𝑎𝑥 . Decorrelated starting points for each simulation in an ensemble were spawned from an NVT simulation performed using a (stochastic) Langevin thermostat [77,78] with time constant 100 fs. Starting configurations were dumped every 100 ps. Note that the production NVT simulations used for evaluating 𝜂(𝑇, 𝑃) were performed using a Nosé-Hoover-style thermostat as stochastic thermostats can strongly alter dynamic properties such as transport coefficients [79].</p><!><p>The shear viscosity of liquid RDX, 𝜂(𝑇, 𝑃), was obtained as a function of temperature and pressure using NVT simulations and the Green-Kubo relation at state points above the predicted melt curve for pressures of 0, 1, 3, 5, 10, 20, and 30 GPa. Equilibrium densities 𝜌(𝑇, 𝑃) used to prepare the NVT simulation cells and predicted 𝜂(𝑇, 𝑃) values are tabulated in the Supporting Information. The total simulation time (𝑡 𝑠𝑖𝑚 ) and block size (𝑡 𝑏𝑙𝑜𝑐𝑘 ) needed to converge (or partially converge) the viscosity is also tabulated. We considered a maximum 𝑡 𝑠𝑖𝑚,𝑚𝑎𝑥 = 160 ns and a maximum 𝑡 𝑏𝑙𝑜𝑐𝑘,𝑚𝑎𝑥 = 4000 ps, which corresponds to a minimum of 𝑁 = 40 samples in the time average of the stress autocorrelation function. It was not possible to fully converge the viscosity for a small number of states near the melt curve, but partially converged viscosity values are included in plots below as they correspond to a reasonable lower bound. These partially converged states are specifically noted in the Supporting Information.</p><p>The shear viscosity is expected to depend exponentially on temperature for a given pressure, following an Arrhenius-like form known as the Andrade equation [80],</p><p>Weighted least-squares fits to Equation 3 were performed to obtain the pre-exponential factor and activation term and are listed in Table 2. While the form is Arrhenius, there is no conclusive physical interpretation of the activation term, nor is there a universal pressure-temperature dependent form for the viscosity of liquids [34]. Similar weighted fits to the Andrade equation were performed on literature values for HMX and TATB. Data for TATB was reported up to 2 GPa and is thus only shown for comparison at the two lowest pressures [21]. The viscosity of HMX at 0 GPa was determined by Bedrov et al. [15] and the values for 1, 3, and 5 GPa were reported in Ref. [24]. Note that the 0 GPa HMX viscosity values did not have uncertainties, so unweighted fits were performed in that case. Both sets of HMX data were determined using the Smith-Bharadwaj FF, although there were several differences in the FF implementation and analysis that are discussed below. Focusing first on the low-pressure response, Figure 5 shows comparisons of the present predictions for RDX to MD predictions for the related explosives HMX and TATB at pressures of 0, 1, 3, and 5 GPa. Several trends are immediately apparent. The viscosities of HMX and RDX exhibit very similar slopes at 1, 3, and 5 GPa, which is matched by their similar activation terms ( 𝐸 𝑎 ). HMX viscosity is approximately one order of magnitude larger than RDX at any given temperature and pressure. In contrast, HMX and RDX have dissimilar slopes at 0 GPa and the activation terms differ by a factor of two. The 0 GPa HMX values were obtained using the Smith-Bharadwaj FF description in which all bonds were held rigid, unlike the higher pressure HMX values in which only C-H bonds were held rigid. An empirical correction factor from the diagonal stress components [81] was also included in the 0 GPa analysis to improve the rate of convergence that was not used for any of the other cases. Both of these differences were previously noted [24], but are more stark when contrasted against the present RDX data that were obtained with fully flexible molecules.</p><p>Comparison of the RDX data across all four pressures and the HMX data above 0 GPa show a general similarity in the activation term, indicating that the earlier 0 GPa data for HMX is an outlier. The determination of RDX viscosity at 0 GPa and 550 K by Izvekov and Rice [7] (open diamond symbol) is in very good agreement with the Andrade fit to our 0 GPa predictions, which serves as an independent validation of our simulation and analysis procedure.</p><p>Comparison against TATB shows similar (if slightly larger) values for the viscosity of TATB compared to RDX, but the pressure dependence is clearly different. At 0 GPa, the magnitude of the activation term (slope) for TATB is larger than for RDX, but the opposite is true at 1 GPa. It is perhaps surprising RDX viscosity is closer to TATB than to HMX, which is more closely related to RDX both in terms of molecular shape and specific chemistry. As alluded to above, there is no general form for the pressure or temperature dependence of liquid shear viscosity. Exponential forms can work well particularly within narrow temperature or pressure intervals [34,35]. These include the Andrade equation, which depends on the exponent of inverse temperature, and a pressure form that is exponential in pressure modulated by the Barus pressure-viscosity coefficient. Combined pressure-temperature exponential forms have been applied to explosive viscosity in the past [33,16], but these forms were not able to reasonably capture the variation of RDX viscosity across the 30 GPa pressure interval considered. Similarly poor descriptions were obtained with the empirical form used previously for HMX up to 5 GPa [24]. (This form was an Andrade equation with power laws for a pressure-dependent prefactor and activation term.) More complex alternatives have been proposed [34,35], although these generally do not have a firm basis in theory.</p><p>Given the difficulty in finding a suitable functional form for the pressure-temperature dependent shear viscosity predictions, we took an empirical approach to arrive at a representation based on the Andrade equation that extrapolates smoothly beyond the states The parameters A, B, C, D, and E were obtained by a global least-squares fit to all the MD data and are collected in Table 3. Plots of the fitted pressuredependent slope (Equation 5) and intercept (Equation 6) are shown in Figure 5. Comparison against the data obtained from independent fits shows that both these forms reasonably capture the MD-predicted dependencies, including the rapid increase and plateau of the intercept with increasing pressure.</p><p>Figure 7 shows the optimized fit of the pressuredependent Andrade equation (Equation 4) for the 𝜂(𝑇, 𝑃) surface of RDX. Inspection of the line plots in panel (a) shows that the MD data for each pressure series is well-represented by the empirical form. Simultaneous changes in slope and intercept are accommodated and monotonic convergence of the intercept (pre-factor) to a constant value is clear. As expected from the functional form, the global surface in panel (b) shows smooth variation. Note that the function is not meaningful in the solid region of the diagram. The (undetermined) boundary between the liquid and vapor phase will also fall on this plot, although this boundary is likely of little consequence in reactive simulations as RDX would almost certainly react before vaporizing.</p><!><p>Accurate grain-scale models for simulating hot spot formation and growth depend on accurate determinations of basic thermodynamic, mechanical, thermal, and chemical material properties. Two of these terms, namely the pressure-dependent melt curve and liquid-phase shear viscosity, have been shown to exhibit a strong influence on predicted hot spot temperatures for HMX and TATB. This motivates an assessment of the RDX melting point, which is poorly constrained above ambient pressure, and the shear viscosity, which has not been characterized as a function of temperature or pressure. We use classical, non-reactive molecular dynamics (MD) simulations and an established force field (FF) model for RDX to predict the melt curve up to 30 GPa and determine the temperature-pressure dependent shear viscosity.</p><p>Predictions for the melting point of RDX were obtained using the phase-coexistence approach. Results obtained here for the melting point at ambient pressure (1 atm) are within 5% of experimental values and are similar to two other determinations made using the same FF but with different MD methods. The melt curve for RDX exhibits a significant pressure dependence, increasing by nearly 1100 K as the pressure increases from 0 to 30 GPa. The melting point of RDX is lower than the melting points of either HMX or TATB at a given pressure.</p><p>Dynamic shear viscosity was determined as a function of temperature at seven pressures ranging from 0 to 30 GPa using equilibrium MD simulations and the Green-Kubo formalism. The temperature dependence at a given pressure exhibits Arrhenius-like behavior and is well-described by the Andrade equation. Comparison against similar results for liquid HMX and TATB at low GPa-range pressures (≤5 GPa) shows that the viscosity of RDX is similar in magnitude to TATB and roughly an order of magnitude lower than that of HMX. The exponential activation terms for RDX and HMX are similar at these pressures, with the difference in magnitude arising due to the pre-exponential factor. Both the activation term and pre-exponential factor exhibit a pressure dependence, which informs an empirical extension of the Andrade equation to obtain a global analytic surface for the viscosity as a function of temperature and pressure. This global function simultaneously captures a wide range of MD data points while taking a form that can be readily incorporated in grain scale models.</p><p>The relative strength of the pressuretemperature functional dependencies of the RDX melt curve and shear viscosity are similar in magnitude to recent determinations for HMX [24] and TATB [21]. Thus, it can be expected that grain scale simulation predictions of hot spot formation and growth in RDX will exhibit similar sensitivity as was found for those explosives [28,29,24]. Both the melt curve and shear viscosity were shown to independently alter peak hot spot temperatures by hundreds of Kelvin in HMX and TATB, which provides strong motivation for including the present results in future grain scale models of RDX.</p><!><p>Non-Coulombic pairwise interactions in the Smith-Bharadwaj force field (FF) [1] were modeled using the Buckingham potential (exp-6), which diverges to minus infinity at very small atom-atom separation distances. This potential has the form 𝑈 𝑜𝑙𝑑 (𝑟) = 𝐴 • 𝑒𝑥𝑝[−𝐵𝑟] − 𝐶 𝑟 6 . (S1)</p><p>Even very infrequent sampling of close contacts at high temperature-pressure conditions can result in simulation instability. We adopted an approach used by others [2,3] in which an additional short-ranged repulsive potential was added to compensate for the divergence at short separation distance. This which includes an additional r -12 factor. The parameter D was set to 5 × 10 −5 kcal/mol for all pair interaction types and was the same value used for a similar FF for TATB [2]. We verified that this modified FF yielded the same equilibrium density for the (1500 K, 10 GPa) state point where both FF versions could be stably applied.</p><p>Figure S1 shows additional validation of the radial distribution functions (RDFs) for each pair interaction type at the (1500 K, 10 GPa) state point that were obtained from the last 1 ns of a 2 ns NVT simulation. The two FF versions predict essentially identical RDFs for each pair type, indicating that the potential of mean force for each interaction type is unperturbed by the additional repulsive potential. The largest difference was seen for the first peak in the N(ring)-C interaction type, which may arise due to a very subtle distortion of the ring.</p>

ChemRxiv

Entrapment of Hydrophobic Drugs in Nanoparticle Monolayers with Efficient Release into Cancer Cells

Gold nanoparticles functionalized with water-soluble zwitterionic ligands form kinetically stable complexes with hydrophobic drugs and dyes. These drugs and dyes are efficiently released into cells, as demonstrated through fluorescence microscopy and cytotoxicity assays. Significantly, there is little or no cellular uptake of particle, making these low toxicity particles promising for delivery applications.

entrapment_of_hydrophobic_drugs_in_nanoparticle_monolayers_with_efficient_release_into_cancer_cells

<!>Supporting Information Available<!>

<p>Drug delivery systems (DDSs) provide an important tool for increasing efficacy of pharmaceuticals through improved pharmacokinetics and biodistribution.1 A wide variety of nanoscale materials such as liposomes, polymeric micelles, and dendrimers, have been employed as drug carriers. 2 Both covalent and non-covalent approaches can be applied to the conjugation of drugs into/onto these DDSs.3 Non-covalent approaches have the capability of employing active drugs, whereas covalent attachment generally requires chemical modification which can cause reduced efficiency of drug release or incomplete intracellular processing of a prodrug.4</p><p>Recently, gold nanoparticle (AuNP) based drug/gene delivery systems have attracted attention due to their functional versatility,5 biocompatibility,6 and low toxicity,7 Recent studies have demonstrated controlled release of payload by intracellular thiols.8 However, controlled dissociation of drugs in active form from covalent AuNP-drug conjugates remains a challenge for clinical applications.2</p><p>Noncovalent incorporation of drugs into AuNP monolayers provides an alternative delivery strategy with the potential for avoiding drug release and prodrug processing issues. The structure of commonly used water-soluble AuNPs is similar to that of unimolecular micelles such as dendrimers, featuring a hydrophobic interior and a hydrophilic exterior.9 The alkanethiol monolayer of the nanoparticle coupled with the radial nature of the ligands10 creates "hydrophobic pockets" inside monolayer of AuNP where organic solutes can be partitioned, as demonstrated by Lucarini and Pasquato.11 We report here the use of these pockets to encapsulate drugs and deliver them with high efficiency to cells.</p><p>The biocompatible AuNPs used in this study features two functional domains: a hydrophobic alkanethiol interior and a hydrophilic shell composed of a tetraethylene glycol (TEG) unit terminated with a zwitterionic headgroup. Particles with this general structure have been shown to minimize non-specific binding with biomacromolecules.12</p><p>We chose three different hydrophobic guest compounds: 4,4-difluoro-4-bora-3a,4a-diaza-s-indacene (Bodipy) as a fluorescent probe,13 and the highly hydrophobic therapeutics tamoxifen (TAF) and β-lapachone (LAP) as drugs (Figure 1). The nanoparticle-payload conjugates (AuNPZwit-Bodipy, TAF, and LAP) were prepared by solvent displacement method.14 First, AuNPZwit (Au core: 2.5 ± 0.4 nm) and guest were dissolved in an acetone/water and the solvent slowly evaporated. The bulk of the excess guest precipitated out and was removed by filtration; the particles were further purified by multiple filtrations through a molecular weight cutoff filter until no free guest was observed, followed by dialysis against buffer. The number of entrapped guest molecules per particle was determined from 1H NMR spectrum and NaCN-induced decomposition experiments (see Supporting Information) and varied depending on size, hydrophobicity (logP), and molecular structure of hydrophobic molecules (Figure 1). The particle/guest complexes are stable in buffer for >1 month and to extended dialysis, a level of kinetic entrapment greater than that observed with dendrimers.4</p><p>The ability of the delivery systems to release their payload was first explored in vitro using AuNPZwit-Bodipy in a two-phase dichloromethane (DCM)-water system, 15 where the dye is quenched by the AuNP, and photoluminescence (PL) only observed upon dye release. In these studies a rapid increase in PL intensity is observed along with transfer of Bodipy into the DCM layer (Figure 1c). Significantly, since no release is observed in monophasic aqueous conditions and no particle was observed in the DCM layer, payload release occurs interfacially.</p><p>Payload delivery to cells using AuNPZwit-Bodipy was determined by confocal laser scanning microscopy (CLSM) using human breast cancer (MCF-7) cells. Efficient delivery of the dye to the cytosol is observed after 2 h incubation with AuNPZwit-Bodipy (Figure 2a–c). Cellular uptake of nanoparticle was studied using transmission electron microscopy (TEM), and inductively coupled plasma mass spectrometry (ICP-MS), using the analogous cationic particle/dye conjugate AuNPTTMA-Bodipy as a positive control. Little cellular uptake of AuNPZwit was observed by either TEM (Figure 2 d,e) or ICP-MS for AuNPZwit-Bodipy (31 ng/well at 4h (Figures 2), 71 ng/well at 24h Figure S3, corresponding to uptake of 0.06% and 0.14% of available particle, respectively), whereas substantial particle uptake was observed with AuNPTTMA-Bodipy (1750 ng/well (4 h), 2150 ng/well (24 h)) Since no free dye was observed during the 24 h incubation of AuNPZwit-Bodipy in medium or serum solution at 37 °C (Figure 1 d), Bodipy delivery presumably occurs via a monolayer-membrane transfer process, consistent with our in vitro studies16</p><p>Demonstration of drug delivery to MCF-7 cells through presumably the same mechanism was determined through cytotoxicity studies of free and encapsulated drugs using an Alamar blue assay (Figure 3). Notably, AuNPZwit itself was non-toxic at 30 μM. In contrast, IC50 values of 4 μM and 4.6 μM were observed using AuNPZwit-LAP and AuNPZwit-TAF, respectively. The delivery process was quite efficient, with the per drug molecule IC50 of AuNPZwit-TAF (46 μM) only three-fold higher than that of TAF (16 μM), and that of AuNPZwit-LAP (6.0 μM) essentially identical to that of LAP (5.2 μM).</p><p>In conclusion, we have demonstrated that hydrophobic dyes/drugs can be stably entrapped in hydrophobic pocket of AuNPs and released into cell by membrane-mediated diffusion without uptake of the carrier nanoparticle. Importantly, the small size of these nanocarriers coupled with their biocompatible surface functionality should provide long circulation lifetimes and preferential accumulation in tumor tissues by the enhanced permeability and retention (EPR) effect.17 Additionally, the noninteracting nature of their monolayer should make these systems highly amenable to targeting strategies. We are currently exploring these applications as well as the role of monolayer and guest structure in the encapsulation process.</p><!><p>Experimental procedures, synthesis of gold nanoparticles and complexes, and TEM of gold nanoparticles. This information is available free of charge via the Internet at http://pubs.acs.org.</p><!><p>a) Delivery of payload to cell through monolayer-membrane interactions. b) Structure of particles and guest compounds: Bodipy, TAF, and LAP, the number of encapsulated guests per particle, and logP of the guests. c) Release of Bodipy from AuNPZwit-Bodipy in DCM-aqueous solution two-phase systems (λex = 499 nm, λem = 517 nm) d) PL intensity AuNPZwit-Bodipy in cell culture medium and 100 % serum, indicating little or no release relative to AuNPZwit-Bodipy in PBS after NaCN-induced release of guest molecules ((λex = 499 nm, λem = 510 nm).</p><p>CLSM images of MCF-7 cell treated with AuNPZwit-Bodipy for 2h: a) green channel b) bright field, and c) overlap. TEM images of fixed cell treated with d) AuNPZwit-Bodipy and e) AuNPTTMA as a positive control, Endosomally trapped AuNPs are marked by arrow. (f) ICP-MS measurement. (200,000 cells/well), indicating low cellular uptake of AuNPZwit (31 ng/well after 4 h)</p><p>Cytotoxicity of AuNPZwit complexes measured by Alamar blue assay after 24h incubation with MCF-7 cells. IC50 of AuNP (NP), equivalent drugs (Drug), and free drugs are shown in table.</p>

PubMed Author Manuscript

THE \xce\xb2-CATENIN/TCF4/SURVIVIN SIGNALING MAINTAINS A LESS DIFFERENTIATED PHENOTYPE AND HIGH PROLIFERATIVE CAPACITY OF HUMAN CORNEAL EPITHELIAL PROGENITOR CELLS

It is clear that the microenvironment or niche plays an important role in determining the fate of stem cells: being stem cells or differentiated. However, the intrinsic pathways controlling the fate of adult stem cells in different niches are largely unknown. This study was to explore the role of \xce\xb2-catenin/Tcf4/survivin signaling in determining the fate of human corneal epithelial stem cells in different media. We observed that the low calcium serum-free media, especially CnT-20, promoted proliferative capacity, colony forming efficiency and stem cell-like phenotype of human corneal epithelial cells (HCECs) when compared with the cells cultured in a high calcium serum-containing medium SHEM. Three key factors in Wnt signaling, \xce\xb2-catenin, Tcf4 and survivin, were found to be expressed higher by HCECs grown in CnT-20 than those cultured in SHEM, as evaluated by real-time PCR, Western blotting and immunostaining. Transfection of siRNA-Tcf4 at 10-50 nM knocked down Tcf4, and also significantly suppressed its down stream molecule survivin at both mRNA and protein levels in HCECs. Furthermore, Tcf4 silencing significantly suppressed the proliferative capacity of HCECs, measured by WST-1 assay, compared with the control groups, untreated or transfected with non-coding sequence siRNA-fluorescein. These findings demonstrate that low calcium serum free media promote ex vivo expansion of corneal epithelial progenitor cells that retain a less differentiated phenotype and high proliferative capacity via \xce\xb2-catenin/Tcf4/survivin signaling, a novel intrinsic pathway. This study may have high impact and clinic implication on the expansion of corneal epithelial stem cells in regenerative medicine, especially for ocular surface reconstruction.

the_\xce\xb2-catenin/tcf4/survivin_signaling_maintains_a_less_differentiated_phenotype_and_high_prol

INTRODUCTION<!>Materials and reagents<!>Human corneal epithelial cell (HCEC) cultures in different media<!>Colony forming efficiency (CFE) and growth capacity<!>RNA interference<!>WST-1 cell proliferation assay for Tcf4 knock-down HCECs<!>Immunofluorescent staining and laser scanning confocal microscopy (LSCM)<!>Total RNA Extraction, Reverse transcription (RT), and Quantitative Real-Time PCR<!>Western Blotting Assay<!>Statistical analysis<!>Proliferative capacity of human corneal epithelial cells promoted by low calcium serum free media<!>Progenitor phenotype of human corneal epithelial cells cultured in low calcium serum free media<!>Expression of \xce\xb2-catenin, Tcf4 and survivin in human corneal epithelial progenitor cells<!>Functional role of Tcf4 signaling in maintaining the proliferative property of human corneal epithelial progenitor cells<!>DISCUSSION

<p>The ocular surface is an ideal region to study epithelial stem cell biology because of the unique spatial arrangement of stem cells and transient amplifying cells [1-4]. The corneal epithelial stem cells have been identified to reside in the basal layer of limbal epithelium over last two decades. Limbal epithelial stem cells exhibit unique characteristics that satisfy the widely accepted criteria for defining adult stem cells, including (1) slow cycling or long cell cycle time during homeostasis in vivo; (2) small size and poor differentiation with primitive cytoplasm; (3) high proliferative potential after wounding or placement in culture; (4) ability for self-renewal and functional tissue regeneration (see review articles by[5-8]). Both intrinsic and extrinsic signals regulate stem cell fate including adult stem cells. Through interaction with intrinsic signals, the extrinsic niche or the stem cell microenvironment is believed to be important in maintaining the "stemness" of the stem cells, including corneal epithelial stem cells [9-12]. For example, it is well known that low calcium, serum-free culture media can provide an ideal niche in vitro to maintain or promote progenitor cell properties, such as proliferative capacity and undifferentiation status [13-16], while high calcium and serum-containing media promote cell differentiation [17-19]. However, the underlining molecular mechanisms by which the niche determines the stem cell fate are far from being completely elucidated.</p><p>Wnt signaling pathway has been recognized to control a variety of functions and properties in various types of stem cells. Wnt signaling can be activated by niche factors to maintain stem cells in a self-renewing state [20-22]. During tissue development and regeneration, Wnt signals ensure the proper balance between proliferation and differentiation [23-25]. Wnt proteins are active in a variety of stem cells, including embryonic, hematopoietic, neural and mammary stem cells, as well as corneal epithelial stem cells [20, 26, 27]. The hallmark of the Wnt signaling pathway is the accumulation of the junctional protein β-catenin in the cytoplasm, which then translocates to the nucleus to trigger the β-catenin/Tcf enhancer factor transcriptional machinery, and upregulate target genes, such as survivin and c-myc [28-30]. A classic example of the importance of this pathway is in the digestive tract, where in the crypt of the colon the loss of transcription factor T cell factor 4 (Tcf4), a key factor of canonical Wnt signaling pathway, leads to depletion of stem cells [30, 31]. After activation by β-catenin/Tcf4 complex, survivin enhances cell proliferation while protecting cells from apoptosis [32, 33]. Recently, Tcf4 and Tcf3 have been found to play a vital role in long-term maintenance and wound repair of both epidermis and hair follicles [34]. However, the role of the Wnt pathway, particularly, β-catenin/Tcf4/survivin signaling in maintaining the properties of adult stem cells has not been elucidated. The purpose of present study was to explore the important role of Tcf4 signaling in determining the fate of corneal epithelial stem cells, using an in vitro culture model with different media providing niche factors: low calcium and serum free versus high calcium and serum containing.</p><!><p>Cell culture dishes, plates, centrifuge tubes, and other plastic ware were purchased from Becton Dickinson and Company (Franklin Lakes, NJ). Nunc Lab-Tec II eight-chamber slides were from Nalge Nunc International Corp (Naperville, IL). Fetal bovine serum (FBS) was from Hyclone (Logan, UT). CnT-20 and CnT-50 progenitor media were from Chemicon International (Temecula, CA). Dulbecco modified Eagle's medium (DMEM), Ham F-12, Keratinocyte-SFM (KSFM) and Defined KSFM (D-KSFM), amphotericin B, gentamicin, 0.25% trypsin/EDTA solution, mouse monoclonal antibody (mAb) against connexin 43 (Cx43), and fluorescein Alexa-Fluor 488 conjugated secondary antibodies (Donkey anti-Goat IgG, Goat anti-rabbit or Goat anti-mouse IgG) were from Invitrogen Corp (Carlsbad, CA). Human AE5/keratin (K) 3 mAb and goat antibodies against human Tcf4 and survivin were from Santa Cruz Biotechnology, Inc (Santa Cruz, CA). Rabbit antibodies against β-catenin and β-actin were from Cell Signaling Technology (Beverly, MA). Human p63 (4A4), integrin β1 and EGFR mAbs were from Lab Vision (Fremont, CA). HRP conjugated secondary antibodies (goat anti-mouse, goat anti-rabbit and rabbit anti-goat for western blot were from Thermo Scientific (Fremont, CA). Ready gels, Precision Plus Protein Unstained Standards and Precision Protein Streptactin-AP Conjugate came from Bio-Rad (Hercules, CA). The BCA protein assay kit was from Pierce Chemical (Rockford, IL). RNeasy® Mini kit, siRNA-F and HiperFect transfection reagent were from Qiagen (Valencia, CA). Ready-To-Go You-Prime First-Strand Beads were from GE Healthcare (Piscataway, NJ). TaqMan® Gene Expression Assay, real-time PCR Master Mix, and Silencer Select® pre-designed small interfering RNA (siRNA) were from Applied Biosystems (Foster City, CA). WST-1 proliferative assay kit was from Roche Molecular Biochemicals (Mannheim, Germany). Human insulin, transferrin, sodium selenite, hydrocortisone, epidermal growth factor (EGF), cholera toxin A subunit, propidium iodide (PI) and all other reagents came from Sigma-Aldrich (St. Louis, MO).</p><!><p>Primary HCECs were cultured from donors' limbal tissue explants using a previously described protocol [35, 36]. In brief, the limbal ring is cut into 12-16 pieces with similar size of approximately 2×2mm each. Two pieces with the epithelium side up were directly put into a well of 6-well plate or one piece per chamber in 8-chamber slides. Low calcium and serum free progenitor cell culture media, CnT-20, CnT-50, KSFM and D-KSFM, and high calcium serum-containing supplemental hormonal epithelial medium (SHEM) were used for cultures at 37°C under 5% CO2 and 95% humidity. The media were changed every 2-3 days.</p><p>Corneal epithelial cell growth was carefully observed and photographed through a Nikon TE200 inverted phase microscope with a Nikon DXM1200 digital camera. Only the epithelial cultures without visible fibroblast contamination were used for this study. When grown to 90% confluence, the cultures were photographed and trypsinized with 0.25%trypsin/0.03%EDTA; and the cells were seeded into a new plate at a density of 2×104 cells/cm2 for serial passages.</p><!><p>To evaluate proliferative capacity of corneal epithelial cells in different culture media, the CFE was assessed in cultures in CnT-20 or SHEM using a previous method [37-39] with modification. Primary human corneal epithelial cells were seeded in triplicate at 500 cells/cm2 into six-well culture plates without 3T3 fibroblasts or any other cells as a feeder layer. Colonies with more than eight viable cells were counted manually under an inverted phase microscopy at days 6 and 8. Experiment was repeated at least three times. The CFE in SHEM or CnT-20 was calculated as a percentage of the number of colonies generated by the number of epithelial cells plated in a well. The growth capacity was evaluated on day 14 when cultured cells were stained with 1% rhodamine.</p><!><p>To explore the functional role of Tcf4 signaling, RNA interference experiments were performed using our previous method [40, 41] with modification. In brief, primary HCECs at a density of 5×104 cells/cm2 were transfected with annealed double-stranded siRNA specific for Tcf4 (siRNA-Tcf4, ID. s13863 containing a pool of 3 target-specific 20-25 nt siRNAs) at different concentrations (10nM, 25nM, 50nM), with a non-coding sequence siRNA-fluorescein (siRNA-F, UUCUCCGAACGUGUCACGU) as a negative control (also serve as visible monitor for transfection efficiency) using fast-forward transfection method with HiperFect reagent according to a manufacturer's protocol. The transfection efficiency in HCECs with different concentrations of siRNA-F (10, 25 and 50 nM) after 24 hours were 81.4±3.5%, 83.5±4.1% and 87.2±4.3%, respectively, as analyzed by flow cytometry. After incubation for additional 24-72 hours, the cells were collected for RNA extraction or protein lysate preparation for further evaluation. The cell viability after transfection for 4 days was more than 90%, as assessed by a 0.2% trypan blue exclusion test and morphological observation.</p><!><p>Primary HCECs were transfected by siRNA-Tcf4 at final concentrations at 10nM, 25nM and 50nM, with a non-coding sequence siRNA-fluorescein (siRNA-F) as a negative control, using HiperFect reagent according to a manufacturer's protocol. In brief, siRNA-Tcf4 mixed with HiPerFect transfection reagent was spotted in the wells of 96-well plate, and incubated for 5-10 min at room temperature. Primary HCECs in CnT-20 were seeded at density of 6,000/well on the top of the siRNA–Hiperfect reagent complex and cultured for 48 hours. WST-1 proliferative assay [42, 43] was assessed according to the manufacturer's protocol. Briefly, 10μl of cell proliferation agent WST-1 was added to each well containing 100 μl cell culture. The cells were incubated for additional 2 hours at 37°C in a 5% CO2 atmosphere. The plate was measured at 450nm with a reference wavelength 690 nm in an Infinite M200 multimode microplate reader (Tecan US, Durham NC). The experiment for cell proliferation assays were repeated 5 times.</p><!><p>Immunofluorescent staining was performed following a previously reported method [44]. In brief, cells were fixed with methanol at 4°C for 10 minutes and permeabilized with 0.2% Triton-X in PBS for 10 minutes. After blocked with 20% of animal serum in PBS for 30 minutes, a primary antibody was applied and incubated for two hours at room temperature. Alexa Fluor 488 conjugated secondary antibody was then applied for one hour in dark, followed by counterstaining with a DNA binding dye propidium iodide (PI, 1μg/mL in PBS) for 5 minutes. A cover slip was applied with Antifade Gel/Mount (Fisher, Atlanta, GA). Slides were examined and photographed with the LSCM (LSM 510, Zeiss, Thornwood, NY).</p><!><p>Total RNA was extracted from HCECs using a Qiagen RNeasy® Mini kit according to manufacturer's protocol, quantified by NanoDrop® ND-1000 Spectrophotometer, and stored at −80 °C. The first strand cDNA was synthesized by RT from 1μg of total RNA using Ready-To-Go You-Prime First-Strand Beads, and the real-time PCR was performed in the Mx3005PTM system (Stratagene) [45, 46]. TaqMan® Gene Expression Assays include β-catenin (Assay ID Hs99999168_m1), Tcf4 (Hs00162613_m1), survivin (Hs00153353_m1), p63 (Hs00186613_m1), EGFR (Hs00193306_m1), integrin β1 (Hs00236976_m1), Cx43 (Hs00748445_s1), K3 (Hs00365080_m1) and GAPDH (Hs99999905_m1). The thermocycler parameters were 50 °C for 2min, 95 °C for 10 min, followed by 40 cycles of 95 °C for 15 s and 60 °C for 1min. A non-template control was included to evaluate DNA contamination. The results were analyzed by the comparative threshold cycle (CT) method and normalized by a housekeeping gene GAPDH [47, 48].</p><!><p>Western blot analysis was performed by a previous method [49, 50]. In brief, HCECs cultured in CnT-20 or SHEM were lysed with RIPA buffer. Equal amount (30μg/well) of cell extract protein, measured by a BCA protein assay kit, was mixed with 6×SDS reducing sample buffer and boiled for 5 minutes before loading. The proteins were separated on an SDS polyacrylamide gel and transferred electronically to PVDF membranes. The membranes were blocked with 5% nonfat milk in TTBS (50 mM Tris [pH 7.5], 0.9% NaCl, and 0.1% Tween-20) for 1 hour at room temperature and incubated with primary antibodies to β-cantenin (1:1000), Tcf4 (1:100), survivin (1:100) and β-actin (1:2000) at 4°C overnight. After three times washes with TTBS, the membranes were incubated with HRP conjugated rabbit anti-goat, goat anti-mouse or goat anti-rabbit IgG (1:2000) for 1h at room temperature. The signal bands were detected with an ECL Plus chemiluminescence reagent (GE Healthcare), and the images were acquired by a Kodak Imaging Station 2000R (Eastman Kodak, New Haven CT).</p><!><p>The Student's t-test or analysis of variance (ANOVA) with Tukey's post-hoc testing was used for statistical comparisons. p≤ 0.05 was considered statistically significant. All of these tests were performed using the GraphPad Prism 5.0 software (Graph-Pad Prism, Inc., San Diego, CA, http://www.graphpad.com).</p><!><p>Primary HCECs were established from limbal explants using five different culture media, four low calcium and serum free progenitor cell culture media (CnT-20, CnT-50, KSFM and D-KSFM) and a high calcium serum-containing SHEM. All media were capable of supporting primary epithelial cell growth. When cells grew to subconfluent (80-90% confluence), serial passages were performed. As shown in Fig. 1. HCECs grown in SHEM, a high calcium and serum containing medium, could not be passaged more than once or twice although SHEM supported rapid and abundant cell growth in the primary culture. In contrast, higher proliferative capacity was observed in HCECs cultured in all low calcium and serum free media. For examples, cells cultured in fully defined CnT-20 were capable of being passaged six or more times, and cells grown in CnT-50 that contains bovine pituitary extracts (BPE) could be serially passaged at least 5 passages. HCECs in defined D-KSFM or BPE-containing KSFM were able to be passaged 3-4 times. CnT-20 was chosen for further studies to represent a low calcium and serum free in vitro niche known to promote epithelial progenitor cells.</p><p>The growth pattern and proliferative capacity of HCECs cultured in CnT-20 or SHEM were further compared. HCECs cultured in CnT-20 spread out from explants as single cells. The cells showed typical epithelial cell morphology with a small round or oval appearance. In comparison, the cells from limbal explants cultured in SHEM expanded like a sheet and stratified after reaching confluence. All cells were tightly packed and highly elongated (Fig. 2A). To evaluate their clonal growth capacity, HCECs were seeded at a density of 500 cells/cm2 in CnT-20 or SHEM without a feeder layer to assess colony forming efficiency (CFE). SHEM could support HCECs to form a few colonies that failed to grow continually without 3T3, a feeder layer (data not shown). In contrast, HCECs in CnT-20 generated a significant number of colonies without use of a feeder layer (Fig. 2B). As shown in Fig. 2C, the CFE of HCECs in CnT-20 reached 2.5±1.3 % at day 6 and 5.4±1.2 % at day 8. Even without any feeders, the colonies in CnT20 expanded rapidly to almost confluence on day 14 (Fig. 2D). The CFE numbers and clonal growth capacity of HCECs in CnT20 was similar to that observed with isolated corneal epithelial progenitor cells grown on a 3T3 feeder layer in previously reported studies by our group [36, 38, 39, 51]. These results indicate that the single small round cells grown in low calcium and serum free defined CnT-20 media primarily consisted of the expanded human corneal progenitor cells, although the cell morphology in CnT-20 medium was also changing with the number of large cells increased in every passage, indicating a gradual differentiation with every passage (Fig. 1).</p><!><p>Phenotypic characterization of the primary HCECs cultured in CnT-20 and SHEM was assessed using immunofluorescent staining with stem cell associated markers, p63, integrin β1 and EGFR, as well as with differentiation markers, connexin 43 and K3. As shown in Fig. 3 A and B, HCECs in CnT-20 possessed a significantly higher percentage of p63-positive (p<0.05, n=3), integrin β1-positive (p<0.01, n=3) or EGFR-positive cells (p<0.05, n=3) than the cells cultured in SHEM. In contrast, the percentages of positively immunoreactive cells for connexin 43 and corneal epithelial specific marker K3 were significantly lower in HCECs cultured in CnT-20 than in SHEM (both p<0.01, n=3).</p><p>The expression of these markers was further confirmed at the transcriptional level by reverse transcription and quantitative real-time PCR, as shown in Fig. 3C. With GAPDH as an internal control, the mRNA expression of stem cell associated markers, p63, integrin β1 and EGFR, were significantly higher by the cells cultured in CnT-20 than in SHEM. In particular, levels of p63 mRNA were 10 fold higher (p<0.001, n=3) in HCECs cultured in CnT20 than those in SHEM. Interestingly, levels of mRNA transcripts for the differentiation markers, connexin 43 and K3, were detected at much lower levels (both p<0.01, n=3) in the cells cultured in CnT-20, compared to cells in SHEM. These data confirmed the progenitor cell phenotype of HCECs cultured in CnT20, as revealed by immunofluorescent staining. Taken together, results of cell growth, passage capacity, colony forming efficiency and phenotypic markers, as shown in Figs 1-3, indicate that this low calcium serum free media can function as an in vitro niche that maintains the progenitor properties of limbal stem cell-derived corneal epithelial cells.</p><!><p>In order to uncover what intrinsic signaling is activated by the extrinsic factors in the niche media, we compared the expression and function of β-catenin, Tcf4 and survivin, the key factors of canonical Wnt signaling [30, 31], in HCECs cultured in CnT-20 with those grown in SHEM. As shown in Fig. 4A, the mRNA of β-catenin was highly expressed by the cells cultured in CnT-20, 4.57±0.75 (p<0.05, n=6) fold to that in SHEM. Tcf4 mRNA was elevated 3.34±0.3 fold (p<0.05, n=6) in the cells cultured in CnT-20. Interestingly, the mRNA expression of survivin, a downstream molecule of Tcf4, was also significantly higher by 5.52±1.97 fold (p<.01, n=6) in HCECs cultured in CnT-20 than in SHEM. Western blot analysis confirmed the expression pattern of β-catenin, Tcf4 and survivin by HCECs cultured in CnT-20 and SHEM, at protein levels. Fig. 4B shows representative immunoblotting results of two pairs of HCEC cell extract samples in the different media. The protein bands of β-catenin (92 kDa), Tcf4 (58 kDa) and survivin (16 kDa) were markedly higher in cells cultured in CnT-20 compared to those in SHEM, while the 45 kDa bands of β-actin protein, an internal control, were not significantly different between the cells in two conditions. Furthermore, fluorescent staining visibly displayed the higher immunoreactivities of β-catenin, Tcf4 and survivin antibodies in HCECs cultured with CnT-20 (Fig. 4C). These results indicate β-catenin/Tcf4/survivin signaling may play a role in maintaining the properties of human corneal epithelial stem cells.</p><!><p>To evaluate functional role of Tcf4 signaling in human corneal epithelial stem cells, Tcf4 gene silencing experiments were performed using HCECs cultured in the corneal epithelial progenitor medium CnT-20 and transfected with a specific siRNA-Tcf4 at different concentrations. A non-coding sequence siRNA conjugated with fluorescein (siRNA-F) was transfected to HCECs in CnT20 as a negative control, as well as a visible monitor for siRNA transfection efficiency. As shown in Fig. 5A, siRNA-F transfection did not significantly alter the mRNA expression of β-catenin, Tcf4 and survivin. In HCECs transfected with siRNA-Tcf4 at 10, 25 or 50 nM, Tcf4 mRNA expression was suppressed dramatically and dose-dependently (P<0.001) by 70-85%, compared with untreated control. Interestingly, levels of survivin transcripts, a downstream molecule of Tcf4, were also suppressed dose-dependently to 53±8, 49±7 and 45±17% (all p<0.05, n=3) at 10, 25 or 50 nM of siRNA-Tcf4, respectively. However, the expression of β-catenin, a Tcf4 molecular partner, was not significantly changed in HCECs transfected with siRNA-Tcf4. The knock down of Tcf4-survivin signaling was confirmed at the protein level by Western blot analysis. As shown in Fig. 5B, the Tcf4 and survivin protein bands were dramatically reduced by Tcf4 silencing at all doses (10-50nM) of siRNA-Tcf4 tested, compared with that in HCECs transfected with non-silencing siRNA-F. In contrast, the bands of β-catenin and housekeeping β-actin were not altered by Tcf4 silencing. Notably, the proliferative capacity of HCECs was impaired by Tcf4 silencing. As measured by WST-1 assay shown in Fig. 5C, the cell proliferative index was significantly suppressed by siRNA-Tcf4 but not by the non-coding sequence siRNA-F in three concentrations (10nM, 25nM and 50nM) of siRNAs tested (all p<0.001, n=5). These results indicate that Tcf4/survivin signaling may determine the proliferative capacity of human corneal epithelial progenitor cells.</p><!><p>The maintenance of corneal epithelial health and corneal tissue repair following trauma both depend on the regenerative capacity of corneal epithelial stem cells, which are located in the basal layer of the limbal epithelium. It is now clear that the niche plays an important role in maintenance of stem cell properties, and the different fate of stem cells, being stem cells or differentiated, can be determined by different niche [1, 6, 12, 52, 53]. It is well known that low calcium, serum free media could provide an ideal niche in vitro to maintain or promote progenitor cell properties by delaying the onset of terminal differentiation of cultured epithelial cells [13, 16, 17, 54]. However, it is not clear how niche extrinsic factors activate intrinsic factors or the signaling pathways that maintain progenitor cells properties. Wnt signaling has been implicated in stem cell and their niche, and it is likely of great importance in the interactions between stem cells and their niche micro-environment [27, 55-57]. In our present study, we explored the potential roles of β-catenin/Tcf4/survivin signaling of the canonical Wnt pathway in maintaining the properties of human corneal epithelial stem cells, which may serve as a representative model of tissue-specific adult stem cells.</p><p>As shown in Figs 1 and 2, all four low calcium, serum free media tested in this study were capable of supporting better growth, more serial passages and higher CFE of HCECs in culture, compare to high calcium, serum-containing medium SHEM, indicating that low calcium, serum free culture niche maintains the proliferative capacity of HCECs. The colony forming efficiency assay is usually used to monitor the proliferative capacity of epithelial progenitor cells, and it requires 3T3 mouse fibroblasts as a feeder layer to support clonal growth of corneal epithelial cells [14, 36, 51, 58]. However, we noted that the low calcium, serum free medium CnT-20 could support clonal growth of HCECs without a feeder layer. The CFE of HCECs cultured in CnT-20 without feeder layer was 5.4±1.2% at day 8, which reached the ranges of CFE rates (3.8-8.7%) of HCECs in SHEM with a feeder layer of 3T3 fibroblasts, as previously reported [35, 36, 38, 39]. This suggests that CnT-20 maintains the proliferative potential of human corneal epithelial cells in culture. Given the clinical use of limbal stem cells in stem cell therapy and corneal regeneration, the identification of culture conditions for HCECs that is serum free and does not require feeder cells is highly significant to the therapeutic ophthalmological field [59].</p><p>We have characterized a unique phenotype of stem cell enriched basal cells at human limbal epithelium and proposed that limbal stem cells are small primitive cells expressing three patterns of molecular markers: (1) exclusively positive for p63, ABCG2 and integrin α9 by a subset of basal cells; (2) relatively higher expression of integrin β1, EGFR, K19 and α-enolase by most basal cells, and (3) lack of expression of nestin, E-cadherin, connexin 43, involucrin, K3 and K12 [35, 39, 44, 51]. In this study, HCECs grown in CnT-20 expressed significantly higher signals of stem cell associated markers, p63, integrin β1 and EGFR, while lower levels of differentiation markers connexin 43 and K3, at both mRNA and protein levels, than the cells in SHEM. Taken together, our results demonstrated that low calcium serum free media maintain the stem cell-like phenotype and high proliferative capacity of human corneal epithelial progenitor cells in vitro.</p><p>Recent progress in cellular and molecular biology has uncovered the crucial role of Wnt signaling in proliferation, differentiation and self-renewal of stem cells. The best known Wnt signaling pathways include the Wnt/β-catenin, Wnt/planar cell polarity (PCP), and Wnt/calcium pathways [60]. The Wnt/β-catenin pathway is often called the "canonical" Wnt pathway, of which β-catenin and Tcf are key regulatory molecules. The hallmark of the Wnt signaling pathway is that the members of the Tcf family bind β-catenin and trigger Wnt target genes, such as survivin, to regulate cell functions [33, 61-63]. Survivin is the smallest member of the inhibitor of apoptosis (IAP) gene family in mammalian cells, an essential mitotic gene, localized to multiple aspects of the mitotic apparatus, and indispensable for several steps in cell division. Although it has been considered as a "cancer gene" since its discovery in 1997, survivin expression has been associated with "stemness" gene signatures of hematopoietic, mesenchymal, neuronal and skin progenitor cells. Recently, survivin has been identified as a direct transcriptional target of Wnt/β-catenin, which involves the recognition of discrete TCF-4-binding elements in the survivin promoter (see review article by [64]). β-catenin/Tcf4/survivin signaling has been found to be important in determining cell development and differentiation [33, 65, 66].</p><p>Using microarray analysis, we have recently observed that Tcf4 was one of the most highly up-regulated genes in human corneal epithelial progenitor cells that were isolated by rapid adhesion to collagen IV. Furthermore, β-catenin/Tcf4 signaling was found to be important in maintaining human corneal epithelial stem cell properties [48]. In our current study, the expression and function of β-catenin, Tcf4 and survivin were evaluated to investigate the underlying mechanism by which the low calcium, serum free medium CnT-20 maintains progenitor cell properties. As shown in Figs 4 and 5, human corneal epithelial progenitor cells cultured in CnT-20 expressed higher levels of these three factors at both gene transcript and protein levels. Silencing Tcf4 by siRNA transfection not only knocked down Tcf4 gene, but also blocked its downstream target molecule survivin at the transcriptional and translational levels, and furthermore significantly suppressed the proliferative capacity of HCECs cultured in CnT-20, confirmed with the WST-1 cell proliferation assay. These results strongly suggest that β-catenin/Tcf4/survivin signaling plays a vital role in maintaining corneal epithelial progenitor cell properties. However, the effect of Tcf4 silencing on the cell differentiation state is not clear and needs to be further studied. It is also important to explore how this signaling pathway is activated and which extrinsic factors contribute to stem cell maintenance.</p><p>In conclusion, these findings demonstrate that low calcium serum free media can provide a microenvironment niche for ex vivo expansion of corneal epithelium progenitor cells that retain a less differentiated phenotype and high proliferative capacity. The ability of this media to maintain the properties of human corneal epithelial progenitor cells is mediated by the β-catenin/Tcf4/survivin signaling pathway. This study may have high impact and clinic implication on the expansion of corneal epithelial stem cells in regenerative medicine, especially for ocular surface reconstruction.</p>

PubMed Author Manuscript

Zymogen and activated protein C have similar structural architecture

Activated protein C is a trypsin-like protease with anticoagulant and cytoprotective properties that is generated by thrombin from the zymogen precursor protein C in a reaction greatly accelerated by the cofactor thrombomodulin. The molecular details of this activation remain elusive due to the lack of structural information. We now fill this gap by providing information on the overall structural organization of these proteins using single molecule FRET and small angle X-ray scattering. Under physiological conditions, both zymogen and protease adopt a conformation with all domains vertically aligned along an axis 76 Å long and maximal particle size of 120 Å. This conformation is stabilized by binding of Ca2+ to the Gla domain and is affected minimally by interaction with thrombin. Hence, the zymogen protein C likely interacts with the thrombin-thrombomodulin complex through a rigid body association that produces a protease with essentially the same structural architecture. This scenario stands in contrast to an analogous reaction in the coagulation cascade where conversion of the zymogen prothrombin to the protease meizothrombin by the prothrombinase complex is linked to a large conformational transition of the entire protein. The presence of rigid epidermal growth factor domains in protein C as opposed to kringles in prothrombin likely accounts for the different conformational plasticity of the two zymogens. The new structural features reported here for protein C have general relevance to vitamin K-dependent clotting factors containing epidermal growth factor domains, such as factors VII, IX, and X.

zymogen_and_activated_protein_c_have_similar_structural_architecture

<!>smFRET and SAXS studies<!><!>smFRET and SAXS studies<!><!>smFRET and SAXS studies<!>Probing the accessibility of aromatic groups and hydrophobic clusters in protein C and APC<!><!>Probing the accessibility of aromatic groups and hydrophobic clusters in protein C and APC<!><!>Discussion<!>Reagents<!>smFRET studies<!>SAXS studies<!>Acrylamide quenching of intrinsic protein fluorescence<!>Reaction with ANS<!>Data availability<!>

<p>Protein C is a glycoprotein with modular architecture similar to that of other vitamin K-dependent factors of the blood coagulation cascade, i.e. prothrombin, factors VII, IX, and X (1). Synthesis of protein C in the liver produces a zymogen form of 419 amino acids comprising the N-terminal Gla domain (residues 1-46), two epidermal growth factor (EGF)-like domains (residues 55-90 and 96-136), and a C-terminal trypsin-like domain (residues 170-419). Three linker regions connect the Gla domain to EGF1 (residues 47-54), the two EGF domains (residues 91-95) and EGF2 to the protease domain (residues 137-169). This last segment contains the activation peptide (residues 158-169) and the dipeptide sequence Lys156–Arg157 that is removed during synthesis to produce a two-chain zymogen where the light chain (residues 1-155) and heavy chain (residues 158-419) remain connected through a disulfide bond between Cys141 and Cys277 (Fig. 1A). It is in this two-chain form that 90% of protein C circulates in the plasma, with the rest being slightly modified but functionally equivalent (2).</p><p>A, schematic representation of protein C comprising the Gla (residues 1-46), EGF1 (residues 55-90), EGF2 (residues 96-136), and protease domains (residues 170-419). Three linkers connect the different domains. A dipeptide consisting of residues Lys156 and Arg157 is removed from the third linker to produce a two-chain zymogen where the light chain (residues 1-155) and heavy chain (residues 158-419) remain connected through the Cys141–Cys277 disulfide bond. Cleavage by thrombin at Arg169 removes the activation peptide (red asterisk, residues 158-169) and produces APC. B, thrombin-catalyzed activation of the AF555/AF647 labeled (black) and unlabeled (gray) protein C S12C/R312C mutant. Activation was monitored by a progress curve using the APC-specific substrate H-D-Asp-Arg-Arg-p-nitroanilide at 405 nm under experimental conditions: 20 mm Tris, 145 mm NaCl, 5 mm EDTA, 0.1% PEG8000, pH 7.5, at 37 °C. C, hydrolysis of the chromogenic substrate S-2366 by the AF555/AF647-labeled (black) and unlabeled (gray) APC S12C/R312C mutant monitored at 405 nm under experimental conditions: 20 mm Tris, 145 mm NaCl, 5 mm CaCl2, 0.1% PEG 8000, pH 7.5, at 25 °C. Saturating amounts of the inhibitor hirudin (250 nm) were added to rule out contamination by thrombin activity.</p><p>Protein C is activated by thrombin upon cleavage at Arg169, resulting in removal of the entire activation peptide and folding of the active site and primary specificity pocket as in other members of the trypsin family (3). Activated protein C (APC) inactivates cofactors Va and VIIIa with the assistance of protein S, down-regulates the amplification and progression of the coagulation cascade, and maintains patency of the capillaries (4, 5). As an anti-inflammatory and cytoprotective agent, APC signals through PAR1 and PAR3 in ways that differ completely from thrombin's activation mechanism and reduces cellular damage following sepsis and ischemia/reperfusion of the brain, heart, lungs, and kidneys (6, 7).</p><p>Activation of protein C by thrombin is highly inefficient and requires the endothelial receptor thrombomodulin to enhance the rate >1,000-fold to a level compatible with physiological function (8–10). An additional improvement of the reaction rate is contributed by the endothelial protein C receptor (4). The molecular mechanism leading to protein C activation remains poorly understood because structural information is limited to Gla-domainless APC (11), with no available data on protein C free or bound to the thrombin-thrombomodulin complex other than computer models (12–15). Of particular importance is establishing whether any conformational change in protein C involves the site of activation around Arg169, which is thought to become exposed upon binding to the thrombin-thrombomodulin complex (16). Furthermore, it is of interest to establish if protein C has intrinsic conformational plasticity as recently observed in prothrombin, where an equilibrium between open and closed forms (17, 18) directs activation to thrombin along two distinct pathways (19–21). The closed form features an intramolecular collapse of kringle-1 onto the protease domain and promotes activation along the meizothrombin pathway (cleavage at Arg320), whereas in the open form the collapse is removed and activation proceeds along the alternative prethrombin-2 pathway (cleavage at Arg271).</p><p>In this study, we use single molecule FRET (smFRET) and small angle X-ray scattering (SAXS) to probe the structural architecture of the zymogen protein C and the protease APC. We find that both proteins assume a nearly identical conformation in solution, with the constitutive domains vertically stacked in a linear arrangement that does not change significantly upon binding of thrombin.</p><!><p>smFRET measurements with protein C and APC labeled with the AF555/AF647 FRET pair at positions C12/C312 across the Gla and protease domains (Fig. 1A) were carried out to evaluate the overall conformational properties of the two proteins. Labeling had no adverse effect on the thrombin-catalyzed activation of protein C, indicating lack of structural perturbations affecting function (Fig. 1B). No perturbation was also detected for APC as established from the catalytic activity of the labeled protein compared with the unlabeled protein (Fig. 1C). smFRET measurements reveal an interprobe distance between the FRET pair that does not change between the zymogen protein C and protease APC. FRET efficiency histograms are consistent with a single population of labeled species with low FRET efficiency and an interprobe distance of 76 Å (Fig. 2, A and B). A very small population of labeled species at high FRET efficiency (E > 0.9) is also notable in the histogram of APC, but due to experimental uncertainties this population was not considered in subsequent analysis. Our results indicate that the overall architecture of protein C does not change significantly during conversion to APC, thereby establishing another significant difference with the behavior recently reported for prothrombin where activation to meizothrombin is linked to transition from a dominant closed form to the elongated open form (17–19). The conclusion is supported by SAXS measurements that reveal envelopes of protein C and APC with identical maximal particle size (Dmax) values of 120 Å (Fig. 3, A–C).</p><!><p>smFRET profiles of protein C and APC free and bound to thrombin. Shown are the FRET efficiency histograms of (A) protein C and (B) APC. Histograms measured in the presence of 5 mm CaCl2 are displayed in gray, whereas those measured in the presence of 10 mm EDTA are in yellow. The concentration of thrombin is indicated. Continuous and discontinuous vertical lines mark the center of the FRET populations in the absence of thrombin in the presence of CaCl2 or EDTA, respectively. Titrations of protein C (C) and APC (D) with thrombin were measured in the presence of 10 mm EDTA. Binding isotherms were constructed by following the change in the peak center of the low FRET population as a function of thrombin concentration. Fit of the data to a single-site model yields values of Kd equal to 0.86 ± 0.2 μm (protein C) and 1.1 ± 0.2 μm (APC). Experimental conditions are: 20 mm Tris, 145 mm NaCl, 0.02% Tween 20, pH 7.5, with either 5 mm CaCl2 or 10 mm EDTA at 20 °C.</p><p>SAXS measurements of protein C and APC. A, overlaid SAXS profiles for protein C (red) and APC (blue) superimpose well, indicating that the two molecules adopt very similar conformations. The good linearity of the Guinier plots in the inset indicates good monodispersity of the samples. Guinier plots are coded in the same color as their respective SAXS profiles, and the black solid lines represent the fit. B, PDDF were calculated from the SAXS data in A. The value of Dmax estimated from PDDF is about 120 Å for both molecules, which is also supported by 3D SAXS envelope models for APC (C) and protein C (D).</p><!><p>The linear arrangement of the domains of protein C and APC under physiological conditions is strongly dependent on the presence of Ca2+. Chelation of the divalent metal with EDTA produces significant broadening of the FRET histograms and a shift toward higher efficiency (Fig. 2, A and B). The EDTA-derived histograms are best interpreted with a double Gaussian distribution, yielding peaks with efficiencies of 0.21 and 0.5. Because broadening of the histograms often results from dynamics on the millisecond time scale that occur during diffusion of the molecules through the confocal volume, we performed burst variance analysis (BVA) to establish whether EDTA has an effect on the conformational flexibility of protein C and APC. BVA tests for dynamics in the millisecond time scale by comparing the expected shot noise-limited standard deviation for a given mean efficiency and the experimentally observed standard deviation (22). When molecules experience significant dynamic fluctuations as they transit through the confocal volume, their FRET efficiencies are characterized by an increased standard deviation from that predicted by short noise only (22). BVA plots for protein C and APC are shown in Fig. 4. There are no significant conformational rearrangements on the millisecond time scale when Ca2+ is present and the main population with FRET efficiency of about 0.1 displays observed standard deviation that closely matches the predicted one (Fig. 4, A and D). In contrast, variations above the predicted standard deviation are noted in the presence of EDTA, especially for molecules with FRET efficiencies in the 0.22-0.6 range (Fig. 4, B and E). Chelation of Ca2+ increases the conformational flexibility of the Gla domain, resulting in conformational transitions in the millisencond time scale. Because protein C undergoes conformational changes as it transits through the confocal volume in the presence of EDTA, the values of 0.21 and 0.5 calculated from the double Gaussian distribution represent apparent rather than true transfer efficiencies (Fig. 2).</p><!><p>BVA plots of protein C (upper panels) and APC (lower panels) measured in the presence of 5 mm CaCl2 (A, D) or 10 mm EDTA without (B, E), and with 5 μm thrombin (C, F). Asterisks above each panel denote the FRET population centers obtained from the respective Gaussian fits shown in Fig. 2. In each panel, the expected standard deviation from shot noise is shown as a black line, with its 99.9% confidence interval appearing as a shaded purple area, whereas the observed standard deviation is shown as a blue dotted line. Experimental conditions are: 20 mm Tris, 145 mm NaCl, 0.02% Tween 20, pH 7.5, with either 5 mm CaCl2 or 10 mm EDTA at 20 °C.</p><!><p>The observation that Ca2+ stabilizes the conformation of the Gla domain is in agreement with results reported by others (23–26). The Gla domain contains multiple binding sites for Ca2+ (27–29) and removal of the cation is known to cause significant changes in structural stability (23). Specifically, the far-UV CD spectrum of the isolated Gla domain of protein C is characterized by a significant loss of helical content and increased percentage of random coil elements in the absence of Ca2+ (23). The smFRET data suggest that binding of Ca2+ to the Gla domain of protein C contributes to the structural integrity of this domain and stabilizes the overall linear arrangement of the entire protein.</p><p>smFRET measurements were also used to monitor changes in the conformational properties of protein C during its interaction with thrombin, used as the catalytically inactive mutant S195A to prevent hydrolysis. Binding of thrombin to protein C was confirmed by independent measurements using fluorescence correlation spectroscopy (data not shown). When measurements are carried out in the presence of Ca2+, the low FRET distribution of protein C remained largely unaffected (Fig. 2A), suggesting that the overall conformation of protein C is already optimized for binding to thrombin and possibly to the thrombin-thrombomodulin complex. Binding of thrombin in the absence of Ca2+ causes protein C to assume a conformation similar to the one in the presence of cation (Fig. 2A). The same effect is observed when thrombin binds to APC (Fig. 2B). BVA plots reveal that the thrombin-bound conformations of protein C and APC, characterized by FRET efficiencies of ∼0.1, display observed standard deviations that map within the upper-limit of the confidence interval of the expected standard deviation (Fig. 4, C and F). We conclude that thrombin dampens the conformational fluctuations of protein C observed in the presence of EDTA. However, a small fraction of molecules with FRET efficiencies >0.2 may still undergo conformational fluctuations as they diffuse through the confocal volume (Fig. 4, C and F).</p><p>Titration of the shift in the FRET distribution measured in the presence of EDTA allows for quantitative measurements of the interaction with thrombin and yields comparable Kd values of 0.9 ± 0.2 μm for protein C and 1.1 ± 0.2 μm for APC (Fig. 2, C and D). Thrombin does not show significant binding preference for the zymogen over the protease. The result is in agreement with previous findings (24, 30) and proves that the activation peptide present in protein C but not APC contributes little to the binding interaction with thrombin at equilibrium. The role of the activation peptide is to control the interaction kinetically (31) by hosting residues that decrease the rate of association between protein C and thrombin, especially in the absence of thrombomodulin (16, 25, 32, 33). Much of this effect is due to caging of Arg169 by the acidic residues in the activation peptide (16).</p><!><p>Acrylamide quenching studies were performed to establish if accessibility of aromatic groups in protein C changes during activation to APC. A graphical representation of the dynamic quenching constant (Equation 1) obtained from analysis of the Stern-Volmer plot is shown in Fig. 5, A and B. The KSV value for protein C is significantly lower than that of APC in the presence of Ca2+, showing that activation is linked to increased accessibility of aromatic groups. The higher KSV value is not due to a more accessible active site in APC because it does not change in the presence of the irreversible active site inhibitor PPACK. A significantly higher value of KSV is also observed in the presence of EDTA for protein C but not APC, suggesting that Ca2+ restricts the accessibility of aromatic residues to acrylamide in the zymogen but not the protease. The effect may be due at least in part to the presence of the activation peptide, whose conformation is known to be influenced by the binding of Ca2+ (34).</p><!><p>Acrylamide quenching of the intrinsic protein fluorescence of protein C and APC. A, Stern-Volmer plots measured in the presence of 5 mm CaCl2 for protein C (black), APC (blue), and APC-PPACK (red), or in the presence of 5 mm EDTA for protein C (green) and APC (purple). B, graphical representation of the Ksv values obtained from fitting the acrylamide quenching data to the Stern-Volmer equation (Equation 1). Error bars are the standard deviation calculated from two independent measurements. Experimental conditions are: 20 mm Tris, 145 mm NaCl, 0.1% PEG8000, pH 7.5, with either 5 mm CaCl2 or 5 mm EDTA at 20 °C.</p><!><p>Differences in the level of exposed hydrophobic clusters between protein C and APC were detected by monitoring the emission maximum of the fluorescent probe ANS, which becomes progressively shifted to lower wavelengths upon complexing to solvent-accessible hydrophobic groups (35). Upon excitation at 375 nm, the emission spectrum of free 8-anilino-1-naphtalenesulfonic acid (ANS) is characterized by a maximum at 520 nm (Fig. 6, A and C). In the presence of protein C, the maximum of ANS is blue-shifted to 513 nm, resulting in a difference of 7 nm relative to the free probe. A more pronounced blue shift of about 15 nm is observed when ANS reacts with APC (Fig. 6, A and C), showing that the level of solvent-accessible hydrophobic clusters increases when the zymogen converts to the mature protease. Again, the increased reactivity of APC with ANS binding to the active site is ruled out by measurements in the presence of PPACK. The level of exposed hydrophobic clusters does not change in the presence of EDTA, for both protein C and APC (Fig. 6, B and C).</p><!><p>Reaction between ANS and protein C or APC. Emission spectra for the reaction with ANS measured in the presence of 5 mm CaCl2 (A) or 5 mm EDTA (B). Colors designate: ANS in the absence (black) or presence of protein C (red), APC (purple), and APC-PPACK (blue dashed). C, graphical representation of the shift in the emission maximum of ANS measured in the presence of protein C or APC. The difference was calculated by subtracting the maxima obtained for free and protein bound ANS. Error bars are the standard deviation calculated from two independent measurements. Experimental conditions are: 20 mm Tris, 145 mm NaCl, 0.1% PEG8000, pH 7.5, with either 5 mm CaCl2 or 5 mm EDTA at 20 °C.</p><!><p>A combination of smFRET and SAXS measurements shows that activation of protein C is not linked to significant changes in the overall structural architecture of the protein. Protein C and APC feature nearly identical SAXS envelopes (Dmax ∼ 120 Å) and interprobe distances for the C12/C312 FRET pair (76 Å) across the Gla and protease domains. The relative arrangement of the auxiliary Gla and EGF domains does not change during the conversion of protein C to APC, in contrast to the conversion of prothrombin to meizothrombin that is accompanied by a drastic relocation of kringle-1, which removes the intramolecular interaction with the protease domain (17–19). Protein C and prothrombin share a common modular assembly, but the N-terminal Gla domain and C-terminal protease domain are connected by two kringles in prothrombin and two EGF domains in protein C and factors VII, IX, and X (36–38). Our results suggest that the presence of intervening EGF domains renders the structure of the zymogen more rigid and similar to that of the active protease. We also note that plasminogen contains kringle domains in its modular assembly and shares with prothrombin a more flexible, intramolecular collapsed architecture that opens up upon activation (39).</p><p>Subtle changes between protein C and APC emerge from analysis of the solvent accessibility of aromatic groups and hydrophobic clusters that are more exposed in the protease. The difference may be due to the presence of the activation peptide in the zymogen, but a definite validation will require solution of the crystal structure of protein C. It is also possible that the structure of APC is inherently more dynamic than that of protein C, especially in the <millisecond time scale, thereby allowing more rapid transient exposure of side chains to the solvent.</p><p>The results from smFRET studies merit attention because they provide clues on the position of the Gla domain relative to the protease domain. Information on the architecture of the Gla domain of protein C is currently lacking as no structure of the zymogen is available and the only deposited structure of APC refers to a construct devoid of Gla domain (11). Our data indicate a distance of 76 Å for the C12/C312 FRET pair, which is in good agreement with previous ensemble FRET measurements that estimated a distance of 94 Å between the active site of APC and a membrane surface to which the Gla domain was bound (40). The crystal structures of factors VIIa (29) and IXa (41) reveal distances of 86 and 71 Å between the analogous residues that were modified for smFRET measurements in our study. Furthermore, our results reveal that protein C and APC assume an elongated conformation in solution, with comparable Dmax values of 120 Å calculated from SAXS measurements. This conclusion is in agreement with computational models of the zymogen and protease whose vertical axis was estimated to be 130-140 Å long (14). We conclude that vitamin K-dependent coagulation factors carrying EGF domains (protein C, factors VII, IX, and X) assume a conformation where all constitutive domains are vertically stacked. The EGF domains function as static spacers that position the protease domain at a certain distance over the membrane surface to which the Gla domain is bound.</p><p>The linear assembly of protein C is stabilized by the binding of Ca2+ under physiological conditions. In the absence of Ca2+, the Gla domain undergoes conformational changes that occur in the millisecond time scale. Binding of thrombin largely restricts the conformational mobility of protein C and favors a conformation very similar to that bound to Ca2+. Our results are in agreement with previous CD measurements with the isolated Gla domain of protein C that have documented a significant loss in total helical content and concomitant increase in random coil-like structures upon removal of Ca2+ from the solution (23). Different levels of Ca2+-dependent conformational rearrangements have been reported for the Gla domain of other vitamin K-dependent proteins (26, 42–45). Recent bioinformatic analyses have also identified that the Gla domain displays significant intrinsic disorder in the Ca2+-free form (46). Therefore, the difference seen in smFRET efficiency for protein C in the presence of EDTA may reflect a partial unfolding of the Gla domain. This is relevant new information on the structural properties of protein C and its mature protease APC that bears on all vitamin K-dependent factors of the coagulation cascade.</p><!><p>Thrombin WT and the catalytically inactive mutant S195A were prepared as reported elsewhere (47). Mutations were introduced into the human protein C plasmid carrying a C-terminal HPC-4 tag using a QuikChange Lightning site-directed mutagenesis kit (Agilent Technologies). Baby hamster kidney cells were transfected with the plasmids of interest using X-tremeGENE 9 DNA transfection reagent (Roche Applied Science) according to a standard protocol supplied by the manufacturer. After an incubation period of 48 h, stably expressing clones were selected by incubating the transfected cells with 1 mg/ml of Geneticin. Following clonal expansion, cells were transferred into large cell factories and the growth media from these factories, collected over a period of several weeks, following centrifugation and filtration, was loaded onto a resin that was coupled to the HPC-4 antibody and the protein was purified as described for prethrombin-1 (31). After the immunoaffinity chromatography step, the sample was diluted to achieve a final NaCl concentration below 50 mm and the protein was loaded onto a 1-ml Q-Sepharose Fast-Flow (GE Healthcare) column attached through its top to a 1-ml HiTrap heparin column (GE Healthcare) equilibrated with 20 mm Tris, 50 mm NaCl, 10 mm EDTA, pH 7.5. After detaching the heparin column, the protein was eluted from the Q-Sepharose Fast-Flow column using a 0.05-1 m NaCl gradient. Protein C was further purified on a size exclusion Superdex 200 column (GE Healthcare) equilibrated with 20 mm Tris, 145 mm NaCl, pH 7.5.</p><p>Protein C (5 μm) was activated with 7 nm thrombin and 50 nm thrombomodulin after overnight incubation at ambient temperature under experimental conditions: 20 mm Tris, 145 mm NaCl, 5 mm CaCl2, 10% glycerol, pH 7.5. The sample was diluted to achieve a final NaCl concentration of 50 mm and the protein was loaded onto a 1-ml Q-Sepharose column attached through its top to 1-ml HiTrap SP HP column (GE Healthcare) equilibrated with 20 mm Tris, 50 mm NaCl, pH 7.0. The upper HiTrap SP HP column, to which the thrombin-thrombomodulin complex predominantly binds, was detached and APC was purified from the Q-Sepharose column using a 0.05-1 m NaCl gradient. Successful activation of protein C and purification from the thrombin-thrombomodulin complex was verified by SDS-PAGE and by monitoring activity with a chromogenic substrate.</p><!><p>Protein C and APC were labeled at engineered Cys residues in the Gla (S12C) and protease (R312C) domains (Fig. 1A) with Alexa Fluor 555-C2-maleamide and Alexa Fluor 647-C2-maleamide (Invitrogen) using a protocol similar to that used for prothrombin (18). Briefly, protein C and APC (16 μm) were incubated for 1 h in the dark with 2.8-fold molar excess DTT (Sigma-Aldrich) in a labeling buffer composed of 20 mm Tris, 350 mm NaCl, pH 7.5. After excess DTT was removed on a ZebaTM spin desalting column (ThermoFisher) equilibrated with labeling buffer, the proteins (12 μm) were incubated with 2.5-fold molar excess Alexa Fluor 555-C2-maleamide and 2.5-fold molar excess Alexa Fluor 647-C2-maleamide for 2 h with gentle shaking and protected from light. Excess label was removed on a size exclusion Superdex 200 column (GE Healthcare) equilibrated with labeling buffer.</p><p>Successful incorporation of the probes at the correct positions was verified by limited proteolysis with thrombin. Samples were run on SDS-PAGE under reducing conditions and visualized on a Typhoon fluorescent scanner after excitation at either 550 or 650 nm. As expected, the label was present in both the light chain (containing residue Cys12) and heavy chain (containing residue Cys312). Labeling at positions Cys12/Cys312 did not produce significant structural perturbations as assessed by rates of protein C activation by thrombin and APC hydrolysis of a chromogenic substrate comparable with WT (Fig. 1, B and C).</p><p>smFRET measurements of freely diffusing protein C and APC molecules (150 pm) were collected on a confocal microscope MicroTime200 using pulsed interleaved excitation (PIE) under experimental conditions: 20 mm Tris, 145 mm NaCl, 0.02% Tween 20, pH 7.5, with either 5 mm CaCl2 or 10 mm EDTA at 20 °C. Raw data were initially processed with the PIE analysis with MATLAB (PAM) software (48) as described (19). After applying the correct γ-factor and the appropriate correction factors for donor leakage and direct acceptor excitation, data were exported and fitted to a Gaussian function using the peak analyzer function in OriginPro 8.1. All measurements were done in triplicate. BVA was carried out with the PAM program (48). Once the correct γ-factor and the appropriate correction factors for donor leakage and direct acceptor excitation were applied, the analysis was performed using a bin number of 20, a confidence interval sampling number of 100, 5 photons per window, and 120 bursts per bin.</p><!><p>SAXS data on protein C and APC were collected at the beamline 12-ID-B of the Advanced Photon Source at the Argonne National Laboratory (Argonne, IL, USA) under experimental conditions: 20 mm Tris, 145 mm NaCl, 5 mm CaCl2, pH 7.5. Scattered X-rays at 14 keV radiation energy were measured using a Pilatus 2M detector with a sample-to-detector distance of 2 m. A flow cell was used to reduce radiation damage. Thirty images were collected for each sample and buffer blank. The scattering vector q=4πλ-1sin ϑ/2 is the momentum transfer defined by the scattering angle ϑ and X-ray wavelength λ. The isotropic 2D images were converted to 1D SAXS profiles, i.e. intensity versus q, followed by averaging and background subtraction using software packages at the beamline. The radius of gyration, Rg, was determined using the Guinier approximation in the low q region (qRg < 1.3) and its linearity served as an initial assessment of data and sample quality. A value of Rg of about 36 Å was obtained for both molecules. Distance distribution functions (PDDF) were calculated from SAXS data using program GNOM (49). PDDF is the Fourier transform of SAXS data and a weighted distance histogram of atom pairs, and provides an estimate of the maximum dimension Dmax for a molecule. The low resolution envelopes were produced using DAMMIF (50) by directly fitting the SAXS profile with q up to 0.40 Å−1. Twenty models were generated for every calculation and then aligned and averaged using DAMAVER (51). The normalized spatial discrepancy values of calculations for APC and protein C are about 1.3 and 1.4, respectively, indicating good convergence for individual models. The SAXS profiles for APC and protein C are almost identical, indicating that they adopt a very similar conformation. The 3D envelope reconstruction reveals that both molecules are elongated with Dmax ∼ 120 Å. SAXS data were deposited in the SASBDB database (codes SASDJC6 for protein C and SASDJD6 for APC).</p><!><p>The accessibility of aromatic groups in protein C and APC was studied by monitoring spectra resulting from the acrylamide-dependent quenching of the intrinsic protein fluorescence. Reactions were carried out by incubating 120 nm protein with a specific concentration of acrylamide for 10 min at 20 °C in a buffer composed of 20 mm Tris, 145 mm NaCl, 0.1% PEG8000, pH 7.5, supplemented with either 5 mm CaCl2 or 5 mm EDTA. A stock solution of 1.5 m acrylamide (Sigma-Aldrich) was prepared in the same buffer. The effect of H-D-Phe-Pro-Arg-CH2Cl (PPACK) (Hematologic Technologies) was studied with APC (1 μm) after incubation for 30 min with 100-fold molar excess PPACK and dilution of the enzyme to 120 nm. Control experiments verified that PPACK completely inhibited APC activity under these conditions.</p><p>Data were collected on a HORIBA FluoroMax-4 spectrofluorometer at 20 °C by monitoring the emission at 340 nm upon excitation at 280 nm in a cuvette with a 0.3-cm path length. Data were analyzed using the Stern-Volmer equation, (Eq. 1)F0F = 1 + KSVQ where F0 is the fluorescence emission in the absence of acrylamide, F the emission at a specific concentration of acrylamide, KSV the dynamic quenching constant, and Q the concentration of acrylamide. All measurements were carried out at least in duplicate.</p><!><p>Protein C (1 μm) and APC (1 μm) were incubated with 80 μm ANS (Sigma-Aldrich) for 1 h at 20 °C. For measurements carried in the presence of PPACK, APC was incubated for 30 min with a 100-fold molar excess inhibitor prior to titration with ANS. Control reactions were carried out with 80 μm ANS. Fluorescence emission spectra were collected in the 420-650 nm range following excitation at 375 nm in a cuvette with a 0.3-cm path length. All reactions were conducted in duplicates at 20 °C under experimental conditions: 20 mm Tris, 145 mm NaCl, 0.1% PEG8000, pH 7.5, with either 5 mm CaCl2 or 5 mm EDTA.</p><!><p>All data described in the manuscript are contained within the manuscript. The amino acid sequence reported in this paper has been submitted to the Small Angle Scattering Biological Data Bank under accession numbers SASDJC6 and SASDJD6.</p><!><p>Author contributions—B. M. S., L. A. P., X. Z., and E. D. C. conceptualization; B. M. S., L. A. P., and X. Z. data curation; B. M. S., L. A. P., X. Z., and E. D. C. formal analysis; B. M. S. and E. D. C. validation; B. M. S. and E. D. C. writing-original draft; B. M. S., L. A. P., X. Z., and E. D. C. writing-review and editing; L. A. P. and X. Z. visualization; E. D. C. resources; E. D. C. supervision; E. D. C. funding acquisition; E. D. C. project administration.</p><p>Funding and additional information—This work was supported in part by National Institutes of Health Research Grants HL049413, HL139554, and HL147821 (to E. D. C.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.</p><p>Conflict of interest—The authors declare that they have no conflicts of interest with the contents of this article.</p><p>epidermal growth factor</p><p>activated protein C</p><p>small angle X-ray scattering</p><p>Phe-Pro-Arg-CH2Cl</p><p>8-anilino-1-naphtalenesulfonic acid</p><p>burst variance analysis</p><p>pair distance distribution functions.</p>

PubMed Open Access

Decoupling the Arrhenius equation via mechanochemistry† †Electronic supplementary information (ESI) available: Details regarding apparatus design, experimental procedures, and software computations. See DOI: 10.1039/c7sc00538e Click here for additional data file.

We identified three different energetic regions that we believe are defining characteristics of most, if not all mechanochemical reactions. For a given ball mill's region, activation energy determines whether a reaction is energetically easy (Region I), challenging (Region II), or forbidden (Region III). In Region II, yield depends exponentially on oscillation frequency. Modifications granted control of the locations of Regions I, II, and III.

decoupling_the_arrhenius_equation_via_mechanochemistry†_†electronic_supplementary_information_(esi)_

Introduction<!>Results and discussion<!>Conclusions<!>Conflict of interest

<p>Mechanochemistry, the field of chemistry relating to mechanically-induced (i.e., grinding or colliding) reactions, has experienced rapid growth within the chemical community.1–10 This growth is caused by a mounting interest in exploiting the unique conditions involved. In conventional solvent reactions, solvent choice may heavily influence the possible reaction pathways. Similarly, the absence of solvent in (many) mechanochemical reactions dictates a defined set of pathways too, some of which are unique. Much of the mechanochemistry literature is devoted to exploring these pathways in various fields of chemistry (e.g., inorganic, organic, organometallic, polymeric, etc.).11–22</p><p>Less attention, however, has been devoted to developing an understanding of how the energetics of these systems work and how they can be controlled.23–27 This is possibly because there are many methods of inducing mechanochemical reactions, which may all have unique factors involved in their energetics. For example, two seemingly disparate methods of mechanochemical activation are planetary milling and vibrational milling. Planetary milling involves the high-speed rotation of drums containing reactants and reagents, as well as some kind of additional grinding media such as dozens or hundreds of milling balls.28 On the other hand, vibrational mills use high-frequency oscillations/shakings of a reaction vial, often with just a single ball, to induce chemistry.15,27,29–34 The vial and balls may be made out of inert materials such as stainless steel (SS) or Teflon (PTFE). Interestingly, copper and nickel vials have demonstrated in situ catalytic behavior.19,35 In addition to the different types of ball mills available (planetary, vibratory, etc.) ball mills from different manufacturers can provide different results as well. For example, mixer mills produced by Retsch36 (e.g., MM 200 and MM 400) and Fritsch37 (e.g., Pulverisette 23) are manufactured with the motor external to the grinding jars. By contrast, the Spex 8000M38 produced by SpexCertiprep has the motor encased in the same environment as the grinding jars which adds additional heat to the milling environment. This difference in thermal energy is mainly due to how the ball mill is constructed, which can influence the environment of the mechanochemical reaction.27,39</p><p>The actual temperature of a milled chemical reaction is still unclear. According to reports in the literature, scientists have reported temperature ranges as low as 40 °C in vibratory mills to as high as 600 °C in planetary mills.31,40,41 However, these reports measured the temperature of the ball and/or vial after the reaction concluded, which doesn't provide information on the temperature of the reaction at impact, which has been theorized to be well over 1000 K.42 Recently, more sophisticated studies have been performed to get a better understanding of the temperature of milling reactions in situ.27,32,34,39 Although these are significant improvements over previous results, these methods still cannot provide the chemical energy created at impact. In order to gain more insight into the energetics of milled chemical reactions, especially to gauge how much of the available energy in the milling process is actually transferred to the molecules upon impact, we studied the Diels–Alder reaction under these unique conditions.43 Although our previous report provided a little more insight into the energetics of these reactions, the precise role of the ball and its characteristics remained unclear: how is this mechanical force compelling chemistry along? Previously in the literature this question has been addressed with reactions where the rate was significantly dependent on the induction period.15,27,32,34,39,44 For example James and coworkers studied the deprotonation of imidazole by zinc oxide to examine the role of oscillation frequency on reaction rate. As they expected, they observed that an increase in oscillation frequency resulted in an increase in the reaction rate for this specific diffusion-limited reaction. The authors envisioned this as resulting from the ball acting as a tool for constantly re-exposing unreacted starting material (i.e., clearing product out of the way such that fresh reagents can interact). However, it is of key importance to determine how this picture holds for reactions that are not diffusion controlled but instead may have a fairly high activation barrier.</p><p>We elected to continue exploring the energetics of the 8000M mill using Diels–Alder reactions. The Diels–Alder reaction is uniquely suited to this energy exploration because it involves two reactants creating a single product in a single concerted step, without the need for any other reagents or catalysts. In this way, its use minimizes potential confounding effects. Furthermore, the activation energy of a Diels–Alder reaction can be readily altered in a straightforward manner by changing the substituents on the reactants. This allows us to directly observe how changes in mechanochemical conditions correspond to changes in yield and accessible activation barriers. The intention of this work is to identify and combine the most important variables for ball milling reactions into a concise and well-defined picture, allowing us control of reaction energetics and predictive capabilities. In addition, the results of these experiments may be useful to unify the ball milling community such that we have a method to calibrate ball mills independent of type and manufacturer.</p><!><p>In order to explore both positive and negative effects on energy (and thus yield), we desired a Diels–Alder reaction yielding approximately 50% under our baseline conditions (stainless steel vial oscillating at 18 Hz). The Diels–Alder reaction of benzoquinone (BQ) with 9,10-dimethylanthracene (9,10-DMA), as outlined in Scheme 1, yielded 41% ± 3% of theoretical yield (all errors reported as standard error of the mean, n = 3). All reactions were run at a 0.5 mmol scale for three hours (pertinent experimental details for this and other reactions are available in the ESI S2†). With an appropriate reaction in hand, we explored the role of oscillation frequency and vial material/hardness (see ESI S2† for specifics on determining oscillation frequency).</p><p>The results of the frequency investigation are presented in Fig. 1. Of critical importance is the observation that the yield doubles every 2 Hz increment over the entire range of tested frequencies. This indicates an exponential dependence on frequency. The doubling effect is reminiscent of the guideline that increasing the temperature of a solution reaction by 10 °C will double the reaction rate. However, the temperature of the vial was monitored throughout the reaction and its average temperature only differed by ∼2 °C when comparing 15 Hz and 21 Hz experiments (details of this measurement are available in the ESI S3†).</p><p>Vial material has also been shown to influence yield. In a prior study, we demonstrated that PTFE vials (soft) produce a lower yield than stainless steel vials (hard).43 In the present study we include a control for any potential catalytic behavior of the metal by creating a vial from heat-treatable steel (see ESI S5† for details on this process). The BQ + 9,10-DMA reaction was run again at 18 Hz in Teflon, hardened steel, unhardened steel, and stainless steel vials. This data is presented in Fig. 2. The hardened steel vial produced 60% yield while the unhardened steel and stainless steel vials both produced 40% yield. All three outperform Teflon, which results in only 6% yield. The difference between the hardened vial and unhardened vial isolates the effect of hardness, indicating a significant role. Furthermore, since the unhardened vial produces a yield that is indistinguishable from the stainless steel vial, we opted to perform subsequent comparison experiments in stainless steel instead of the unhardened steel, as they are commercially available.</p><p>The underlying cause of the frequency and hardness effects remains unclear. It is possible we are increasing the amount of "chemically usable energy" released during the impact. It is also possible we are merely increasing the rate of molecular collisions between reactants by affecting the mixing. It is also possible we are affecting both. To get some insight on this point, we expanded our data to an entire series of Diels–Alder reactions spanning a range of activation energies.</p><p>The transition state geometries and energies for the Diels Alder series, were calculated using the mPW1PW91/6-31+G(d,p) level of theory and basis set. Linder and Brinck studied the effect of theory and basis set on Diels–Alder transition state calculations and found good agreement (generally ±1.5 kcal mol–1) with benchmark CCSD(T)/6-31+G(d)//CCSD/6-31+G(d) calculations obtained with the Gaussian 09 program suite.45–48 Two of our calculated transition state energies (9,10-DMA + MA and 9-H + MA) have been previously determined experimentally in xylenes and match within 1.0 kcal mol–1.49 The combinations of these reactants (all solids at the vial's temperature during milling) produce six different Diels–Alder adducts, all with different activation energies (Scheme 2). Computational details are available in the ESI S5.†</p><p>Each of these six reactions were subjected to nine different ball-milling conditions (permutations of various vial materials and oscillation frequencies). Fig. 3 contains these results (note that the 18 Hz data for E a = 12.8 kcal mol–1 was present in Fig. 1 and 2). For ease of reference in discussing the results, we gave the label "Region I" to the energy range encompassing reactions that, regardless of the frequency (i.e. 15–21 Hz) we chose or vial material used (i.e. Teflon, stainless steel, hardened steel). On the opposite end of the figure, we gave the label "Region III" to the energy range containing reactions that produced no observable yield after three hours of milling, regardless of frequency chosen or vial material used (we extended reactions times to 16 hours and still didn't observe any product). We labeled the intermediate range "Region II." Note that the positions of Regions I, II, and III are not universally fixed. Indeed, they are a function of several variables, such as how long the reactions were run, but in practice they are useful as they highlight the area of greatest sensitivity to mechanochemical conditions (Region II). The reactions associated with Region III are theoretically attainable given significantly longer milling times, but from a practical perspective, we are focusing reaction times that are applicable to conventional ball mills. Furthermore, given the differences in ball mills, type and manufacturer, it is important to be able to develop a nomenclature that universally describes the limitations of unmodified ball mills.</p><p>Regions I, II, and III display noteworthy characteristics that can be explained by an argument rooted in the Arrhenius equation (eqn (1)). First, we must appreciate that even if the ball wasn't present, the molecules would still presumably be in equilibrium with the temperature of the vial itself. This temperature is 36 °C for all six reactions conducted at an oscillation frequency of 15 Hz. As mentioned previously, when changing the frequency from 15 Hz to 21 Hz, the temperature of the vial increases by just 2 °C. With this in mind, one could argue that in "Region I" (E a ≤ 11.9 kcal mol–1), a large enough fraction of molecules possess enough energy that we observe quantitative conversion even when the collisions between molecules are at the minimum for our conditions (PTFE and 15 Hz, where you may imagine the mixing caused by the ball is very inefficient). Thus, in the case of "Region I" we are led to assume that the energy profile term (e–Ea/RT) of eqn (1) is sufficiently large such that relatively few collisions are needed to reach quantitative yield. In "Region II" (11.9 kcal mol–1 < E a < 14.9 kcal mol–1), we could attribute increases in yield to increases in frequency factor ("A") due to improved mixing as we use harder materials or higher frequencies. Therefore, although the energy profile term is less favorable due to a higher activation energy, it can be offset by an increased collision rate/frequency factor (i.e., speeding up the mill). "Region III" encompasses reactions that are not accessible given practical frequency limitations on our ball mill and our given time frame. Thus, we propose calling Region I a "thermally-driven region," Region II a "collisionally-driven region," and Region III a "energetically limited region."1</p><p>Since the main energy source of these reactions would come from the overall temperature of the vial, this theory predicts that an overlay of a theoretically calculated Boltzmann (or Boltzmann-like) distribution should correlate reasonably well with the observed results. This suggests that oscillation frequency and hardness would merely act upon the thermal distribution. This overlay (assuming R = 0.001986 kcal K–1 mol–1) is presented in Fig. 3's inset, demonstrating outstanding agreement. Furthermore, comparing this to "Region II" in Fig. 3 well delineates the additional benefit of the increased collision rate acting on the thermally available energy. In effect, this means controlling A in the Arrhenius equation. It is important to note that an equivalent control over collision rate cannot be had in well-stirred solutions (e.g., solutions with no gradient in concentration or temperature).</p><p>At this point, one can see that a ball-milling approach to solvent-free reactions affords chemists separate access points to reaction rates: thermal and collisional. Theoretically speaking, the upper end of the collisional frequency effect may be limited only by the lifetime of molecular vibrations. However, the current engineering and design of ball mills forces a practical upper limit to the maximum collisional frequency (i.e., when we attempted 22 Hz for extended times the milling apparatus broke, other mills may differ in frequency limitations). Because there are currently practical limitations on A, we investigated the effect of changes in the system's temperature on the rate of reaction. We hypothesized that changing the vial temperature will let us choose between the Boltzmann distributions in Fig. 4. Shrewd selection of Boltzmann distribution should effectively control the locations of Regions I, II, and III, assuming the milling time is unchanged.</p><p>To this end, we made several modifications to the ball mill. For testing the effect of a temperature decrease, the ball mill was interfaced with a cooling unit (see ESI S3† for details). The yield of BQ + 9,10-DMA served as a comparison point. Under normal operating conditions using a hardened vial oscillating at 21 Hz the peak operating temperature was 38 °C. When cooled during operation, the system maintained a peak operating temperature of about 22 °C. This modest drop of 16 °C caused the yield to plummet from 92% to 12%. The significance of this result cannot be overstated. Because oscillation frequency was unchanged, it is reasonable to attribute the drop in yield to indicate a significant dependence on temperature as opposed to any change in the pre-exponential factor. This greatly bolsters the Arrhenius-based argument. Simultaneously, it suggests that for an activation energy of 12.8 kcal mol–1, the impact of the ball with the wall does not provide a significant amount of chemically usable energy. The current method for cooling is undesirable for practical purposes, but has been useful as a proof of concept. Development of an improved cooling method with access to all Boltzmann distributions in Fig. 4 is currently in progress.</p><p>Having decreased the temperature of the system, increasing the temperature is important not only as another test of the theory, but also because it would shift all regions towards higher energies. Thus, we would obtain access to reactions that were previously inaccessible to us on a reasonable timescale. To increase the temperature, a heating band (BB010004, http://www.instrumentation-central.com/) was wrapped around an aluminum rod holding the vial. Coupling the band with a Variac allowed temperature control of the vial with a fair amount of precision (±2 °C) at any targeted temperature. See ESI S3† for details regarding the reproducibility and reliability of this setup. The two reactions of the conditionally-forbidden region (III) (and a third reaction for which the mill could not reach quantitative yield in 3 h) were each tested at a variety of temperatures. The results of these experiments (18 Hz, SS) are displayed in Fig. 5. All three reactions now readily reach quantitative conversion within three hours. These results make clear that we have successfully identified a way to shift the regions of Fig. 3 to higher activation energies. For the sake of comparing with solution, the reaction of 9-H + BQ (E a = 17.4 kcal mol–1) is done at ∼140 °C (refluxing xylenes), and we observed quantitative conversion at just 100 °C.50 We expect that in the solvent-free ball milling conditions here, reactants are free of fruitless collisions with solvent molecules, allowing a significant boost in reaction rate. Lastly, it is important to note that the milling and heating must occur simultaneously, as we observed in a separate experiment that if the 9-H + BQ reaction is milled for three hours with no extra heating followed by subsequent heating in an oven for three hours at a comparable temperature, the yield drops from 100% to 19%.</p><p>Given the resounding success of the Arrhenius equation when applied to a vibratory ball mill, we propose a straightforward path for energetically converting conventional, solution-based syntheses to ball mill ones. If selectivity is not a concern, the vial temperature should be high enough to expand "Region I" to encompass the reaction. If selectivity is needed to avoid side reactions, "Region II" should be targeted. Once a reaction is in the collisionally-driven region (Region II), fine adjustments can be made such as a change in oscillation frequency or vial and ball material, which will have a significant effect on reaction rate and selectivity.</p><!><p>Mechanochemistry has continued to grow in popularity despite a poor understanding of the inner workings of its energetics. Our experiments suggest that we can conceptualize vibrational ball mills (and likely many other forms of mechanochemistry) as collision-facilitating devices that act upon molecules existing in a thermally-derived energy distribution. In this way, conducting chemical reactions in a variable-frequency, variable-temperature ball mill allows chemists to decouple both halves of the Arrhenius equation: the frequency factor ("A") and the energy profile term (e–Ea/RT). Coarse adjustment of temperature and optional fine adjustment of oscillation frequency dictate the energetics and thus serve as a way to translate established conventional syntheses to ball mill conditions. This proposal is straightforward, yet powerful in its predictive capabilities. It has succeeded in getting access to reactions that were previously inaccessible in the mill due to prohibitively high activation energies.</p><p>Finally, we would like to argue heuristically that the new access point to the frequency factor grants mechanochemistry a unique prospect with respect to selectivity. Conventionally, a chemist decreases temperature to increase selectivity, a result of narrowing the Boltzmann distribution. However, it may be the case that in the process of obtaining the desired selectivity, that either (A) the rate of the target reaction is slowed so much as to render the reaction impractical or (B) solubility problems arise. At this point, the chemist would require a catalyst, which may be toxic, expensive, and/or inconvenient. In the ball mill, however, our theory predicts that the selectivity should be unchanged by oscillation frequency. In this way, we can recover the original reaction rate by increasing the oscillation frequency without sacrificing selectivity. Maximum selectivity would be observed by operating the highest practical oscillation frequency upon the lowest practical temperature to produce an acceptable reaction rate. We are currently investigating this possibility.</p><!><p>The authors declare no competing financial interest.</p>

PubMed Open Access

Linker modification reduced the renal uptake of technetium-99m-labeled Arg-Ala-Asp-conjugated alpha-melanocyte stimulating hormone peptide

The purpose of this study was to examine the biodistribution of 99mTc-RAD-Arg-(Arg11)CCMSH in B16/F1 melanoma-bearing C57 mice to determine whether the replacement of the Lys linker with an Arg linker could decrease the renal uptake of 99mTc-RAD-Arg-(Arg11)CCMSH. 99mTc-RAD-Arg-(Arg11)CCMSH exhibited rapid and high tumor uptake (17.98 \xc2\xb1 4.96% ID/g at 2 h post-injection) in B16/F1 melanoma-bearing C57 mice. As compared to 99mTc-RAD-Lys-(Arg11)CCMSH, the replacement of the Lys linker with an Arg linker dramatically decreased the renal uptake of 99mTc-RAD-Arg-(Arg11)CCMSH by 68, 62, 73 and 64% at 0.5, 2, 4 and 24 h post-injection, respectively. Flank B16/F1 melanoma lesions were clearly imaged at 2 h post-injection using 99mTc-RAD-Arg-(Arg11)CCMSH as an imaging probe.

linker_modification_reduced_the_renal_uptake_of_technetium-99m-labeled_arg-ala-asp-conjugated_alpha-

<p>Melanocortin-1 (MC1) receptor is a G protein-coupled receptor which is over-expressed on human and mouse melanoma cells.1–5 Over the past several years, we have been developing a novel class of radiolabeled alpha-melanocyte stimulating hormone (α-MSH) peptides to target MC1 receptors for melanoma imaging.5–8 Specifically, the cyclic RXD {Arg-X-Asp-dTyr-Asp} motif (X = Gly, Ala, Ser, Val, Thr, Nle, Phe or dPhe) was conjugated to [Cys3,4,10, d-Phe7, Arg11]α-MSH3-13 {(Arg11)CCMSH} peptide via a lysine linker to generate a series of RXD-Lys-(Arg11)CCMSH peptides. Interestingly, single amino acid at the X position displayed a profound effect in melanoma targeting and clearance properties. First of all, we found that the switch from RGD to RAD dramatically improved the MC1 receptor binding affinity of RAD-Lys-(Arg11)CCMSH as compared to RGD-Lys-(Arg11)CCMSH in M21 and B16/F1 melanoma cells. 5,6 The stronger MC1 receptor binding yielded higher melanoma uptake for 99mTc-RAD-Lys-(Arg11)CCMSH than 99mTc-RGD-Lys-(Arg11)CCMSH. 6 Moreover, the residues of Ser, Val and Thr resulted in more favorable melanoma targeting and clearance properties than the residues of Nle, Phe or dPhe.7,8 However, all 99mTc-RXD-Lys-(Arg11)CCMSH peptides exhibited high non-specific renal uptake. Thus, it is desirable to reduce the renal uptake to facilitate their potential applications.</p><p>Despite the profound effect of single amino acid (Gly, Ala, Ser, Val and Thr) at the X position in 99mTc-RXD-Lys-(Arg11)CCMSH peptides, the positively-charged amino acid residues were same among the 99mTc-RXD-Lys-(Arg11)CCMSH peptides. Specifically, each 99mTc-RXD-Lys-(Arg11)CCMSH peptide has three arginine residues and one lysine linker. In our previous reports, L-lysine co-injection successfully reduced the renal uptake of 99mTc-RXD-Lys-(Arg11)CCMSH peptides by 37–55% at 2 h post-injection.6–8 Meanwhile, it was reported that the replacement of Lys with an Arg dramatically decreased the renal uptake of 99mTc-(Arg11)CCMSH and 99mTc-RGD-Arg-(Arg11)CCMSH by 41–64% at 2 h post-injection.9–11 Therefore, we hypothesized that the replacement of the Lys linker with an Arg linker would decrease the renal uptake of 99mTc-RXD-Lys-(Arg11)CCMSH peptides. Thus, as a proof-of-principal study, we prepared and evaluated the biodistribution property of 99mTc-RAD-Arg-(Arg11)CCMSH to examine our hypothesis in this study.</p><p>Firstly, RAD-Arg-(Arg11)CCMSH was synthesized using fluorenylmethyloxycarbonyl (Fmoc) chemistry, purified by reverse phase-high performance liquid chromatography (RP-HPLC) and characterized by electrospray ionization mass spectrometry according to our published procedure.12 The schematic structure of RAD-Arg-(Arg11)CCMSH is presented in Figure 1. The competitive binding study of RAD-Arg-(Arg11)CCMSH was determined in B16/F1 melanoma cells. The competitive binding curve of RAD-Arg-(Arg11)CCMSH is presented in Figure 2. The MC1 receptor binding affinity of RAD-Arg-(Arg11)CCMSH was 0.22 nM which was comparable to that of RAD-Lys-(Arg11)CCMSH (0.26 nM) in B16/F1 cells. The receptor binding result indicated that the replacement of the Lys linker with an Arg linker retained its nanomolar MC1 receptor binding affinity. 99mTc-RAD-Lys-(Arg11)CCMSH was readily prepared with greater than 95% radiolabeling yield and was separated from its excess nonlabeled peptide by high performance liquid chromatography (HPLC). The radiochemical purity of 99mTc-RAD-Lys-(Arg11)CCMSH was greater than 98%. The specific activity of 99mTc-RAD-Lys-(Arg11)CCMSH was 8.406 × 109 MBq/g.</p><p>Secondly, cellular internalization and efflux properties of 99mTc-RAD-Arg-(Arg11)CCMSH were examined in B16/F1 melanoma cells. The cellular results are presented in Figure 3. 99mTc-RAD-Arg-(Arg11)CCMSH exhibited rapid cellular internalization and extended cellular retention. Approximately 77.81 ± 4.46% of the 99mTc-RAD-Arg-(Arg11)CCMSH activity internalized at 20 min post incubation, whereas 81.26 ± 4.81% of the 99mTc-RAD-Arg-(Arg11)CCMSH activity internalized after 2 h incubation. The efflux results demonstrated that 72.07 ± 1.61% of the 99mTc-RGD-Arg-(Arg11)CCMSH activity remained inside the cells 2 h after incubating cells in culture medium. 99mTc-RAD-Arg-(Arg11)CCMSH displayed similar rapid internalization and extended retention pattern as 99mTc-RAD-Lys-(Arg11)CCMSH.11</p><p>Thirdly, the melanoma targeting and pharmacokinetic properties of 99mTc-RAD-Arg-(Arg11)CCMSH were determined in B16/F1 melanoma-bearing C57 mice. The biodistribution results are presented in Table 1. 99mTc-RAD-Arg-(Arg11)CCMSH exhibited rapid and high tumor uptake in melanoma-bearing mice. The tumor uptake was 17.98 ± 4.96 and 14.07 ± 2.90% ID/g at 2 and 4 h post-injection. In peptide blocking study, approximately 87% of the tumor uptake of 99mTc-RAD-Arg-(Arg11)CCMSH was blocked by 10 µg (6.1 nmol) of non-radiolabeled NDP-MSH at 2 h post-injection (p<0.05), demonstrating that the tumor uptake was specific and MC1 receptor-mediated. Whole-body clearance of 99mTc-RGD-Arg-(Arg11)CCMSH was rapid, with approximately 75% of the injected radioactivity cleared through the urinary system by 2 h post-injection (Table 1). Normal organ uptake of 99mTc-RGD-Arg-(Arg11)CCMSH was generally lower than 1.9% ID/g except for kidneys after 2 h post-injection. High tumor/blood and tumor/muscle uptake ratios were demonstrated as early as 0.5 h post-injection (Table 1). The renal uptake of 99mTc-RAD-Arg-(Arg11)CCMSH reached its highest value of 41.24 ± 4.48% ID/g at 0.5 h post-injection and decreased to 11.99 ± 2.29% ID/g at 24 h post-injection. In peptide blocking study at 2 h post-injection, the renal uptake was not significantly (p = 0.063) different with or without the peptide blockade, suggesting that the renal uptake was not receptor-mediated.</p><p>As compared to 99mTc-RAD-Lys-(Arg11)CCMSH,11 99mTc-RAD-Arg-(Arg11)CCMSH exhibited comparable melanoma uptake at 2, 4 and 24 h post-injection. The comparable melanoma uptake between 99mTc-RAD-Arg-(Arg11)CCMSH and 99mTc-RAD-Lys-(Arg11)CCMSH was attributed to the similar receptor binding affinities between RAD-Arg-(Arg11)CCMSH and RAD-Lys-(Arg11)CCMSH (0.22 vs. 0.26 nM). Despite that 99mTc-RAD-Arg-(Arg11)CCMSH and 99mTc-RAD-Lys-(Arg11)CCMSH displayed similar distribution pattern in normal organs, the renal uptake of 99mTc-RAD-Arg-(Arg11)CCMSH was dramatically lower than that of 99mTc-RAD-Lys-(Arg11)CCMSH. The renal uptake of 99mTc-RAD-Lys-(Arg11)CCMSH was 3.1, 2.6, 3.7 and 2.8 times the renal uptake of 99mTc-RAD-Arg-(Arg11)CCMSH at 0.5, 2 ,4 and 24 h post-injection, respectively. The comparable melanoma uptake and dramatically decreased renal uptake of 99mTc-RAD-Arg-(Arg11)CCMSH resulted in higher tumor to kidney uptake ratios as compared to 99mTc-RAD-Lys-(Arg11)CCMSH.</p><p>Finally, the melanoma imaging property of 99mTc-RAD-Arg-(Arg11)CCMSH was examined in a B16/F1 melanoma-bearing C57 mouse in this study. The representative whole-body single photon emission computed tomography (SPECT)/CT image is presented in Figure 4. Flank melanoma tumors were visualized clearly by 99mTc-RAD-Arg-(Arg11)CCMSH at 2 h post-injection. 99mTc-RAD-Arg-(Arg11)CCMSH exhibited high tumor to normal organ uptake ratios except for kidney, which was consistent with the biodistribution results (Table 1). The urine collected from the imaging mouse was analyzed for the metabolites by HPLC. The urinary HPLC profile of 99mTc-RAD-Arg-(Arg11)CCMSH is shown in Figure 6. 99mTc-RAD-Arg-(Arg11)CCMSH remained intact at 2 h post-injection.</p><p>In conclusion, the replacement of the Lys linker with an Arg linker dramatically decreased the renal uptake of 99mTc-RAD-Arg-(Arg11)CCMSH while retaining its high melanoma uptake. B16/F1 melanoma lesions were clearly visualized by SPECT/CT using 99mTc-RAD-Arg-(Arg11)CCMSH as an imaging probe. High melanoma uptake and decreased renal uptake of 99mTc-RAD-Arg-(Arg11)CCMSH suggests that the replacement of the Lys linker with an Arg linker may be useful in reducing the renal uptake of other 99mTc-RXD-Lys-(Arg11)CCMSH peptides in future studies.</p><p>The experimental details are presented in References and notes. 13–16</p><p>This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.</p>

PubMed Author Manuscript

Synthesis, Characterization, and Reactivity of a Uranium(VI) Carbene Imido Oxo Complex**

We report the uranium(VI) carbene imido oxo complex [U(BIPMTMS)(NMes)(O)(DMAP)2] (5, BIPMTMS=C(PPh2NSiMe3)2; Mes=2,4,6-Me3C6H2; DMAP=4-(dimethylamino)pyridine) which exhibits the unprecedented arrangement of three formal multiply bonded ligands to one metal center where the coordinated heteroatoms derive from different element groups. This complex was prepared by incorporation of carbene, imido, and then oxo groups at the uranium center by salt elimination, protonolysis, and two-electron oxidation, respectively. The oxo and imido groups adopt axial positions in a T-shaped motif with respect to the carbene, which is consistent with an inverse trans-influence. Complex 5 reacts with tert-butylisocyanate at the imido rather than carbene group to afford the uranyl(VI) carbene complex [U(BIPMTMS)(O)2(DMAP)2] (6).

synthesis,_characterization,_and_reactivity_of_a_uranium(vi)_carbene_imido_oxo_complex**

<!>

<p>There is burgeoning interest in covalent uranium–ligand (UL) multiple bonding because of the ongoing debate regarding the level and nature of covalency that these bonds may exhibit.[1] Uranium UL compounds containing covalent terminal monocarbene, -imido, -nitride, and -chalcogenide linkages are well known.[2–5] Homoleptic UL2 compounds are also well represented; in addition to a modest range of uranium biscarbene and bisimido complexes,[2b,d, 6] the bisoxo uranyl unit accounts for more than 50 % of all structurally characterized uranium complexes.[7] Recently, progress has been made preparing heteroleptic ULL′ complexes with carbene–oxo,[2k] imido–oxo,[8] nitrido–oxo,[9] and heavier chalcogen–oxo compounds.[10] Although both rare and challenging to prepare, these compounds are of interest with respect to the inverse trans-influence (ITI),[11] where strong donor ligands adopt trans geometries in contrast to d-block analogues that tend to adopt cis geometries. Concerning three-ligand multiple-bond linkages to uranium, examples are limited to homoleptic systems, such as uranium trioxide and triscarbenes,[2c] or heteroleptic UL2L′ systems with a maximum of two different types of multiply bonded ligands, such as a uranyl–carbene.[2f] Remarkably, no heteroleptic ULL′L′′ uranium complexes containing three different multiple bond linkages to uranium have ever been reported. Furthermore, even for d-block complexes where metal–ligand (ML) multiple bonding is more favorable, homoleptic combinations are so dominant that there are no examples of MLL′L′′ multiply bonded complexes containing heteroatoms from different element groups; the only example of a MLL′L′′ complex from hundreds of examples of ML multiple bond complexes is the all-chalcogen complex [W(C5Me5)(O)(S)(Se)][PPh4], reported over a decade ago.[12] This paucity may reflect the difficulties of constructing different covalent ML multiple bonds at a metal center whilst avoiding decomposition of previously installed multiple bonds. Although MLL′L′′ complexes utilizing heteroatoms from different element groups are yet to be reported, their synthesis would establish new synthetic strategies, give structure–bonding insights, and allow competitive reactivity studies to be investigated.</p><p>Herein, we describe the synthesis of a uranium(VI) carbene imido oxo complex, which is the first example of a metal complex to exhibit formal covalent multiple-bond interactions to three different ligands with heteroatoms from different element groups, and we describe its structure, bonding, and preliminary reactivity.</p><p>The starting material [U(BIPMTMS)(Cl)(μ-Cl)2Li(THF)2] (1, BIPMTMS=C(PPh2NSiMe3)2)[2j] was treated with two equivalents of benzyl potassium to afford, after workup and recrystallization, the brown uranium(IV) carbene dialkyl complex [U(BIPMTMS)(CH2Ph)2] (2) in 72 % yield.[13] Although a number of uranium carbene derivatives have now been reported, dialkyls were unknown.[2] Treatment of 2 with mesitylamine, Scheme 1,[13] gave the uranium(IV) carbene imido complex [{U(BIPMTMS)(μ-NMes)}2] (3) as brown crystals in 92 % yield. We tested the oxidation of 3 with common oxygen-atom-transfer reagents and found that whilst morpholine N-oxide, pyridine-N-oxide, and trimethylamine-N-oxide all gave intractable products, tetramethylpiperidine-N-oxide (TEMPO) effected clean oxidation to afford the black uranium(VI) carbene imido oxo complex [{U(BIPMTMS)(NMes)(μ-O)}2] (4) as a crystalline product in 57 % yield. Complex 4 was treated with two equivalents of DMAP to give the uranium(VI) carbene imido oxo complex [U(BIPMTMS)(NMes)(O)(DMAP)2] (5) as black crystals. Complex 5 can also be prepared in 49 % yield from a one-pot reaction of 3 with TEMPO and two equivalents of DMAP.</p><p>Synthesis of compounds 3–5. (Compound 2 was prepared from 1 and KCH2C6H5).</p><p>The characterization data for compounds 2–5 are consistent with their formulations. The 31P NMR spectrum of 5 has a resonance signal at δ=−22 ppm, shifted from δ=−35 ppm for 4. Despite exhaustive attempts we could not locate the carbene resonances in the 13C NMR spectra of 4 or 5 in the range δ=−200 to +1000 ppm and no folded-in resonances could be detected; 2D 13C-31P NMR experiments showed only one cross-peak for the P-phenyl ipso carbon atoms. In contrast, in a cerium(IV) BIPMTMS carbene complex, this method easily located the carbene resonance at δ=+325 ppm.[14] The FTIR spectra of 4 and 5 exhibit strong bands at 837 and 900 cm−1, which we attribute to bridging and terminal oxo groups, respectively. The UV/Vis electronic absorption spectra of 2–5 are dominated by LMCT (ligand-to-metal charge transfer) absorptions that tail in from the UV region to the visible and the NIR regions are generally featureless (4 and 5) or exhibit very weak f→f absorptions (2 and 3). The profile of the experimental UV/Vis absorption spectrum of 5 is reproduced well by SAOP/ZORA/TZP TD-DFT calculations, with the absorptions in the λ=400–750 nm range arising principally from LMCT transitions involving the carbene and imido lone pairs to vacant uranium 5f orbitals.[13] The uranium(IV) formulations of 2 and 3 were confirmed by SQUID magnetometry.[13] The magnetic moment of 2 is 2.6 μB at 298 K and this falls to 0.8 μB at 1.8 K. For 3, the magnetic moment at 298 K is 3.4 μB (2.4 μB per uranium center) and this falls to 1.04 μB at 1.8 K (0.7 μB per uranium center). The magnetic moments of 2 and 3 both tend to zero and are consistent with uranium(IV) which is a magnetic singlet at low temperature. We find no evidence of magnetic coupling between the two uranium(IV) centers in 3, but coupling between uranium(IV) centers is rarely observed.[15]</p><p>Compounds 2–5 have been characterized by single-crystal X-ray diffraction.[13] The structure of 5 (Figure 1b) confirms the monomeric formulation. In this structure, the uranium center is coordinated to terminal carbene, imido, and oxo groups with two coordinated molecules of DMAP completing a pentagonal-bipyramidal coordination sphere. Notably, the oxo and imido groups adopt axial positions in a T-shaped motif with respect to the carbene. However, unlike uranyl which typically exhibits O-U-O angles of more than 172°, the N-U-O angle is distorted significantly from linearity at 167.14(9)°. This angle is close to the angle of 161° measured in gas-phase UO3 which also adopts a distorted T-shaped geometry.[16] The N-U-O angle in 5 is slightly closer to linearity than in 4 (160.49(11)°; Figure 1a), but this most likely reflects the increase in coordination number at uranium in 5 (seven-coordinate) compared to 4 (six-coordinate). This is supported by the significantly different U—NDMAP bond lengths in 5 (2.592(2) and 2.611(2) Å) that are consistent with a more congested coordination environment in 5 compared to 4. Also, the N-U-O bond angle in 4 may be distorted because of the bridging oxo groups. The U—O bond length in 5 is 1.814(2) Å, which is approximately 0.14 Å shorter than in 4 presumably as a result of its terminal nature. The U—Nimido linkage is essentially linear (U-N-Cipso ∡=174.2(2)°) and the U—Nimido bond length of 1.921(2) Å in 5 is comparable to 4. The U—O and U—N bond lengths in 5 are each approximately 0.1 Å longer than the analogous distances in [U(NtBu)(O)(I)2(OPPh3)2],[8c] perhaps reflecting the presence of the BIPMTMS carbene. The U—Ccarbene bond length of 2.400(3) Å in 5 is indistinguishable from the analogous bond length in 4 (2.408(3) Å) and is essentially the same as the analogous distances in 2 and 3 (2.351(4) and 2.396(10) Å, respectively).[13] This similarity may reflect the constraints imposed on the carbene by residing in a pincer ligand, but also that with two π-donor ligands already coordinated to uranium this metal ion is electron-rich. A similar effect has been observed in the uranyl(VI) carbene complex [UO2{C(PPh2S)2}(C5H4N)2].[2f] Note that the imido rather than the carbene is trans to the oxo in 4 and 5, an observation which can be rationalized by an ITI effect. For actinyls, the semi-core 6pz orbital hybridizes with and transfers charge to 5f orbitals. This transfer leaves a hole in the 6pz orbital directed to the trans position so that the ligand bonds more strongly to compensate.[11] Taking the oxo as the reference group, the ligand that in principle can donate trans electron density most strongly, and hence compensate for the 6p hole the most, is the imido group, which is experimentally observed.</p><p>Single-crystal X-ray structures of a) [{U(BIPMTMS)(NMes)(μ-O)}2] (4), b) [U(BIPMTMS)(NMes)(O)(DMAP)2] (5), and c) [U(BIPMTMS)(O)2(DMAP)2] (6). Displacement ellipsoids set at 40 % probability. Hydrogen atoms, any lattice solvent, and minor disorder components are omitted for clarity. Selected bond lengths [Å]: 4: U1–C1 2.408(3), U1–N1 2.408(3), U1–N2 2.374(3), U1–N3 1.943(3), U1–O1 1.953(2), U1–O1A 2.337(2); 5: U1–C1 2.400(3), U1–N1 2.554(2), U1–N2 2.577(2), U1–N3 1.921(2), U1–N4 2.592(2), U1–N5 2.611(2), U1–O1 1.814(2); 6: U1–C1 2.383(3), U1–N1 2.606(2), U1–N2 2.600(2), U1–N3 2.564(3), U1–N4 2.594(3), U1–O1 1.794(2), U1–O2 1.785(2).</p><p>We conducted DFT calculations on complex 5 which compare well to the experimental solid-state data and we conclude the calculations represent a qualitative model of the electronic structure of 5. Donation of electron density from the ligands to the uranium center in 5 is suggested by calculated charges of +3.66, −1.91, −1.24, and −0.90 for the uranium, carbene, imido, and oxo centers, respectively. The BIPMTMS P- and N-centers exhibit calculated charges of +1.56 and −1.43, respectively. The calculated charges suggest that the dipolar resonance form of BIPM dominates in this complex.[17] The P—N and P—Ccarbene Nalewajski–Mrozek (NM) bond indices are calculated as 1.09 and 1.10, respectively. Multiple-bond interactions to uranium from the carbene, imido, and oxo groups are suggested by NM bond indices of 1.23, 2.34, and 2.68, respectively. For comparison, the formally dative imino and pyridine U—N NM bond indices average 0.69 and 0.40, respectively. Uranium BIPM carbenes exhibit NM bond indices in the range 1.2–1.5 for the U—C interaction,[2] and the imido and oxo bond indices are consistent with threefold bonding interactions. Examination of the Kohn–Sham orbitals of 5 reveals a frontier orbital manifold that exhibits σ- and π-interactions involving the carbene, imido, and oxo donors. However, these orbitals are extensively delocalized across each donor group and the uranium center, precluding an assessment of ITI effects; this contrasts to calculations on 6 (see below) where the orbitals are more localized as discrete U—C or [O—U—O]2+ combinations.[13]</p><p>To develop a more chemically intuitive bonding picture of 5 we examined the uranium carbene, imido, and oxo bonding interactions by natural bond orbital (NBO) analysis.[13] The uranium–carbene σ-bond is composed of 15 % U and 85 % C character. From this σ-bond, the uranium component contains 0.4 % 7s-, 0.3 % 7p-, 19.6 % 6d-, and 79.7 % 5f-orbital contributions whereas the carbon component is composed of 16.6 % 2s- and 83.4 % 2p-orbital contributions. The uranium–carbene π-bond is composed of 18.4 % U and 81.6 % C contributions. The carbon component of this bond is essentially 100 % 2p-orbital hybridized, reflecting the π-character of this orbital, whereas the uranium component comprises 0.4 % 7s-, 0.2 % 7p-, 5.5 % 6d-, and 93.9 % 5f-orbital contributions. The two uranium–imido π-bonds are essentially identical and are composed of 23.3 % U and 76.7 % N contributions. The uranium component comprises 11.7 % 6d-and 88.3 % 5f-orbital contributions with no 7s or 7p components whereas the nitrogen component comprises essentially 100 % 2p-orbital character, in agreement with the π-bonding nature of these orbitals. No formal U—Nimido σ-bond was indicated by the NBO calculations. The U—O π-bonds are returned as being primarily localized on the oxygen whereas the U—O σ-bond is identified by NBO as being composed of 23.3 % U and 76.7 % O character. The uranium component has 1.5 % 7s-, 0.3 % 7p-, 9.3 % 6d-, and 88.9 % 5f-orbital character whereas the oxygen contributions are 12.7 % 2s and 87.3 % 2p. The calculations suggest that uranium principally employs 5f rather than 6d orbitals in the multiple bonds to the carbene, imido, and oxo centers in 5 as has been determined in other uranium–ligand multiple bonds.[2–4]</p><p>To provide a topological analysis of the UL interactions in 5, we used Bader's quantum theory of atoms in molecules (QTAIM). In QTAIM, a chemical bond is defined by the presence of a line of locally maximum electron density [ρ(r)] along a bond path between two atoms and by a bond critical point (BCP) representing the minimum in the electron density along the locally maximal line. For a covalent bond, ρ(r) at the BCP between two nuclei is usually greater than 0.1 and the electronic energy-density term H(r) is usually negative for a covalent bond. The calculated ρ(r) and H(r) values for the U—C, U—N, and U—O 3,−1 BCPs are 0.092/−0.031, 0.185/−0.109, and 0.247/−0.186, respectively. The corresponding values for the uranium–imino and uranium–pyridine dative bonds average 0.0478/−0.004, respectively. The ellipticity of a BCP provides quantification of the σ/π character of a bond; for a σ- or σ-/2π-bond, which present cylindrical contours of electron density, the ellipticity is approximately 0, and for a σ-/π-bond the ellipticity is greater than 0 arising from the asymmetric electron-density distribution which is perpendicular to the bond path. The ellipticities of ethane, benzene, ethene, and acetylene are calculated to be 0.00, 0.23, 0.45, and 0.00, respectively.[18] Group 6 carbonyl complexes exhibit ellipticities of approximately 0,[19] whereas M—C interactions exhibit ellipticities in the range 0.20–0.62.[20] The calculated ellipticity for the U—C bond in 5 (0.21) is comparable to the C—C bonds in benzene. For the U—C interactions in complexes [U(BIPMTMSH)(Cl)3(THF)],[2e] [U(BIPMTMS)(I)2(Cl)],[2j] [UOCl2(BIPMTMS)],[2k] and [U(C5H5)3C(H)PMe3],[21] we previously calculated ellipticities of 0.04, 0.35, 0.38, and 0.26, respectively. Where only a spherical σ-bonding interaction is possible in the first of this series the ellipticity is approximately 0, but for the remaining complexes the ellipticities are similar to those calculated for 5. The U—N bond ellipticity (0.11) in 5 is smaller than the U—C interaction and its deviation from zero most likely reflects conjugative effects to the N-aryl ring.[18] The ellipticity for the uranium–oxo bond (0.03) suggests a triple-bond interaction.[18]</p><p>Whilst we note that dipolar U+—L− resonance structures will contribute to the overall bonding picture of the uranium–ligand multiple bonds in 5, the combined computational data are in agreement; all identify multiple bond combinations for all of the uranium–carbene, -imido, and -oxo linkages that are polarized but which involve more than one electron-pair per heteroatom and are thus multiple in nature.</p><p>Our preliminary investigations on the reactivity of complex 5 have shown it to be reactive. Complex 5 was allowed to react with tert-butylisocyanate to afford the uranyl carbene complex [U(BIPMTMS)(O)2(DMAP)2] (6; Figure 1c) [13] as black crystals in 67 % yield with concomitant elimination of tert-butylmesitylcarbodiimide (Scheme 2).[13] The identity of the carbodiimide by-product was confirmed by comparison of the NMR spectra to literature data.[22] Although the resonance signals in the 31P NMR spectra of 5 and 6 are within 0.3 ppm of each other (δ≈−22 ppm), the reaction of 5 with tert-butylisocyanate proceeds via an intermediate that we could not isolate. This intermediate exhibits a 31P NMR resonance at δ=−44 ppm which suggests the formation of a [2+2]-cycloaddition product.[6d]It is germane to note that all previous attempts to prepare complex 6, by deprotonation of [UO2(Cl)(BIPMTMSH)(THF)][23] with a wide range of bases, or oxidation of carbene precursors, failed and instead afforded pentavalent or hexavalent uranyl methanides.[2k,l]</p><p>Synthesis of complex 6 from complex 5.</p><p>To conclude, by installing carbene, imido, and oxo groups at a uranium center by salt elimination, protonolysis, and two-electron oxidation, it has been possible to prepare a complex with three formal covalent multiply bonded ligands where the coordinated heteroatoms derive from different element groups. Computational analyses suggest formal U—C double bond and triple-bonding interactions for the imido and oxo linkages. In all cases, the computational data suggest the dominance of uranium 5f rather than 6d orbitals in the three multiple bonds. The delocalization of the frontier orbitals involved in the uranium–carbene, -imido, and -oxo interactions suggests that the intuitive formulation of 5 as a carbene N—U—O uranyl analogue is not appropriate. This conclusion is also consistent with the preliminary reactivity study of 5 which has enabled the preparation of a previously inaccessible uranyl carbene complex through N for O metathesis reactivity at the imido group,[3o, 6d] rather than at the carbene.</p><!><p>Supporting information for this article is available on the WWW under http://dx.doi.org/10.1002/anie.201403892.</p>

PubMed Open Access

An engineered eukaryotic protein glycosylation pathway in Escherichia coli

We performed bottom-up engineering of a synthetic pathway in E. coli for the production of eukaryotic trimannosyl chitobiose glycans and the transfer of these glycans to specific asparagine residues in target proteins. Glycan biosynthesis was enabled by four eukaryotic glycosyltransferases, including the yeast uridine diphosphate-N-acetylglucosamine transferases Alg13 and Alg14 and the mannosyltransferases Alg1 and Alg2. By including the bacterial oligosaccharyltransferase PglB from C. jejuni, glycans were successfully transferred to eukaryotic proteins.

an_engineered_eukaryotic_protein_glycosylation_pathway_in_escherichia_coli

<p>N-linked protein glycosylation is the most common post-translational modification in eukaryotes, affecting many important protein properties1. N-linked glycosylation is not limited to eukaryotes, however, as bona fide N-linked glycosylation pathways are found in proteobacteria2 and can be transferred to E. coli3. There are several notable differences between bacterial and eukaryotic N-glycosylation systems. First, bacteria assemble oligosaccharides on undecaprenyl pyrophosphate (Und-PP) in the cytoplasmic membrane whereas eukaryotes use dolichyl pyrophosphate (Dol-PP) in the ER membrane. Second, the N-X-S/T consensus sequence for N-glycosylation in eukaryotes appears to be extended to D/E-X−1-N-X+1-S/T (X−1, X+1 ≠ P) in bacteria4 with few exceptions5,6. Third, bacterial N-glycans are completely distinct from any known eukaryotic glycan7. As a result, glycoproteins derived from existing bacterial expression systems are restricted to bioconjugate vaccines8,9 or glycoproteins that require extensive in vitro modification10. The construction of a eukaryotic glycosylation pathway in E. coli that generates human-like N-glycans remains an elusive challenge despite much speculation7,11.</p><p>To address this challenge, we focused on engineering E. coli to produce mannose3-N-acetylglucosamine2 (Man3GlcNAc2) glycans. We chose Man3GlcNAc2 because it is: (i) the core structure common to all human N-glycans; (ii) the predominant N-glycan produced by baculovirus-insect cells, carrot root plant cells, and Tetrahymena thermophila, all of which yield glycans that are fit for pre-clinical and clinical products; and (iii) the minimal glycan required for a therapeutic glycoprotein currently on the market12. To generate Man3GlcNAc2 on the cytoplasmic membrane of E. coli, a synthetic pathway was designed (Fig. 1).</p><p>The first step in this pathway involved WecA, an endogenous glycosyltransferase (GTase) that transfers GlcNAc-1-phosphate to undecaprenyl phosphate (Und-P). To extend the glycan, several heterologous GTases from Saccharomyces cerevisiae were selected because these can be solubly expressed in E. coli13–15 and in some cases the expressed enzymes are active in vitro13,14. Specifically, for addition of the second GlcNAc residue to GlcNAc-PP-Und, we chose the S. cerevisiae β1,4-GlcNAc transferase that is comprised of the Alg13 and Alg14 subunits. In yeast, Alg14 is an integral membrane protein that functions as a membrane anchor to recruit soluble Alg13 to the cytosolic face of the ER membrane15, where synthesis of GlcNAc2-PP-Dol occurs. Consistent with their localization in yeast, both Alg13 and Alg14 localized in the membrane fraction of E. coli while Alg13 was also detected in the soluble fraction (Supplementary Fig. 1). For the subsequent steps, we employed S. cerevisiae β1,4-mannosyltransferase Alg1, which specifies the addition of the first mannose to the glycan14, and the bifunctional mannosyltransferase Alg2, which carries out the addition of both an α1,3- and α1,6-mannose in a branched configuration13. Like Alg13/14, both Alg1 and Alg2 localized in the membrane fraction of E. coli (Supplementary Fig. 1).</p><p>To determine if enzyme co-expression was capable of producing a functional Man3GlcNAc2 biosynthesis pathway, we constructed plasmid pYCG that encoded a synthetic gene cluster comprised of ALG13, ALG14, ALG1 and ALG2 (Supplementary Fig. 2). To increase the availability of the GDP-mannose substrate for Alg1 and Alg2, GDP-mannose dehydratase (GMD) that converts GDP-mannose to GDP-4-keto-6-deoxymannose in the first step of GDP-L-fucose synthesis was deleted from E. coli strain MC4100. To assay glycan synthesis, we exploited the fact that bacterial cell surfaces can display engineered oligosaccharides in their lipopolysaccharide layer16,17. This approach depends upon the O-antigen ligase WaaL, which catalyzes the transfer of Und-PP-linked oligosaccharides to lipid A. These oligosaccharides are shuttled to the cell surface where they can be conveniently labeled16,17. Upon labeling with fluorescent concanavalin A (ConA), a lectin that binds terminal α-mannose, MC4100 gmd::kan cells expressing the synthetic pathway but not empty-vector control cells became highly fluorescent (Fig. 2a). The fluorescence was clearly localized on the cell surface (Supplementary Fig. 3a). In the absence of ALG1 or ALG2, cell fluorescence was significantly diminished (Supplementary Fig. 3b) confirming that these enzymes were required for producing surface-associated α-mannose residues. Likewise, when the synthetic pathway was expressed in MC4100 gmd::kan that also lacked waaL, cells were minimally fluorescent (Fig. 2a) confirming WaaL-dependent transfer of α-mannose-containing oligosaccharides to lipid A. Importantly, a native E. coli flippase (e.g., Wzx) must be involved since WaaL uses Und-PP-linked oligosaccharides that are present on the periplasmic face of the cytoplasmic membrane18.</p><p>To verify the glyan structure, lipid-linked oligosaccharides (LLOs) were extracted and characterized by matrix-assisted laser desorption/ionization tandem time-of-flight (MALDI-TOF/TOF) analysis. The MALDI-MS spectrum revealed Hex3HexNAc2 as the primary oligosaccharide, consistent with the expected Man3GlcNAc2 glycan. In addition, Hex2HexNAc2 and Hex4HexNAc2 were detected (Fig. 2b). The MALDI-MS spectrum of LLOs isolated from MC4100 gmd::kan ΔwaaL cells also revealed Hex3HexNAc2 as the primary oligosaccharide (Supplementary Fig. 4). This confirmed that the lack of cell surface labeling observed for these cells was a result of the waaL deletion and not the inability to synthesize oligosaccharides. Finally, released glycans analyzed by 1H NMR spectroscopy were consistent with the eukaryotic core glycan Manα1–3(Manα1–6)-Manβ1–4-GlcNAcβ1–4-GlcNAc (Supplementary Figs. 5–7). NMR analysis also revealed a residue with H-1 (5.080 ppm) and H-2 (4.065 ppm) chemical shifts indicating that the fourth hexose residue was likely Man linked to one of the branching Man residues (Supplementary Fig. 5). The presence of a putative Man4GlcNAc2 was surprising because elongation of Man3GlcNAc2 is attributed to the bifunctional Alg1113. It should be noted, however, that both Man3GlcNAc2-PP-Dol and Man4GlcNAc2-PP-Dol accumulated in a S. cerevisiae ALG11 mutant19, suggesting that Alg1 or Alg2 may catalyze Man4GlcNAc2-PP-Dol production in vivo.</p><p>To transfer Man3GlcNAc2 glycans to secretory glycoproteins in vivo, we focused our attention on PglB from C. jejuni (PglBCj) because it is the best characterized bacterial OTase20 and can utilize diverse Und-PP-linked oligosaccharides as substrates2,3,8,9. For glycoprotein targets, we chose (i) E. coli maltose binding protein (MBP) which is a native periplasmic protein and (ii) anti-β-galactosidase single-chain antibody fragment called scFv13-R4 that was modified with an N-terminal co-translational export signal from E. coli DsbA17. These proteins were each modified C-terminally with four tandem repeats of the bacterial glycan acceptor motif DQNAT17. MC4100 gmd::kan ΔwaaL cells were transformed with plasmids encoding one of these target proteins and the Man3GlcNAc2 pathway with PglBCj (Supplementary Fig. 2). The MBP4x-DQNAT and scFv13-R44x-DQNAT produced in these cells, but not in cells carrying an inactive PglBCj mutant3, was bound by ConA (Fig. 3a and Supplementary Fig. 8). When these target proteins were first treated with peptide:N-glycosidase F (PNGase F), an amidase that specifically cleaves between a reducing-end GlcNAc and asparagine, ConA binding was eliminated (Fig. 3a). To further confirm that glycans were linked specifically to asparagines in target proteins, a version of scFv13-R4 with a single C-terminal DQNAT sequon was digested with Pronase E and the resulting glycopeptides were identified using MS21. The major ion seen at m/z 1282 was consistent with Man3GlcNAc2-Asn, wherein the asparagine residue underwent β-elimination during the permethylation procedure (Fig. 3b)21. MS analysis of the PNGase F-released glycans from glycosylated scFv13-R44x-DQNAT revealed Hex3HexNAc2 as the predominant glycoform along with a lesser amount of Hex4HexNAc2 (Fig. 3c). MS2 sequencing of the glycan at m/z 1171 confirmed the biantennary trihexosyl structure (Supplementary Fig. 9a). When PNGase F-released glycans were treated with α-exomannosidase to specifically hydrolyze terminal α-mannose residues, HexHexNAc2 emerged as the major glycoform at the expense of both Hex3HexNAc2 and Hex4HexNAc2 (Supplementary Fig. 9b). Finally, 1H NMR analysis on PNGase F-released glycans was consistent with Manα1–3(Manα1–6)-Manβ1–4-GlcNAcβ1–4-GlcNAc (Supplementary Figs. 10 and 11).</p><p>We next attempted to transfer Man3GlcNAc2 to eukaryotic glycoproteins including: (i) the Fc domain of human IgG1 at its conserved N297 glycosylation site, (ii) bovine ribonuclease A (RNaseA) at its N34 acceptor site, and (iii) the placental variant of human growth hormone (hGHv) at its N140 glycosylation site. The genes encoding these proteins were cloned downstream of an N-terminal DsbA export signal or full-length MBP in the case of hGHv. Since the N-X-S/T consensus motif in eukaryotes is extended to D/E-X−1-N-X+1-S/T in bacteria4, we mutated the native glycosylation motifs in the Fc (QYNST, residues 295–299) and hGHv (IFNQS, residues 138–142) to DQNAT. Likewise, we used an RNaseA variant with an S32D substitution22. Expression of these target proteins in cells carrying the pYCG-PglBCj plasmid yielded clearly glycosylated proteins (Supplementary Fig. 12a and b). It should be noted that RNaseA glycosylation was unexpected because the acceptor site is located in a structured domain that is not glycosylated by PglBCj in vitro22 Hence, our data indicate that PglBCj can glycosylate residues in both unstructured and structured regions of eukaryotic acceptor proteins in vivo.</p><p>Since it does not have native glycosylation pathways, our engineered E. coli strain is the only platform for glycoprotein expression that offers bottom-up synthesis of precise glycan structures by expression of diverse GTases and OTases. Despite our success, however, there remain some important challenges that need to be overcome for the practical application of this technology. For example, an acidic group at the -2 position to the asparagine seems to be a common prerequisite of PglB homologs for efficient glycosylation4. Relaxed acceptor site specificity has been reported for C. lari and Desulfovibrio desulfuricans PglB homologs5,6. However, this has only been shown for one very unique site (271DNNNST276) in the C. jejuni AcrA acceptor protein. PglBCl did not glycosylate the wild-type CH2 domain of a human IgG15. In our hands, PglBCj and PglBCl were able to transfer Man3GlcNAc2 to extended sites (Supplementary Fig. 12c) but not to minimal glycosylation sites in engineered or eukaryotic target proteins (data not shown). Another issue is that only a small fraction (<1%) of each expressed protein was glycosylated under the conditions tested here. With that said, the yield of glycosylated proteins has reached up to ~50 μg/L in our hands and might be further improved by increasing expression in the periplasm, relieving enzymatic and metabolic bottlenecks, and/or optimizing the glycosylation enzymes. Along these lines, simple optimization strategies have previously been used to generate nearly 25 mg/L of bacterial glycoproteins in E. coli9. We anticipate further improvements will be achieved by applying new glyco-display technologies including cell surface and phage display systems17,23,24. Such methods will be needed to create bacterial OTase variants that efficiently glycosylate minimal N-X-S/T acceptor sites. Alternatively, novel bacterial OTases with distinct properties6 or single-subunit eukaryotic OTases25 could prove useful. Overall, the engineering of defined glycosylation pathways in E. coli sets the stage for further engineering of this host for the production of vaccines and therapeutics with even more structurally complex human-like glycans. Moreover, glycoengineered E. coli has the potential to serve as a model genetic system for deciphering the "glycosylation code" which governs the non-template driven synthesis of diverse glycans and their specific attachment to proteins.</p>

PubMed Author Manuscript

Identification of Multiple Structurally-Distinct, Nonpeptidic Small Molecule Inhibitors of Protein Arginine Deiminase 3 Using a Substrate- Based Fragment Method

The protein arginine deiminases (PADs) are a family of enzymes that catalyze the post-translational hydrolytic deimination of arginine residues. Four different enzymologically active PAD subtypes have been characterized and exhibit tissue-specific expression and association with a number of different diseases. In this Article we describe the development of an approach for the reliable discovery of low-molecular weight, nonpeptidic fragment substrates of the PADs that then can be optimized and converted to mechanism-based irreversible PAD inhibitors. The approach is demonstrated by the development of the first potent and selective inhibitors of PAD3, a PAD subtype implicated in the neurodegenerative response to spinal cord injury. Multiple structurally distinct inhibitors were identified with the most potent inhibitors having >10,000 min\xe2\x88\x921 M\xe2\x88\x921 kinact/KI values and \xe2\x89\xa510-fold selectivity for PAD3 over PADs 1, 2, and 4.

identification_of_multiple_structurally-distinct,_nonpeptidic_small_molecule_inhibitors_of_protein_a

INTRODUCTION<!>RESULTS AND DISCUSSION<!>Synthesis of Guanidine Substrate Library<!>Guanidine Library Screening Assay Method<!>Step 1: Hit Substrate Identification<!>Substrate evaluation<!>Hydantoin substrate synthesis and optimization<!>N-benzyl amide substrate synthesis and optimization<!>Step 3: Conversion of Substrates to Inhibitors<!>Inhibitor isozyme selectivity<!>CONCLUSION

<p>The protein arginine deiminases (PADs) are a family of enzymes that catalyze the post-translational hydrolytic deimination of arginine residues (Figure 1A).1–3 Several functionally active PAD subtypes, PAD1-4, have been characterized,4–7 and though the primary structure of mammalian PADs is highly conserved, the human isozymes exhibit tissue-specific expression patterns.3 Dysregulated PAD activity has been associated with multiple human diseases, including PAD1 for psoriasis,8 PAD2 for multiple sclerosis,9–12 and PAD4 for autoimmune disorders13 and certain cancers.14 Additionally, PAD3 has been implicated in the neurodegenerative response to spinal cord injury.15</p><p>The irreversible inhibitor Cl-amidine (Figure 1B) represents one of the most advanced PAD inhibitors.5,16–17 Due to its low MW, reasonably hydrophobic character, and nonpeptidic structure, Cl-amidine has shown activity in animal models18 and has contributed to an improved understanding of the role of PADs in different diseases. However, Cl-amidine shows modest isozyme selectivity, with greatest potency against PAD1 and only poor activity against PAD2 and PAD3.19 The lack of selectivity and moderate potency of Cl-amidine complicates deciphering the pharmacology of targeting the different isozymes. While more potent and selective larger peptidic inhibitors of PADs have been identified,20–22 their activity in cells and animals has not been reported, and their peptidic nature poses challenges for proteolytic stability, cell permeability, and rates of metabolic clearance. The identification of low MW, nonpeptidic, and isozyme-selective PAD inhibitors should facilitate a more thorough understanding of the individual roles of each PAD isozyme.</p><p>We have previously reported on a fragment-based approach for the discovery of enzyme inhibitors termed substrate activity screening (SAS).23 The SAS method consists of the identification of nonpeptidic substrate fragments,24 substrate optimization, and conversion of optimized substrates to inhibitors. The key advantage of this substrate-fragment discovery approach is that substrate hits are only identified upon productive binding and processing by the enzyme catalytic machinery. This approach minimizes undesirable false positives commonly observed in inhibitor screens, such as those due to small molecule micelle formation25–26 or the presence of trace reactive impurities. The comparative ease of synthesis and assay of substrates relative to inhibitors are additional advantages. We have successfully used this approach for the identification of selective low molecular weight inhibitors of therapeutically relevant proteases27–32 and phosphatases,33–35 and other labs have implemented related strategies to target kinases.36–37</p><p>Herein, we report on the development of the SAS method for the identification of low MW, nonpeptidic substrates and inhibitors of PADs. Moreover, we report on the identification of multiple structurally distinct and selective small molecule inhibitors of PAD3, for which potent and selective compounds have not previously been reported.38</p><!><p>The SAS method for the development of PAD inhibitors consists of three steps (Scheme 1): (1) a library of diverse, low molecular weight guanidines are screened for substrate activity using a colorimetric assay; (2) the identified weakly-cleaved guanidine substrates are optimized by analogue synthesis and subsequent screening; and (3) the efficiently-cleaved substrates are converted to inhibitors by direct replacement of the guanidine with the chloroacetamidine warhead, a known mechanism-based pharmacophore.5,39</p><!><p>More than 200 guanidine substrates were prepared by solution-phase parallel synthesis from primary amine starting materials. A subset of primary amines was selected using 2D extended connectivity analysis from thousands of commercially available amines with molecular weights below 300 Da. Each of the amines was converted into the corresponding guanidines using a one-step guanylation reaction (see Supporting Information). To achieve further substrate diversity, several additional guanidine substrates, containing a variety of heterocyclic scaffolds, were synthesized and included for screening. Subsequent to the identification of hit substrates, analogs of representative hits were also prepared. All guanidine library members were purified by preparative-scale reverse-phase chromatography and assayed for purity using LCMS and NMR spectroscopy.</p><!><p>The guanidine library was screened against PAD3 using a colorimetric coupled assay for the detection of urea-containing compounds.40 Briefly, PAD-mediated substrate turnover results in the formation of an ammonium ion and a urea product. In the presence of strongly acidic conditions and elevated temperatures, reaction of a urea functionality with diacetyl monoxime results in the formation of a chromogenic product that can be detected at 540 nm (Figure 3). This coupled assay was adapted for screening in 96-well plates and spectrophotometric plate readers to enable high throughput screening of the guanidine library. To serve as a background control each guanidine substrate was also submitted to the assay conditions without enzyme.</p><!><p>The guanidine substrate library was initially screened at 1 mM of substrate and 400 nM PAD3. From this screen, multiple distinct substrate classes were identified as weakly-cleaved substrate hits. For each of these hits the Km values were determined to be greater than 10 mM, and thus their relative cleavage efficiency accurately correlates with kcat/Km. (Table 1). Both indole substrate 1a and hydantoin substrates 3a and 4a incorporate known drug pharmacophores with multiple potential sites for diversification. The highest detected relative cleavage efficiency for 5a is also surprising because the amide carbonyl and NH are out of register relative to the placement of these functionalities in physiological Arg-based peptide substrates. Substrate 5a is moreover an attractive starting point for further optimization because it does not contain any chiral centers, and therefore straightforward introduction of alkenes and other conformational constraints within the alkane chain could be possible. Notably, these types of conformational constraints have proven beneficial in the development of sub-type selective histone deacetylase (HDAC) inhibitors.41 Based on these characteristics, we chose to pursue optimization of the hydantoin, benzyl hydantoin, and benzylamide scaffolds. Although benzodiazepine substrate 2a was not chosen for optimization, it represents another possibility for small molecule inhibitor development.</p><!><p>The Km values were determined for representative substrates and in all cases were >5 mM. Because substrate assays were performed at 1 mM, well below the substrate Km values, the relative substrate cleavage efficiencies directly correspond to the catalytic efficiencies (kcat/Km) of the substrates.24–30</p><!><p>Hydantoin derivatives were synthesized by addition of HArg( Pbf)-OMe to an isocyanate, followed by cyclization to give hydantoins using basic conditions that also ensure racemization of the methine proton. Racemic rather than enantiomerically pure substrates were synthesized because we established that rapid epimerization at the hydantoin stereocenter occurred at physiological pH and under the assay conditions (see Supporting Information).</p><p>Table 2 shows the relative kcat/Km of select substrates and depicts the optimization of a weakly-cleaved initial substrate hit (3a) to a substrate that is cleaved 17- fold more efficiently (11a). While methylation of the hydantoin at N1 completely abolished activity (6a), phenyl substitution of the hydantoin at N3 resulted in a slight improvement in cleavage efficiency and provided a site for further variation (7a). Evaluation of several phenyl substituted derivatives resulted in the identification of the 4-methoxyphenyl benzamide analogue 9a, cleaved with ~two-fold greater cleavage efficiency than the initial hit. Analogues 10a and 11a led to significant increases in cleavage efficiency.</p><p>Table 3 shows the relative kcat/Km of selected select substrates for the optimization of substrate hit 4a to substrate 15a cleaved almost three times more efficiently. Substrate 12a with meta-phenyl substitution showed a modest increase in kcat/Km. Further substitution upon this phenyl ring was therefore evaluated. Both meta-fluoro (13a) and ortho-chloro (14a) substituents increased substrate activity, and the combination of these substitutions showed a cumulative effect, leading to the most efficiently cleaved substrate in this series, 15a.</p><!><p>Derivatives of the N-benzyl-amide fragment 5a were synthesized by a carbodiimide-mediated coupling reaction between N, N′-di-Boc-protected γ-aminobutyric acid and various substituted benzylamines (see Supporting Information). As with the phenyl hydantoin series, methyl substitution of the amide NH resulted in a dramatic decrease in substrate activity (16a). Several substituents were introduced at the α-benzylic position, with the phenyl group (17a) resulting in more than a two-fold increase in cleavage efficiency as compared to 5a. Separately, substitutions on the benzyl aromatic ring were investigated, with the phenyl substituted substrate 18a being cleaved three times more efficiently than the original hit 5a. Substitutions around the secondary phenyl ring were also tolerated, most notably 19a, the most efficiently cleaved substrate in the series. The α-methyl substituted enantiomers 20a and 21a were also of interest because they showed strong chiral discrimination with the more active stereoisomer 21a being cleaved four times more efficiently than its enantiomer 20a.</p><!><p>Inhibitors were prepared by replacing the guanidine present in the identified substrates with the known chloroacetamidine irreversible inhibitor pharmacophore. Each substrate with the highest relative kcat/Km in the three substrate classes was converted to its corresponding inhibitor (Tables 2 – 4). The optimal N-phenyl hydantoin inhibitor 11b showed a kinact/KI of 5800 (min−1 M−1) towards PAD3 (Table 2), the most efficiently-cleaved N-benzyl amide substrate 19a resulted in inhibitor 19b with a kinact/KI of 13220 (min−1 M−1) (Table 4), and the most efficiently-cleaved N-benzyl hydantoin substrate 15a was converted to 15b, which was the most potent inhibitor to be identified with a kinact/KI of 17400 (min−1 M−1) (Table 3). These novel, nonpeptidic inhibitors represent distinct structural motifs capable of PAD3 inhibition and serve as useful templates for further optimization.</p><p>Additionally, many of the less efficiently cleaved substrates in each series were also converted to inhibitors to enable an assessment of the correlation of substrate cleavage efficiency to inhibitor activity (Tables 2–4). Within each compound series the relative cleavage efficiency and inhibitory potency correlated reasonably well. The most efficiently cleaved substrate also resulted in the most potent inhibitor for each series. However, correlation did not extend across the three series. For example, substrate 11a (Table 2) was the most efficiently cleaved substrate from all of the compound series, but it did not result in the most potent inhibitor. In fact, substrate 15a, which corresponded to most potent inhibitor 15b (Table 3), was ~two-fold less efficiently cleaved than 11a.</p><p>For a related series of substrates and mechanism-based inhibitors, the log[Km/kcat] often linearly correlates with log[KI] for the corresponding inhibitors incorporating stable transition state analogs.43–44 However, substrate and inhibitor correlation is often more complex. In some cases inhibitors better correlate with the corresponding substrate's ground-state binding (Km).35,45 For irreversible inactivators such as those employed in this study, inhibition might correlate better with the kcat term.46 Unfortunately, because the substrates reported here are not soluble at the high concentrations required to accurately measure Km, separate kcat and Km terms could not be determined.</p><!><p>The most potent inhibitor in each compound series was evaluated for isozyme selectivity (Table 5). Inhibitors 11b, 15b and 19b each were highly selective over PAD1 but showed more modest 5–6- fold selectivity over PADs 2 and 4. However, two of the more potent inhibitors in the N-benzyl hydantoin and N-benzyl amide series, 14b and 18b, respectively, showed ≥10-fold selectivity not only over PAD1 but also over PADs 2 and 4.47 Given the potency and selectivity observed for 14b and 18b, these two structures are particularly promising for biological studies as well as for further inhibitor development.</p><!><p>Low molecular weight, non-peptidic and selective inhibitors of the PAD isozymes have the potential to be powerful pharmacological tools for evaluating the roles of PADs in a number of disease states. This report describes the first discovery of PAD3 selective small molecule inhibitors. We have successfully implemented a substrate- based fragment discovery method for identifying PAD inhibitors by screening a library of guanidines to identify substrates, optimizing substrate structure for cleavage efficiency and then conversion to inhibitors by replacement of the guanidine by the chloroamidine inhibitor pharmacophore. This method enabled the rapid identification of three distinct classes of small molecule inhibitors. Inhibitor 14b, with a kinact/KI of 15600 towards PAD3, represents the most selective PAD3 inhibitor reported in the literature.</p>

PubMed Author Manuscript

Conversion of Wood into Hierarchically Porous Charcoal in the 200gram-scale using Home-built Kiln

Wood-to-charcoal is crucial in developing new materials at the lab-scale for relevant applications, such as pollutant removal from water. Unfortunately, laboratory carbonization methods are costly and produce charcoal on the gram-scale. This work presents a simple-to-build and simple-to-operate home-made kiln that carbonizes Eucalyptus wood chips (yield of 30  1%) and produces charcoal on the 200-gram scale.Solid particles had the typical structure, composition, and chemical behavior of charcoal obtained from wood. We believe that his carbonization process eases the charcoal synthesis required for the development of new charcoal-based materials.

conversion_of_wood_into_hierarchically_porous_charcoal_in_the_200gram-scale_using_home-built_kiln

Introduction<!>Materials and Methods<!>Carbonization of Eucalyptus Wood<!>X-ray Diffraction Analysis (XRD).<!>Energy dispersive x-ray spectroscopy (EDX).<!>Fourier Transformed Infrared Spectroscopy Measurements (FTIR) measurements.<!>Chemical Characterization of Solid Product.<!>Data Analysis.<!>3.Results and Discussion<!>Limitations and Strengths of this Study.<!>Implication and Future Research.<!>Conclusion

<p>Conversion of wood into charcoal at a laboratory scale is crucial in developing new materials for everyday uses, such as solid-phase extraction, heterogeneous catalysis, or water potabilization. Charcoals made from wood can have a hierarchical pore structure with macropores (50 nm < d < 100 m) that facilitate mass transfer throughout the material. Charcoals may also have mesopores (2 nm < d < 50 nm) and micropores (d < 2 nm) that increase the specific surface area up to 1000 m 2 .g -1 . Pore walls in charcoals have remarkable chemical stability under many relevant industrial processes (e.g., heterogeneous catalysis, pollutant removal from water) and contain diverse unsaturated organic groups containing mainly C. However, some chemical groups are also containing O and H in a much smaller proportion. The surface chemistry of charcoals can be finetuned by changing chemical groups at the interphase between pore walls and fluid within pores. The wall develops a higher affinity for specific molecules in the solution.</p><p>At a laboratory scale, wood has been converted into charcoal mainly with relatively expensive processes. These processes involve hardware such as glass (quartz) tubes heated with tubular ovens either under vacuum or the flow of inert gas (e.g., He, N2) that produces charcoal on the 1-gram scale. So, the cost of production and gram-scale outputs have been limiting the research of charcoal-based materials. Costs have been impeding access to equipment sometimes. Other times, a few grams of charcoal have not been enough to synthesize new materials and adequately test them afterward.</p><p>This work shows that a cheap, home-built kiln converts Eucalyptus wood chips into charcoal at the hundred-gram scale. During the study, we sought evidence by first comparing the yield of our carbonization process and then the properties (inner structure, composition, and chemical behavior in water) of the solid product with those reported in the literature.</p><!><p>Chemicals. NaNO3 (BIOPACK). Distilled water. HNO3 65% solution (Cicarelli). NaOH (S). Crystal violet (S) (Sigma, analytical grade).</p><p>Chips of Eucalyptus Wood. Eucalyptus branches were collected in the campus of CETMIC. After cutting leaves, branches were chipped (Oy Santasalo-Sohlberg Ab, Helsinski 50) (Scheme 1). Wood chips were stored in an open box until treatment in a home-built kiln.</p><!><p>Chips with Home-built Kiln. Scheme 1 describes the carbonization process. The small metallic cylinder (A) was filled with wood chips.</p><p>After covering with the big metallic cylinder (B), both cylinders were turned upside down, avoiding wood chips leaving the small cylinder. After filling the space between both cylinders with Eucalyptus branches and adding ethanol (50 mL), branches were set on fire. Elements C, D, and F (Scheme 1) were immediately positioned. Every 30 min, Eucalyptus branches were added to keep the fire burning. Two hours after the ignition of the fire, elements C, D, and F were removed, the fire was put out with distilled water (2 L). After pouring distilled water, elements A and B cooled down to room temperature.</p><p>Once cooled down to room temperature, elements A and B were turned upside down (ashes and partially combusted branches fell). The big cylinder was removed, and the black solid in the small cylinder was dried (electric furnace, 60 °C, 1 day). The dried, black solid was milled (10-L, plastic jar with screw cap; 25 ceramic balls (3 cm); 30 min) and sieved (ASTM #30 & #60, 250-600 m). Sieved particles -from now on, charcoal for the sake of simplicity-were stored in closed plastic bags until use after mixing all three samples (synthesis was performed by triplicate). A thorough description of elements A to F is given in Supplementary Information. The pore size distribution was calculated from the derivative of the accumulated specific pore volume. The position of the maxima of the distribution curves was estimated with the fit of Gaussian curves.</p><!><p>Diffractogram was obtained from dried powders (Philips PW-3710; Cu-Kα radiation λ=0.154 nm, 35 kV, 40 mA, step 0.04°, 2 s.step -1 ). Reflexes were assigned after comparing the diffractogram with reflexes from a database (Open Crystallographic Database, http://www.crystallography.net/cod/, software X'Pert HighScore).</p><p>Thermal Gravimetric & Differential Thermal Analysis (TGA-DTA). Measurement was performed on dried charcoal (Rigaku Thermo Plus II; drying 100 °C 2 h, 100 to 1000 °C at 10 °C.min -1 under air flow).</p><!><p>The dried powder was analyzed with EDX (JEOL JCM-6000 Neo Scope) to obtain an elemental composition and mapping of elements. Results were expressed as mass %.</p><!><p>Spectra of dried black powder were obtained (Bruker IFS 66).</p><!><p>Point of Zero Charge (PZC) Measurements. The addition method used to obtain the PZC of the black powder has been described previously [1] . Briefly, 10 flasks (10 mL, glass, screw cap) were filled with aqueous NaNO3 solution (0. Closed flasks were mechanically agitated (Decalab Rotolab-25, 24 h, room temperature 25 °C). The aqueous solution's pH was measured after exposure to charcoal. PZC was determined from ∆pH (pH values of solution before and after contact with charcoal) vs. initial pH. All experiments were done in triplicate.</p><p>Removal of Crystal Violet from solution in batch systems. Black powder´s ability to remove crystal violet from solution was investigated in batch systems (S/L ratio 10 g.L -1 ) starting at two different concentrations (C1 = 7.9 ± 0.4 ppm; C2 = 123.0 ± 0.4 ppm). Crystal violet solution (10.00 mL) and charcoal (100.0 mg) were added to 8 flasks (15 mL, glass, screw cap). Closed flasks were mechanically agitated (Decalab Rotolab-25, 25 °C) for t = X h (X = 0.25, 0.50, 0.75, 1, 3, 5, 7 y 24 for C1; X = 0.50, 0.75, 1, 3, 5, 7, 24 and 48 for C2). After centrifugation (10,000 rpm, 15 min), the concentration of crystal violet in aqueous solution was determined spectrophotometrically (HP 8453, 590 nm). All experiments were done in triplicate.</p><!><p>Yield of carbonization. The yield of the carbonization process was calculated (Equation 1) from the mass of charcoal (m) and of wood chips (m0):</p><p>Percentage of crystal violet removed in batch systems. Percentage of removal of crystal violet from solution in batch systems was calculated (Equation 2) considering the initial concentration Ci and the concentration at the time X h (Ct):</p><!><p>The process with Home-built Kiln compatible with Carbonization. Eucalyptus wood chips convert into a black solid with a yield of 30  1% (n = 3) (see Data in SI). The low dispersion indicates a reproducible process. The mean value is within yield values (28 -36%) reported for wood carbon with standard laboratory carbonization equipment [2][3][4] . In contrast, the combustion of Eucalyptus wood has much lower yields (1.9 -2.9%) [5,6] .</p><p>Hence, the yield of the process presented here is compatible with a carbonization process.</p><p>Solid Material having Wood-like Structure and Hierarchical Pores. After grinding and sieving the black solid, a particulate material (250 -600 m) was obtained. These particles have a wood-like structure [7][8][9][10] with hierarchical intraparticle pores. Intraparticle pores seem to have a trimodal size distribution. The smallest pores seem to peak at ca. 65 nm, mid-size pores at 2.5 m, and big pores at 50 -100 m (Figure 1 & Figure 2). SEM-images show big pores seem, but Hg porosimetry isotherms do not. The presence of big intraparticle pores seems to overlap with interparticle pores. This inner structure is also characteristic of charcoal obtained from wood with conventional carbonization equipment. Though being a complex process, the carbonization of wood starts with water evaporation below 200 °C. It follows with other condensation reactions that set free more water and other C-containing species. During this process, the lignin present in cell walls seems to preserve the inner structure while cellulose and hemicellulose largely dehydrate. Solid Material with Composition of Charcoal. The black particulate material has much C (78.8%), some O (14.6%), and a small amount of Ca and K (4.2% and 2.4%, respectively) (EDX). These values may somewhat overestimate the actual elemental composition, as H remains undetected with EDX. However, H contributes little (< 3%) to the composition of carbonized Eucalyptus wood [11,12] . For example, if we consider the presence of 3% H, the corrected composition in terms of C and O would be 76.5% and 14.2%, respectively. So, the composition trend remains unaltered for C and O. This trend in composition has been reported for carbonized Eucalyptus wood [11][12][13] .</p><p>The solid experienced a drastic mass loss (93.4%) when heated under airflow below 600 °C (Figure 9). Charcoal has been reported to burn when heated in air [14] similarly. This mass loss should originate in the combustion of most black solid.</p><p>FTIR spectra show the presence of alkane (3000-2840 (w), 1250-1000 (m) cm -1 ), alkene (2962-2853 (w), 1650-1500 (m), 1513-1495 (s) cm -1 ), hydroxyl (3700-3200 (br) cm -1 ), and carbonyl (1700 (m) cm -1 ) chemical groups in the solid (Figure 3). These chemical groups may form the backbone of the black particles and are typical for charcoal obtained from wood.</p><p>Besides, the black solid contained micrometer-sized CaCO3 and CaC2O4 crystals (Figure 4) scattered throughout the solid particles (EDX, Figure 8 SI). Those crystals are present in Eucalyptus wood. Despite the thermal treatment experienced in the kiln, those crystals also appear in the black solid. These inorganic particles, which are likely trapped within the hierarchical pore structure, have been seen in charcoals obtained with standard laboratory equipment. Solid Material with Surface Chemistry of Charcoal. When immersed in an aqueous solution, its pH significantly increases for solutions with an initial pH value at 3 -9, slightly increases for pH = 2, and does not change for solutions with initial pH = 10 -11 (Figure 5). The pH increment peaks for a solution with initial pH = 4. Also, solutions with initial pH = 4 -9 end up with final pH values between 9 and 10. In solid particles, weak basic chemical groups at the surface seem to determine their chemical behavior. At least some basic groups are so weak that they can remove H from water at pH > 7. This chemical behavior has been reported for charcoal [15] . Furthermore, black particles effectively remove (92-97%) crystal violet from aqueous solutions (ca. 1 to 100 ppm initial concentration) (Figure 6). Crystal violet is an organic cation in aqueous solutions sometimes used as a probe molecule to remove pollutants from water [16][17][18] . The drop in crystal violet in this study's solutions seems to occur because a chemical bonding with surface groups in the black solid has already been observed for charcoal [19,20] . In few words, the home-built oven enables converting Eucalyptus wood chips into charcoal with a simple process.</p><!><p>The solid's elemental composition was not fully determined. EDX used in this study is unsuitable for the determination and quantification of light elements like H. Because of unsaturated organic groups in carbon, the amount of H is generally small. Though H ought to be quantified to determine the elemental composition of the charcoal, this limitation does not affect the fact that the home-built kiln successfully carbonized wood chips.</p><p>Despite its simplicity, the carbonization process presented in this work was highly reproducible. Three independent carbonizations had a yield with low dispersion (30  1%). The drift method and the removal of crystal violet were also performed in triplicate.</p><p>Mean values with low dispersions were obtained. The reproducibility of the experiments seems to be a valuable strength.</p><!><p>We speculate that this study may ease the scientific research with charcoal. Charcoal is now accessible without expensive lab equipment running either in vacuum or with inert gases. Furthermore, charcoal can now be quickly produced in the hundred-gram scale instead of the usual gram-scale (production two-order of magnitude higher!). An easy and chip process is now available to produce charcoal for studies in, for instance, energy storage, heterogeneous catalysis, environmental remediation, and water potabilization [21] . The carbonization process has variables that may be explored to modify the properties of the charcoal (e.g., impregnation of wood chips with ZnCl2 or H3PO4; variation of biomass).</p><!><p>In summary, this study presents a simple, cheap, and fast carbonization method that enables the production of charcoal on a hundred-gram scale. Though proven with Eucalyptus wood chips, the carbonization process may quickly adapt to other types of woods and plant parts. By easing the carbonization of biomass, this study may facilitate scientific research with charcoal worldwide.</p>

ChemRxiv

Antibacterial activity of Cu(II) and Co(II) porphyrins: role of ligand modification

In this study, we report antibacterial activity of metalloporphyrins; 5, 10, 15, 20-tetrakis (para-X phenyl)porphyrinato M (II) [where X = H, NH2 and COOMe for M = Cu and X = COOH and OMe for M = Co]. The activity study of the as-synthesized metalloporphyrins toward two Gram-positive (S. aureus and S. pyogenes) and two Gram-negative (E. coli and K. pneumoniae) bacteria showed a promising inhibitory activity. Among the complexes under study, the highest antibacterial activity is observed for 5, 10, 15, 20-tetrakis (p-carboxyphenyl)porphyrinato cobalt (II), with inhibition zone of 16.5 mm against Staphylococcus aureus (S. aureus). This activity could be attributed to the high binding ability of COOH group to cellular components, membranes, proteins, and DNA as well as the lipophilicity of the complex. Moreover, consistent with literature report, the study revealed that metalloporphyrins with electron withdrawing group at para-positions have better antibacterial activity than metalloporphyrin which possess electron donating group at para position.

antibacterial_activity_of_cu(ii)_and_co(ii)_porphyrins:_role_of_ligand_modification

Introduction<!><!>Antibacterial activity testing<!>Media preparation and sterilization<!>Inoculation of test plates<!>Sample injection and incubation<!>Synthesis and photophysical properties<!><!>Antibacterial activity<!><!>Antibacterial activity<!>Conclusion<!>

<p>Metalloporphyrins are assumed to have extra ordinary importance in recent years as agents for photodynamic therapy, optoelectronic devices, sensors, molecular logic devices and artificial solar energy harvesting and storage schemes [1]. Taking into account a great number of infections resulting from different bacterial species and the growing antibacterial resistance, the development of compounds with high antibacterial activities and novel mechanism of action is an urgent need [2–4]. As a consequence, researchers are designing novel, convenient, robust and inexpensive strategies for combating microorganisms with minimal invasive consequences [5, 6]. In this regard, natural and synthetic metalloporphyrins are among relatively low toxic molecules (either in vitro or in vivo) and are capable of effecting microbial and viral pathogens through the large number of different mechanisms [7]. In addition, the possibility of structural modifications place these molecules into a group of compounds that present a sustainable source for discovery of novel procedures, materials and agents active against a wide range of pathogenic microorganisms [7]. Modification of porphyrin ligand at the peripheral positions provokes tunable shape, size and symmetry which have suitable applications in materials and therapeutics [8]. The most common structural modification of synthetic porphyrins is made at the meso-position to achieve target molecules with required properties in biomedical applications such as photo diagnosis, cancer therapy and as antibacterial agents [9, 10]. Nowadays, an ever increase in the mortality rate throughout the world is linked with infectious diseases with multiple resistances to antibiotics and the lack of effective treatments [2–4].</p><p>Porphyrin based systems have been reported as potential antibacterial agents against Gram-positive and Gram-negative bacteria species for decades [11–21]. They were used to treat different kinds of bacteria including bacillus subtilis, Escherichia coli, mycobacterium smegmatis, and actinobacillus [22–24]. The activities are based on their ability to catalyse peroxidase–oxidase reactions, generate reactive oxygen species (ROS) by absorbing light and partition into lipids of bacterial membranes [2, 25]. However, in most cases, much attention has been paid to ionic porphyrins (cationic [4, 15, 16, 26–33] and anionic [34–36] presumably because of their ability to strongly bind with cellular components and better activity than the neutral ones [15, 28, 37]. But, ionic porphyrins are very limited and studies involving neutral porphyrins to treat bacterial infections are becoming attractive. In line with this, antibacterial activity has been reported against Staphylococcus aureus, Mycobacterium smegmatis and Yersinia enterocolitica by using neutral porphyrins with the alkyl substituents at the β-pyrrolic positions [15, 33, 38]. However, there is no intensive report or documantation regarding neutral porphyrins for treating antibacterial infections.</p><!><p>A schematic diagram for bacterial growth inhibition by metalloporphyrins</p><!><p>The metal salts, ligands and their metal complexes were evaluated for in vitro antibacterial activities against strains of the two Gram-negative bacterial strains such as Escherichia coli (E. coli) and Klebsiella pneumoniae (K. pneumoniae); two Gram positive bacterial strains such as Staphylococcus aureus (S. aureus) and Streptococcus pyogenes (S. pyogenes) bacterium by disc diffusion method. In this method, activity of the test compounds was expressed by measuring the diameter of zone of inhibition. The plates were observed for zones of inhibition after 24 h, and incubation at 37 °C. The diameters of the zone of inhibition produced by the complexes were compared with a standard antibiotic drug Gentamycin. All the bacterial strains used in the experiment were received from microbiology laboratory, Bahir Dar University.</p><!><p>The Culture media (Mueller Hinton) were prepared according to the manufacturer's guideline (suspend 38 g in 1 L of distilled water). The mass of the culture medium was weighed and dissolved in distilled water. The mixture was stirred with a sterilized glass rod and tightly covered with an aluminum foil and then the culture medium was autoclaved for 15 min at 121 °C. Next to that,the agar was allowed to cool in order to maintain the media in a molten stage. Petri dishes were dried in lower humidity by keeping them in a laminar flow hood. The freshly prepared and cooled Muller–Hinton agar was spread at the surface of petri dishes.</p><!><p>A small volume, about 0.1 mL of the bacterial suspensions were inoculated onto the dried surface of Muller–Hinton agar plate and streaked (swabbed) by the sterile cotton swab over the entire sterile agar surface. This procedure was repeated by streaking two more times, rotating the plate approximately 60 °C each times to ensure an even distribution of inoculums and the rim of the agar was swabbed. The lid was left ajar for 3–15 min, to allow for any excess surface moisture to be absorbed before applying the samples on the respective well.</p><!><p>Anti-bactericidal activities of each reagents and synthesized complexes were evaluated by the disc diffusion method. Agar were prepared by using a sterilized cork borer with 6 mm diameter, 4 mm deep and about 2.5 cm apart to minimize overlapping of zones. Then holes of 6 mm diameter were punched carefully using a sterile cork borer. The metal salts of each complex, DMSO, the ligands, and their metal complexes were carefully injected to the respective disc in duplicate. The reference antibiotic agent disc (gentamycin) was dispensed via sterile pair of forceps onto the surface of the inoculated agar plate and pressed down to ensure complete contact with the agar surface. It was allowed to diffuse for about 40 min before incubation and then the plates were incubated at 37 °C for 24 h. After 24 h incubation, the antibacterial activity was evaluated by measuring the diameter of inhibition zones in millimeter. The test was carried out in duplicate and the results were recorded as mean ± standard deviation.</p><!><p>The metalloporphyrins employed in this study were synthesized by following reported methods [39–44]. The detail synthetic procedure, characterization data and photophysical properties of as synthesized compounds is shown in supporting information.</p><!><p>Antibacterial activiy (mean IZ diameter (mm) ± SD) of metalloporphyrins, corresponding ligands, metal salts, and gentamycin with concentration 500 mg/L</p><!><p>All the complexes under study showed better antibacterial growth inhibition activity than the corresponding porphyrin ligand. This is a clear indication for the involvement of metal ions as potential candidates in bacterial growth inhibition. The justification for enormous antibacterial activity of transition metal complexes of porphyrins is based on overtone concept and chelation theory. The solubility of the complexes in lipid is an important factor to control the antibacterial activity [45–48]. Based on the overtone concept of cell permeability, the passage of the materials which are only lipophilic is favored by the lipid membrane that surrounds the cell. On the other hand, the dramatic decrease in polarity of metal ions because of an overlap of orbital of ligand and partial sharing of the positive charge of the metal ion with donor groups can be blearily explained by employing chelation theory. Moreover, this phenomenon increases the π-electron delocalization all over the whole porprhyrin ring and enhances the lipid solubility behavior of the complexes. Presumably, an increase in lipid-solubility of the porphyrin ligands upon metallation makes the complexes easily move across the bacterial cell. This process inhibits the metals to bind with the enzymes in microorganisms. In addition, the respiration process of the cell could be interrupted and thereby block the synthesis of biomolecules, which limit over enlargement of organism [49].</p><!><p>Antibacterial activiy (mean IZ diameter (mm) ± SD) of cobaltporphyrins, at different concentrations</p><p>Bar graph of 5, 10, 15, 20-tetrakis (p-X phenyl)porphyrinato cobalt (II), where X = COOH and OMe on the same concentration pattern</p><p>Antibacterial activity (mean IZ diameter (mm) ± SD) of copperporphyrins, at different concentrations</p><p>Inhibition of; a Staphylococcus aureus (S. aureus) and b Streptococcus pyogenes (S. pyogenes) by 5, 10, 15, 20-tetrakis (p-carboxyphenyl)porphyrinato cobalt (II)</p><p>Inhibition of; a Escherichia coli (E. coli) and b Klebsiella pneumoniae (K. pneumoniae) by 5, 10, 15, 20-tetrakis (p-carboxyphenyl)porphyrinato cobalt (II)</p><!><p>Though antimicrobial activity of porphyrin derivatives of natural origin with COOH groups at β-pyrrolic positions have been reported so far [50–57], metalloporpyhrins with p-COOH at meso position of phenyl ring is not reported. Moreover, consistent with the report by Ke and coworkers, the electron withdrawing substituents enhance antibacterial activity attributing to increasing lipophilicity and polarity of the complex [15, 28]. Generally, the metal complexes containing electron withdrawal groups (with COOH and –COOMe showed better activities than the metal complex containing electron donating groups namely –NH2 and –OMe.</p><!><p>In general, antibacterial activity of metalloprphyrins with different peripheral substituents is reported. The study indicated that all the complexes under study have promising antibacterial activity toward two Gram-positive (Staphylococcus aureus (S. aureus) and Streptococcus pyogenes (S. pyogenes) and two Gram-negative [Escherichia coli (E. coli) and Klebsiella pneumoniae (K. pneumoniae)] bacterial species. It is also found that bacterial growth inhibition by metallopophyrins is higher than the corresponding metal salt or DMSO. Increasing the concentration of the complexes slightly increases the inhibition activity. Among the complexes under study, the highest antibacterial activity is observed for CoTPPCOOH, which could be attributed to the high binding ability of COOH group to cellular components, membranes, proteins, and DNA as well as the lipohilicity of the complex. Moreover, consistent with literature report, the study revealed that metalloporphyrins with electron withdrawing group at para-positions have better antibacterial activity than metalloporphyrins which possess electron donating group at para position. The result finally concludes that metalloporphyrin derivatives are promising candidates for antibacterial activity.</p><!><p>Deoxyribonucleic acid</p><p>Tetra phenyl porphyrin</p><p>Metal with oxidation state of 2</p><p>Methyl</p><p>Methoxy</p><p>Methoxycarbonyl</p><p>Reactive oxygen species</p><p>Degree centigrade</p><p>Hour</p><p>Dimethylsulphoxide</p><p>Inhibition zone</p><p>Standard Deviation</p><p>Staphylococcus aureus</p><p>Streptococcus pyogenes</p><p>Escherichia coli</p><p>Klebsiella pneumonia</p><p>Gram</p><p>Milligram</p><p>Liter</p><p>Milliliter</p><p>Microgram</p><p>Millimeter</p><p>Publisher's Note</p><p>Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.</p>

PubMed Open Access

Study of the geometry of open channels in a layer-bed-type microfluidic immobilized enzyme reactor

This paper aims at studying open channel geometries in a layer-bed-type immobilized enzyme reactor with computer-aided simulations. The main properties of these reactors are their simple channel pattern, simple immobilization procedure, regenerability, and disposability; all these features make these devices one of the simplest yet efficient enzymatic microreactors. The high surface-to-volume ratio of the reactor was achieved using narrow (25–75 μm wide) channels. The simulation demonstrated that curves support the mixing of solutions in the channel even in strong laminar flow conditions; thus, it is worth including several curves in the channel system. In the three different designs of microreactor proposed, the lengths of the channels were identical, but in two reactors, the liquid flow was split to 8 or 32 parallel streams at the inlet of the reactor. Despite their overall higher volumetric flow rate, the split-flow structures are advantageous due to the increased contact time. Saliva samples were used to test the efficiencies of the digestions in the microreactors.Graphical abstract Supplementary InformationThe online version contains supplementary material available at 10.1007/s00216-021-03588-x.

study_of_the_geometry_of_open_channels_in_a_layer-bed-type_microfluidic_immobilized_enzyme_reactor

Introduction<!>Materials<!>Flow simulation<!>Preparation of the PDMS microreactor<!>Enzymatic digestion of saliva samples<!>In-solution digestion<!>Digestion in μ-IMER [25]<!>Capillary electrophoresis<!>Increasing the surface-to-volume ratio<!><!>Increasing the surface-to-volume ratio<!>Effect of curves<!><!>Application of the microreactor<!><!>Application of the microreactor<!>Conclusions<!>

<p>The aim of proteomic research is to provide a comprehensive characterization of the proteome. Although the complete protein profiling of living organisms still faces analytical challenges, the emergence of shotgun proteomics greatly enhanced the success of such endeavours [1]. In a typical "shotgun" workflow, the sample is mixed with a proteolytic enzyme (e.g. trypsin) to cleave the proteins into smaller peptides; the resulting peptides are then separated by high-performance liquid chromatography (HPLC) [2] or capillary electrophoresis (CE) [3] and finally analysed by tandem mass spectrometry (MS/MS). The standard, in-solution digestion protocol suffers from long incubation times (2–16 h) as the enzyme can only be used in high dilution (protein:trypsin = 20–100:1) to minimize autolysis [4]. Extended digestion times can be considered the bottleneck for fast, high-throughput MS-based analysis. Trypsin autolysis, however, can be suppressed effectively by binding the enzyme to solid supports, offering accelerated digestions (~ few minutes) [5]. To this end, the development and application of microfluidic immobilized enzyme reactors (μ-IMERs) have attracted much attention. Enzymes are most commonly attached to supports via adsorption [6], covalent bond [7], and bioaffinity linkage [8]. An important feature of such microfluidic devices (MD) is their high surface area-to-volume ratio (S/V), which can boost digestion efficiency by increasing the occurrence of enzyme–substrate interactions.</p><p>Despite the inherently high specific surface area of MDs, considerable efforts have been made to further increase the S/V ratio by integrating porous membranes [6, 9], particles [10, 11], monoliths [12–14], and magnetic microspheres [15, 16] into the channels. Each strategy has their benefits and drawbacks. Accommodation of particles in a microfluidic channel is in the interest of chromatography experts, as well. However, particle retention is an exhaustive task both in conventional and microfluidic platforms; furthermore, packing heterogeneity cannot be prevented, leading to peak broadening. An elegant way to circumvent this issue is to form collocated monolith support structures (COMOSS) [17] or pillars [18] by microfabrication techniques adapted from the microelectronic industry. In this way, excellent homogeneity was achieved while maintaining a sufficiently high S/V ratio. Although the authors used this pillar array platform for LC separations, the underlying motivation is valid for the development of μ-IMERs, as well.</p><p>The present study focuses on wall-coated (or layer-bed type) μ-IMERs and the possibilities of advancing catalytic performance in such systems. Wall-coated μ-IMERs are the simplest and probably possess the lowest enzyme loading capacity in comparison with packed or monolithic μ-IMERs, as in this case only the inner wall of the microchannel serves as solid support for immobilization. More recently, the utilization of multi-channel capillaries has attracted attention [19, 20] as possible candidates for enzyme housing. Such a capillary embodies an array of microchannels, surmounting the shortcomings of single-channel wall-coated IMERs in terms of S/V ratio. In order to increase the available surface area, the pillar array structure mentioned above can be utilized [21, 22] as well as the modification of the inner wall (e.g. sol–gel matrix [23], cross-linked enzyme membrane [24]). The long diffusion length can also be minimized by decreasing the channel diameter, as realized by Foret et al., who have developed an excellent enzymatic reactor inside a 10-μm-inner-diameter (i.d.) capillary [7]. While most immobilization techniques described in the literature require a multi-step manipulation of the channel surface, our group has recently proposed the method of spontaneous trypsin adsorption directly on the channel surface of a polydimethylsiloxane (PDMS) microchip, providing one of the simplest μ-IMER setups, where trypsin is adsorbed directly on the unmodified PDMS channel walls via hydrophobic interactions [25]. Previously, the quasi-irreversible adsorption of proteins onto PDMS had been confirmed utilizing surface plasmon resonance spectroscopy [26]. A brief evaluation of trypsin adsorption can be found in the supplementary material as Fig. ESM-1. PDMS is often at the centre of dispute as to whether it has any real advantage over other sturdier materials, but the very problems usually associated with PDMS (adsorptivity, hydrophobicity, porosity) are actually exploited in the IMER presented. The digestion with this microreactor using a layer-bed-type immobilized enzyme reactor (empty channels) requires less than 10 min, while in-solution digestion takes 16 h [25, 27].</p><p>The goal of the current work was to study such open channel geometries in a layer-bed-type immobilized enzyme reactor system with computer-aided simulations. The curves of the channel may either have a considerable or a negligible effect, mainly depending on the channel diameters and the flow rate. Since in microfluidic enzyme reactors applied for proteomic studies the required volumetric flow rate and the channel width/height are in a relatively narrow range, the effect of the curves was examined for our particular IMER design. The simulations allow us a better understanding of laminar liquid flow in empty channel geometries, granting us the possibility to fully harness the proteolytic potential of these systems. To put the designed microchannels to a test, protein samples bearing clinical significance and challenging complexity were used for digestions.</p><!><p>Analytical-grade reagents were used. Urea, dithiothreitol (DTT), iodoacetamide (IAM), NH4HCO3, and formic acid (FA) stock solutions (all Sigma products, St. Louis, MO, USA) were prepared in double-deionized water (Elix-3, Millipore, Darmstadt, Germany). Porcine pancreas trypsin (Type IX-S, lyophilized powder, Sigma) solutions were freshly prepared before each experiment. Human saliva was digested to examine the efficiency of the μ-IMERs. Phosphate buffer (PB) electrolyte, isopropanol, methanol, and acetonitrile were purchased from VWR (Radnor, PA, USA).</p><p>For microchip fabrication, SU-82025 photoresist and SU-8 developer solution (1-methoxy-2 propyl) acetate) were acquired from Microchem (Newton, MA, USA). The PDMS silicone elastomer kit (Sylgard 184) was purchased from Dow Corning (Midland, MI, USA).</p><!><p>COMSOL Multiphysics (Burlington, MA, USA) software was utilized to simulate liquid flow behaviour in the microchannels. This is a finite element (FEM) analysis-based simulation software. Software version 5.3a was used with laminar flow and transport of diluted species modules to simulate liquid flow behaviour. The flow velocity was fixed at 3 × 10−3 m/s, the diffusion coefficient was set to 6.1 × 10−11 m2/s, and the concentration of one of the inlets was set to 4.3 × 10−5 mol/L, in all cases. The mesh size was set to extremely fine.</p><!><p>The microfluidic chips were fabricated by means of soft photolithography [28]. The channel patterns designed with AutoCAD software were printed with a high-resolution printer (Keppont Ltd., Debrecen, Hungary). A 3″ silicon wafer was coated with negative-type photoresist (SU-8) using a spincoater (3000 rpm, 30 s). Following soft bake (95 °C, 15 min), the photoresist-coated wafer was exposed to UV light (365 nm, 10 min) through the printed lithographic mask. After post-exposure bake (95 °C, 5 min), the unexposed areas were dissolved by rinsing with SU-8 developer solution. The attained SU-8 pattern on the wafer served as the mould, from which inverse replicas were made using PDMS. A mixture containing PDMS oligomer and curing agent in a 10:1 ratio was poured onto the mould. The mould was then placed into a vacuum chamber to eliminate air bubbles. After curing (65 °C, 60 min), the PDMS was stripped off the mould and cut to size, and holes (300 μm diameter) were pierced at the ends of the channels (for liquid connections) and finally irreversibly sealed onto another PDMS slab following oxygen plasma treatment (PDC-32G, Harrick, Ithaca, NY, USA).</p><p>A peristaltic pump (IPC, Ismatec, Cole-Palmer, IL, USA) was used to transfer the reagents and protein samples through the microreactors. The peristaltic pump's tubing (ID, 0.19 mm; Tygon, Cole-Palmer, IL, USA) was connected to the inlet port of the channel.</p><!><p>Human saliva samples were obtained from a healthy female volunteer. Sampling was carried out according to the spitting method [29] after having abstained from food and beverage consumption for at least 2 h. Whole saliva collected into a 2-mL-volume Eppendorf tube was centrifuged (2700 × g, 20 min). Supernatant was freeze-dried. The pretreatment of saliva samples for the tryptic digestion was executed as follows: ~ 1 mg lyophilisate was dissolved in 6.6 μL 25 mM NH4HCO3. Twenty microlitres of 8 M urea solution (30 min, room temperature) was added to unravel the tertiary structure of the proteins. For the reduction of the disulfide bonds, 2.66 μL 100 mM DTT was used (1 h, 37 °C). Adding 2.66 μL 200 mM IAM (alkylating agent) to the solution (45 min, room temperature in the dark), the recombination of disulfide bonds was precluded. Finally, 133 μL 25 mM NH4HCO3 was added to the mixture. The prepared samples were stored at − 20 °C until digestion. Digestions were performed the conventional way (in-solution) and via μ-IMER.</p><!><p>6.6 μL freshly prepared 1 mg/mL trypsin solution was pipetted into the saliva samples. Reaction was stopped after overnight incubation (16 h, 37 °C) by the addition of 1% FA to a 0.1% FA final concentration.</p><!><p>For immobilization, freshly prepared 20 mg/mL trypsin solution was flushed through the PDMS channels (2 μL/min, 10 min). Unbound trypsin was removed from the channel by flushing with 25 mM NH4HCO3 (2 μL/min, 10 min). For the digestion, 10-μL aliquots of the sample were transported through the microreactor at a flow rate of 0.65 μL/min (contact time, ~ 2 min; room temperature). The peptide mixture at the outlet was collected for the subsequent CE-UV and CE-MS measurements. 1.5 μL 1% FA was added to the effluent sample to inhibit accidental tryptic activity (in case of trypsin leaching).</p><!><p>The separation of peptides was performed with a 7100 model CE instrument (Agilent, Waldbronn, Germany) using on-capillary UV and MS (maXis II UHR ESI-QTOF MS instrument, Bruker, Bremen, Germany) detection. Hyphenation was achieved with a CE-ESI Sprayer interface (G1607B, Agilent). Sheath liquid was transferred with a 1260 Infinity II isocratic pump (Agilent). CE instrument was operated by OpenLAB CDS Chemstation (Agilent) software and the MS instrument was controlled by otofControl version 4.1 (build: 3.5, Bruker).</p><p>Fused silica capillaries of 90 cm × 50 μm i.d. (Polymicro, Phoenix, AZ, USA) were used for the separations. Sample solutions were introduced at the anodic end of the capillary; the applied voltage was 25 kV. The capillaries were preconditioned with the background electrolyte (BGE) for 5 min and postconditioned with methanol, acetonitrile, and BGE for 2 min each. In the case of CE-UV determinations, the BGE used was phosphate buffer (PB) (100 mM, pH = 2.2); hydrodynamic sample introduction (50 mbar, 10 s) was used for sample injection, and peptides were detected at λ = 200 nm. For CE-MS measurements, the BGE consisted of formic acid (FA) (1 M, pH = 1.9), and larger sample loading was performed (50 mbar, 120 s). The sheath liquid was isopropanol:water = 1:1 with 0.1% FA and applied with 6 μL/min flow rate. MS was operated in positive ionization mode; 0.5 bar nebuliser pressure, 180 °C dry gas temperature, 4 L/min dry gas flow rate, 4500 V capillary voltage, 500 V end plate offset, 3 Hz spectra rate, and 50–2200 m/z mass range were applied. Collision-induced dissociation (CID) was used to produce fragment ions. The MS/MS spectra rate was 1–4 Hz. Na-formate calibrant injected after each run enabled internal m/z calibration. Electropherograms were processed by OpenLAB CDS Chemstation (Agilent) software; mass spectra were processed by Compass DataAnalysis version 4.4 (build: 200.55.2969, Bruker). Protein identification was carried out with Byonic software (Protein Metrics, Cupertino, CA, USA). Digestion specificity was set to fully specific allowing up to 1 missed cleavage; precursor mass tolerance was 10 ppm. Carbamidomethylation at Cys as fixed modification, deamidation at Asn and Gln, and formylation at N-term were set as variable modifications.</p><!><p>For enhancing the efficiency of an enzyme reactor, the main parameters to be tuned are the total surface and the surface-to-volume ratio (S/V) of the channel system as well as the residence time of the analyte in the close vicinity of the wall covered by the enzyme; i.e. these parameters are to be maximized. There are several ways to largely increase the total surface and the S/V of a channel system. Recently, the arrangement of pillars [22] or integration of a highly porous medium [6–9, 12–14] in a channel has seemed to be very efficient and is preferred. However, the simplest channel pattern is obviously an empty channel, where the total surface and the S/V can easily be adjusted by changing the dimension (mainly the width and the length) of the channel. In a recent study, we have shown that S/V values and total surface similar to those of channels with pillars can be achieved in empty channels as well [22]. The pillars in a microfluidic system do not cause turbulence or diversion of the liquid stream, which could allow components to travel close to the wall covered by the immobilized enzyme (Fig. ESM-2).</p><!><p>The dependence of surface-to-volume ratio (S/V) and total surface (S) on the channel width (a), diffusion time (tD) as a function of channel width (Dalbumin = 6.1 · 10−11 m2/s) (b), and the time it takes to transport 10 μL sample solution through the channels of different widths (p, 2 bar; L, 20 cm; liquid, water) (c)</p><!><p>Wider channels are less advantageous for μ-IMERs because the analyte molecules are more restricted in reaching the surface immobilized with the enzymes. In μ-IMERs, the flow is strongly laminar, so the molecules can approach the wall only with diffusion. Proteins (as the analytes for μ-IMERs) are large molecules; therefore, their diffusion rate is very small (e.g. D = 6.1.10−11 m2/s for albumin). While in a 100-μm-wide channel it takes 20 s for a molecule of albumin to travel from the middle to the surface, only 1.3 s is required in a 25-μm-wide channel (Fig. 1b). That means if the albumin sample is pumped through the μ-IMER faster than 20 s, a small portion of the protein has no chance for enzyme digestion.</p><p>However, the application of too narrow channels would limit the volume of the sample to be digested. Ten microlitres volume of sample is minimally needed for an MS measurement even if it is hyphenated with HPLC or CE. Figure 1c shows the time needed to transport 10 μL sample solution through the channels of different widths. Using 10 μm channel width, the sample should be transported for more than 10 h, while only 15 min is needed in a 25-μm-wide channel and only a few seconds in a 100-μm-wide channel (p, 2 bar; L, 20 cm). Based on the considerations above, there is no reason to decrease the channel width below 25 μm or increase it over 100 μm; therefore, the recommended range of the channel width is 25–75 μm.</p><p>Since the essential requirement for the efficient operation of a μ-IMER is that components should reach the wall, the diffusion/radial motion of the transported components in the channel was studied by COMSOL simulations. We studied a straight channel where water and 43 μM albumin were introduced into the inlet as two parallel streams at 1:1 ratio ("each above each") and pumped these liquid streams with the same speed (3 mm/s). COMSOL simulations (Fig. ESM-3 and Fig. ESM-4) show the concentration distribution of a protein across channels of varying widths (10, 25, 50, 100 μm). Four cross-sectional channel segments of 50 μm width at 0.005, 1, 9.95, and 19.95 mm distance were magnified for better visibility. The simulations demonstrate that while the concentration distribution of the component is completely homogeneous in the channel of 10 μm width (all components can contact the wall) after 1 mm length, in the channel of 100 μm width the distribution is not homogeneous even after 20 mm length (a considerable part of the components cannot reach the wall). In the case of the 25- and 50-μm-wide channels, the mixing of the solutions becomes completed within a 20-mm distance.</p><p>Mixing in the channel is defined by the flow rate, as well. Concentration distributions corresponding to three different linear flow velocities (0.3–3–30 mm/s) in a 25-μm-wide channel differ considerably (Fig. ESM-5 and Fig. ESM-6). Although the linear speeds examined (commonly used in microfluidics) are largely different, the flow is still strongly laminar (Re, 0.0084–0.84) in each case. When the speed is only 0.3 mm/s, the concentration distribution becomes homogeneous after a length of 1 mm. Not surprisingly, similar homogeneous distribution can be obtained after 10 mm with 3 mm/s flow rate because the residence times of the component in the channel are identical. Theoretically, in straight channels, no difference can be expected in the digestion efficiency between long channel–high flow rate and short channel–low flow rate arrangements. However, the generation of a very low flow rate in a stable, constant way can count as a technical difficulty (e.g. 0.3 mm/s equals 0.009 μL/min in a 25-μm-wide and deep channel, which is a lower rate than the lowest available flow rate with a typical peristaltic pump (~ 0.2 μL/min).</p><!><p>Fluid flow in microfluidic channels is known for its laminar behaviour, where streamlines do not cross paths. Special channel configurations, however, induce passive mixing [31–42]. In such cases, the geometric features or obstructions in the channel are the source of the mixing phenomenon. It has been reported that curves in a microchannel generate a spiralling fluid flow, the magnitude of which can be characterized by the Dean number [31–35]. The mixing effect of such secondary flow is pronounced in systems where Re > 10 [34]. Raising the Re (hence, the flow rate) increases the value of the Dean number, which is favourable if a homogenous concentration distribution is desired. Some studies show a considerable effect of the curves on the disturbance of liquid flow [34–36]. However, other papers claim that the curves have only a minor effect on the dynamics/flow profile of the liquid [34, 37], therefore, different types (passive or active) of mixing/diverting are still required for efficient mixing [38, 39]. In addition to the Dean vortices, expansion vortices can also form in channels where there is an abrupt increase in cross-sectional area at the curvature, creating a multivortex field [43]. Therefore, the impact curvatures have on mixing is highly variable. Depending primarily on the channel diameter and the flow rate, curves may have either an appreciable or a negligible effect. In microfluidic enzyme reactors applied for proteomic studies, the required volumetric flow rate and the channel width/height can be in a relatively narrow range, because (i) at least 10–50 μL sample should be gained at the reactor outlet, (ii) the proteins present in the solution should reach (diffuse to) the wall of the reactor, and (iii) the digestion (residence time of the sample in the channel) should take less than a few minutes. Therefore, it is not obvious that the application of curves could lead to advantages for digestion in a microfluidic IMER. This is why we intended to study the effects of the curves.</p><p>In the case of microfluidic IMERs, two—counteractive—key objectives have to be fulfilled: (i) the exploitation of the radial motion of components toward the channel wall (i.e. an intrinsic passive mixing due to curvilinear channel geometry) and (ii) low flow rate for longer residence time (for increased contact time between enzyme and substrate). In view of these aims, finding a delicate balance between these two contributing factors was a priority. Our systems can be described with Re < 1 (in the case of 0.65 μL/min, the Reynolds numbers are 0.62 and 0.21 for the 25-μm- and the 75-μm-wide channels, respectively), thus only slightly facilitating the development of Dean vortices. Since the Re can be higher if the flow rate is increased, in order to provide sufficient time for enzyme–substrate interaction, the channel length should be increased. Based on these theoretical considerations, with the aid of COMSOL simulations, we studied the extent of the impact curves in the microfluidic IMERs have on the mixing effect and thus the digestions.</p><p>Firstly, as a simple case, a straight channel (20 mm long, 100 μm wide) with a single 180° bend was used for studying the effect of one curvature on mixing/diffusion. The concentration distributions at the different positions (fore part, middle, and rear part) of the curve and close to the inlet and outlet ends of the channel were compared (Fig. ESM-7). Although no visually obvious differences in the distributions can be seen in the curvature section, the concentration gradient diagrams demonstrate a little change of the concentration differences in the cross section of the channel in the curvature compared to that obtained for the straight channel before the curve. The concentration differences along the cross section of the channel before and after the curve (0.78 mm length) were 0.026 mM and 0.022 mM, respectively. By contrast, regarding the 9.55-mm-long straight part of the channel, the concentration differences were 0.043 mM and 0.026 mM, respectively. Comparing the change in concentration distributions for each case, it can be concluded that the curvature induces a mixing effect almost three times larger than a straight channel per unit length. Similar conclusions can be reached if the changes of concentration gradients are compared in the case of a straight channel and a channel including a curve (Fig. ESM-8).</p><!><p>COMSOL simulations on the effect of incorporating several curves in the microchannel. The top panel shows the initial concentration gradient at L = 0.1 mm for all cases. The panel below shows concentration distributions at the end of the channel for four different cases: straight channel and channel including one, four, and eight curves. Plots on the right indicate the concentration distribution along the cross section at positions shown on the left (L = 0.1 mm and L = 20 mm). Water and albumin were introduced at the inlet at 1:1 ratio. Four channel segments were magnified for better visibility. Values at the bottom mark the distance from the entry point. (L, 2 cm; Dalbumin = 6.1.10–11 m2/s; v, 3 mm/s)</p><p>The gradual change in concentration distribution at different positions (length) of the channel containing 8 curves. The concentration differences were calculated based on cut lines of simulations, which were positioned before, in the middle, and after the curves (similarly as in Fig. ESM-7). The dotted line indicates the concentration differences in the case of a straight channel. Parameters were the same as in Fig. 2, except v, 10 mm/s</p><p>Images of three different microfluidic chips (on the right). The channel system of each microchip was filled with red food dye; certain sections were magnified for better visibility. Channel parameters: w, 25 μm; L, 20 cm; number of channels, 1, 8, and 32 for design A (a), B (b), and C (c), respectively</p><!><p>In order to reduce the applicable flow rate in the channel (hence increase the residence time of the component), the liquid flow was split to 8 parallel streams at the inlet of the reactor and the liquid passed through the 8 parallel channels, which were merged at the outlet (Fig. 4b). In this design, the overall volumetric flow rate was identical to that of the previous design, but the rate could be reduced by a factor of 8 in each channel reactor (if the same pressure is applied). A further factor of 4 can be achieved in lowering the flow rate in the channel, if four identical reactor units including 8–8 parallel channel reactors (shown in Fig. 4b) are connected. In this arrangement, the sample is introduced in the centre part of the chip and it is split first to 4 parts and then to additional 8–8 parts (Fig. 4c).</p><p>The efficiency of the enzyme reactors of different designs was studied by digesting human saliva samples. Saliva is a mixture of different proteins and other biological compounds. Fig. ESM-9 shows the electropherograms of human saliva digested via standard in-solution procedure (16 h, 37 °C) and on-chip using a microchip (design B in Fig. 4b; channel width, 75 μm; 2 min contact time; 25 °C). On the electropherograms, a large number of peaks with similar migration times were obtained. Although the signal intensities for the corresponding peaks differ, this probably only means that the enzymes immobilized and being present freely in solution cleave a given bond with different probability. Because the number of the components obtained in the digested sample is similar (55–62) and no residues of undigested proteins were detected, the developed enzyme reactor can be considered useful for peptide mapping. Very similar digestions were obtained with the other two microchip designs.</p><!><p>Comparison of peptide maps obtained from reactor-based digestions using microchip A (design B; w, 25 μm) and microchip B (design B; w, 75 μm). CE conditions: fused silica capillary; i.d., 50 μm; Leff, 71.5 cm; BGE, 100 mM PB (pH, 2.2); U, 30 kV; sample introduction, 50 mbar, 10 s; λ, 200 nm</p><p>Sequence coverage (SC%) values of the proteins identified in human saliva by CE-MS/MS using in-solution and on-chip digestions. Only 14 proteins identified with at least two unique peptides are shown</p><!><p>For the reproducibility study of on-chip digestion, 5–5 consecutive digestions were carried out for all three platforms (in-solution and chip design B (w1, 25 μm; w2, 75 μm)). In the case of chip-based digestions, the 5–5 digestions were performed using different reactors (for the evaluation of the reliability of enzyme immobilization). The precision of each digestion strategy was evaluated in terms of the scattering of SC% values (Fig. 6), which ranged between 0 and 39%.</p><!><p>The goal of this paper was to study the design, efficiency, and applicability of a simple PDMS microfluidic chip used for rapid protein digestion. PDMS is a material frequently preferred for the fabrication of microfluidic chips; its superior adsorptive ability has recently been utilized for nonspecific adsorption of enzymes in order to prepare enzymatic microreactors. In our previous work [25], it was shown that the wall of the empty, long channel system in a PDMS chip can be used directly as a solid support for trypsin immobilization and thus to develop IMERs with simple open channel geometry. There have been numerous impressive IMER constructions published so far, most of which enable the efficient or complete proteolysis of the substrates of interest. The significance of this particular IMER setup lies in its simplicity; the immobilization procedure only consists of trypsin solution being transported through the native PDMS microchannels, excluding laborious, multi-step procedures, where each step might entail sources of error. It is obvious, however, that such IMERs do not exhibit the lifetime and enzyme loading capacity characteristic of other published IMERs which utilize monolithic or packed-bed microchannels or even porous layer open tubular reactors. The limitations in the longevity of the IMER are resolved by the straightforward and fast (~ 10 min) immobilization process. Furthermore, by appropriately manipulating the liquid flow in wall-coated μ-IMERs, it is possible to mitigate the problem of relatively low enzyme load and diffusion-limited mass transfer in order to achieve successful proteolytic cleavage. Our objective was to study such open-channel geometries in a layer-bed-type immobilized enzyme reactor system with computer-aided simulations. The COMSOL simulation software proved to be a useful tool for designing and optimizing the channel pattern and for giving us an insight into the effects the channel width, channel length, and curves as well as flow rates have on the radial diffusion and mixing in the channel. The simulations obtained for the different channel designs and configurations well supported the identification of the relevant aspects that have to be taken into account for achieving optimal conditions for improving the performance of the microfluidic IMERs and allow us a better understanding of laminar liquid flow in empty channel geometries.</p><p>S/V values largely increase with the narrowing of the empty channel; however, two notable drawbacks discourage us from endorsing such narrow channels (w ≤ 10 μm): (i) the lack of sophisticated microfluidic facilities in an average laboratory hampers the successful fabrication of these small features and (ii) the transportation of a 10-μL-volume sample solution through the reactor can be unreasonably prolonged (~ 10 h). The duration of transportation might be significantly accelerated (~ 3.5 s) using 100-μm-wide channels, albeit at the cost of decreased digestion efficiency, since thorough mixing in the channel is inhibited due to the laminar flow. Therefore, the recommended range of the channel width was found to be 25–75 μm. The simulation clearly showed that curves support (slightly) the mixing of solutions in the channel even in the strong laminar flow conditions (Re < 1); thus, it is worth including several curves in the channel system.</p><p>In our previous work [22], we integrated micropillars into the channel in order to increase the S/V ratio and the total surface (S) of the reactor. The use of empty channels instead of the array of pillars can be considered as a simplification, but our goal and the key point of the developments of IMERs was to study whether the S/V ratio of an empty channel pattern can be increased to the degree of the pillars' pattern. In the case of a 25-μm-wide empty channel, an S/V ratio similar to the channels containing 25-μm pillars and 25-μm interpillar distance can be achieved, but the total surface is much smaller in an empty channel. In the present study, it has been shown that both the S/V and the S can be similarly high if multiple empty channels are applied (split and merged between the inlet and outlet ports, respectively). With the proposed empty channel reactor (w, 25 μm; L, 20 cm; number of channels, 8), complete digestion of the proteins can be achieved just like with the reactor including pillars or with the classical in-solution method. The S/V-increasing effect of the pillar arrangement is only significant if the channel is tightly "packed". However, technical difficulties (e.g. light scatter during photoresist exposure or pattern collapse due to PDMS flexibility [44]) pose limitations on pillar density. Precise and refined micropillar patterning onto harder materials can be carried out utilizing other, more sophisticated technologies, but such processes are generally time consuming and more expensive. On the other hand, the proposed serpentine-like channels are considerably easier to prepare and the split-flow architecture ensures a sufficiently high total surface. Another advantage of the empty channels is that the hydrodynamics in the empty channel could be more easily monitored and also tuned by the application of curves (with proper numbers and arch dimension). The proposed μ-IMER provided similar efficiencies for the digestion of saliva as the standard in-solution digestion procedure. We have shown that the flow rate and channel geometry play a crucial role in achieving efficient proteolysis. Split-flow structures are especially advantageous because of the increased contact time, despite the overall higher volumetric flow rate. Based on our investigations (both theoretical and experimental), it can be concluded that in the 25–75-μm range, no specific value for channel width can be selected as the gold standard, since the desired digestion efficiency can be attained by fine-tuning flow conditions and channel parameters.</p><!><p>(PPTX 4927 kb)</p><p>Publisher's note</p><p>Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.</p>

PubMed Open Access

Modeling Protein–Glycosaminoglycan Complexes: Does the Size Matter?

Docking glycosaminoglycans (GAGs) has been challenging because of the complex nature of these long periodic linear and negatively charged polysaccharides. Although standard docking tools like Autodock3 are successful when docking GAGs up to hexameric length, they experience challenges to properly dock longer GAGs. Similar limitations concern other docking approaches typically developed for docking ligands of limited size to proteins. At the same time, most of more advanced docking approaches are challenging for a user who is inexperienced with complex in silico methodologies. In this work, we evaluate the binding energies of complexes with different lengths of GAGs using all-atom molecular dynamics simulations. Based on this analysis, we propose a new docking protocol for long GAGs that consists of conventional docking of short GAGs and further elongation with the use of a coarse-grained representation of the GAG parts not being in direct contact with its protein receptor. This method automated by a simple script is straightforward to use within the Autodock3 framework but also useful in combination with other standard docking tools. We believe that this method with some minor case-specific modifications could also be used for docking other linear charged polymers.

modeling_protein–glycosaminoglycan_complexes:_does_the_size_matter?

Introduction<!>Comparing the Performances of MM/PBSA and MM/GBSA Free Energy Decomposition Calculations for GAG Ligands in AA and CG Representations Complexed with Proteins<!>Protein Structures<!>Peptide Structures<!>GAG Structures<!>Molecular Docking<!>Coarse-Grained Model Parameters for a Docked GAG Oligomer Elongation<!>Molecular Dynamics<!>Binding Free Energy Calculations<!>MM/PBSA Calculations for Protein–GAG Complexes: AA vs CG Representation of a GAG<!><!>MM/PBSA Calculations for Protein–GAG Complexes: AA vs CG Representation of a GAG<!>All-Atom Simulations<!><!>CG Parameters Obtained from All-Atom MD Simulations<!>Bonded Parameters<!><!>Nonbonded Parameters (Charges, Lennard-Jones Parameters)<!><!>Mixed AA/CG Simulations: CG Elongation of a GAG<!><!>Mixed AA/CG Simulations: CG Elongation of a GAG<!><!>Mixed All-Atom/Coarse-Grained Simulations Based on Per-Residue Energy Analysis<!>Energy Prediction for GAG Elongation<!><!>Energy Prediction for GAG Elongation<!>Single Pseudoatom as an Extension of the GAG Molecule<!>Conclusions<!><!>Notes

<p>Human cells express multiple polymers that display a variety of functions. One particular class of those polymers are glycosaminoglycans (GAGs). They are long periodic linear and negatively charged polysaccharides that by interacting with proteins play an immense role in the extracellular matrix processes. Depending on their sulfation pattern and charge densities, GAGs manifest different conformational and binding properties.1 GAGs are built of repeating disaccharide units. Each of them consists of an amino sugar and an uronic acid or galactose.2 Depending on the sulfation pattern and monosaccharide composition, GAG disaccharide units can display 408 variants,3 of which 202 can be found in mammals.4,5 While some of the protein–GAG interactions are specific, most of them are considered as nonspecific and/or electrostatically driven due to the high negative charge of those polysaccharides directly correlating with the binding affinities.6 Among many proteins, there are two major protein groups that GAGs can interact with. One of them are growth factors,7,8 and the second group are chemokines.9−11 In the case of growth factors, GAGs are able to influence the cell signaling and the activity of the proteins by changing their conformation or by oligomerization facilitation of their receptors by binding and clustering multiple fibroblast growth factors (FGFs) at the same time.12,13 For example, in the case of vascular endothelial growth factor (VEGF), a key player in cancer, arthritis, angiogenesis, and regenerative processes,14 global conformational changes induced by heparin (HP) binding influence its binding capability to its receptor on the cell membrane.7 HP and heparan sulfate are also able to bind to transforming growth factor β (TGF -β1),15,16 a protein that is responsible for the regulation of the proliferation, adhesion, differentiation, and cell migration.17 Depending on the sulfation pattern, hyaluronan derivatives influence TGF-β1 activity and its binding to its receptor.18,19 The second mentioned group of the protein that interacts with GAGs are chemokines.10,20 This is mostly a proinflammatory group of proteins that belongs to cytokines. They may influence cells in different manners: some of the chemokines can alter metastasis tumor growth and angiogenesis by either promoting or inhibiting it.21 GAGs by interacting with IL-8 can alter the ability to activate leukocytes.22−24 GAGs can also affect pro-/anti-inflammatory functions of IL-10.25,26 It was also shown that HP may interact with CXCL-14,11 and by doing that, it increases migratory potential on monocytic THP-1 cells.27 Many computational studies on GAGs show their promising potential in the examination of the protein–GAG interactions. The following studies successfully investigated effects of the GAGs binding on a variety of different proteins, such as CXCL-14,11 VEGF,7 CXCL-8,9,24,28 a Proliferation Inducing Ligand (APRIL),29 IL-10,25,26,30 CXCL-12,31 acidic fibroblast growth factor (FGF-1),32 or protein–ion–GAG complexes.33</p><p>Even though computational studies seem to be very successful and helpful in protein–GAG investigations, there are still a lot of challenges that have not been fully overcome yet. One of them is docking long GAG molecules. Usually, GAGs dp4 or 6 (dp stands for degree of polymerization) are used in molecular docking. This is caused by the fact that most of the docking software can handle only a limited number of torsional degrees of freedom for the docked molecules. The number of torsional degrees of freedom is often given by the number of rotatable bonds in the ligand. For example, when using Autodock3, which is the most accurate docking tool for GAG docking,34 to dock GAGs of a higher degree of polymerization, a user cannot include all torsional degrees of freedom and needs to manually pick the most relevant ones not to overcome the limit of 33. The more the torsional degrees of freedom are active in a docking procedure, the more accurate docking results should theoretically be possible to obtain. Thus, using very long GAG molecules (e.g., dp10 or higher) heavily hampers docking performance and makes it unfeasible and/or unreliable. However, there are some ways to overcome this issue. In a fragment-based approach, trimeric GAG fragments are docked on the protein's surface, and afterward, they are assembled based on structural overlaps.35 While this method is of great benefit for a number of protein–GAG complexes, it has some flaws, e.g., when GAG is located in a way that some of the oligosaccharide units are in close proximity to the negatively charged amino acid residues (contributing to unfavorable interactions), this method may fail to dock trimeric fragments nearby such residues and thus fail to produce properly docked longer GAG fragments. Perhaps the best method to dock long GAG molecules so far is replica exchange with repulsive scaling method.36,37 This method is rather independent of the length of the GAG both in terms of docking predicting power and computational resources requirement (although, this method could demand heavy computational resources—no matter how long the GAG is—depending on the protein size in the complex). This method, while being promising for GAG docking in the vast majority of cases, may experience difficulties to dock GAG molecule into an enzymatic pocket of the protein.37 One more argument in disfavor of the above-mentioned specific GAG docking approaches is the fact that they bring in some considerable complexity compared to standard docking methods and may be complicated to handle especially for nonexperts in the molecular modeling and researchers not familiar with the mentioned technical solutions.</p><p>Given all that, we aimed to propose a straightforward approach to dock longer GAG molecules without creating unnecessary technical complications while maintaining docking quality. The approach is based on four simple steps: (1) to dock a short (hexameric) GAG; (2) to add more GAG units in the coarse-grained (CG) representation to the previously docked ones manually, e.g., using programs that prepare molecular dynamics (MD) input files like LEaP program from the AMBER suite; (3) to run a molecular dynamics simulation to find an ensemble of GAG conformations for the whole GAG molecule; and (4) to calculate binding free energy. Combining molecular docking with molecular dynamics approaches to predict a complex structure between a receptor and a ligand was previously shown to be a more powerful approach than the usage of the molecular docking alone for other molecular systems.38,39 Moreover, the application of molecular dynamics approach allows for the scoring of docking poses with the use of more accurate free energy calculation schemes than it is usually done within molecular docking software and that, in addition, takes into account movements in the molecular system (this aspect is partially or completely neglected in classical docking scoring schemes). In particular, molecular mechanics/Poisson–Boltzmann surface area (MM/PBSA) and its approximation molecular mechanics/generalized Born surface area (MM/GBSA), both based on the use of the implicit solvent model,40 showed previously to be able to rank experimental binding poses41 and the modeled binding poses22,42 for a number of protein–GAG systems in accordance with the experimental data. Apart from this, the per residue free energy decomposition scheme implemented within these methods allows us to dissect individual free energy contributions of the particular residues to the binding affinity allowed and properly rank the effects of point mutations on the binding affinity in the protein–GAG systems.43,44 Also, recently, it was shown that the MM/GBSA scoring could be useful in distinguishing a native binding pose from other ones for this type of complexes.37</p><p>Therefore, our method combining molecular docking, molecular dynamics, and molecular dynamics-based free energy calculation schemes is expected to be more effective than classical docking approaches because of its conceptual superiority, in particular when applied to a GAG ligand that represents numerous challenges for conventional docking protocols.</p><p>The study consists of several parts. First, MM/PBSA and MM/GBSA methods to calculate binding free energies are applied to a dataset of protein–GAG experimental structures. The results for all-atom (AA) and coarse-grained (CG) GAGs modeled using previously obtained CG parameters that describe several GAG chemical moieties as different beads45 are compared, and the general applicability of these free energy calculation approaches for a CG GAG model is justified. Furthermore, short GAGs from the X-ray structures available for two proteins and GAG docked poses obtained with three peptide receptors are elongated and simulated using a conventional AA approach and the corresponding binding energies are calculated. Then, a new, essentially more simplified, CG model of GAG is introduced. In this model, each GAG monosaccharide unit is represented just by a single pseudoatom. These pseudoatoms are used to substitute the parts of the GAG that are not in contact with the protein/peptide receptor based on the AA simulations. These systems with CG parts are simulated, and the differences between the obtained free binding energies in AA and CG simulations are discussed. Finally, we aimed to propose a model that allows us to calculate free binding energy of a GAG of a given length without simulating the GAG containing an elongated part explicitly using Coulomb and Hückel models of electrostatics. We also attempted to approach the interactions of these GAGs with the protein using only one CG bead to model the elongated part.</p><p>We believe that the method for modeling protein complexes with long GAGs proposed in the study with the introduction of some minor changes should also be applicable to most other charged linear polysaccharides or biopolymers like, for example, nucleic acids.</p><!><p>Short (10 ns) MD simulations (see the protocols in the Molecular Dynamics section) were performed for a dataset of nine protein–GAG X-ray structures obtained from the PDB with the following PDB IDs: 1GMN (receptor: NK1; ligand: HP dp5), 1HM2 (receptor: chondroitinase AC lyase; ligand: dermatan sulfate dp4), 1LOH (receptor: hyaluronate lyase; ligand: hyaluronic acid dp6), 1OFM (receptor: chondroitinase B; ligand: chondroitin sulfate-4 dp4), 2D8L (receptor: rhamnogalacturonyl hydrolase; ligand: desulfated chondroitin sulfate dp2), 2NWG (receptor: CXCL-12; ligand: two HP dp2 bound to two different binding sites), 3ANK (receptor: glucuronyl hydrolase mutant D175N; ligand: chondroitin sulfate-6 dp2), 3OGX (receptor: peptidoglycan recognition protein; ligand: HP dp2), 3OJV (receptor: FGF-1 in complex with the ectodomain of FGFR1c; ligand: HP dp6). The dataset included both enzymatic and nonenzymatic proteins previously shown to be characterized by significantly different binding properties46 and GAGs of different types and lengths. Two series of the simulations were performed: in the first one, GAGs were described by all-atom model (AA), while in the second one, GAGs were simulated using the coarse-grained representation with the parameters obtained previously (CG).45 In this model, specific GAG chemical groups were represented by pseudoatoms, spherical particles described by an integer charge corresponding to the charge of the respective chemical groups and Lennard-Jones parameters. In brief, in this CG representation constructed to be compatible with the AMBER package,47 several pseudoatom types were selected to model the pyranose sugar ring (without hydroxyl group substitutes), N-acetyl, sulfate, and carboxyl groups, as well as glycosidic oxygen atoms. The bonded parameters (bonds, angles, dihedral angles) were obtained by the Boltzmann inversion approach from the corresponding AA simulations: the distributions of the parameters corresponding to the atomic groups defining pseudoatoms were analyzed, and the corresponding force field parameters fitting the distributions were extracted to define the new atomic types using the AMBER formalism. The charges were assigned empirically, while the Lennard-Jones potential parameters for pseudoatoms were calculated using the potential of mean force approach. Molecular mechanics/Poisson–Boltzmann surface area (MM/PBSA) calculations with default parameters for the whole trajectories of the binding free energies as well as per residue decomposition analysis was performed for the whole obtained trajectories.</p><p>Furthermore, the dynamic molecular docking approach (DMD)48 was applied to the structures obtained from the PDB with the following PDB IDs: 1BFB (receptor: FGF-1; ligand: HP dp4), 1BFC (receptor: FGF-1; ligand: HP dp6), 2NWG (receptor: CXCL-12; ligand: HP dp2), 3C9E (receptor: cathepsin L; ligand: chondroitin sulfate-4 dp6), 2JCQ (receptor: CD44; ligand: hyaluronic acid dp7). In these simulations, the GAG molecules were treated as CG, and the obtained results were compared with the AA DMD results for the same protein–GAG complexes from the original DMD work.48 In brief, the DMD approach uses targeted molecular dynamics protocol to dock a GAG ligand to a protein receptor by applying an additional potential to move a ligand from a distant starting position (beyond the cutoff of nonbonded interactions) to the predefined binding site on the receptor surface. DMD performance was compared for AA and CG ligand models of GAGs. The details for the applied protocols can be found in the original DMD work. The following parameters were included for this comparative analysis: RMSatdtop: structural difference between the best scored docked structure and the corresponding experimental structure; RMSatdbest: structural difference between the docked structure, which is the most similar structure to the corresponding experimental structure and the corresponding experimental structure; Rankbest; rank of the docked structure, which is the most similar structure to the corresponding experimental structure; RMSatd: mean structural difference between all docked structures and the corresponding experimental structure; RMSatdtop cluster: mean structural difference between all docked structures from the cluster of solutions with the highest scores and the corresponding experimental structure; r(ΔGtotal ∼ RMSatd): Pearson correlation coefficient for total free binding energy and RMSatd of all docked structures; r(ΔGelect ∼ RMSatd): Pearson correlation coefficient for in vacuo electrostatic free binding energy component and RMSatd of all docked structures; number of correctly predicted residues; number of correctly charged predicted residues; and number of correctly predicted uncharged polar residues were referenced to the 10 protein residues with the highest impacts on binding according to the per residue decomposition for the corresponding X-ray structures.</p><!><p>The following X-ray experimental structures from PDB was used in this work: 1AMX, 2AXM (FGF-1 with HP dp4 and dp6, respectively, monomeric form was used; dp stands for degree of polymerization),49 1BFB, 1BFC (FGF-2 with HP dp4 and dp6, respectively).13</p><!><p>The structure of the N-terminal fragment of the APRIL protein (ALA-VAL-LEU-THR-GLN-LYS-GLN-LYS-LYS-GLN) was adopted from Marcisz et al.29 The structures of both peptides GLY-LYS-GLY-LYS-GLY and LYS-GLY-GLY-GLY-LYS (called InLYS and OutLYS, respectively) were constructed using xleap tool from AMBER suite.47 Afterward, in the case of both peptides, 100 ns MD runs (described in the Molecular Dynamics section) were performed in AMBER to obtain most probable peptide conformations. The APRIL-derived peptide was chosen to represent a naturally existing GAG binding epitope, while InLYS and OutLYS, peptides were artificially constructed as short positively charged model peptides with the difference in the sequential and spatial distance between the GAG binding positively charged LYS side chains.</p><!><p>All of the full-atom GAG structures—HP dp4 and dp6, dp10, dp16—were constructed from the building blocks of the sulfated GAG monomeric units' libraries22 compatible with AMBER16 package. 47GLYCAM06 force field50 and literature data51 were the sources of GAGs' charges.</p><!><p>Since there are no available experimental structures of the peptides with HP, for all three peptides, Autodock352 was used as it was previously described to yield the best results for protein–GAG complexes.34,41 Entire peptides were covered using maximum gridbox size (126 Å × 126 Å × 126 Å) with a 0.375 Å grid step. The size of 300 for the initial population and 105 generations for termination conditions were chosen. A total of 1000 independent runs with Lamarckian genetic algorithm was used, and 9995 × 105 energy evaluations were performed. DBSCAN algorithm53 was used for clustering. RMSatd metric was used for clustering, which accounts for equivalence of the atoms of the same atomic type. This metric was reported to be more appropriate for GAG docking than classical root-mean-square deviation (RMSD) for periodic ligands.48</p><!><p>Obtained in this work, CG parameters compatible with AMBER format were obtained by the Boltzmann inversion approach and saved as the Parameter modification file (file.frcmod, see the Supporting Information). These parameters are described in the Results and Discussion section. These new parameters were obtained to be used for the MD simulations of the docked GAG in the AA representation that was further elongated by CG units. Each monomeric unit was represented by a single pseudoatom.</p><!><p>Experimental structures of protein–GAG, the docked structures of peptide–GAG complexes, and the corresponding structures with elongated GAGs were further analyzed by the MD approach. All of the MD simulations were performed using AMBER16 software package.47 The ff14SB force field parameters were used for the protein and peptide molecules, while GLYCAM06j-1 parameters were used for GAGs. 8 Å water layer from solute to box's bordes in shape of truncated octahedron was used to solvate complexes. Even in the case of HP dp16, this size of the layer was verified to be sufficient enabling the whole GAG molecule to always remain in the periodic box unit during the MD simulation. Na+/Cl– counterions were used to neutralize the net charge of the system. Preceding the production MD runs, energy minimization was made. A total of 500 steepest descent cycles and 103 conjugate gradient cycles with 100 kcal mol–1 Å–2 harmonic force restraint were performed. It continued with 3 × 103 steepest descent cycles and 3 × 103 conjugate gradient cycles without any restraints and followed by heating up the system to 300 K for 10 ps with harmonic force restraints of 100 kcal mol–1 Å–2 with the Langevin thermostat (γ = 5 ps–1). Afterward, the system was equilibrated at 300 K and 105 Pa in isothermal isobaric ensemble for 500 ps with the Langevin thermostat (γ = 5 ps–1) and Berendsen barostat (taup = 1 ps). Then, the actual MD runs were carried out using the same isothermal isobaric ensemble for 100 ns. Particle mesh Ewald method for treating electrostatics and SHAKE algorithm for all of the covalent bonds containing hydrogen atoms were implemented in the MD simulations. For both AA and CG simulations, the integration step of 2 fs was used.</p><p>Although we used short 10 ns MD simulations for a dataset of the experimental structures with short GAGs in the first part of our work (see the Comparing the Performances of MM/PBSA and MM/GBSA Free Energy Decomposition Calculations for GAG Ligands in AA and CG Representations Complexed with Proteins section), here we used 100 ns for all modeled complexes with elongated GAGs with the purpose of obtaining more proper sampling of the GAG conformational space when starting from a docked/modeled structures that cannot be verified by experimental data.</p><!><p>For the free energy and per residue energy decomposition calculations, MM/GBSA (molecular mechanics generalized Born surface area) model igb = 254 from AMBER16 was used with default parameters on the whole trajectories (100 ns) obtained from MD simulations. Linear interaction energy (LIE) analysis was performed with a dielectric constant of 80 and noncalibrated weights (both α and β were set to 1), performed by CPPTRAJ scripts on the same frames as the MM/GBSA.</p><!><p>Prior to analyzing the elongated AA-GAG ligands bound to the proteins with the CG part, which represents the focus of this study, we performed MM/PBSA calculations of the binding free energies for nine nonredundant representative protein–GAG complexes where the full GAGs are modeled by with the AA and CG approaches. The aim of these calculations was to find out if the MM/PBSA method yields the results for a system containing a CG part that are in agreement with the data obtained for a conventional AA system. The CG parameters used to obtain the data provided in this subsection were described in detail in the work of Samsonov et al.45 The data are summarized in Table 1. Pearson and Spearman correlations for ΔGelect, ΔGvdW, and ΔGtotal are 0.997, 0.645, and 0.920; and 0.988, 0.503, and 0.758, respectively, suggesting that CG approximation, as it would be expected, affects van der Waals energy components but retains a very similar description of the systems in terms of the electrostatics. Since the electrostatic interactions are dominating in the protein–GAG systems, the total binding free energies were very similar as well. This suggests that the introduction of the CG part of a GAG that only interacts with the protein receptor via electrostatic interactions could be properly described by the MM/PBSA or MM/GBSA calculations compatibly with similar calculations for AA GAG representation. However, this conclusion should be taken with care: even if the effects of van der Waals description inaccuracies originated from the CG model do not directly affect electrostatic component of binding, they affect the general flexibility of the bound molecule. CG GAGs were shown to be indeed in general less flexible than the AA ones in the original work on this CG model.45 Therefore, the introduction of the CG description affects GAG conformational space and, as a consequence, the whole structural organization of the bound GAG. This, in turn, results in the indirect effect of the modified van der Waals interactions on the electrostatics of the system influencing the binding affinity.</p><!><p>ΔGelect, ΔGvdW, and ΔGtotal are in vacuo electrostatic, van der Waals, and total MM/PBSA binding free energy values, respectively.</p><!><p>The relative mean differences between AA and CG absolute energy values (normalized by the AA corresponding values) are 30, 5, and 18% for ΔGelect, ΔGvdW, and ΔGtotal, respectively (clear outlier 3ANK is excluded). For all components, the values obtained with CG approach are overestimated in comparison to the ones from the AA approach. Per-residue free energy decomposition also shows systematic agreement for the AA and CG approaches when analyzing the individual impacts of the protein residues (Table S1). At the same time, there are no correlations in the per residue values obtained for the GAG residues. Furthermore, we compared the performances of the DMD docking approach using both AA and CG GAG representations (Table S2). The results obtained for the CG GAG model are slightly worse but, in general, quite similar to the ones obtained for the AA GAG model in the original DMD study.45 All of these analyses suggest that the CG description of a GAG molecule complexed with a protein is consistent with the AA representation in terms of application of the MM/PBSA. This served as a premise for our further step in this study: in particular, for the proposition of even a more simplistic CG model for a GAG part that does not establish direct contact with a protein receptor. In this model, the interactions between this CG part of a GAG and the protein could be described as purely electrostatics-driven.</p><!><p>To obtain the reference data for the CG model, development and testing AA MD simulations were performed. For this, the available experimental structures of FGF-1 (PDB ID: 1AXM, 2AMX) and FGF-2 (1BFB, 1BFC) with HP dp4 and dp6 were used. These complexes could be successfully obtained by many conventional docking programs including AD3 (RMSD ∼2.5 to 3.5 Å for the best scored docked poses).34,41 Since the experimental structures with the peptides are not available, HP dp4 and dp6 were docked to all of the peptides: N-terminal part of the APRIL protein, InLYS, and OutLYS (all targets described in the Materials and Methods section). It is important to mention that in this work, we did not aim to improve the docking quality for short GAGs but to estimate the effect of the GAG elongation and to understand if this elongation could be described properly using a mixed AA/CG GAG model. AA representation of GAGs was used as a reference for our analysis.</p><p>Since MM/PBSA and MM/GBSA approaches yielded essential correlation in protein–GAG systems (see an example in Figure S1), we further used only the MM/GBSA approach for these longer simulations since this approach is significantly faster.</p><p>We clearly observe that longer GAGs bind stronger independently of the analyzed system and the type of the receptor (protein or peptide) (Table 2). This is an expected net effect of the electrostatic interactions that become stronger with the increase of the GAG negative charge upon its elongation. Since the net charge of a GAG binding site on the protein/peptide surface always corresponds to the extent of the positive electrostatic potential,41 an elongation of any GAG ligand bound to any of its receptors would render the interactions stronger. Although the specific binding unit of GAGs is relatively short according to the available PDB structures of protein–GAG complexes,41 natural GAGs in the extracellular matrix are very long, reaching molecular weights up to over 100 kDa,5 rendering the energetic effect originating in a GAG long chain to be important to take into account when the corresponding modeling is performed. Except the 2AXM, the difference between dp6 and dp16 in terms of binding free energy was 20% or higher (on average 24%). One more highlight of this comparison is that the energy discrepancy between dp6 and dp10 was 2 times higher than that between dp10 and dp16 despite addition of more sugar ring units in the case of dp10 to dp16 elongation. A very large increase in terms of binding strength was observed upon the elongation from dp4 to dp6, indicating that experiments with dp4 GAGs may strongly underestimate the binding strength of longer GAGs. Taken into account how often dp4/dp6 GAGs are used as models in computational studies, it is worth checking and rethinking those standards prior to applying dp4-based protocols to any new system.</p><!><p>In the case of one MD simulation, dissociation was observed. The first value indicates energies w/o mention of MD run, and the second value indicates those when taking it into account.</p><!><p>The new parameters described below were obtained from the AA MD simulations to be used for the CG elongation of the docked GAG in the AA representation as described in the following subsections. This new model was particularly developed for the purpose of elongating those parts of bound GAG chains that do not establish direct contact with the protein these GAGs are interacting with, and, therefore, it is thought to account only for electrostatics. Containing a single new atomic type corresponding to a whole GAG monomeric unit, this model is conceptually different and much more simple than the old one.45 It is completely nonspecific for any chemical modifications of GAG residues since it is constructed to account primarily for electrostatic interactions and could be used for all negatively charged monosaccharide residues allowing for a straightforward modification of the residue point charge when needed. This is not the case for the old model that, on the contrary, was developed to consider specific electrostatic and van der Waals interactions for particular monosaccharide units. In terms of the required computational expenses, MD simulations with the new model would be faster if only a CG GAG would be simulated. However, in the presence of a protein, an AA-GAG part, and explicit water molecules, the benefit in terms of the computational time reduction is rather negligible.</p><p>The new Z1 atomic type constructed corresponds to a CG pseudoatom describing a complete residue unit (monosaccharide unit with the charge of −2) and, therefore, glycosidic linkages are omitted between monosaccharide units in this CG model.</p><!><p>Bonded parameters (bonds, angles, and dihedral angles) were obtained from the AA MD simulations. For the calculations of equilibrium values and harmonic constants for bonds, angles, and dihedral angles (Tables 3–5), the Boltzmann inversion approach was used.55 In the case of dihedral angles (Table 5), periodicity was set to 1 or 3 depending on the number of maxima/minima of the potential per 360°, and the amplitude was obtained as the difference between the global minimum and the highest energetical barrier between global and local minima. In the case of artifacts observed during simulations, particular parameters were manually refined.</p><!><p>Force constant.</p><p>Equilibrium bond length.</p><p>Force constant.</p><p>Equilibrium angle value.</p><p>Factor by which the torsional barrier is divided.</p><p>Barrier height divided by a factor of 2.</p><p>Phase shift angle in the torsional function.</p><p>Periodicity of the torsional barrier.</p><!><p>The charge of the pseudoatom of the monomeric unit of the HP was set accordingly to the number of sulfate and carboxyl groups, which is −1 per group in the unit. In the case of Lennard-Jones parameters, the RvdW (van der Waals radius) and EDEP (energy well depth) values were empirically assigned to the doubled and equal values obtained for the internal pyranose ring in our previous CG model of GAGs, respectively (Table 6).45</p><!><p>van der Waals radius.</p><p>Energy well depth.</p><!><p>To evaluate our CG model (Figure 1) of the HP, MD simulations with CG atoms were performed and compared to all-atom MD simulations. In AA runs, we observed that the core of GAG—the part that is especially the closest to the binding side of the protein/peptide—is in the closest proximity of the protein and barely moved. In contrast, it is the lateral parts of the GAGs that tend to move freely (Figure 2). It suggests that interactions between those parts and the protein are even less specific and thus almost purely electrostatics-driven. Therefore, we believe that replacing lateral parts of the GAGs with CG model units should not substantially affect the nature of the interactions established between the analyzed molecules.</p><!><p>Graphical representation of all-atom (left) and mixed (right) model of dp16 heparin in complex with FGF-2. Protein is in cartoon representation (yellow); all-atom and CG GAGs are in licorice and van der Waals sphere representation, respectively (cyan).</p><p>Graphical representation of the MD run of complex of APRIL peptide (orange cartoon) with HP dp16 (licorice). The color scheme from red to blue indicates heparin conformations ranging from the beginning to the end of the MD simulation.</p><!><p>First, we compared the convergence of MD simulations for the AA and CG approaches in terms of the structural flexibility and energetics (Figures S2 and S3, respectively). In most of the cases, the convergence in terms of RMSD was observed already after 20 ns. Clearly, the flexibility of the AA GAGs is significantly higher than in the mixed AA/CG model. For MM/GBSA binding free energy, the converge is already reached after 10 ns of the simulation, and there are slightly higher variations of the energy observed for the AA simulation, while there are no differences in the time needed for the convergence. The trends of the convergence observed here should not be expected to be the same for other protein–GAG or peptide–GAG complexes. Indeed, in other systems, MD simulations may take longer or shorter to converge. Nevertheless, the goal of the MD simulations performed in this study is not to reach a convergence but to show that the transition from AA to CG representation of the GAG part does not substantially affect the results of the free energy calculations in the same system.</p><p>Starting positions of the molecules from all-atom simulations were taken. Original dp6 part of the HP was not modified, and only atoms that were manually added to build dp16 were replaced with CG pseudoatoms for HP rings. Additionally, the user can use the script (Supporting Information) for automatic addition of pseudoatoms. Then, MD simulations with a GAG represented as AA in the binding core and as CG in its lateral parts were performed, and the results obtained from MM/GBSA energy analysis from mixed model simulations of dp16 HP are listed in Table 7. Average difference obtained from energy analysis of mixed CG/AA model compared to the AA model was 5.6%. Compared to the difference that is a consequence of using shorter GAGs, which is on average 24% (dp6) and 39% (dp4) underestimation of the value, it is a substantial improvement. In the case of the mixed model, most of the values were also underestimated (compared to the AA model): 7% for the N-terminal fragment of APRIL, 3% for the FGF-2 and OutLYS peptide, and 1% for the InLYS peptide. However, the binding free energy calculations showed 14% overestimation in the case of the FGF-1/HP complex. Additional energy analysis was performed in the form of LIE calculations and is described in the Supporting Information (Table S3).</p><p>During MD runs of both AA and mixed AA/CG models, we observed similar motions of the GAGs molecules with respect to the protein/peptide, which suggests that the used CG model also properly reflects the dynamics of the system (Figure 3).</p><!><p>Graphical representation of the MD run of complexes of APRIL peptide (cartoon) with all-atom (left, orange) and mixed model (right, green) HP dp16 (licorice). The color scheme from red to blue indicates heparin conformations ranging from the beginning to the end of the MD simulation.</p><!><p>The division of the modeled GAG chain into AA and CG parts for the further MD analysis could be done by analyzing the free energy properties of the binding poses instead of using visual inspection of AA MD followed by the manual selection of the residues to substitute. For this, we performed per-residue energy analysis of the complexes from AA MD simulations. This procedure allows us to define the particular contributions of the individual GAG units to binding a protein or a peptide. Then, only the residues with "weak" contributions to the binding energies were selected and further modeled by the CG approach. The threshold was set to −0.5 kcal mol–1, and any residue with energy value less favorable than this value was replaced. The idea behind such a procedure to substitute only the monosaccharide units with less substantial contributions in terms of binding energy is related to our goal to use the CG model for residues that are further away from the binding region and so less affecting the binding. Interestingly, the obtained error was higher (on average 10% of free energy difference compared to the AA simulation) when the residues were picked based on per-residue free energy decomposition than when the elongation was completed independently of such calculations (Table 7).</p><!><p>Furthermore, we aimed to extrapolate binding energies obtained from the analysis of the dp6 GAG to calculate them for the elongated GAG molecules without performing any further MD simulations. First, we proposed an equation based on Coulomb's law to calculate the factor (depicted as W factor) that would allow us to obtain the binding energy of the complex containing GAGs of any length. Such an approach assumes that only electrostatic interactions are substantial for the added GAG part. We also proposed a script (see the Supporting Information) that would automatically calculate the binding energy of the elongated fragment of the GAG when given two files (pdb file of a bound GAG molecule and a receptor) and predefined W factor.</p><p>To calculate the W factor for the particular GAG residue, we use the following equationwhere W is the factor, ΔGres is the energy obtained from per-residue energy decomposition from MM/GBSA analysis, and ∑i/j is the sum of reciprocities of the distances between GAG residues and all of the positively/negatively charged residues of the protein.</p><p>Each positive and negative residue is taken into account if it is within the cutoff of nonbonded interactions in the corresponding MD simulation. The W factor for the whole complex is the mean of the W factors for each of the GAG residues calculated from the simulations with HP dp16, and its usage for HP dp16 energy prediction would, therefore, yield the same energies as the ones obtained from the MD simulation.</p><p>The W factors and their distribution (Figure 4) for the peptide–GAG complexes were very similar for the peptides: −3.35, −3.31, and −3.33 kcal mol–1 e–1 for InLys, OutLys, and N-terminal fragment of the APRIL protein, respectively. In contrast, in the case of protein complexes, they differed substantially in terms of mean of the W factors (0.65 and −0.50 kcal mol–1 e–1 for FGF-1 and FGF-2, respectively), and their distribution (Figure 4). It indicates that bigger and therefore more complex systems need an individual approach each time they are analyzed. However, in the case of simple and short systems (e.g., small peptide and GAG) individual approach is not necessary and the binding energy could be calculated directly using W factor of −3.33 kcal mol–1 e–1. In this case, performing MD simulations and binding energy analysis for longer GAG variants is not needed.</p><!><p>Plot of W value probability densities calculated from MD runs (5 MD runs for each individual complex) for HP dp16 and short peptides (top) or proteins (bottom) used in this study.</p><!><p>Then, similarly to the previously described procedure, the Debye–Hückel equation (ΔG ∼ e–ϰr/r, where r is the distance and ϰ is the reversed Debye screening length) was used to calculate the W factor. In this approach, electrostatics screening in the electrolyte solution is taken into account. Physiological value of the ionic strength (0.15 M) was used in the calculations. The obtained data also suggested that W is very similar for all three peptides: 86.40, 89.13, and 85.50 kcal mol–1 e–1 for APRILpep, OutLYS, and InLys, respectively. The calculated values for the protein–GAG systems were essentially different for the two systems and compared to the peptides: −0.83 and 20.00 kcal mol–1 e–1 for 2AXM and 1BFC, respectively.</p><p>Therefore, the energies could be, in principle, predicted for HP using a specific W factor for each system (in the case of three peptides, W factors are essentially the same), and such predictions applied for longer GAGs with this particular W factor would yield similar values to those in the MD simulations. However, for proteins, it is not possible to make such predictions a priori without performing MD simulations that are needed to define the W factor.</p><p>Based on these results, we believe that the difference in W profiles for two proteins obtained by calculations based on two dissimilar physics-based models is originated in the different charge distribution topology, protein surface geometry, and thus resulting electrostatic screening effects that do not allow us to find the same uniform factor for distinct protein receptors.</p><!><p>Furthermore, we aimed to design a model where only a single pseudoatom would function as an elongated lateral part of the bound GAG. Unfortunately, among the different parameters that were used, none yielded promising results in terms of reliably obtaining binding energies for the complexes compared to the ones from AA simulations, both when compared energies from MM/GBSA and LIE analysis (Table S3). Some artifacts were also observed when pseudoatom had a high negative charge (−5 or lower) causing the interruption of the MD simulation. We believe that this approach does not have broad applicability. It is rather unlikely to propose parameters for a pseudoatom that would work consistently for the complexes with different electrostatic properties and geometry topologies. Additionally, one would need to propose a complete library of parameters for pseudoatoms distinct for every different length of an elongated GAG part that pseudoatom is replacing. The possible reason for this could be that an attempt to approximate an elongated molecule with a spherical particle could probably be physically inappropriate in terms of molecular symmetry.</p><!><p>While docking long GAG molecules may require additional laborious technical work than docking shorter (dp4/6) GAG oligomers, it is definitely worth the effort. In our approach, we use Autodock3 to find the best starting poses for the dp6 GAGs34,41 that can be used for further GAG elongation. At the same time, it is important to mention that our approach is not limited to any special docking software. We expect that carbohydrate- and GAG-specific docking programs as Vina-Carb56 or GlycoTorch Vina,57 respectively, which also belong to the family of Autodock programs, would perform similarly or even outperform Autodock3 for obtaining the initial structures of protein/peptide complexes with short GAGs that are to be further elongated using the procedure proposed in this manuscript. In this procedure, we elongate a docked GAG using the CG model for the monosaccharide units and use it in conventional MD simulations. In this study, it was proven that elongating GAGs substantially increases the binding energy of the complex. While it is not a linear increase of binding strength, it is still substantial when dp16 is compared to dp4 or dp6. We consider that GAG elongation using a CG model for the monosaccharide units provides nearly equivalent outcome as the AA elongation, resulting only in a 5.6% difference in assessed binding energies, without introducing excessive technical complications. This suggests that a straightforward description of electrostatic interactions of the GAG parts not establishing direct contacts with their protein target is sufficient to describe the energetics of the system accurately enough. Binding energies obtained when using our script that elongates a GAG molecule (Supporting Information) and the CG model that are provided in this work are more accurate than using shorter GAGs with a standard AA approach. This method can be utilized by any user of AMBER and standard docking software like Autodock3 in a straightforward manner. It is a great advantage that with this approach, a user can specify the length of the extended lateral part of GAG to properly satisfy his needs. We also believe that this method with minor modifications could be implemented to other linear polysaccharides or negatively charged linear polymers like nucleic acids, in general.</p><!><p>MM/PBSA free binding energy per residue decomposition analysis; DMD docking in protein–GAG systems: comparison of AA and CG GAG representations; LIE energy calculations; comparison of MM/GBSA and MM/PBSA binding free energies; RMSD of the bound HP dp16 (to investigated proteins/peptides) in AA and AA/CG representation; elongation script; script for automatic calculation of the energy of elongated GAG molecule; and frcmod file (PDF)</p><p>ci1c00664_si_001.pdf</p><p>The authors declare no competing financial interest.</p><!><p>All of the PDB files were downloaded from the RCSB Protein Data Bank (https://www.rcsb.org). AMBER libraries for GAG residues are accessible on the website of GAG Computational Group under "Libraries" (http://www.comp-gag.org/downloads). The following free pieces of software were used: Autodock3 for molecular docking (autodock.scripps.edu), Open Babel for formats conversion (http://openbabel.org), VMD for molecular visualization (https://www.ks.uiuc.edu/Research/vmd), and scripts for docking analysis (https://gehrcke.de/code/). All of these softwares are freely accessible on their websites. The differences between the input scripts used in molecular docking and the default values are listed on the website of GAG Computational Group under "Scripts/protocols" (http://www.comp-gag.org/downloads). All MD simulations were performed by the commercial AMBER16 molecular dynamics package (https://ambermd.org) purchased by the University of Gdańsk. The XLEAP and CPPTRAJ modules of AMBER16 were used for adding terminal residues to the polysaccharide chains and postprocessing of trajectories, respectively. MM/GBSA free binding energy calculations were performed with the MM/GBSA module of AMBER16.</p>

PubMed Open Access

The association between spinal cord trauma-sensitive miRNAs and pain sensitivity, and their regulation by morphine

Increased pain sensitivity is a common sequela to spinal cord injury (SCI). Moreover, drugs like morphine, though critical for pain management, elicit pro-inflammatory effects that exacerbate chronic pain symptoms. Previous reports showed that SCI results in the induction and suppression of several microRNAs (miRNAs), both at the site of injury, as well as in segments of the spinal cord distal to the injury site. We hypothesized that morphine would modulate the expression of these miRNAs, and that expression of these SCI-sensitive miRNAs may predict adaptation of distal nociceptive circuitry following SCI. To determine whether morphine treatment further dysregulates SCI-sensitive miRNAs, their expression was examined by qRT-PCR in sham controls and in response to vehicle and morphine treatment following contusion in rats, at either 2 or 15 days post-SCI. Our data indicated that expression of miR1, miR124, and miR129-2 at the injury site predicted the nociceptive response mediated by spinal regions distal to the lesion site, suggesting a molecular mechanism for the interaction of SCI with adaptation of functionally intact distal sensorimotor circuitry. Moreover, the SCI-induced miRNA, miR21 was induced by subsequent morphine administration, representing an alternate, and hitherto unidentified, maladaptive response to morphine exposure. Contrary to predictions, mRNA for the pro-inflammatory interleukin-6 receptor (IL6R), an identified target of SCI-sensitive miRNAs, was also induced following SCI, indicating dissociation between miRNA and target gene expression. Moreover, IL6R mRNA expression was inversely correlated with locomotor function suggesting that inflammation is a predictor of decreased spinal cord function. Collectively, our data indicate that miR21 and other SCI-sensitive miRNAs may constitute therapeutic targets, not only for improving functional recovery following SCI, but also for attenuating the effects of SCI on pain sensitivity.

the_association_between_spinal_cord_trauma-sensitive_mirnas_and_pain_sensitivity,_and_their_regulati

1. Introduction<!>2.1. Subjects<!>2.2 Surgery<!>2.3. Drug Administration<!>2.4. Assessment of locomotor function<!>2.5. Sensory Reactivity<!>2.6. Isolation of RNA for qRT-PCR<!>2.7. Quantitative RT-PCR<!>2.8. Data analysis<!>3.1. Quantification of morphine-induced changes in spinal cord miRNA expression<!>3.2. Impact of injury severity and functional recovery on miRNA expression in spinal cord<!>3.3. Relationship between SCI-sensitive miRNA expression and sensory reactivity<!>3.4. Expression of miR-1 in vascular tissues following SCI<!>3.5. MiRNA regulation of pain neuro-circuitry associated cytokines in spinal cord<!>4. Discussion<!>5. Conclusion

<p>The effects of spinal cord injury (SCI) can be severely debilitating, especially the chronic presence of neuropathic pain that affects close to two-thirds of all spinal cord-injured patients (Anderson, 2004; Siddall and Loeser, 2001). One of the primary treatments for chronic pain following SCI is administration of opiate analgesics such as morphine, which are also given during the acute phase of spinal injury for pain mitigation (Dworkin et al., 2003; McCarberg, 2004). However, administration of morphine in the acute phase of a spinal contusion injury in rats has been shown to significantly attenuate locomotor function and increase expression of chronic pain symptoms (Hook et al., 2007; Hook et al., 2009). The increased chronic pain associated with morphine administration has been attributed to the activation of astrocytes and microglia by morphine and inhibition of analgesic effects by increased expression of pro-inflammatory cytokines TNF, IL-1, and IL-6 (Cui et al., 2006; Johnston et al., 2004; Raghavendra et al., 2002; Song and Zhao, 2001; Tai et al., 2006). Accordingly, the application of an interleukin-1 receptor antagonist prior to intrathecal morphine prevents the maladaptive effects of morphine on neuropathic pain and functional recovery (Hook et al., 2011).</p><p>While these results are promising, there is limited knowledge of the remodeling of pain neuro-circuitry and the activation of potentially maladaptive gene regulatory networks following morphine treatment in SCI. MiRNAs are short, 18 to 25 nucleotide-long non-coding small RNAs that play an important regulatory role in gene expression by inhibiting protein translation or targeting mRNA transcripts for degradation (Alvarez-Garcia and Miska, 2005; Zamore and Haley, 2005). Individual miRNAs are able to coordinate gene networks to affect a specific cellular endpoint by simultaneously controlling expression of several hundred mRNAs. Furthermore, miRNAs have been implicated in physiological processes relevant to neuropathic pain, including innate immune responses involved in wound inflammation (Roy and Sen, 2011; Williams et al., 2008) and the regulation of inflammation in response to morphine treatment in human monocyte-derived macrophages (Dave and Khalili, 2010).</p><p>MiRNAs are also dysregulated following SCI (Liu et al., 2009; Nakanishi et al., 2010; Strickland et al., 2011; Yunta et al., 2012). We previously reported that miR124, miR129, and miR1 were significantly down-regulated following spinal cord contusion, while both miR21 and miR146a were transiently induced, and changes in miR146a and miR129-2 expression significantly explained the variability in injury severity (Strickland et al., 2011). We hypothesized that the temporal expression of miR146a, miR21, and other SCI-sensitive miRNAs may be further dysregulated following administration of morphine in the acute phase of a spinal contusion injury. Such dysregulaton could constitute an important role in both inflammation and the progression of neuropathic pain during the chronic phase of injury. Moreover, these miRNAs may contribute to the regulation of the nociceptive circuitry and chronic effects of morphine exposure. To assess these possibilities, the current study investigated the relationship between SCI-sensitive miRNA expression and morphine administration, evaluated at 2 and 15 days post-SCI to explore the effects of acute morphine treatment in both the acute and chronic phases of SCI. In addition, we also investigated the relationship between miRNA expression and the variance in baseline and morphine-attenuated pain sensitivity following SCI.</p><!><p>All of the experiments were reviewed and approved by the institutional animal care committee at Texas A&M University and all NIH guidelines for the care and use of animal subjects were followed. Male Sprague–Dawley rats (Rattus norvegicus) obtained from Harlan (Houston, TX) at approximately 90-110 days old (300-350 g) were individually housed in Plexiglas bins [45.7 (length) × 23.5 (width) × 20.3 (height) cm] with food and water continuously available. To facilitate access to the food and water, extra bedding was added to the bins after surgery and long mouse sipper tubes were used so that the rats could reach the water without rearing. Subjects were weighed on the same days that they were assessed for locomotor function, and were checked daily for signs of autophagia and spasticity. A subject was classified as having spasticity if the limb was in an extended, fixed position and was resistant to movement. Bladders were manually expressed in the morning (8:00-9:30 hrs) and evening (18:00-19:30 hrs) until subjects regained bladder control, which was operationally defined as three consecutive days with an empty bladder at the time of expression. The rats were maintained on a 12 hr light/dark cycle and tested during the last 6 hrs of the light cycle.</p><!><p>Subjects were anesthetized with isoflurane (5%, gas). Once a stable level of anesthesia was achieved, the inspired concentration of isoflurane was lowered to 2-3% and an area extending approximately 4.5 cm above and below the injury site was shaved and disinfected with iodine. A 7.0 cm incision was made over the spinal cord. Next, two incisions were made on either side of the vertebral column, extending about 3 cm rostral and caudal to the T12-T13 segment. The dorsal spinous processes at T12 were removed (laminectomy), and the spinal tissue exposed. The dura remained intact. For the contusion injury, the vertebral column was fixed within the MASCIS device (Constantini and Young, 1994; Gruner, 1992) and a moderate injury was produced by allowing the 10 g impactor (outfitted with a 2.5 mm tip) to drop 12.5 mm. Sham controls received a laminectomy, but the cord was not contused with the MASCIS device. Following surgery, the wound was closed with Michel clips. T12 level contusion models have been routinely used by members of our group to define spinal cord learning circuits and molecular mechanisms involved with recovery of function (Brown et al., 2011; Ferguson et al., 2008; Hook et al., 2011). Lesions at this level result in well-defined and replicable sensory-motor deficits, and we therefore chose to utilize contusion at this level to also examine changes in miRNA expression.</p><p>An intrathecal cannula was also inserted into the subarachnoid space immediately after the contusion injury. For this procedure, a 15 cm long polyethylene (PE-10) cannula, fitted with a 0.23 mm (diameter) stainless steel wire (SWGX-090, Small Parts), was threaded 2 cm under the vertebrae immediately caudal to the injury site. The cannula tip terminated over the S2-S3 spinal segments, so that morphine was delivered to the lumbosacral regions mediating hind-paw pain reactivity. To prevent cannula movement, the exposed end of the tubing was secured to the vertebrae rostral to the injury using an adhesive (Cyanoacrylate). The wire was then pulled from the tubing and the wound was closed using Michel clips.</p><p>To help prevent infection, subjects were treated with 100,000 units/kg Pfizerpen (penicillin G potassium) immediately after surgery and again 2 days later. For the first 24 hrs after surgery, rats were placed in a recovery room maintained at 26.6 °C. To compensate for fluid loss, subjects were given 2.5 ml of saline after surgery. Michel clips were removed 14 days after surgery.</p><!><p>After baseline assessments of locomotor and sensory function, rats were assigned to a morphine or vehicle treatment group. Twenty-four hours or 14 days after injury, the rats were given an intrathecal infusion of morphine sulfate (90 μg i.t., Mallinckrodt, Hazelwood, MI) dissolved in 2 μL of distilled water. This dose of morphine was chosen based on previous studies that demonstrated that this high dose is required to achieve strong anti-nociception and block behavioral reactivity to nociceptive stimulation after SCI, as well as being detrimental to long-term recovery of locomotor function (Hook et al., 2009; Hook et al., 2011). Control subjects were treated with 2 μL of vehicle. These drug injections were followed by a 10 μL injection of saline, to flush the catheter. Locomotor ability was assessed with the BBB scale prior to and after drug treatment. This behavioral index was used to ensure that injury severity was balanced across groups prior to drug treatment.</p><!><p>Locomotor behavior was assessed using the Basso, Beattie and Bresnahan (BBB) scale (Basso et al., 1995), in an open enclosure (99 cm diameter, 23 cm deep). This 21-point scale is used as an index of hindlimb functioning after a spinal injury. Using this scale, no movement of the hindlimbs (ankle, knee or hip) is designated a score of 0, and intermediate milestones include slight movement of one joint (1), extensive movement of all three joints (7), occasional weight supported stepping in the absence of coordination (10), and consistent weight supported stepping with consistent forelimb-hindlimb coordination (14). Higher scores reflect consistent limb co-ordination and improved fine motor skill. Baseline motor function was assessed on the day following injury and prior to drug treatment. Because rodents often freeze when first introduced to a new apparatus, subjects were acclimated to the observation fields for 5 min per day for 3 days prior to surgery. Each subject was placed in the open field and observed for 4 min. Care was taken to ensure that the investigators' scoring behavior had high intra- and inter-observer reliability (all r's > 0.89) and that they were blind to the subject's experimental treatment.</p><!><p>For the assessment of morphine efficacy, nociceptive reactivity was assessed immediately before, and 30 minutes after intrathecal morphine administration. Thermal reactivity was assessed using radiant heat in the tail-flick test. Subjects were placed in the restraining tubes with their tail positioned in a 0.5 cm deep groove, cut into an aluminum block, and allowed to acclimate to the apparatus (IITC Inc., Life Science, CA) and testing room for 15 min. The testing room was maintained at 26.5°C. Thermal thresholds were then assessed. Thermal reactivity was tested using a halogen light that was focused onto the rat's tail. Prior to testing, the temperature of the light, focused on the tail, was set to elicit a baseline tail-flick response in 3-4 sec (average). This pre-set temperature was maintained across all subjects. In testing, the latency to flick the tail away from the radiant heat source (light) was recorded. If a subject failed to respond, the test trial was automatically terminated after 8 s of heat exposure. Two tests occurred at 2-minute intervals, and the second test's tail-flick latencies were recorded. To confirm that subjects did not respond in the absence of the stimuli, blank trials were also performed. A 'false alarm' was recorded if subjects made a motor or vocalization response during the blank tests. The blank trials were performed 1 min before or after each sensory test (in a counterbalanced fashion). No false alarms were recorded.</p><!><p>Subjects were euthanized (100 mg/kg of pentobarbital, i.p.) 24 hrs after drug treatment. A 5 mm segment of spinal cord from the injury site, hearts, and carotid arteries were collected, snap frozen in liquid nitrogen, and stored at −80 °C until further analysis. Specimens were crushed and RNA was isolated using the TRIzol (Invitrogen; Carlsbad, CA) protocol. Total RNA was then quantified using a NanoDrop 2000 Spectrophotometer (Thermo Scientific; Wilmington, DE) and stored at -80 °C.</p><!><p>MiRNA expression data was measured with quantitative reverse transcription (qRT)-PCR for miRNAs, based on the protocol of the miRCURY™ LNA microRNA Universal RT-PCR system (EXIQON; Woburn, MA). RNA samples were converted to cDNA, and qRT-PCR was performed using a 7900HT Fast Real-Time PCR System (Applied Biosystems, Foster City, CA). Forward and reverse primers (EXIQON; Woburn, MA) for hsa-miR124, hsa-miR1, hsa-miR21, hsa-miR129-2, and hsa-miR146a were used for PCR amplification, and real time data was normalized using U6 RNA. Similarly, mRNA expression of IL6R was measured using qRT-PCR for mRNAs, based on the protocol for PerfeCTa® SYBR® Green SuperMix with ROX™ (Quanta Biosciences; Gaithersburg, MD). RNA samples were converted to cDNA using qScript™ cDNA SuperMix (Quanta Biosciences; Gaithersburg, MD), and qRT-PCR was performed using a 7900HT Fast Real-Time PCR System (Applied Biosystems, Foster City, CA). Forward and reverse primers (Integrated DNA Technologies; Coralville, IA) for IL-6R were used for PCR amplification, and real time data was normalized using glyceraldehyde 3-phosphate dehydrogenase (GAPDH). Relative miRNA and mRNA expression was determined by calculating the mean difference between cycle threshold of either the miRNA from the U6 small nuclear RNA (U6SNB) normalized control, or the IL6R mRNA from the GAPDH normalized control for each sample [cycle threshold (CT)] within each sample group (samples with same miRNA ID, time, and condition parameters) and expressed as - CT for relative change in expression. Sample means that were greater than ± 2 standard deviations from the mean CT, or ± 3 standard deviations from the mean CT after exclusion, were considered outliers and removed from the analysis. Of the 218 data points in 36 groups used in the analysis, only 14 data points were excluded according to this criteria, and no more than one data point in any individual experimental group was excluded. Fold change in miRNA/mRNA expression was determined by calculating the difference between the mean CTs of sham, morphine, and vehicle sample groups at the same time point (- CT), and expressed as a baseline-corrected percentage of fold-change ([2-ΔΔCT-1]*100).</p><!><p>All data were analyzed using SPSS software version 18 (SPSS; Chicago, IL). MicroRNA expression, verified by qRT-PCR, was analyzed by multivariate analysis of variance (ANOVA) using Pillai's trace statistic, and further analyzed using post hoc univariate ANOVA and Fisher's least significant difference (LSD) test. Other data were analyzed using ANOVAs followed by post hoc LSD using planned comparisons. In all cases, the a priori value was set at 0.05. Data were expressed as mean ±SEM, as indicated in the figure legends.</p><p>Correlations between initial injury severity and either miRNA or IL6R mRNA expression, and between tail-flick latency and either miRNA or IL6R mRNA expression, were determined by Pearson's product-moment correlation using – CT values of either the miRNAs or IL6R, initial BBB scores, day 13 BBB scores, and tail-flick latency as separate independent variables. All partial correlations were corrected for time only. The a priori value was set at 0.05, and data were expressed as the mean difference in cycle threshold change of either each miRNA relative to the cycle threshold of U6 controls (- CT= CTU6 - CTmiRNA), as indicated in the legend of Figures 2, 3, and 4, or IL6R expression relative to the cycle threshold of GAPDH controls (- CT= CTGAPDH - CTIL6R), as indicated in the legend of Figure 7.</p><!><p>We previously reported that miR1, miR21, miR124, miR129-2, and miR146a were significantly dysregulated following spinal cord contusion (Strickland et al., 2011). To determine whether morphine treatment further dysregulates SCI-sensitive miRNAs, their expression was examined by qRT-PCR in sham controls and in response to vehicle and morphine treatment following contusion, at either 2 or 15 days post-SCI (Figure 1). Multivariate ANOVA of qRT-PCR data indicated a significant main effect of time (Pillai's Trace Statistic, F(5,19) =21.094; p<0.005) and treatment (F(10,40)=3.742; p<0.005), and a significant interaction effect between time and treatment (F(10,40) =6.478; p<0.005). Post hoc univariate ANOVAs indicated a main effect of treatment on miR21, miR129-2, and miR146a (F(2,23) =65.038; p<0.005, F(2,23) =17.265; p<0.005, and F(2,23) =16.163; p<0.005, respectively), and a significant interaction effect of time and treatment on miR21 and miR146a expression (F(2,23) =8.229; p<0.005 and F(2,23) =10.055; p<0.005, respectively). In addition, post hoc LSD t-tests indicated that both miR21 and miR146a expression is increased following spinal cord trauma, irrespective of morphine treatment (pmiR21<0.001 and pmiR146a <0.001; Fig. 2a, b). In contrast, miR129-2 expression was down-regulated following both morphine and vehicle administration (pmiR129-2<0.001 and pmiR129-2<0.001, respectively), and there was no change in the expression of either miR1 or miR124 (Figure 2a, b). Post hoc planned comparisons indicated that there was a significant increase in expression of miR21 in response to morphine treatment relative to vehicle at 15 days post-SCI (Student's two-tailed t-test, p<0.05; Figure 2c, d). In addition, Pearson's product-moment correlations indicated a statistically significant correlation between miR21 expression and time (Pearson's r=0.61, p<0.005). Controlling for time, partial correlation analysis showed that miR21 expression was significantly correlated with the expression of miR1, miR124, and miR146a (Pearson's r=0.51, p<0.03, Pearson's r=0.58, p<0.01, and Pearson's r=0.83, p<0.001, respectively).</p><!><p>We previously observed a significant relationship between the expression of both miR129-2 and miR146a and injury severity, as indicated by initial BBB scores following surgery. Accordingly, we assessed whether this relationship would be replicated by the new data set. Indeed, there was a statistically significant correlation between BBB scores at 24 hours post-SCI and the expression of both miR129-2 and miR146a across both time points (Pearson's r=0.66, p<0.001 and Pearson's r=-0.55, p<0.001, respectively; Figure 3a,b). Contrary to our previous findings, we also found a statistically significant correlation between the expression of miR21 and initial BBB scores (Pearson's r=-0.88, p<0.001; Figure 3c). Additionally, we analyzed the relationship between recovery of function, as indicated by Day 13 BBB scores, and miRNA expression at 15 days post-SCI. Similar to initial injury severity, there were statistically significant correlations between BBB scores at 13 days post-SCI and expression of miR129-2, miR146a, and miR21 (Pearson's r=0.50, p<0.05, Pearson's r=-0.70, p<0.001, and Pearson's r=-0.80, p<0.001, respectively; Figure 4a-c).</p><!><p>In order to assess the possibility that pain sensitivity is associated with the expression of SCI-sensitive miRNAs, thermal thresholds were measured by the tail flick-latency test, following SCI, both before and after treatment of morphine. Accordingly, there were statistically significant correlations between baseline tail-flick latency and expression of miR1, miR124, and miR129-2 regardless of time point (Pearson's r=-0.41, p<0.05, Pearson's r=-0.43, p<0.05, and Pearson's r=-0.41, p<0.05, respectively; Figure 5a-c), and as predicted these correlations were abolished following morphine administration.</p><!><p>We previously observed a significant decrease in miR1 expression following spinal cord contusion (Strickland et al., 2011), but did not find a significant change in expression following SCI and administration of the tail-flick test. Given the positive correlation of miR1 expression with sensory reactivity and the capability of miR1 to inhibit angiogenesis (Stahlhut et al., 2012), added to the fact that miR1 is heavily expressed in the cardiovascular system (Fichtlscherer et al., 2011; Jakob and Landmesser, 2012), it is possible that the change in expression of miR1 at the injury site may represent a broader response of vascular and cardiac tissues to SCI. We therefore assessed miR-1 expression in heart and carotid arterial tissue. Quantitative RT-PCR indicated that there was not a significant change in miR1 expression in either of these tissues following SCI (Figure 6), suggesting that the changes in miR-1 expression previously observed (Strickland et al., 2011) are localized within the injury site rather than due to adaptive changes in the vascular system.</p><!><p>As SCI-induced miRNA is further dysregulated following morphine administration, we assessed the extent to which changes in miR21 expression were associated with alterations in known cytokine mRNA targets. MiR21 has previously been shown to regulate the interleukin-6 receptor, IL6R (Frankel et al., 2008). To determine if IL6R was additionally dysregulated by alteration of pain neuro-circuitry, the expression of mRNA transcripts for IL6R was examined by qRT-PCR in sham controls and in response to vehicle and morphine treatment following contusion injury. An ANOVA revealed a main effect of both time and treatment on IL6R mRNA (F(1,28) =4.533; p<0.05, and F(2,28) =10.881; p<0.01, respectively), and post hoc LSD t-tests indicated that IL6R expression is increased following spinal cord trauma, irrespective of morphine treatment (pContused Vehicle vs. Sham <0.0001 and pContused Morphine vs. Sham <0.001; Figure 7a, b). In addition, there were significant correlations between IL6R mRNA expression and miR21, miR124, miR129-2, and miR146a (Pearson's r=0.63, p<0.001, Pearson's r=-0.38, p<0.05, Pearson's r=-0.59, p<0.001, and Pearson's r=0.57, p<0.001, respectively). Partial correlations controlling for the effects of time also indicated that the expression of IL6R mRNA was significantly correlated with expression of both miR124 and miR146a (Pearson's r=-0.54, p<0.05, and Pearson's r=0.47, p<0.05, respectively). Furthermore, IL6R expression was predictive of nociceptive sensitivity, as there was a statistically significant correlation between baseline tail-flick latency and IL6R expression across both time points (Pearson's r=0.59, p<0.01; Figure 8a), indicating that pain sensitivity decreased as IL6R expression increased, and this correlation was abolished following morphine administration. Finally, there was a statistically significant correlation between BBB scores at 24 hours post-SCI and IL6R expression, indicating that variation in initial injury severity is predictive of variation in IL6R expression (Pearson's r=-0.66, p<0.01; Figure 8b), but there was not a significant correlation between day 13 BBB scores and IL6R mRNA expression (p=0.49).</p><!><p>Morphine treatment plays an important role in pain mitigation in both the acute and chronic phase of spinal cord trauma. However, its deleterious effects on functional recovery and exacerbation of chronic pain systems, through activation of innate immune responses, provide a substantial challenge towards clinical efforts to minimize the short and long term effects of SCI (Dworkin et al., 2003; Hook et al., 2007; Hook et al., 2009; Hook et al., 2011; McCarberg, 2004; Scholz and Woolf, 2007; Watkins et al., 2005; Watkins et al., 2007). Given their dysregulation following SCI and ability to coordinate the expression of large gene networks, it is likely that miRNAs are involved in SCI-induced remodeling of nociceptive circuitry, and consequently, the effects of opiates like morphine on this circuitry. Therapeutic manipulation of these miRNAs could attenuate the maladaptive effects of acute morphine administration on both the acute and chronic phases of SCI by suppressing activation of inflammation and inhibiting the synaptic remodeling that increases both neuropathic pain and morphine tolerance.</p><p>In the current study, we investigated the regulation of SCI-sensitive miRNAs following morphine treatment. Specifically, we replicated our previous observation that miR129-2 is decreased following spinal cord contusion, while miR21 and miR146a were induced, irrespective of morphine treatment (Strickland et al., 2011). Interestingly, in this study, neither miR1 nor miR124 were significantly altered in SCI animals. One possible explanation that merits further investigation is that the measurement of pain sensitivity itself exposed SCI animals to nociceptive stimuli, which may have normalized miR1 and miR124 expression. This possibility is supported by reports showing that peripheral inflammation, which is also known to trigger nociception, results in persistent elevation in miR1 in the dorsal horn of the spinal cord (Kusuda et al., 2011). Additionally, both miR1 and miR124 are predicted to play a central role in the late phase gene repression associated with long-term potentiation-dependent neuronal plasticity (Ryan et al., 2012), and therefore, the normalization of these miRNAs may be important for activation of neural circuitry associated with pain as well.</p><p>We found that morphine exposure significantly induced miR21 expression and resulted in a near-significant increase in miR146a expression 15 days after SCI. This is the first evidence for the potential involvement of miRNAs in a post-trauma response to morphine. Though it remains to be determined, it is intriguing to hypothesize that miRNA-regulated gene networks may mediate morphine's effects on plasticity in the spinal cord following SCI. In addition, changes in expression of miR21 were significantly explained by the variance in expression of miR1, miR124, and miR146a, suggesting that these miRNAs may be co-regulated. Moreover, while the expression of miR21 and miR146a was significantly correlated with locomotor behavioral performance following SCI, their expression did not predict pain sensitivity. Consequently, the induction of these miRNAs by morphine is unlikely to be related to the analgesic actions of morphine, but rather, to its pro-inflammatory actions. Several lines of evidence support this hypothesis. Firstly, the expression of both miR146a and miR21 is induced by pro-inflammatory cytokines (Loffler et al., 2007; Nakasa et al., 2008; O'Connell et al., 2007; Taganov et al., 2006; Tili et al., 2007), and reciprocally, both miRNAs target mRNA transcripts of pro-inflammatory cytokines or their receptors (Frankel et al., 2008; Li et al., 2010a; Nahid et al., 2009). Secondly, both miR21 and miR146a play an important role in the innate immune response (Li et al., 2010a; Li et al., 2010b; Moschos et al., 2007; Nahid et al., 2009; Taganov et al., 2006), because increased expression of both miR21 and miR146a negatively regulates innate immune signaling. Finally, miR21 and miR146a have been found to be highly expressed in activated astrocytes, and their overexpression results in attenuation of astrocytic hypertrophy and suppression of the astrocyte-mediated inflammation response, respectively (Bhalala et al., 2012; Iyer et al., 2012). As a result, it is more likely that these miRNAs serve as a component of morphine-induced inflammation rather than being involved in the morphine-triggered signaling cascade driving analgesia.</p><p>Interestingly, we observed that although the interleukin receptor IL6R was dysregulated following SCI, it was not additionally dysregulated by morphine administration. Moreover, contrary to our prediction, the transcript for IL6R, the identified miR21 target, was persistently up-regulated following SCI, suggesting a dissociation between miRNA-target gene networks following SCI, as has been observed in other models of neural damage (Pappalardo et al., 2013). In silico analyses (MirWalk, (Dweep et al., 2011)) indicate that aside from miR21, other SCI miRNAs including miR124, miR129-2 and miR146a are predicted to to target IL6R. While changes in expression of IL6R were significantly explained by the variance in all four SCI-sensitive miRNAs, the expression of miR124 and miR129-2 were negatively correlated with IL6R expression. It is therefore likely that while IL6R regulation involves a network of miRNAs, miR124 and miR129-2 may serve a more conventional role as negative regulators of gene expression.</p><p>We also observed that variations in IL6R mRNA expression explained a statistically significantly portion of the variance in initial injury severity, suggesting that decreased expression of IL6R, and presumably lower levels of inflammation, was associated with a less severe initial injury. Indeed, this relationship is likely to be causal in nature, because application of an anti-IL6R antibody immediately after injury has been shown to significantly decrease tissue damage, increase axonal regeneration, and improve functional recovery (Mukaino et al., 2010). Finally, since variability in IL6R mRNA expression also explained a statistically significant proportion of the variability in pain sensitivity (58.5%), it is possible that IL6R also plays a significant role in remodeling inflammation-driven pain neuro-circuitry following SCI and opiate exposure. While this initial analysis focused specifically on IL6R, in silico analyses (DIANA-mirPath/microT-CDS, (Vlachos et al., 2012)) predict that 3 out of 4 SCI miRNAs, miR21, miR-129-2 and miR-146a, are likely to collectively target cytokine-cytokine receptor pathways (KEGG pathway hsa04060, FDR-corrected p<0.02) that mediate inflammation and pain-sensitivity. It will be important to determine the extent to which inflammatory pathways are generally susceptible to miRNA regulation following SCI.</p><p>We were able to reaffirm our previous finding that expression changes of both miR129-2 and miR146a were significantly explained by the variance of initial injury severity (Strickland et al., 2011). In addition, we found that fluctuations in miR21 expression were significantly explained by the variance in BBB scores 24 hours post-SCI, and that changes in miR129-2, miR146a, and miR21 were also significantly explained by the variance in functional recovery at 13 days post-SCI. Similar to miR146a, miR21 is a negative regulator of inflammation, but it also has anti-apoptotic activity (Roy and Sen, 2011; Sathyan et al., 2007). As such, its up-regulated expression may result in increased neuronal survival following SCI by inhibiting inflammation and cell death, factors which are likely to result in better locomotor function and higher initial BBB scores. One therapeutically relevant prediction from these data that merits further investigation is that application of miR21 or miR146a mimetics prior to or concurrent with morphine administration in the acute phase of SCI may attenuate opioid mediated inflammation, and could also reduce apoptosis (protecting the potential for recovery) in this early phase of injury.</p><p>Finally, we observed that miRNA expression at the injury site explained a significant proportion of the sensitivity to nociceptive stimuli, i.e., the response of sensory-motor circuitry distal to the injury site. Not surprisingly, these correlations between pain sensitivity and miRNA expression were abrogated by morphine administration, due to the analgesic effects of morphine. While these correlations do not necessarily imply causality, they do advance the possibility that there is a degree of molecular integration between the injury site and more caudal, but anatomically intact neuro-circuitry. A mechanism for cross-talk between miRNAs at the injury site and caudal neuro-circuitry is unclear at this time, but the possibility that these interactions serve an adaptive function is intriguing, and needs further investigation.</p><!><p>Our evidence shows that acute morphine administration results in a very specific and limited profile of changes in SCI-sensitive miRNA expression during the chronic phase of injury. Interestingly, expression changes are confined to miRNAs, like miR21, that are both induced by and inhibitors of pro-inflammatory cytokines, suggesting that morphine may potentially interfere with miRNA-mediated negative feedback pathways. Like miR21, IL6R mRNA was persistently up-regulated after SCI, and its expression negatively correlated with initial injury severity, suggesting that IL6R and miR21 may be a coordinately regulated, adaptive response to SCI. Furthermore, miR1, miR124, and miR129-2 expression at the injury site appears to play a substantial role in the regulation of distal nociceptive signaling pathways. Collectively, these data suggest that SCI-sensitive miRNAs constitute an important component of emerging pain-sensitivity and inflammation, and consequently, these miRNAs could provide therapeutic targets following SCI.</p>

PubMed Author Manuscript

Neurons expressing the aryl hydrocarbon receptor in the locus coeruleus and island of Calleja major are novel targets of dioxin in the mouse brain

The aryl hydrocarbon receptor (AhR) acts as a receptor that responds to ligands, including dioxin. The AhR–ligand complex translocates from the cytoplasm into the nucleus to induce gene expression. Because dioxin exposure impairs cognitive and neurobehavioral functions, AhR-expressing neurons need to be identified for elucidation of the dioxin neurotoxicity mechanism. Immunohistochemistry was performed to detect AhR-expressing neurons in the mouse brain and confirm the specificity of the anti-AhR antibody using Ahr−/− mice. Intracellular distribution of AhR and expression level of AhR-target genes, Cyp1a1, Cyp1b1, and Ahr repressor (Ahrr), were analyzed by immunohistochemistry and quantitative RT-PCR, respectively, using mice exposed to 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD). The mouse brains were shown to harbor AhR in neurons of the locus coeruleus (LC) and island of Calleja major (ICjM) during developmental period in Ahr+/+ mice but not in Ahr−/− mice. A significant increase in nuclear AhR of ICjM neurons but not LC neurons was found in 14-day-old mice compared to 5- and 7-day-old mice. AhR was significantly translocated into the nucleus in LC and ICjM neurons of TCDD-exposed adult mice. Additionally, the expression levels of Cyp1a1, Cyp1b1, and Ahrr genes in the brain, a surrogate of TCDD in the tissue, were significantly increased by dioxin exposure, suggesting that dioxin-activated AhR induces gene expression in LC and ICjM neurons. This histochemical study shows the ligand-induced nuclear translocation of AhR at the single-neuron level in vivo. Thus, the neurotoxicological significance of the dioxin-activated AhR in the LC and ICjM warrants further studies.Supplementary InformationThe online version contains supplementary material available at 10.1007/s00418-021-01990-1.

neurons_expressing_the_aryl_hydrocarbon_receptor_in_the_locus_coeruleus_and_island_of_calleja_major_

Introduction<!>Animals<!>RT-PCR<!>Western blotting<!>Chemical treatment<!>Quantitative RT-PCR<!>Immunohistochemistry<!>Intracellular localization of AhR<!>Statistical analysis<!><!>AhR localization in neurons of the developing brain<!><!>Discussion<!>

<p>The aryl hydrocarbon receptor (AhR), a ligand-activated transcription factor, exists in a wide range of animal species, including humans and rodents (Hahn 2002). Ligand-bound AhR translocates from the cytoplasm into the nucleus and enhances the expression of AhR-target genes, such as Cyp1a1 and Cyp1b1, whereas these are not activated in Ahr−/− mice (Mimura and Fujii-Kuriyama 2003). To date, evidence has accumulated to indicate the presence of endogenous and exogenous substances that act as AhR ligands (Barroso et al. 2021). An intake of indole-3-carbinol, a dietary AhR ligand, induces expression of Cyp1a1 mRNA in the small intestine of the mouse (Li et al. 2011). Thus, intracellular localization of AhR can be directly linked to its transcriptional function.</p><p>Orthologues of the mammalian Ahr gene regulate neuronal growth in Caenorhabditis elegans and Drosophila (Huang et al. 2004; Kim et al. 2006; Qin and Powell-Coffman 2004; Smith et al. 2013). In rodents, Ahr transcripts are detected in various brain regions, including the cerebral cortex, cerebellum, hippocampus, and olfactory bulb (Kimura and Tohyama 2017; Petersen et al. 2000). In particular, Ahr−/− mice show learning and memory impairments possibly due to atypical proliferation and morphology of hippocampal neurons (de la Parra et al. 2018; Latchney et al. 2013), implying that ligand-activated AhR is involved in the regulation of neuronal growth and brain function.</p><p>Exposure to dioxin, an exogenous AhR ligand, causes disease conditions, such as cleft palate and hydronephrosis, in Ahr+/+ mice but not Ahr−/− mice (Mimura et al. 1997), showing that AhR is required for induction of dioxin toxicity. Disruption of neuronal migration and neurite elongation is observed in neurons expressing AhR constitutively in the mouse brain (Kimura et al. 2017), suggesting that AhR overactivation impairs neuronal growth and neural circuit structure. Indeed, perinatal dioxin exposure enhances AhR-target gene expression and alters neuromorphology in the mouse brain (Kimura et al. 2016, 2015). Furthermore, dioxin exposure adversely affects a variety of cognitive and neurobehavioral functions in humans (Nishijo et al. 2014; Patandin et al. 1999; Rogan et al. 1988) and rodents (Endo et al. 2012; Haijima et al. 2010; Kakeyama et al. 2014; Kimura et al. 2020; Kimura and Tohyama 2018). These results enabled us to speculate that brain neurons having a greater amount of AhR are strongly associated with dioxin neurotoxicity.</p><p>To understand the mechanism of dioxin neurotoxicity, we utilized histological experiments for the identification of AhR-rich neurons and analysis in intracellular AhR dynamics at the single-neuron level. We examined the expression of Ahr transcript and AhR protein in the mouse brain, identified AhR-expressing neurons immunohistochemically, and evaluated the nuclear translocation of AhR in dioxin-exposed mice.</p><!><p>The experimental protocols were approved by the Animal Care and Use Committee of the University of Tokyo and that of the National Institute for Environmental Studies. Pregnant female and adult male C57BL/6J mice were purchased from CLEA Japan (Tokyo, Japan). Ahr−/− mice with a B6.129S-Ahr < tm1Yfk > mouse strain (BRC01710) (Mimura et al. 1997) were provided by RIKEN BioResource Research Center (Tsukuba, Japan). Ahr+/− male and female B6.129S-Ahr < tm1Yfk > mice were bred to obtain Ahr−/− progeny. These mice were housed singly and in groups (three per cage), respectively, in an animal facility at a temperature of 22–24 °C and humidity of 40–60% on a 12:12-h light/dark cycle (lights on from 08:00 to 20:00). Laboratory rodent chow (Lab MR Stock; Nosan, Yokohama, Japan) and distilled water were provided ad libitum. Offspring were selected for transcript and protein expression analyses as described in sections of RT-PCR, western blotting, quantitative RT-PCR, and immunohistochemistry below. The number of animals used for these analyses is described in the legends to figures.</p><p>To produce and maintain the Ahr−/− mouse strain, genotyping of the Ahr gene was performed as follows: genomic DNA was extracted from tail tips by lysis in 50 mM Tris–HCl (pH 8.0), 100 mM NaCl, 20 mM ethylenediaminetetraacetic acid (EDTA), 1% sodium dodecyl sulphate, and proteinase K (Wako Pure Chemicals, Osaka, Japan) at 55 °C for 4 h. The lysate was centrifuged at 17,400×g at 4 °C for 3 min. The genomic DNA in the supernatant was purified using phenol and chloroform, followed by washing with 70% ethanol. The genomic DNA (dissolved in Tris–EDTA buffer) was used as the template for PCR using the Takara LA Taq PCR kit (Takara Bio, Kusatsu, Japan) on a Veriti thermal cycler (Applied Biosystems, Foster City, CA, USA). The amplification conditions were as follows: 94 °C for 5 min, followed by 35 cycles of 94 °C for 30 s, 55 °C for 30 s, and 72 °C for 35 s. The PCR primers to amplify the genomic Ahr locus were 5′-GCCCGAGTCTCCTCTGTCG-3′/5′-CTCACGGCAGCGGAGATCT-3′ for the wild-type Ahr allele and 5′-GCCCGAGTCTCCTCTGTCG-3′/5′-CGCCGAGTTAACGCCATCAA-3′ for the Ahr-null allele. The 25-μl reaction contained 400 nM of each primer, 1× GC buffer II, 320 μM deoxynucleoside triphosphate (dNTP) mixture, and 0.5 U of LA Taq DNA polymerase. PCR products were separated by electrophoresis on agarose gels, which were stained with Midori Green Advance (Nippon Gene, Tokyo, Japan). The PCR products of the wild-type Ahr allele and Ahr-null allele were expected to be 439 and 671 bp in size, respectively.</p><!><p>Developing mice at P3 and P5 were decapitated, and several organs, including the brain, liver, lung, kidney, thymus, and spleen, were quickly removed and stored at −80 °C until RT-PCR analysis. The total RNA was isolated from each organ using an RNeasy Mini Kit (Qiagen, Tokyo, Japan). The cDNA for a given mRNA was synthesized using oligo-dT and random hexamers with a Primescript RT reagent kit (Takara Bio). Expression levels of Ahr and Gapdh transcripts were determined using a Veriti thermal cycler (Applied Biosystems) with a KOD Plus kit (Toyobo, Osaka, Japan). The amplification conditions were as follows: 95 °C for 1 min, followed by 35 cycles of 95 °C for 15 s, 55 °C for 15 s, and 68 °C for 30 s. The PCR primers for amplifying the murine Ahr and Gapdh transcripts were 5′-AGGATTTGCAAGAAGGAGAG-3′/5´-TTGGTTCGAATTTCCAGGAT-3´ and 5′-ACCCAGAAGACTGTGGATGG-3′/5′-CACATTGGGGGTAGGAACAC-3′, respectively. The 20-μl reaction solution contained 400 nM of each primer, 1× KOD Plus buffer, 200 μM dNTP mixture, 1 mM MgSO4, and 0.5 U of KOD Plus DNA polymerase. PCR products were separated by electrophoresis on agarose gels, which were stained with Midori Green Advance (Nippon Gene). The PCR products of the Ahr and Gapdh transcripts were expected to be 508 and 171 bp in size, respectively.</p><!><p>Developing mice at P3, P5, and P14 were decapitated, and several organs (brain, liver, lung, kidney, thymus, and spleen) were quickly collected and stored at −80 °C until western blotting analysis. Protein was extracted at 4 °C in an ice bath unless stated otherwise. Each type of organ was homogenized with 4 mM HEPES–NaOH buffer, pH 7.3, containing 0.32 M sucrose and 1% protease inhibitor cocktail (Sigma-Aldrich, St. Louis, MO, USA), using a Potter-type homogenizer. The homogenates were centrifuged at 1000×g at 4 °C for 10 min, and the supernatants were used for western blotting. Protein concentration in the supernatants was measured with the Quick Start Bradford Protein Assay (BioRad, Hercules, CA, USA). Proteins in the supernatants were separated on a 7.5% polyacrylamide gel and blotted onto immobilon-P transfer membranes (Millipore, Bedford, MA, USA). The proteins adsorbed to membranes were allowed to react with mouse monoclonal anti-AhR antibody (1:1000; sc-398877, Santa Cruz Biotechnology, Santa Cruz, CA, USA) in Tris-buffered saline, pH 7.4, containing 0.1% Tween-20 (TBST), overnight at 4 °C, followed by incubation in TBST containing anti-mouse IgG-horseradish peroxidase (HRP)-conjugated antibody (1:5000; 7076S, Cell Signaling Technology, Beverly, MA, USA), for 1 h at room temperature. Chemi-Lumi One (Nacalai Tesque, Kyoto, Japan) was used to visualize the protein bands, which were detected on Hyperfilm ECL (GE Healthcare Ltd., Tokyo, Japan) and developed and fixed with GBX developer and GBX fixer (Kodak, Rochester, NY, USA), respectively. Following deactivation of endogenous HRP by incubation in TBST containing 15% hydrogen peroxide for 30 min at room temperature, the membranes were immersed in TBST containing rabbit polyclonal anti-GAPDH antibody (1:5000; ab9485, Abcam, Cambridge, UK), overnight at 4 °C, followed by incubation in TBST containing anti-rabbit IgG-HRP-conjugated antibody (1:5000; 7074S, Cell Signaling Technology). Then, targeted protein bands were visualized in the same manner as described for AhR detection. The intensity of AhR and GAPDH bands was measured using ImageJ software (National Institutes of Health, Bethesda, MD, USA).</p><!><p>2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD; purity > 99.5%) was purchased from Cambridge Isotope Laboratory (Andover, MA, USA). Corn oil and n-nonane were purchased from Wako Pure Chemicals and Nacalai Tesque, respectively. Twelve-week-old male C57BL/6J mice were divided to control and TCDD groups, and they were orally administered with vehicle (corn oil containing 0.6% n-nonane) or TCDD dissolved in vehicle (20 μg/kg body weight).</p><!><p>Brain and liver tissues of 12-week-old male mice treated with vehicle or TCDD were collected quickly and stored at −80 °C until analysis. Total RNA was isolated from the brain and liver using an RNeasy Mini Kit (Qiagen). The cDNA for a given mRNA was synthesized using oligo-dT and random hexamer primers with a PrimeScript RT reagent kit (Takara). Gene expression levels were determined quantitatively using a LightCycler System (Roche Molecular Biochemicals, Indianapolis, IN, USA) with Thunderbird SYBR qPCR Mix (Toyobo). The genes and primers are summarized in Supplementary Table 1. No-template reactions were analyzed in every PCR to monitor for cross-contamination. To verify the specificity of amplification, melting curve analyses of the products were performed at the end of every PCR. The Cyp1a1, Cyp1b1, and Ahr repressor (Ahrr) mRNA expression levels were calculated using the ΔΔCt method and normalized to the 18S rRNA expression.</p><!><p>Developing and adult mice were transcardially perfused with 4% paraformaldehyde in 0.1 M phosphate-buffered saline (PBS, pH 7.4) under anesthesia with sodium pentobarbital (conducted at the University of Tokyo) or three types of mixed anesthetic agents containing medetomidine hydrochloride, midazolam, and butorphanol (at the National Institute for Environmental Studies). Brains were collected, fixed in 4% paraformaldehyde overnight, immersed in a series of 0.1 M PBS containing 5%, 15%, and 30% sucrose, frozen in Tissue-Tek O.C.T. compound (Sakura Finetek, Tokyo, Japan), and stored at −80 °C until histological sectioning. Frozen brains were sliced in the sagittal plane using a cryostat (Model 3050S; Leica Microsystems, Tokyo, Japan). Brain sections were cut at 50 μm thickness for immunofluorescence analysis.</p><p>Brain tissue sections were immunohistochemically stained for AhR, tyrosine hydroxylase (TH), dopamine β-hydroxylase (DBH), or NeuN. In brief, the tissue sections were washed in PBS containing 0.1% Triton X-100 (PBST), soaked in 0.01 M citrate buffer (pH 6.0) (Muto Pure Chemicals, Tokyo, Japan), and incubated at 90 °C (developing mouse brains) or 65 °C (adult mouse brains) in a water bath for 10 min. The sections were blocked with PBST containing 5% bovine serum albumin (A3059; Sigma-Aldrich) (blocking solution) and allowed to react with mouse monoclonal anti-AhR antibody (1:500; sc-398877, Santa Cruz Biotechnology) and rabbit polyclonal anti-TH antibody (1:1000; ab112, Abcam), anti-DBH antibody (1:1000; 22,806, Immunostar, Hudson, WI, USA), or anti-NeuN (1:1000; ab177487, Abcam) in blocking solution overnight at 4 °C. Then, the signals of AhR and TH, DBH, or NeuN were visualized with the respective secondary antibodies anti-mouse IgG AlexaFluor 488 (Life Technologies, Gaithersburg, MD, USA) and anti-rabbit IgG AlexaFluor 568 (Life Technologies) in PBST (1:1000). Furthermore, the nucleus was stained with PBST containing Hoechst 33342 (1:1000; Dojin Laboratories, Kumamoto, Japan), followed by mounting with VECTASHIELD (H-1400; Vector Laboratories, Burlingame, CA, USA) for confocal microscopy. Immunostained images were captured using an inverted Leica DMi8 microscope, equipped with the Leica TCS SP8 confocal module (Leica Microsystems). Specific objective lens (HC PL APO CS 10×/NA = 0.40 and HC PL APO CS2 20×/NA = 0.75; Leica Microsystems) and LAS X 3.1.5 software (Leica Microsystems) were used to capture images (x = 2048 pixels and y = 2048 pixels, bit depth = 8 in each RGB color).</p><!><p>Cellular morphology and immunostaining intensity of AhR- and TH-double-positive cells in the locus coeruleus (LC) and AhR- and NeuN-double-positive cells in the island of Calleja major (ICjM) were analyzed by applying the ImageJ software to the confocal microscopy images. In analyses of AhR- and TH-double-positive cells in the LC, we outlined the nucleus and soma of TH-positive cells [i.e., noradrenergic (NA) neurons] and measured their nuclear and soma sectional areas, and then calculated the nuclear area percentage by dividing the nuclear area by the soma area in each cell. In addition, we determined AhR immunostaining intensity per nucleus and soma (AhRNuc intensity and AhRSoma intensity, respectively) of TH-positive cells. After the background subtraction of AhR-stained images, the AhRNuc intensity percentage was calculated by dividing the AhRNuc intensity by the AhRSoma intensity. In order to analyze the intracellular localization of AhR, AhRNuc intensity percentage data were normalized depending on the soma size of TH-positive cells. We divided the AhRNuc intensity percentage by the nuclear area percentage to calculate the ratio in locus coeruleus-noradrenergic (LC-NA) neurons (ratioLC−NA, thereafter). In order to compare ratioLC−NA of individual mice between the control and TCDD groups, we used the distribution of ratioLC−NA values in each mouse as a surrogate parameter and analyzed the percentage of ratioLC−NA that was divided by the arbitrarily chosen value of 0.2. For each mouse, 57 to 103 cells in developing mice and 51 to 96 cells in adult mice were subjected to analyses in intracellular localization of AhR in TH-positive cells in the LC. In analyses of AhR- and NeuN- double-positive cells in the ICjM, we outlined the nucleus of NeuN-positive cells and measured nuclear sectional areas and AhRNuc intensity in each cell. To adjust the variability of luminance among images, AhRNuc intensity was normalized to the mean value of immunostained AhR intensity in the whole ICjM area. The parameter ratio in ICjM neurons (ratioICjM, thereafter) was calculated by dividing AhRNuc intensity by the nuclear area in each cell. Furthermore, we analyzed the percentage of the ratioICjM that was divided by the arbitrarily chosen value of 0.005. The numbers of cells that were used to analyze the intracellular localization of AhR in NeuN-positive cells in the ICjM ranged from 65 to 263 cells in developing mice and from 133 to 291 cells in adult mice.</p><!><p>Statistical analysis was performed using BellCurve for Excel software (Social Survey Research Information Co., Ltd., Tokyo, Japan). Protein expression, cellular morphology, immunostaining intensity, and ratio values were analyzed using Student's t-test or one-way analysis of variance (ANOVA), followed by the Tukey–Kramer post hoc test, and p-values < 0.05 were considered statistically significant. Because the mention of F- and p-values for each statistical analysis in the main text is very complicated, statistically significant differences are shown by asterisks in each figure.</p><!><p>Ahr transcript and AhR protein expression in developing mouse organs. a RT-PCR–amplified Ahr (508 bp) and Gapdh (171 bp) transcripts were detected in the brain, liver, lung, kidney, thymus, and spleen of mice at P3 and P5 (n = 4 mice at each stage). NC negative control. b Representative images showing AhR and GAPDH proteins detected by western blotting in the brain, liver, lung, kidney, thymus, and spleen of male mice at P3. c Quantitative analysis of AhR band intensity in six organs (n = 8 mice/tissue). d Images showing AhR and GAPDH proteins detected by western blotting in the brains of male and female mice at P5. e Quantitative analysis of AhR band intensity in male and female mouse brains (n = 4 mice/group) at P5, showing no significant difference in AhR protein amounts between sexes. AhR band intensity was normalized to GAPDH band intensity. f Images showing AhR and GAPDH proteins detected by western blotting in the brain, liver, and lung of Ahr+/+ and Ahr−/− mice at P5 (n = 1 in each genotype). No AhR protein was detected in these tissues of Ahr−/− mice, demonstrating the specificity of the antibody. Circles represent individual mouse data. Values are shown as the mean ± SD. Asterisks (** and ***) denote statistical significance at p < 0.01 and 0.001, respectively, by one-way ANOVA with the Tukey–Kramer post hoc test</p><p>AhR expression in mouse LC-NA neurons at P5, P7, and P14. a Diagram displaying the location of the LC (left) and the metabolic pathway of noradrenaline synthesis in LC-NA neurons (right). b, c Brain tissue sections immunostained with anti-TH or DBH and anti-AhR antibodies. AhR was detected in TH-expressing neurons in the LC at P5, P7, and P14 (n = 3 mice at each stage) (b). Additionally, AhR was also found in neurons expressing DBH in the LC at P14 (n = 3 mice) (c). Scale bar = 100 μm. d Representative images showing the LC of Ahr+/+ and Ahr−/− mice at P14 (n = 3 mice in each genotype), demonstrating antibody specificity. Arrowheads mark TH-negative cells with nonspecific anti-AhR antibody staining. Scale bar = 100 μm</p><p>Intracellular localization of AhR in mouse LC-NA neurons at P5, P7, and P14. a Immunostained LC-NA neurons double-positive for TH and AhR. The cellular boundary of TH-stained cells was clearly observed, which enabled the measurement of the nuclear and soma areas of individual LC-NA neurons. Scale bar = 10 μm. b–f Quantitative analyses of AhR-expressing LC-NA neurons. The nuclear area was significantly larger at P5 than at P7 and P14 (b), whereas the soma area was significantly larger at P14 compared to P5 and P7 (c), indicating that marked changes in cellular morphology occur between P5 and P14. The nuclear area percentage that represents the nuclear area normalized to the soma area in each neuron was significantly decreased during the period from P5 to P14 (d). The AhR immunostaining intensity per nucleus (AhRNuc intensity) was normalized to that of the soma in each neuron (AhRNuc intensity percentage). The AhRNuc intensity percentage at P5 was significantly higher than that at P7 and P14 (e). To normalize the developmental-stage-related changes in cellular morphology, the ratio calculated by dividing AhRNuc intensity percentage by nuclear area percentage (ratioLC−NA) served as an index to evaluate the nuclear AhR. No significant difference in ratioLC−NA was found across developmental periods (f). Values are shown as the mean ± SD. Circles represent individual cell data (215, 228, and 200 cells from 3 mice each at P5, P7, and P14, respectively). Asterisks (** and ***) denote statistical significance at p < 0.01 and 0.001, respectively, by one-way ANOVA with the Tukey–Kramer post hoc test</p><p>AhR expression in mouse ICjM neurons at P5, P7, and P14. a Diagram (left) illustrating the location of the ICjM anterior to the commissural fiber (cf). A representative low-magnification image (right) of Hoechst-stained brain tissue at P5 is shown. Scale bar = 200 μm. b Representative images of immunostained sections of the ICjM at P5, P7, and P14 show distinct AhR signals in cells expressing NeuN, a marker of mature neurons (n = 3 mice at each stage). Scale bar = 100 μm. c Representative images showing the ICjM of Ahr+/+ and Ahr−/− mice at P14 (n = 3 mice in each genotype). Immunostained AhR signals were not observed in Ahr−/− mice, demonstrating the specificity of the antibody. Scale bar = 100 μm</p><p>Intracellular localization of AhR in mouse ICjM neurons at P5, P7, and P14. a Immunostained ICjM neurons double-positive for NeuN and AhR. Because the cellular boundary could not be distinctly visualized, the soma areas of each neuron were not characterized. Scale bar = 10 μm. b–d Quantitative analyses of AhR-expressing ICjM neurons. The nuclear area of each neuron gradually increased with brain development (b). The AhR immunostaining intensity per nucleus (AhRNuc intensity) was significantly increased during the period from P5 to P14 (c). The ratioICjM representing AhRNuc intensity normalized to the nuclear area was significantly higher at P14 than those at P5 and P7, and no significant difference was observed between P5 and P7 (d). Values are shown as the mean ± SD. Circles represent individual cell data (711, 528, and 332 cells from 3 mice at each stage, P5, P7, and P14, respectively). Asterisks (***) denote statistical significance at p < 0.001 by one-way ANOVA with the Tukey–Kramer post hoc test</p><!><p>To study AhR dynamics in other brain regions, we analyzed the AhR expression in the cerebral cortex, cerebellum, hippocampus, and olfactory bulb at P14. Although AhR in these regions was observed by western blotting, distinct immunohistochemical signals were not detected (Supplementary Fig. 1a–c).</p><!><p>Expression level of AhR-target genes in the brain and liver of TCDD-exposed mice. a Schematic of the TCDD experiment. Twelve-week-old mice were orally exposed to TCDD (20 μg/kg body weight), and their brains and livers were sampled 24 h later. The liver was used as a positive control to confirm increased expression of the AhR-target genes Cyp1a1, Cyp1b1, and Ahrr, which are drastically enhanced by TCDD exposure. b–d Body weight b and organ size of the brain c and liver d in the control and TCDD groups. Upon TCDD exposure, no changes in these weights were observed between the two groups. e, f Expression levels of Cyp1a1, Cyp1b1, and Ahrr mRNAs in the brain e and liver f. In the TCDD group, the expression of the three AhR-target genes in the brain and liver was significantly increased. Circles represent individual mouse data (n = 4 mice/group). Values are shown as the mean ± SD. Asterisks (** and ***) denote statistical significance at p < 0.01 and 0.001, respectively, by Student's t-test</p><p>Nuclear translocation of AhR in LC-NA neurons of TCDD-exposed mice. a Representative images showing fluorescent signal (FS) of TH and AhR in LC-NA neurons in the control and TCDD groups. Heatmap (HM) images represent the relative intensity of immunostained AhR signals. The fluorescent intensity scale for HM images is shown below the images. Scale bar = 10 μm. b Scatter plot of nuclear area percentage (x-axis) and AhRNuc intensity percentages (y-axis) in LC-NA neurons in the control (black) and TCDD (red) groups. The AhRNuc intensity percentage of the TCDD group was significantly higher than that of the control group without a significant change in the nuclear area percentage. c The ratioLC−NA was significantly higher in the TCDD group than in the control group. Values are shown as the mean ± SD. Circles represent individual cell data (382 and 398 cells from 6 mice each in the control and TCDD groups, respectively). Asterisks (***) denote statistical significance at p < 0.001 by Student's t-test</p><p>Nuclear translocation of AhR in ICjM neurons of TCDD-exposed mice. a Representative images showing fluorescent signal (FS) of Hoechst and AhR in ICjM neurons in the control and TCDD groups. Heatmap (HM) images represent the relative intensity of immunostained AhR signals. The fluorescent intensity scale for HM images is shown below the images. Scale bar = 10 μm. b Scatter plot of nuclear area (x-axis) and AhRNuc intensity (y-axis) in ICjM neurons in the control (black) and TCDD (red) groups. The AhRNuc intensity of the TCDD group was significantly higher than that of the control group without a significant change in the nuclear area. c The ratioICjM was significantly higher in the TCDD group than in the control group. Values are shown as the mean ± SD. Circles represent individual cell data (1423 and 1398 cells from six mice each in the control and TCDD groups, respectively). Asterisks (***) denote statistical significance at p < 0.001 by Student's t-test</p><!><p>The AhR–ligand complex translocates into the cellular nucleus to enhance the expression of AhR-target genes, which in turn may induce developmental and physiological responses or toxicities. Thus, the types of brain neurons expressing AhR must be characterized for understanding the impacts of AhR ligands on the nervous system. In the present study, we immunohistochemically identified two neuronal populations having AhR in the mouse LC and ICjM and quantitatively analyzed nuclear translocation of AhR at the single-neuron level. Although AhR expression in neurons and glias has been reported in humans and rodents (Bravo-Ferrer et al. 2019; de la Parra et al. 2018; Rothhammer et al. 2018, 2016; Shackleford et al. 2018), the specificity of the AhR antibodies in these studies has not been verified by immunohistochemistry using AhR-null tissues. After confirming the specificity of the AhR antibody using Ahr−/− mouse brains, we unequivocally demonstrated the presence of AhR in LC-NA and ICjM neurons (Figs. 2d, 4c). Furthermore, a significant increase in nuclear AhR was found in these neurons of TCDD-exposed mice (Figs. 7, 8 and Supplementary Figs. 2, 3), which is consistent with gene expression changes (Fig. 6e). Thus, our immunohistochemical analysis is considered to be robust. We describe three implications of the presence of neuronal AhR in the LC and ICjM below.</p><p>First, loss-of-function and gain-of-function experiments reveal that AhR regulates neurogenesis, neuronal migration, and neurite elongation in C. elegans (Huang et al. 2004; Qin and Powell-Coffman 2004; Smith et al. 2013), Drosophila (Kim et al. 2006), and mice (de la Parra et al. 2018; Kimura et al. 2017; Latchney et al. 2013). Thus, it is plausible that AhR is involved in the growth of LC-NA and ICjM neurons. Although it has been reported that neurogenesis of ICjM occurs in rat fetuses (Bayer 1985) and that ICjM structure is formed in mice at P2 (Hsieh and Puche 2013), the molecular mechanisms regulating the growth of ICjM neurons remain largely unclear. Since, in the present study, altered AhR dynamics was found in the nuclei of ICjM neurons during development (Fig. 5), further studies on the role of AhR in ICjM neurons could help understand the mechanism of the ICjM formation. On the other hand, no distinct AhR immunostaining was observed in the cerebral cortex, cerebellum, hippocampus, and olfactory bulb, where AhR protein was detected by western blotting (Supplementary Fig. 1). One plausible explanation would be that AhR abundance in LC-NA and ICjM neurons is greater than that in other neurons, suggesting the possibility of AhR as a marker for specified neuronal populations.</p><p>Second, the intracellular dynamics of AhR is essential for understanding how AhR ligands impact cellular activities. AhR ligands contained in diet and gut microbiota metabolites have been reported to regulate various physiological systems. Treatment with indole-3-carbinol enhances the immune capacity of Ahr+/+ mice but not Ahr−/− mice (Kiss et al. 2011; Li et al. 2011). Furthermore, indole-3-aldehyde produced by lactobacilli has AhR agonistic activity and protects against candidiasis and colitis in an AhR-dependent manner (Zelante et al. 2013). In particular, a large body of evidence suggests that gut microbiota influences neuronal activities and brain functions (Mayer et al. 2015). However, the molecular mechanisms linking gut microbiota with brain neurons are not fully understood, although several pathways via which microbiota affect brain function have been proposed (Cryan and Dinan 2012). Our present study provides experimental evidence that oral exposure to TCDD significantly increases nuclear translocation of AhR in LC-NA and ICjM neurons (Figs. 7, 8 and Supplementary Figs. 2, 3), suggesting that other AhR ligands might also influence the AhR dynamics and signaling activation in these neurons.</p><p>Third, cognitive impairments and neurobehavioral abnormalities have been reported in humans and laboratory animals perinatally exposed to dioxin (Endo et al. 2012; Haijima et al. 2010; Kakeyama et al. 2014; Kimura et al. 2020; Kimura and Tohyama 2018; Nishijo et al. 2014; Patandin et al. 1999; Rogan et al. 1988). However, the brain regions and neuronal populations responsible for those neurotoxic effects remain still uncharacterized. Our immunohistochemical analysis demonstrated that LC-NA and ICjM neurons are targets of dioxin (Figs. 7, 8 and Supplementary Figs. 2, 3). LC-NA neurons elongate their axons into a wide range of brain regions and regulate a variety of brain functions (Waterhouse and Navarra 2019). For example, in rodents, the LC is involved in sleep/awake states (Aston-Jones and Bloom 1981), stress response (Ziegler et al. 1999), behavioral flexibility (McGaughy et al. 2008), fear memory (Soya et al. 2013), and everyday memory (Takeuchi et al. 2016) as well as infant attachment learning (Moriceau et al. 2009). Remarkably, mouse offspring born to dams exposed to TCDD show abnormalities related to attachment behavior in infancy (Kimura and Tohyama 2018) and executive function and emotion in adulthood (Endo et al. 2012; Haijima et al. 2010), suggesting that these phenotypes could be caused by impaired growth of LC-NA neurons. Additionally, a change in the number of midbrain dopaminergic neurons in TCDD-exposed mice (Tanida et al. 2009) has been reported. ICjM neurons receive axonal projections from midbrain dopaminergic neurons (Fallon et al. 1978) and express dopamine receptors (Mengod et al. 1991; Sokoloff et al. 1990). The dopaminergic circuit plays a role in reward-related behavior (Ikemoto 2007), and rats afflicted with drug addiction show an increase in ICjM neuronal activity (Prast et al. 2014; Singh et al. 2006), suggesting an involvement of the ICjM in reward-related behavior. Thus, it is plausible that dioxin adversely affects the growth of both ICjM and dopaminergic neurons, leading to an abnormality in reward-related behavior. Collectively, our histological results highlight the need for studies focusing on LC-NA and ICjM neurons to understand the molecular mechanisms of dioxin neurotoxicity.</p><!><p>Supplementary file1 (PDF 494 KB)</p><p>Publisher's Note</p><p>Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.</p><p>Change history</p><p>6/18/2021</p><p>A Correction to this paper has been published: 10.1007/s00418-021-01997-8</p>

PubMed Open Access

The relationship between vitamin D and estimated glomerular filtration rate and urine microalbumin/creatinine ratio in Korean adults

The present study was conducted to assess the association between 25-hydroxyvitamin D [25(OH)D], estimated glomerular filtration rate (eGFR) and urine microalbumin/creatinine ratio (uACR) in Korean adults. Data on 4,948 adults aged ≥20 years from the Korean National Health and Nutrition Examination Survey V-3 (2012) were analyzed. After adjusting for the related variables (except age), the odds ratios (ORs) of vitamin D deficiency with the normal group as a reference were significantly higher in the decreased eGFR plus elevated uACR group [3.089 (95% CI, 1.722–5.544)], but not in the elevated uACR [1.247 (95% CI, 0.986–1.577)] and decreased eGFR group [1.303 (95% CI, 0.789–2.152)]. However, when further adjusting for age, the ORs of vitamin D deficiency with the normal group as a reference were significantly higher in the elevated uACR group [1.312 (95% CI, 1.035–1.662)], decreased eGFR group [1.761 (95% CI, 1.062–2.919)] and the decreased eGFR plus elevated uACR group [3.549 (95% CI, 1.975–6.365)]. In conclusion, vitamin D deficiency was positively associated with the elevated uACR and decreased eGFR. In addition, vitamin D level decreased greatly when decreased eGFR and elevated uACR appeared simultaneously.

the_relationship_between_vitamin_d_and_estimated_glomerular_filtration_rate_and_urine_microalbumin/c

Introduction<!>Study subjects<!>General characteristics and blood chemistry<!>Glomerular filtration rate and urine microalbumin/creatinine ratio and serum 25(OH)D<!>Statistical analysis<!>Clinical characteristics of the research subjects<!>Clinical characteristics of the subjects according to decreased eGFR, elevated uACR and decreased GFR plus elevated uACR<!>Comparison of 25(OH)D levels and odds ratios of vitamin D deficiency according to decreased eGFR, elevated uACR and decreased GFR plus elevated uACR<!>Discussion<!>Conclusion<!>Conflict of Interest<!>

<p>Chronic kidney disease (CKD) is a global public health problem with 20 million adult Americans currently living with CKD in various stages of CKD: >400,000 individuals with end-stage kidney disease and >300,000 individuals requiring maintenance hemodialysis.(1–3) CKD is defined by an estimated glomerular filtration rate (eGFR) <60 ml/min/1.73 m2; a decrease in eGFR is a risk factor for cardiovascular disease (CVD) and is correlated with cardiovascular mortality and morbidity in high-risk groups.(4,5) Albuminuria is a well-known predictor of CKD progression and is considered to be an early sign of glomerular damage used as a risk factor for end-stage renal disease in people with diabetes mellitus.(6,7)</p><p>Vitamin D from the diet or dermal synthesis from sunlight is biologically inactive [25-hydroxyvitamin D, 25(OH)D], which is metabolized to the biologically active 1,25 dihydroxyvitamin D [1,25(OH)2D] through enzymatic conversion in the kidney.(8,9) 25(OH)D usually functions as storage due to its relatively long half-life of 2–3 weeks, and the total vitamin D status in the human body is generally estimated through measurements of serum 25(OH)D.(10) Vitamin D is known to be involved in calcium and phosphate absorption in the intestines, and maintains sufficient concentrations of circulating calcium and phosphate levels and normal mineralization of bone by providing the minerals to bone-forming sites.(11,12)</p><p>Recently, vitamin D has receiving an attention on concerning its effect on CKD and CVD.(13,14) It is important to monitor eGFR and the urine microalbumin/creatinine ratio (uACR) in patients with CKD and progressive CVD. In particular, when a decrease in eGFR is combined with an increase in uACR, CVD mortality rates in patients with CKD increase greatly.(15) The Republic of Korea has recently become known as a country that has a severe vitamin D deficiency problem,(16) and the burden of CKD and CVD are also increasing. Therefore, our objective in this study was to assess the association between vitamin D and eGFR and uACR in Korean adults using data from the fifth Korean National Health and Nutrition Examination Survey (KNHANES V-3; 2012) to be representative of the Korean population.</p><!><p>This study was performed using data from the Korean National Health and Nutrition Examination Survey (KNHANES V-3). KNHANES V-3 were each conducted for 1 years (2012), using a rolling sampling survey that involved a complex, stratified, multistage, probability cluster survey of a representative sample of the non-institutionalized civilian population in South Korea. The survey was composed of three parts: a health interview survey, a health examination survey and a nutrition survey. Each survey was conducted by specially trained interviewers. The interviewers were not provided with any prior information regarding specific participants before conducting the interviews. Participants provided written informed consent to participate in this survey, and we received the data in anonymized form. In the KNHANES V-3 (2012), 8,958 individuals over age 1 were sampled for the survey. Among them, of the 6,665 subjects who participated in the KNHANES V-3, we limited the analyses to adults aged ≥20 years. We excluded participants 1,717 subjects whose data were missing for important analytic variables, such as serum 25(OH)D, urine microalbumin and creatinine level, or various blood chemistry tests; pregnant women; and a high uACR (uACR; ≥3,000 mg/g) indicative of nephrotic-range albuminuria because a previous study found an association between altered vitamin D metabolism and nephrotic syndrome.(17) Finally, 4,948 subjects were included in the statistical analysis. The KNHANES V-3 study has been conducted according to the principles expressed in the Declaration of Helsinki. (Institutional Review Board No, 2010-02CON-21-C). All participants in the survey signed an informed written consent form. Further information can be found in "The KNHANES V-3 (2012) Sample", which is available on the KNHANES website. The official website of KNHANES (http://knhanes.cdc.go.kr) is currently operating an English-language information homepage. The data of the respective year are available to everyone at the free of charge. If the applicant enters simple subscription process and his/her email address in the official website of KNHANES, the data of the respective year can download to free of charge. If additional information is required, the readers can contact the department responsible for data (Su Yeon Park, [email protected]).</p><!><p>Research subjects were classified by sex (men or women), smoking (non-smoker or ex-smoker or current smoker), alcohol drinking (yes or no) and regular exercise (yes or no). In the smoking category, participants who smoked more than one cigarette a day, those who had previously smoked but do not presently smoke, and those who never smoked were classified into the current smoker, ex-smoker, and non-smoker groups, respectively. Alcohol drinking was indicated as "yes" for participants who had consumed at least one glass of alcohol every month over the last year. Regular exercise was indicated as "yes" for participants who had exercised on a regular basis regardless of indoor or outdoor exercise. (Regular exercises was defined as 30 min at a time and 5 times/weeks in the case of moderate exercise, such as swimming slowly, doubles tennis, volleyball, badminton, table tennis and carrying light objects; and for 20 min at a time and 3 times/weeks in the case of vigorous exercise, such as running, climbing, cycling fast, swimming fast, football, basketball, jump rope, squash, singles tennis and carrying heavy objects). Anthropometric measurements included measurement of body mass index (BMI) and waist measurement (WM), as well as final measurements of systolic blood pressure (SBP) and diastolic blood pressure (DBP). Blood chemistries included measurements of total cholesterol (TC), high density lipoprotein cholesterol (HDL-C), triglycerides (TGs), fasting blood glucose (FBG), blood urea nitrogen (BUN), serum creatinine (Crea), urine microalbumin, urine creatinine and 25-hydroxyvitamin D.</p><!><p>Glomerular filtration rate (GFR) was estimated from the simplified equation developed using MDRD data: eGFR = 186.3 × (serum creatinine in mg/dl)−1.154 × age−0.203 × (0.742 for women) × (1.212 if African American).(18) Decreased eGFR was classified as eGFR <60 ml/min/1.73 m2. Urine microalbumin was measured with a turbidimetric assay (Albumin; Roche, Germany) using a Hitachi Automatic Analyzer 7600 (Hitachi, Japan). The urine creatinine was measured with a colorimetric assay (CREA; Roche, Indianapolis, IN) using a Hitachi Automatic Analyzer 7600. Elevated uACR was classified as uACR ≥30 mg/g. Serum 25(OH)D levels were measured with a radioimmunoassay (25-hydroxy-vitamin D 125I RIA Kit; DiaSorin, Still Water, MN) using a 1470 Wizard Gamma Counter (Perkin Elmer, Turku, Finland). To minimize the analytical variation, serum 25(OH)D levels were analyzed by the same institute, which carried out a quality assurance program through the analysis period. Serum 25(OH)D levels were classified as either vitamin D deficiency [25(OH)D <15.0 ng/ml] and vitamin D sufficiency [25(OH)D ≥15.0 ng/ml].(19)</p><!><p>The collected data were statistically analyzed using SPSS WIN ver. 18.0 (SPSS Inc., Chicago, IL). The distributions of the participant characteristics were converted into percentages, and the successive data were presented as averages with standard deviations. The distribution and average difference in clinical characteristics according to vitamin D sufficiency and vitamin D deficiency were calculated using chi-squared and an independent t test. In the case of the analysis of covariance test (ANCOVA) for the serum 25(OH)D, the 2 models constructed were: 1) adjusted for alcohol drinking, SBP, DBP, BMI and WM; 3) further adjusted for TC, TGs, HDL-C, BUN and FBG; 2) further adjusted for age. In the case of logistic regression for odds ratio of vitamin D deficiency, the 4 models constructed were: 1) non-adjusted; 2) adjusted for alcohol drinking, SBP, DBP, BMI and WM; 3) further adjusted for TC, TGs, HDL-C, BUN and FBG; 4) further adjusted for age. The significance level for all of the statistical data was set as p<0.05.</p><!><p>The clinical characteristics of the research subjects are shown in Table 1. Serum 25(OH)D, eGFR and uACR were 20.38 ± 4.58 ng/dl, 89.78 ± 17.01 ml/min/1.73 m2 and 19.30 ± 102.13 mg/g, respectively, in subjects with vitamin D sufficiency (n = 3,034). The prevalence rates of decreased eGFR and elevated uACR were 3.5% (n = 107) and 8.4% (n = 254), respectively. Serum 25(OH)D, eGFR and uACR were 11.92 ± 2.15 ng/dl, 94.75 ± 19.14 ml/min/1.73 m2 and 21.39 ± 130.56 mg/g, respectively, in subjects with vitamin D deficiency (n = 1,914). The prevalence rates of decreased eGFR and elevated uACR were 2.7% (n = 51) and 8.9% (n = 170), respectively.</p><!><p>Clinical characteristics of the subjects according to decreased eGFR and elevated uACR are shown in Table 2. eGFR and uACR were 93.26 ± 16.44 ml/min/1.73 m2 and 5.48 ± 5.46 mg/g for the normal group, 92.69 ± 17.60 ml/min/1.73 m2 and 132.28 ± 204.03 mg/g for the elevated uACR group, 52.78 ± 6.75 ml/min/1.73 m2 and 10.38 ± 7.51 mg/g for the decreased eGFR group, and 46.85 ± 12.91 ml/min/1.73 m2 and 411.93 ± 726.23 mg/g for the decreased eGFR plus elevated uACR group, respectively. Variables with significant difference in normal, elevated uACR, decreased eGFR and decreased GFR plus elevated uACR group are current drink (p<0.001), SBP (p<0.001), DBP (p<0.001), BMI (p<0.001), WM (p<0.001), TGs (p<0.001), HDL-C (p<0.001), BUN (p<0.001), Crea (p<0.001), FBG (p<0.001) and age (p<0.001). However, gender (p = 0.373), current smoke (p = 0.249), regular exercise (p = 0.112) and TC (p = 0.807) were not significant.</p><!><p>Comparison of odds ratios (ORs) of vitamin D deficiency according to decreased eGFR, elevated uACR and decreased GFR plus elevated uACR are shown in Table 3 and 4. After adjusting for the related variables (except age), the ORs of vitamin D deficiency with the normal group as a reference were significantly higher in the decreased eGFR plus elevated uACR group [3.089 (95% CI, 1.722–5.544)], but not in the elevated uACR [1.247 (95% CI, 0.986–1.577)] and decreased eGFR group [1.303 (95% CI, 0.789–2.152)]. However, when further adjusting for age, the ORs of vitamin D deficiency with the normal group as a reference were significantly higher in the elevated uACR group [1.312 (95% CI, 1.035–1.662)], decreased eGFR group [1.761 (95% CI, 1.062–2.919)] and decreased eGFR plus elevated uACR group [3.549 (95% CI, 1.975–6.365)]. 25(OH)D levels (M ± SE) were 17.20 ± 0.08 ng/dl (95% CI, 17.04–17.36) for the normal group, 16.62 ± 0.30 ng/dl (95% CI, 16.04–17.21) for the elevated uACR group, 16.41 ± 0.60 ng/dl (95% CI, 15.23–17.59) for the decreased eGFR group and 13.82 ± 0.75 ng/dl (95% CI, 12.34–15.29) for the decreased eGFR plus elevated uACR group (p<0.001) (Table 4).</p><!><p>In the present study, an investigation into the association between vitamin D and eGFR and uACR in Korean adults was carried out using data from the KNHANES V-3 conducted in 2012. Vitamin D deficiency was positively associated with the elevated uACR, and decreased eGFR and vitamin D level decreased greatly when decreased eGFR and elevated uACR appeared simultaneously.</p><p>Vitamin D deficiency is found in various populations worldwide in high proportions and was associated with diabetes, hypertension and insulin resistance.(20,21) In particular, vitamin D deficiency patients with CKD have been associated with a higher risk of cardiovascular events and mortality and accelerate a progression of kidney disease.(22,23) Ravani et al.(23) suggested that serum 25(OH)D is an independent inverse predictor of renal disease progression and death in patients with earlier stages of CKD. However, among the research on the association between vitamin D and eGFR or uACR, previous results have been inconsistent. Park et al.(24) reported that 25(OH)D was positively associated with eGFR (p<0.001) and negatively associated with uACR (p = 0.043) in Korean adults. In contrast, O'Seaghdha et al.(25) reported that 25(OH)D was not associated with either eGFR (ptrend = 0.3) or uACR (ptrend = 0.9) in the Framingham Heart Study. In the present study, after adjusting for the related variables (except age), the association between vitamin D and elevated uACR and decreased eGFR group was not significant, and these results were similar to the study of O'Seaghdha et al. However, when further adjusting for age, the ORs of vitamin D deficiency with the normal group as a reference were significantly higher in the elevated uACR group [1.312 (95% CI, 1.035–1.662)] and decreased eGFR group [1.761 (95% CI, 1.062–2.919)], and these results were similar to the study of Park et al. Age is a strong risk factor of albuminuria and CKD.(26,27) In our results, the prevalence of elevated uACR and decreased eGFR levels were increased as an increase of age, but the prevalence of vitamin D deficiency was decreased (Fig. 1). Vitamin D was increased as an increase of the age because the outdoor activity in the Korean elderly is higher than in the younger.(28) However, aging affects the formation of 1,25(OH)2D, the active form of vitamin D. Although vitamin D [25(OH)D] increases, production of 1,25(OH)2D is reduced by 50% as a result of a decline in renal function according to increase of age.(29) Therefore, some studies emphasized that need to measure both 25(OH)D and 1,25(OH)2D in vitamin D deficiency.(10,30,31)</p><p>We examined the ORs of vitamin D deficiency when the decreased eGFR and elevated uACR occurred simultaneously. It is important to monitor uACR levels in populations with CKD. Albuminuria is an unequivocal surrogate marker for CKD progression as well as future cardiovascular events and its reduction is used as a treatment goal for these diseases.(32) In addition, albuminuria is an early warning sign of diabetic nephropathy (DN), and DN is associated with an elevated risk of progression toward ESRD as well as increase of cardiovascular events and mortality.(33–35) In the present study, the ORs of vitamin D deficiency in the decreased eGFR plus elevated uACR group [3.549 (95% CI, 1.975–6.365)] was very higher than the elevated uACR group [1.312 (95% CI, 1.035–1.662)] or decreased eGFR group [1.761 (95% CI, 1.062–2.919)]. We thought these results that the synergistic interaction between the decreased eGFR and elevated uACR. Levey et al.(18) suggested that the synergistic interaction between the decreased eGFR and elevated uACR. They reported that the progressive CKD in the decreased eGFR plus elevated uACR group (at least 9.4 times, up to 57 times) was higher than the elevated uACR group (at least 0.4 times, up to 8.1 times). In particular, the OR of end stage renal disease (ESRD) for the decreased eGFR plus elevated uACR group (at least 40 times, up to 2,286 times) was greatly higher than the elevated uACR group (at least 3.8 times, up to 67 times).</p><p>Vitamin D from the diet or skin synthesis is biologically inactive and is converted to 25(OH)D in the liver. And then, 25(OH)D is further hydroxylated in the kidneys to form 1,25(OH)2D which is the biologically active form of vitamin D.(36) However, in patients with CKD, there is high rate of prevalence of vitamin D deficiency because the reduced ability to convert the active form 1,25(OH)2D.(37) Therefore, if renal function decreases rapidly by the synergistic interaction between the decreased eGFR and elevated uACR, the frequency of vitamin D deficiency may increase. On the other hand, renal function may decrease as vitamin D deficiency. Vitamin D is known to suppress the renin gene transcription,(38) and administration of vitamin D preparations such as calcitriol and paricalcitol inhibit renin expression and consequently reduce angiotensin II expression.(39) Angiotensin II is a key mediator of proteinuria, raises efferent glomerular arteriole resistance and induces transforming growth factor (TGF)-β1, which inhibits cell proliferation and increases apoptosis in the kidney,(40,41) and so the reduction of angiotensin II by vitamin D may be a mechanism to counter these effects. NF-κB is involved in the regulation of inflammatory cytokines that may promote inflammation and fibrogenesis in kidney disease.(42) In mice with obstructive nephropathy, administration of paricalcitol was found to block NF-κB and attenuate tubule-interstitial inflammation.(43) Cohen-Lahav et al.(44) reported that vitamin D upregulates IkappaBalpha (IκBα) levels by increasing mRNA stability; an increase in IκBα levels reduces nuclear translocation of NF-κB and thereby downgrades its activity. It is unclear whether the decrease of renal function increased the incidence of vitamin D deficiency, or vitamin D deficiency decreases the renal function. Furthermore, the association between the decrease of renal function and vitamin D is still debated. In the relationship between vitamin D and CKD and albuminuria, the result may differ according to the country and ethnicity, study population and the use of different reference GFR methods. In Asian, MDRD and CKD-EPI Equations for Taiwanese and Japanese adults is modified for their studied population. However, there is no definite modified model for Korean adults yet. Therefore, research is necessary to modify the MDRD and CKD-EPI Equations for the Korean adults.</p><p>There are a few limitations in the present study. First, season is the most important determinant of serum 25(OH)D levels, but the data of the KNHANES V-3 study did not specify serum 25(OH)D levels according to season. Second, serum calcium concentration and daily intake of vitamin D are important determinants of serum 25(OH)D levels, but these were not measured as part of the KNHANES V-3 study. Therefore, serum calcium concentration and daily intake volume of vitamin D could not be used as an adjustment variable. Third, parathyroid hormone (PTH) is an important determinant of serum vitamin D levels as increased PTH promotes calcium influx into adipocytes, where intracellular calcium enhances lipogenesis.(45) Therefore, serum vitamin D levels could change depending on serum PTH. However, in the data from the KNHANES V-3 study, there are no measurements of PTH of the participants (adults ≥20 years of age). The serum 25 (OH)D levels for each season, along with calcium and PTH levels, should be included as variables of vitamin D status in future studies. Fourth, because this was a cross-sectional study, the ability to establish a causal relationship between vitamin D and uACR and eGFR was limited. Therefore, more accurate results might be obtained by performing a cohort study by adding these variables.</p><!><p>The present study investigated the association between serum 25(OH)D and urine microalbumin/creatinine ratio and estimated glomerular filtration rate in Korean adults using data from the KNHANES V-3 conducted in 2012. Vitamin D deficiency was positively associated with the elevated uACR and decreased eGFR. In addition, vitamin D level decreased greatly when decreased eGFR and elevated uACR appeared simultaneously.</p><!><p>We have not received any financial support or other benefits from commercial sources for the work reported in the manuscript. None of the authors have financial interests that could create a potential conflict of interest or appearance of a conflict of interest with regard to this work.</p><!><p>Comparisons of the vitamin D deficiency, decreased eGFR, and elevated uACR according to age. Decreased eGFR: eGFR <60 ml/min/1.73 m2; Elevated uACR: uACR ≥30 mg/g; vitamin D deficiency: 25(OH)D <15.0 ng/dl. The prevalence of elevated uACR (p<0.001) and decreased eGFR (p<0.001) levels were increased as an increase of age, but the prevalence of vitamin D deficiency (p<0.001) was decreased as an increase of age.</p><p>Clinical characteristics of the research subjects</p><p>αeGFR, estimated glomerular filtration rate; βuACR, urine microalbumin/creatinine ratio; γBMI, body mass index; δWM, waist measurement; εSBP, systolic blood pressure; ζDBP, diastolic blood pressure; ηTC, total cholesterol; θTGs, triglycerides; κHDL-C, high density lipoprotein cholesterol; λFBG, fasting blood glucose; µBUN, blood urea nitrogen; πCrea, serum creatinine; σ25(OH)D, 25-hydroxyvitamin D. Vit. D sufficiency: 25(OH)D ≥15.0 ng/dl; Vit. D deficiency: 25(OH)D <15.0 ng/dl.</p><p>Clinical characteristics of the subjects according to decreased eGFR, elevated uACR and decreased eGFR plus elevated uACR</p><p>αNormal: eGFR ≥60 ml/min/1.73 m2 and uACR ≤30 mg/g. βElevated uACR: uACR ≥30 mg/g, γDecreased eGFR: eGFR <60 ml/min/1.73 m2, ζDecreased eGFR plus Elevated uACR: eGFR <60 ml/min/1.73 m2 and uACR ≥30 mg/g.</p><p>Comparisons of vitamin D deficiency odds ratios according to decreased eGFR, elevated uACR and decreased eGFR plus elevated uACR</p><p>25(OH)D, 25-hydroxyvitamin D; Vit. D deficiency, 25(OH)D <15.0 ng/dl; eGFR, estimated glomerular filtration rate; uACR, urine microalbumin/creatinine ratio. Model 1 [odds ratio (OR), 95% CI)], Non-adjusted; Model 2 [OR, 95% CI], adjusted for alcohol drinking, SBP, DBP, BMI and WM; Model 3 [OR, 95% CI], Model 2 further adjusted for TGs, HDL-C, BUN and FBG; Model 4 [OR, 95% CI], Model 3 further adjusted for age.</p><p>Comparisons of 25(OH)D levels according to decreased eGFR, elevated uACR and decreased eGFR plus elevated uACR</p><p>25(OH)D, 25-hydroxyvitamin D; eGFR, estimated glomerular filtration rate; uACR, urine microalbumin/creatinine ratio. Model 1 (Mean ± SE, 95% CI), alcohol drinking, SBP, DBP, BMI and WM; Model 2 (Mean ± SE, 95% CI), Model 1 further adjusted for TGs, HDL-C, BUN and FBG; Model 3 (Mean ± SE, 95% CI), Model 2 further adjusted for age.</p>

PubMed Open Access

Conformational selection or induced-fit? A critical appraisal of the kinetic mechanism\xe2\x80\xa0

For almost five decades, two competing mechanisms of ligand recognition \xe2\x80\x93 conformational selection and induced-fit - have dominated our interpretation of ligand binding in biological macromolecules. When binding/dissociation events are fast compared to conformational transitions, the rate of approach to equilibrium, kobs, becomes diagnostic of conformational selection or induced-fit based on whether it decreases or increases with the ligand concentration, [L]. However, this simple conclusion based on the rapid-equilibrium approximation is not valid in general. Here we show that conformational selection is associated with a rich repertoire of kinetic properties, with kobs decreasing or increasing with [L] depending on the relative magnitude of the rate of ligand dissociation, koff, and the rate of conformational isomerization, kr. We prove that, even for the simplest two-step mechanism of ligand binding, a decrease of kobs with [L] is unequivocal evidence of conformational selection, but an increase of kobs with [L] is not unequivocal evidence of induced-fit. Ligand binding to glucokinase, thrombin and its precursor prethrombin-2 are used as relevant examples. We conclude that conformational selection as a mechanism for ligand binding to its target may be far more common than currently believed.

conformational_selection_or_induced-fit?_a_critical_appraisal_of_the_kinetic_mechanism\xe2\x80\xa0

<!>Materials and Methods<!>Results<!>Discussion<!>

<p>The specific encounter between a ligand and a host target is fundamental to the chemistry of all biological activities. Understanding the molecular mechanism of how ligands recognize their targets and how those interactions are regulated remains a central issue to biochemistry and biophysics and a critical prerequisite for our ability to rationally design effective drugs and new therapeutics (1). In its simplest incarnation, binding of ligand L to its biological target E can be cast in terms of the single step reaction scheme (2) where kon (M−1s−1) is the second-order rate constant for ligand binding and koff (s−1) is the first-order rate of dissociation of the E:L complex into the parent species E and L. The strength of the interaction is quantified by the equilibrium association constant Ka (M−1) defined as the ratio kon/koff, or equivalently by the equilibrium dissociation constant Kd (M) defined as the inverse of Ka, or koff/kon. Scheme 1 provides an important starting point for any discussion of ligand binding, but offers little insight into the mechanism of recognition. Basically, the scheme assumes that the binding interaction is a rigid body collision between the ligand and its target, with no conformational change involved. In this case, the system approaches equilibrium according to an observed rate constant, kobs, that increases linearly with [L]. The set of differential equations associated with Scheme 1 is in fact (1)(d[E]/dtd[E:L]/dt)=(−kon[L]koffkon[L]−koff)([E][E:L]) and the non-zero eigenvalue of the 2×2 matrix above gives the kobs measured experimentally as (2)−λ1=kobs=koff+kon[L] A plot of kobs vs [L] is linear with intercept koff and slope kon, from which the value of the equilibrium association constant Ka or Kd can be easily derived.</p><p>Scheme 1 needs extension in the more realistic scenario of a binding interaction that involves conformational transitions. In this case, the rate of approach to equilibrium is no longer a linear function of [L]. Two limiting schemes become of interest as special cases of a more general scheme that links ligand binding to conformational transitions (3, 4), as shown in Figure 1. In the first case (Scheme 2), the target exists in distinct conformations in equilibrium and the ligand selects the one with optimal fit, i.e., The species E* is added to reflect a pre-existing equilibrium between two forms, E* and E, of which only E can interact with the ligand L. The rate constants kr and k−r refer to the transitions from E* to E and backward, with the ratio r=k−r/kr quantifying the population of E* relative to E. This is the simplest form of the celebrated Monod-Wyman-Changeux model of allosteric transitions (5). In the second case (Scheme 3), the conformation of the target changes after ligand binding to provide an optimal fit, i.e., In this case, the rate constants kr and k−r refer to the transition from E*:L to E:L and backward, with the ratio r=k−r/kr quantifying the population of E*:L relative to E:L. Scheme 3 is the simplest form of the alternative Koshland-Nemethy-Filmer model of allosteric transitions (6) based on the induced-fit hypothesis (7). Schemes 2 and 3 have long been considered as mutually exclusive and distinguishing between them continues to dominate discussions in several systems of interest to biology and chemistry (4, 8–10).</p><p>Under the "rapid equilibrium approximation", binding and dissociation steps in Schemes 2 and 3 are assumed to be fast compared to conformational changes (11, 12) and the dependence of kobs on [L] becomes diagnostic of the mechanism involved (Figure 1). In the case of Scheme 2, the kobs decreases hyperbolically with [L] according to the equation (3)kobs=kr+k−r11+Ka[L]=kr+k−rKdKd+[L] A plot of kobs vs [L] is an inverse hyperbola with asymptotic values of k−r+kr for [L]=0 and kr for [L]=∞. The mid-point between these values defines the equilibrium constant Kd. The dependence of kobs on [L] reflects the decrease in the number of species from two (E* and E) to one (EL) as [L] increases, with the rate of approach to equilibrium shifting from the reversible E*-E interconversion at low [L], with a value k−r+kr, to the irreversible E* to E conversion at high [L] (rate kr). In the case of Scheme 3, the kobs increases hyperbolically with [L] according to the equation (4)kobs=k−r+krKa[L]1+Ka[L]=k−r+kr[L]Kd+[L] A plot of kobs vs [L] is a rectangular hyperbola with asymptotic values of k−r for [L]=0 and k−r+kr for [L]=∞, and again the mid-point between these values defines the equilibrium constant Kd. In this case the dependence of kobs on [L] reflects the increase in the number of species from one (E*) to two (E*L and EL) as [L] increases, with the rate of approach to equilibrium shifting from the irreversible EL to E*L conversion at low [L], with a value k−r, to the reversible E*L-EL interconversion at high [L] (rate k−r+kr). The widely used rapid equilibrium approximation has fueled the notion that conformational selection (Scheme 2) and induced-fit (Scheme 3) can easily be distinguished from the kinetics of approach to equilibrium (11). In turn, the preponderance of experimental systems found to obey eq 4 relative to eq 3 has been cited as evidence that induced-fit is the dominant mechanism of recognition in ligand binding to proteins (11, 13). In this study we analyze the kinetics of approach to equilibrium for Schemes 2 and 3 without simplifying assumptions on the rate constants and show how information on the mechanism of recognition can be extracted from the plot of kobs vs [L]. Our analysis calls for caution in the use of the rapid equilibrium approximation in the analysis of experimental data.</p><!><p>Prethrombin-2 and thrombin were expressed in E. coli and purified from inclusion bodies, essentially as described (14, 15). Both proteins were expressed with the S195A substitution, which renders the protein catalytically inert while leaving its binding properties intact (16, 17). Stopped-flow fluorescence measurements were conducted on an Applied Photophysics SX20 spectrometer using 1:1 mixing in a total volume of 60 µL. For Na+, K+, FPR and VPR the intrinsic fluorescence of thrombin was monitored with an excitation wavelength of 283 nm and a cutoff filter of 305 nm. The active site inhibitor p-aminobenzamidine (PABA) has a strong fluorescence signal at 380 nm when excited at 330 nm and shows extraordinary sensitivity to its binding environment in the active site of trypsin-like proteases (18, 19), thus these experiments were conducted by exciting at 330 nm with a 375 nm cutoff filter, as described previously (19). Final thrombin concentrations were 50 nM (Na+ binding), 75 nM (VPR and FPR binding), 100 nM (K+ binding) and 1 µM (PABA binding). All thrombin binding experiments were conducted in the presence of 5 mM Tris, pH 8.0 at 15 °C, 0.1 % PEG8000, with ionic strength maintained constant at 400 mM with choline chloride. Prethrombin-2 (75 nM) used essentially the same buffer with pH 8.0 at the temperature of interest. Individual kinetic traces were determined by averaging a minimum of four traces each from three independent ligand titrations. Traces were fit to a single exponential equation, with the quality of the fit determined by evaluation of the residuals. The kobs values, taken from the single exponential fits, were plotted against the ligand concentration [L] and these plots were used for all subsequent fitting to various kinetic schemes. Best-fit parameter values were derived by non-linear least squares with Mathematica.</p><!><p>Consider Scheme 2 and the set of differential equations associated with it (5)(d[E*]/dtd[E]/dtd[E:L]/dt)=(−krk−r0kr−k−r−kon[L]koff0kon[L]−koff)([E*][E][E:L]) The two non-zero eigenvalues of the 3×3 matrix of kinetic rate constants are (6)−λ1,2=k−r+kr+koff+kon[L]±(koff+kon[L]−k−r−kr)2+4k−rkon[L]2 In the general case, when no assumption is made on the relative rates of binding and conformational transition, the kinetics of approach to equilibrium depend on two exponentials, each associated with an observed rate constant defined by the solutions of eq 5. The larger eigenvalue, −λ1, defines a kobs that becomes increasingly fast as [L] increases. Detection of the contribution of this eigenvalue to the kinetics of approach to equilibrium may be difficult with standard transient kinetics and may require the use of ultra-rapid techniques like continuous flow and T-jump. The smaller eigenvalue, −λ2, typically defines the evolution of the system over a time scale accessible to rapid kinetics technique like stopped-flow. It is this eigenvalue that directly relates to the kobs accessible to experimental measurements. Elementary rearrengements of eq 6 show that under the rapid equilibrium approximation, koff+kon[L]>>k−r+kr, the larger eigenvalue −λ1 is simply koff+kon[L] and the smaller eigenvalue −λ2 is given by eq 3 and decreases hyperbolically with [L]. We are interested in the behavior of eq 6 in the general case.</p><p>Consider how eq 6 depends on [L] and especially its limiting values for [L]=0 and [L]=∞ (Table 1). As [L] increases, −λ1 grows linearly as kon[L] but −λ2 approaches the asymptotic value of kr. The limit for [L]=0 depends on the difference koff−k−r−kr. When the difference is positive (koff>k−r+kr), then −λ1=koff and −λ2=k−r+kr. If the difference is negative (k−r+kr>koff), then −λ1=k−r+kr and −λ2=koff. Basically, for [L]=0, −λ1 and −λ2 respectively assume the larger and smaller value between koff and the sum k−r+kr. When koff exceeds the sum k−r+kr, the situation is analogous to the rapid equilibrium approximation where −λ1=koff+kon[L] and −λ2 is given by eq 3. However, when k−r+kr exceeds koff, the kinetics become dependent on the relative magnitude of koff and kr. When koff>kr the value of −λ2=kobs decreases with [L]. Although this situation is analogous to the case koff>k−r+kr, there is an important distinction insofar as the asymptotic value for [L]=0 does not give k−r+kr as in the rapid equilibrium approximation but koff. Hence, when k−r+kr>koff>kr, Scheme 2 produces a dependence of kobs vs [L] that can be mistaken with that observed under the rapid equilibrium approximation (Figure 2), with the asymptotic value of kobs for [L]=0 becoming incorrectly assigned as the sum k−r+kr instead of koff. It also follows from inspection of eq 6 that the value of −λ2=kobs does not change with [L] when koff=kr (Figure 2). The lack of dependence of kobs on [L] was originally reported in the first rapid kinetics study of Na+ binding to thrombin (20) and interpreted in terms of an extended version of Scheme 2. It is now clear that Scheme 2 used in its general form easily accounts for a value of kobs that is independent of [L] and no extension is necessary. Finally, when koff<kr the value of −λ2=kobs actually increases with [L] and mimics the dependence observed in the case of induced-fit (Figures 1 and 2). In the general case, conformational selection (Scheme 2) produces a gamut of kinetic properties that include those uniquely pertaining to induced-fit (Scheme 3) under the assumption of rapid equilibrium.</p><p>Unlike Scheme 2, Scheme 3 always produces eigenvalues that increase with [L]. Scheme 3 is associated with the set of differential equations (7)(d[E*]/dtd[E*:L]/dtd[E:L]/dt)=(−kon[L]koff0kon[L]−kr−koffk−r0kr−k−r)([E*][E*:L][E:L]) The two non-zero eigenvalues associated with the 3×3 matrix of kinetic rate constants are (8)−λ1,2=k−r+kr+koff+kon[L]±(koff+kon[L]−k−r−kr)2+4krkoff2 There is a basic similarity between eqs 6 and 8, that only differ in the last term under the square root expression. However, this difference results in two eigenvalues that always increase with [L]. Unlike Scheme 2, there is no finite value of the rate constants in Scheme 3 that makes −λ2 independent of [L]. As [L] increases, −λ1 grows linearly as kon[L] as in the case of Scheme 2, but −λ2 approaches the asymptotic value of k−r+kr. The limit for [L]=0 does not depend on the relative values of koff and the sum k−r+kr (Table 1) as in Scheme 2 and is always less than the value reached for [L]=∞. This implies that Scheme 3 always produces values of kobs that increase with [L], as opposed to Scheme 2 that produces values of kobs that decrease, increase or are independent of [L] depending on the relative values of koff and kr.</p><p>A plot of kobs decreasing with [L] or independent of [L] is unambiguous proof of conformational selection (Scheme 2), but a plot of kobs increasing with [L] may also be associated with conformational selection (Figure 2) and cannot be considered unambiguous proof of induced-fit (Scheme 3). Glucokinase illustrates this scenario directly. The enzyme was originally assumed to bind glucose at a single site according to induced-fit (21) based on the analysis of rapid kinetics data where both the larger and smaller eigenvalues were resolved experimentally (Figure 3). The larger eigenvalue produces a kobs that increases linearly with [L] and the smaller eigenvalue shows a hyperbolic increase with [L] as expected of Scheme 3 under the rapid equilibrium approximation. However, the data in Figure 3 cannot be considered unequivocal evidence of induced-fit because they fit accurately to Scheme 2 without the rapid equilibrium approximation with values of the rate constants k−r=7.16 s−1, kr=0.8 s−1, koff=0.34 s−1 and kon=540 M−1s−1. It comes as no surprise, then, that more recent kinetic measurements of glucose binding to glucokinase support conformational selection over induced-fit (22) and it is now accepted that glucokinase exists in alternative conformations in equilibrium prior to the binding of any ligands (23). This conclusion is strongly supported by recent X-ray structures of glucokinase that reveal how ligand binding does not result in conformational changes of the enzyme but only in stabilization of the E form (24).</p><p>Figure 4 shows the values of kobs as a function of [L] for the case of different ligands binding to the clotting protease thrombin (25). The active site inhibitor PABA shows an inverse hyperbolic dependence of kobs vs [L] conforming to conformational selection according to Scheme 2. Under the rapid equilibrium approximation for PABA binding, analysis of the data in Figure 2 gives values of kr=74 s−1 and k−r=344 s−1 for the E*-E interconversion, along with a value of Kd=53 µM for the equilibrium association constant, that translates into a value of Kd,app=Kd(1+r)=300 µM in agreement with equilibrium titration measurements (18). The kinetic features of PABA binding to thrombin show evidence of the E*-E equilibrium in the free form of the enzyme. As for glucokinase, this conclusion is directly supported by X-ray structural data (19). Thrombin crystallizes in the free form in two alternative conformations, one (E form) with the active site open and accessible to substrates and inhibitors like PABA, and the other (E* form) with the active site occluded by a collapse of the 215–217 segment that precludes binding of substrates or inhibitors like PABA. The E* and E forms have been detected crystallographically in the same protein construct for several thrombin mutants (19), as well in numerous other members of the trypsin family (26). Indeed, conformational selection is the dominant mechanism of ligand recognition in this large family of biologically important enzymes (10).</p><p>Because the E*-E equilibrium exists in solution independent of any binding event, its presence should be detected with exactly the same kinetic signatures regardless of the ligand under study. Specifically, different ligands influenced directly by the E*-E equilibrium should produce the same dependence of kobs on [L] as seen for PABA (Figure 4), with identical asymptotic values for [L]=0 and [L]=∞. A key property of thrombin is its ability to bind monovalent cations like Na+ and K+ to a site near the primary specificity pocket that is obliterated by the repositioning of an Arg side chain when the enzyme switches from the E to the E* form (25). Consistent with the scenario supported by structural data, binding of K+ produces a dependence of kobs on [L] that closely resembles that observed for PABA (Figure 4), thereby supporting the conclusion that the asymptotic values of kobs for [L]=0 and [L]=∞ reflect the properties of the E*-E interconversion. However, binding of Na+ deviates from K+ and PABA, especially in the asymptotic value of kobs for [L]=0 that defines the value of kr+k−r (Table 1). Why do K+ and Na+ produce different asymptotic values of kobs for [L]=0 if they bind to the same site (27, 28) and according to the same kinetic mechanism? A fourth ligand, the chromogenic substrate VPR binding to the active site like PABA, produces a dependence of kobs on [L] that departs even more drastically from the profile seen for PABA and K+ (Figure 4). In this case, the value of kobs actually increases with [L]. The properties of Scheme 2 in the general case rationalize the seemingly disparate behavior observed in the binding of different ligands to thrombin (Figure 4). Binding of PABA, Na+ and K+ is consistent with the E*-E equilibrium, with only E enabling binding at the active site and the cation binding site, as indicated by structural biology (19). For all ligands, the value of kobs levels off around 70 s−1 at high [L] and increases to 400 s−1 for [L]=0 for PABA and K+, and 130 s−1 for Na+. When kobs decreases with [L], the value for [L]=0 always measures whichever is smaller between koff and the sum k−r+kr (Table 1). Hence, the value of 130 s−1 measured for Na+ binding at [L]=0 should not be assigned to k−r+kr but to koff for Na+ dissociation, and the value of 400 s−1 measured for PABA and K+ binding is most likely the sum k−r+kr. These values enable assignement of kr=70 s−1 and k−r=340 s−1 (Table 2), making the E* form 5-fold more populated than E under the experimental conditions of the rapid kinetics measurements. Assignment of the asymptotic value of kobs at [L]=0 for Na+ binding as the sum k−r+kr would give values of kr=60 s−1 and k−r=67 s−1, thereby making E* and E equally populated. This is the conclusion drawn in previous studies where the value of kobs for [L]=0 was interpreted as kr+k−r under the assumption of rapid equilibrium for Na+ binding (19, 29, 30). The analysis presented here demonstrates that previous studied have underestimated the contribution of E* and supports the need for measurements involving different ligands, under the same solution conditions, to fully explore the range of kobs values in the limits [L]=0 and [L]=∞ when the larger eigenvalue is not accessible to experimental measurements. Once the values of kr and k−r are assigned from measurements of PABA and K+ binding, all other rate constants in eq 6 can be assigned for Na+ (Table 2). The value of kon=3.4·104 M−1s−1 is in agreement with that determined from continuous flow ultra-rapid kinetics (31) and the value of Kd,app=Kd(1+r)=27 mM agrees with equilibrium titrations of Na+ binding (32, 33). In the case of the chromogenic substrate VPR, the value of kobs at high [L] levels off around 70 s−1, again in keeping with the rate for the E* to E transition kr, but the dependence of kobs is drastically different and shows an increase with [L]. This behavior could be easily interpreted in terms of an induced-fit under the rapid equilibrium approximation. On the other hand, Scheme 2 offers an alternative explanation in terms of conformational selection with koff much smaller than kr, resulting in values of kon=1.7·107 M−1s−1 and Kd,app=Kd(1+r)=0.75 µM consistent with previous kinetic (34) and equilibrium (17) measurements. Binding of four different ligands (PABA, K+, Na+ and VPR) to thrombin support the same mechanism of conformational selection, even though the kinetic signatures differ drastically in the four cases.</p><p>Screening multiple ligands is a specific example of the more general strategy of altering the relative rates of kr and koff to differentiate between potential mechanisms and detect signatures of conformational selection. Perturbations that alter the values of koff relative to kr for the same ligand produce similar results. We demonstrate this principle with prethrombin-2 (15), an inactive zymogen precursor of thrombin, binding to the chromogenic substrate FPR at different temperatures. Because temperature may affect the values of koff and kr to different extent, it may be ideally suited to detect drastic changes in the dependence of kobs vs [L]. At 15 °C, FPR binding displays the same hyperbolically increasing dependence of kobs on [L] seen with VPR in thrombin, with asymptotes of 45 s−1 at [L]=0 and 156 s−1 at [L]=∞. Both the [L]=0 and the [L]=∞ asymptotes, reflecting koff and kr respectively, increase with temperature but to different extent. At 25 °C, the asymptotes increase to 142 s−1 and 255 s−1 in the limits of [L]=0 and [L]=∞. However, at 35 °C, koff becomes at least 550 s−1 and faster than kr (365 s−1) causing the dependence of kobs on [L] to invert and decrease hyperbolically with [L] (Figure 4). Under the rapid equilibrium approximation, this type of behavior could only be explained by the mechanism of binding switching from induced-fit at or below 25 °C to conformational selection at 35 °C, a very unrealistic scenario. On the other hand, Scheme 2 in the general case explains the binding of FPR to prethrombin-2 with a single mechanism, conformational selection, with the dependence of kobs on [L] changing drastically based on the relative values of kr and koff at different temperatures, in keeping with the simulations reported in Figure 2. Hence, conformational selection not only applies to the mature enzyme thrombin, but also to its inactive precursor prethrombin-2, thereby confirming kinetically well established results from structural biology for thrombin (15) and trypsin-like proteases in general (10, 26).</p><!><p>A previous study by Galletto et al. (35) merged the mechanisms in Schemes 2 and 3 to produce a three-step scheme of ligand binding where a pre-existing equilibrium is followed by ligand-induced conformational changes. Because of the algebraic complexity of the expressions involved, the properties of such scheme were analyzed under separation of time scales. Important conclusions were reached about the dependence of kobs on [L] and how the extended scheme would account for kinetic properties often associated with induced-fit. A more recent study by Hammes et al. (4) offered a description of Schemes 2 and 3 in terms of fluxes along selected pathways of a more general scheme encompassing conformational transitions and ligand binding (Figure 1). A continuum version of that scheme has been discussed by Zhou (9). Again, a richer repertoire of kinetic properties was identified for conformational selection. Our study reveals the kinetic properties of Schemes 2 and 3 and derives mathematical expressions to extract rate constants from the analysis of experimental data. Conformational selection is shown to produce kinetic properties such as an increase of kobs with [L] that overlap with those of induced-fit, even though no such mechanism is present in Scheme 2. On the other hand, induced-fit (Scheme 3) is unable of kinetic properties pertaining uniquely to conformational selection, i.e., a kobs decreasing with [L] or independent of [L].</p><p>When the rapid equilibrium approximation is invoked, both Scheme 2 and 3 depend on three independent parameters (eqs 3 and 4) that can be resolved from a plot of kobs vs [L]. If the approximation is not used, both schemes depend on four independent rate constants but the plot of kobs vs [L] only contains information on three independent variables, i.e., the asymptotic values at [L]=0 and [L]=∞ and the mid-point of the transition. In this case, it is not possible to unequivocally resolve all the independent rate constants kr, k−r, kon and koff. The value of kr for Scheme 2 is always defined by kobs for large [L], but the value at [L]=0 is the smaller between kr+k−r and koff. The mid-point of the transition is not necessarily related to Kd. The situation is even more complex in the case of Scheme 3 where the value of kobs for [L] large is kr+k−r, but the value for [L]=0 depends on three rate constants (Table 1) and again the mid-point of the transition does not depend only on Kd. Determination of all four independent rate constants is always possible when both the larger and smaller eigenvalues are accessed experimentally, as shown in the case of glucokinase (Figure 3). However, this approach may not be feasible in general. Use of different ligands binding with different rate constants becomes essential, as shown in the present study for thrombin. Although the approach in the general case may fail to provide the exact magnitude of the rate constants involved in the scheme, it is nonetheless instrumental in avoiding potentially incorrect conclusions based on the rapid equilibrium approximation. Particularly important is the need to assign the mechanism of binding when kobs increases with [L], which cannot be linked unambiguously to induced-fit as shown by our analysis and previous work (4, 35). In a recent review, Tummino and Copeland pointed out that induced-fit is by far the most common mechanism of recognition documented experimentally, with conformational selection being confined to a handful of cases (11). This conclusion was based on the results of kinetics analyzed under the rapid equilibrium approximation. Conformational selection can explain an increase of kobs with [L] commonly assigned to induced-fit, but induced-fit can never account for a decrease of kobs with [L] that unequivocally identifies conformational selection. We therefore suspect that the preponderance of induced-fit as a mechanism of ligand binding needs to be critically reconsidered. Indeed, several proteins like alkaline phosphatase (36), chymotrypsin (12, 37), thrombin (20, 29), meizothrombin-desF1 (38), glucokinase (22), trypsin (39), the immunoglobulin IgE (40), clotting proteases factor Xa and activated protein C (30) obey conformational selection, and so does DNA in its B to Z transition (41). Structural biology offers unequivocal evidence of multiple conformations in pre-existing equilibrium for maltose-binding protein (42), trypsin-like proteases in general (10, 26) and RNA (43).</p><p>Our work offers a clear explanation for the apparent infrequency of conformational selection (11). There is a lower limit on the equilibrium dissociation constant Kd where the characteristic inverse hyperbola of conformational selection will be detectable. As discussed earlier, the strength of a ligand binding interaction is controlled by the interplay between kon and koff rates, with the diffusion limit setting the upper bound for kon at around 6.5·108 M−1s−1 (44). Because kon is limited and koff must exceed kr for conformational selection to be detected, only ligands binding with (9)Kd>krkon would produce a plot of kobs that decreases hyperbolically with [L]. If the value of kr is in the range of 100 s−1, conformational selection can only be detected for values of Kd>150 nM, provided the ligand binds at the diffusion-limited rate. Thus, not only is sampling multiple ligands important to establishing the kinetic mechanism of binding, but experiments should be designed to sample conditions where ligand binding is relatively weak.</p><!><p>This work was supported in part by the National Institutes of Health Research Grants HL49413, HL73813, HL95315 and HL112303.</p>

PubMed Author Manuscript

Observation and rationalization of nitrogen oxidation enabled only by coupled plasma and catalyst

Heterogeneous catalysts coupled with non-thermal plasma (NTP) are known to achieve reaction yields that exceed the contributions of the individual components. Rationalization of the enhancing potential of catalysts, however, remains challenging because the background contributions from NTP are often non-negligible. Here, we first demonstrate nitrogen (N 2 ) oxidation by radio frequency plasma and platinum (Pt) combination at conditions in which nitric oxide (NO) yield from plasma or Pt is vanishingly small. We then develop reactor models based on reduced NTP-and surface-microkinetic mechanisms to identify the features of each that lead to the synergy between NTP and Pt. At experimental conditions, NO yields from NTP and thermal catalysis are suppressed by radical reactions and inhibited by high N 2 dissociation barrier, respectively. Pt catalyzes NTP-generated radicals and vibrationally excited molecules to produce NO. The model construction further illustrates that the optimization of yield and energy efficiency involves tuning of plasma species, catalysts properties, and the reactor configurations to couple the two. These results provide unambiguous evidence of the benefits of combining plasma and catalysts and open approaches to design the coupled system.

observation_and_rationalization_of_nitrogen_oxidation_enabled_only_by_coupled_plasma_and_catalyst

Introduction<!>Plasma-catalytic N 2 oxidation experiments<!>Plasma-catalytic N 2 oxidation models<!>Intrinsic catalytic rates<!>Plasma vs plasma-catalytic NO production<!>Optimal plasma-catalytic N 2 oxidation regimes<!>Conclusions<!>Experimental details<!>Computational details<!>Plasma catalysis

<p>The ability of a non-thermal plasma (NTP) and heterogeneous catalyst combination to achieve reaction yields that exceed the contributions of the individual components is well documented. 1,2 Often reactions are explored at conditions at which background NTP yields are non-negligible. Disentangling the gain achieved by combining NTP and catalyst from the background contributions of NTP and catalyst alone, and inferring the origins of yield enhancements, are thus significant practical challenges. [3][4][5] Here, we demonstrate NTP-catalytic nitrogen oxidation:</p><p>in a reactor configuration and at conditions in which neither catalyst nor NTP yields significant product. Observed product yields are thus the result of the mutual action of NTP and catalyst. Further, we demonstrate a modeling strategy to integrate and isolate NTP and catalyst contributions to observed performance. These models recover and rationalize the observed product yields and provide a foundation for system optimization.</p><p>Reaction 1 is highly endothermic, and at ambient conditions in air, the equilibrium lies far to the left (Supplementary Fig. 1). The equilibrium shifts towards NO with increasing temperature, a fact exploited in the Birkeland-Eyde (B-E) process for thermally fixing N 2 at high temperatures achieved within a thermal plasma. 6 At these high temperatures, N 2 and O 2 are partially atomized, and NO is produced by O and N radical reactions with N 2</p><p>and O 2 , respectively, in the so-called Zeldovich mechanism. 7 Similar processes are at play in conventional combustion of fuel in air, motivating interest in NO decomposition catalysts for environmental protection. 8 However, catalytic N 2 oxidation under thermal conditions is unknown.</p><p>Non-thermal N 2 /O 2 plasmas are known and have been observed to generate NO at bulk temperatures much below those necessary for thermal N 2 oxidation. NO yields can even exceed those expected based on the bulk thermodynamic equilibrium. 9 These NTPs contain radicals, ions, and vibrationally excited molecules, and kinetic models that incorporate the reactions of these excited species have been applied to microwave, 10,11 pulsed-power glidingarc, 12 glow discharge 13 and stationary plasmas. 14 These models recover observed NO yields and even the densities of intermediates. 10,12,13 These models suggest that the same Zeldovich mechanism is at play in NTPs as in thermal oxidation. 15,16 NO yields have been reported to increase over a non-zero background when a radio frequency, microwave or dielectric barrier discharge plasma is combined with a catalyst. [17][18][19] Patil et al showed that the extent of that increase is catalyst-dependent, in experiments comparing a variety of metal oxides within a packed-bed, dielectric barrier discharge reactor. 19 Similarly, introduction of WO 3 and MoO 3 down-stream of either an inductively coupled high frequency plasma or microwave plasma was observed to increase NO yields over background. 18,20 In all cases, the observed improvements are with respect to significant background NO yields. Models to qualitatively or quantitatively explain these observations have not been reported.</p><p>These observations leave open the question of the mechanisms by which NO production is increased and even the extent to which that increase can be attributed to surface catalytic reactions. To disentangle the NTP and catalytic contributions to N 2 oxidation, here we report N 2 oxidation experiments at N 2 /O 2 mixing ratios at which NTP-only and Pt-catalyzed NO yields are each vanishingly small. We demonstrate that introduction of a Pt catalyst positioned post-discharge from the N 2 -O 2 radio frequency plasma results in a substantial increase in NO production. To interpret and quantify this evident synergy between NTP and catalyst, we represent the experimental system as a series of coupled reactors to predict NO yields in the absence and presence of a Pt catalyst. We describe the plasma chemistry using a combination of vibrational heating of the diatomics and atomic nitrogen generation and parameterize surface reactions with density-functional-theory (DFT)-computed data. At O 2 mole fractions on the order of 10 The results highlight approaches to exploiting plasma-catalytic chemical synthesis as well as modeling strategies for identifying optimal plasma-catalyst combinations.</p><!><p>We measured NO production via N 2 oxidation in a radio frequency plasma reactor at low O 2 -to-N 2 pressure ratios, with and without downstream catalyst. The reactor consists of an inductive coil connecting to a radio frequency power supply and a matching network and a quartz tube with a heating mantle (Fig. 1a). A porous Pt film deposited on a tubular YSZ membrane is used as the catalyst. Catalyst microstructure consists of a network of percolated particles of the order of micron (Fig. 1b) with thickness of approximately 14 micron (Fig. 1c).</p><p>N 2 -O 2 mixtures are introduced to the reactor at 100 SCCM, 5 mbar and ambient temperature.</p><p>The N 2 -O 2 plasma is generated in the area near the coil and activated species flow toward the heating mantle, which is kept at 873 K. 21 to 10 −2 . Also reported is the thermodynamic equilibrium production of NO at 873 K. NO production at zero plasma power (i.e. thermal catalysis) is zero within the detection limit of the instrument. Thus thermal catalysis is ineffective at these operating conditions. Moreover, NO production via plasma (80 W) is also ineffective. NO yields are less than 20 and 60 ppm without and with the YSZ tube with a standard deviation of 40-50%.</p><p>However, NO production increases significantly upon placing a Pt catalyst in the middle of the heating mantle (Fig. 1d). NO yields exceed thermal equilibrium across the entire gas composition range explored and vary non-linearly with O 2 mole fraction, maximizing unambiguous evidence of NO production dependent upon both plasma and Pt catalyst and sensitivity to exact plasma composition. X-ray diffraction (Supplementary Fig. 2a) and Xray photoelectron spectroscopy (Supplementary Fig. 2b) observations confirm that the bulk and surface of the Pt catalyst are unmodified by plasma exposure.</p><!><p>To rationalize plasma-catalytic NO production and its unusual dependence on O 2 mole fraction, we created microkinetic models for the thermal catalytic, non-thermal plasma, and coupled systems. Fig. 2 illustrates the relevant physical processes in each case, including surface activation of thermalized gas molecules, reactions of vibrationally excited molecules and of radicals presnt in an NTP, and reactions at the interface between the two, respectively.</p><p>Catalytic reactions occur at surfaces and thus are characterized by rates per surface site, or turnover frequency. Plasma-phase reactions occur in an (inhomogeneous) bulk phase. The relative contributions of the two to observed yield are thus dependent on the relative number of active site and volume of the reactor. Here, we first consider intrinsic rates over a catalyst in the absence and presence of relevant concentrations of plasma-generated, excited species.</p><p>We then couple the two through an integral reactor series parameterized to be representative of the reactor of Fig. 1a and incorporating plasma-only and plasma-catalytic steps.</p><!><p>Catalytic N 2 oxidation is the reverse of the more widely studied catalytic NO decomposition reaction. The overall reaction energy and free energy are both about 1.8 eV because the reaction conserves molecules and therefore ∆S • ≈ 0. Thermal catalytic N 2 oxidation is therefore endergonic. We adopt as the thermal catalytic mechanism reactions indicated to be relevant to NO decomposition on Pt. [22][23][24] Fig. 3a summarizes the potential energy surfaces for N 2 oxidation over models for a Pt terrace (Pt(111)) and a step (Pt(211)), extracted from previously reported DFT results. 24 The initial N 2 activation step is both endothermic and has high barrier on terraces and even step sites on Pt. We supplement this reaction scheme with two additional steps to incorporate the potential adsorption of plasma-generated radicals, consistent with their observed relevance to plasma-wall chemistry: 15,25</p><p>where * represents a surface active site. Kinetic parameters for all surface reaction steps are detailed in Supplementary Table 1.</p><p>We parametrize a mean-field microkinetic model to predict the intrinsic steady-state NO turnover frequency (TOF) over Pt at conditions consistent with experiment (Fig. 1).</p><p>TOFs are computed at fixed N 2 , O 2 , N and O presssures and at zero conversion. 26 N radical densities are estimated from N 2 dissociation fractions measured at similar plasma conditions and O 2 dissociation fractions assumed to be one order of magnitude greater than N 2 . 27-31 N 2 and O 2 are assumed to have the same vibrational temperature, and the vibrational energy distributions and the consequent effects on the kinetics of N 2 and O 2 dissociative adsorption are modeled following Mehta et al (Supplementary Methods and Supplementary Fig. 3). 32 Fig. 3b compares TOFs at a Pt step as a function of conditions. Thermal TOFs (black line) over Pt(211) increases with T but are vanishingly small at even the highest temperature.</p><p>Oxygen atoms are the most abundant surface intermediate across this regime (Supplementary Fig. 4a), and by degree of rate control 33 analysis (Supplementary Fig. 5a), N 2 dissociative adsorption is rate-controlling, both consistent with the high N 2 dissociation barrier. The green region of Fig. 3b shows TOFs at N 2 and O 2 vibrational temperatures from 3000 to 10 000 K. TOFs are increased substantially relative to thermal-only catalysis, especially at the lowest bulk temperatures, and apparent activation energies are diminished. Predicted surface coverages are unchanged from the thermal case (Supplementary Fig. 4b) but the rate-controlling step changes to NO desorption at lower bulk and O 2 adsorption at higher bulk temperatures (Supplementary Fig. 5b), consistent with N 2 dissociation rates enhanced by vibrational excitation. The orange region of Fig. 3b shows TOFs at N radical density artificially set to 2 to 8 orders of magnitude less than the N 2 density and O/O 2 fraction to 10N/N 2 . N and O radicals have a similar to even greater enhancing impact on NO TOFs, most notably at the lowest bulk temperatures. NO is the most-abundant surface species at bulk temperatures below 800 K (Supplementary Fig. 4c), and NO desorption becomes rate controlling (Supplementary Fig. 5c), both reflecting the assumed barrierless accommodation of O and N by the Pt surface. At higher bulk temperatures, surface coverages tend to zero and rates become controlled by oxygen adsorption. Both vibrational excitation and radical adsorption relax the rate limitations of N 2 dissociation and are particularly effective at lower bulk temperatures.</p><p>Rates derived from kinetic parameters appropriate to a Pt terrace lead to similar general observations (Supplementary Fig. 6). Absolute Pt terrace TOFs, however, are significantly less than Pt steps at all but the highest vibrational temperatures or N densities, consistent with the greater reaction barriers on the terrace.</p><!><p>The above microkinetic models predict absolute, per active site rates at given conditions. 34 Experiments most directly provide access to product yields rather than reaction rates. To compare plasma-only to plasma-catalytic yields, we develop well-mixed, isothermal integral reactor models appropriate to the plasma afterglow region and the Pt catalyst bed, respectively (Fig. 4a and b and see details in Supplementary Methods). Bulk temperatures are assumed to be 873 K to correspond with experiments.</p><p>We describe non-catalytic NO oxidation in the afterglow (Fig. 4a) using the Zeldovich mechanism, consistent with previous experiments and simulations of non-thermal radio frequency, gliding arc and microwave N 2 /O 2 plasmas: [10][11][12]15,16,27,35</p><p>Rate constants are from experimental measurements (Supplementary Table 3). 15,36,37 To capture the influence of vibrational exciations on the rate of Reaction 3a, we reduce the reaction barrier by an amount commensurate with the degree of plasma-induced vibrational excitation (Fig. 2, middle). 16 We take the reactor volume to be consistent with the length of the heating mantle and flow rates consistent with experiment.</p><p>To describe plasma-catalytic NO oxidation yields over the porous Pt catalyst (Fig. 4b), we treat this region as a sum of coupled contributions of a non-catalytic bulk-phase, described using the same Zeldovich parameters, and a surface-catalytic phase, described using the Pt step parameters. The relative contributions of these two phases is a function of the reactor volume and number of active sites (Supplementary equation ( 11)). The total volume is taken to be that of the porous Pt catalyst bed, and flow rate and free volume-to-active site ratios are taken to be consistent with the experimental setup.</p><p>Fig. 4c reports the plasma-only, non-catalytic NO yields, plotted as NO partial pressure for ready comparison to experiment, as a function of assumed vibrational temperature and inlet atomic N density as descriptors because of their relevance in plasma N 2 oxidation. [10][11][12]15,16,27,35,38 The partial pressures of other species are reported in Supplementary Fig. 4d reports corresponding NO yields in the plasma-catalytic regime as a function of the same parameters. Complete results are summarized in Supplementary Fig. 7. NO yields are of similar orders of magnitude overall but less sensitive to vibrational temperature and more sensitive to N radical density. These differences become more evident when plotted as a ratio of yields, as shown in Fig. 4e. The plasma-catalyst combination is expected to have a negligible to deleterious effect on NO yields across a large swath of N radical densities and vibrational temperatures. In contrast, the combination has dramatically greater yields than the plasma-only reactor at relatively high N radical densities and especially at the lower limit of vibrational temperatures, conditions typical of low-P radio frequency and microwave plasmas. 11,28,29 The origins of this catalytic influence are revealed by an analysis of relative steady-state reaction fluxes normalized to the overall NO production rate. 39 and upper right corner of Fig. 4d), the absolute catalyzed TOF reaches its maximimum of 0.22 s −1 . Catalyzed rates are dominated by the large fluxes of N radicals to the catalyst and their subsequent reactions to NO or recombination to N 2 . These surface reactions become more effective at directing N into channels that produce NO than is the homogeneous phase, where the reverse Zeldovich reaction depletes NO. As a result, this regime is yellow in Fig. 4e. At high N radical density but lower vibrational temperature (Fig. 5a and upper left corner of Fig. 4d), surface reactions remain dominated by N adsorption and reaction to NO or N 2 .</p><p>Despite the fact that the TOF in this quadrant (0.13 s −1 ) is less than in the upper right, the plasma-catalyst combination is most effective in yielding NO, because the catalyst is most effective here in shunting N radicals away from the unproductive reverse Zeldovich reaction.</p><p>NO yields are maximized at high T vib and high N radical densities and in the plasmaonly and plasma-catalytic reactors, respectively (Fig. 4e). In both of these regimes, a large fraction of plasma-generated N radicals ultimately return to N 2 either homogeneously or at the catalyst surface, diminishing the energy efficiency of NO production. To compare the theoretical energy consumption of the two reactors as a function of reaction species, we calculate the enthalpy required to reach T vib and P N from reactants at the bulk gas temperature, and normalize the energy to the NO production (details in Supplementary Methods). Supplementary Fig. 10 shows the plasma-only reactor is most energy efficient at higher T vib , consistent with previous reports. 16 The catalyzed plasma reactor is more energy efficient overall and is particularly efficient at low T vib and intermediate P N , where the energy deposited into dissociated N 2 is most effectively directed into NO. The predicted energy consumption is two orders of magnitude lower than plasma only in this regime. The minimum predicted energy consumption is 2.9 MJ/mol NO , which exceeds the minimum energy (about 0.3 MJ/mol NO ) using established estimates of the efficiencies of known processes in a N 2 -O 2 NTP. 16,40,41 These results illustrate that the quest for optimal NO yields and optimal energy efficiency may lead to different target plasma regimes.</p><!><p>Thermal N 2 oxidation rates and thus NO yields are negligibly small at the conditions of In the plasma-catalytic experiment of Fig. 1, the thin Pt film was placed at the center of the heating mantle. The plasma-generated species can react homogeneously in the space ahead of the catalyst bed, over the bed itself, and potentially even post-catalyst. To better represent the experimental configuration and to rationalize the dependence of integral NO yield on the gas composition, we construct a series of pre-catalyst plasma reactor, plasmacatalytic reactor, and post-catalyst reactor (Fig. 6a). The first reactor is fed with N 2 -O 2 gases at P N = 2 × 10 −3 P N2 and T vib = 6000 K, measured in radio frequency plasmas at conditions similar to Fig. 1. 28,29 We predict NO yields across a wider gas composition range than the experiments in Fig. 1. is shown in Supplementary Fig. 12). As a result, the catalyst is ineffective in promoting (or inhibiting) NO production, as observed in the experiments. These results illustrate how optimal catalyst characteristics are a function of reaction conditions.</p><!><p>Conventional heterogeneous catalyst design typically focuses on catalytic rates, as those (in concert with reaction conditions) tend to determine both productivity and energy efficiency. Coupling of NTPs to heterogeneous catalysts opens the potential to increase intrinsic rates 32 and, as illustrated here and elsewhere, 42 to achieve yields that exceed the limits of bulk thermodynamic equilibrium. Quantification of these effects has been challenged by the background influence of plasma-and/or catalyst-only reactions. We illustrate here an example of an NTP-catalyst combination that yields products at conditions at which neither alone is appreciately effective, providing unambiguous evidence for the benefits of the combination in the context of nitrogen fixation to NO.</p><p>The reactor model construction here illustrates the additional degrees of freedom that enter into optimizing an NTP-catalyst combination. Under conditions at which relevant reactions occur in the homogeneous and surface-catalyzed phases, optimal product yields do not necessarily correspond with optimal surface-catalyzed rates, and yield optimization involves tuning of catalyst properties, tuning of plasma-generated species, and tuning of the coupling of the two. Here that coupling is a function of the distance and volume of the pre-catalyst space, the volume of the catalyst bed, as well as catalyst site density, which become additional design variables. Further, as shown here, yield and energy-efficiency do not necessarily correspond to the same regime of operation. These additional degrees of freedom both reflect the challenges and opportunities in selecting NTP and catalyst combinations optimally suited for target applications, especially at the small reactor scales most suited to NTPs.</p><!><p>Schematic of our plasma reactor is shown in Fig. 1 (in two modes of operation). The setup consists of an inductive coil which is connected to the matching network of a radio frequency generator (13.56 MHz, 300 W maximum power rating Huttinger PFG 300 RF).</p><p>The coil encloses a quartz tube of 40 mm outer diameter and 700 mm length, mounted on both ends to two vacuum flanges which also serve as mechanical support for quartz tube.</p><p>The reactor is equipped with one inlet (for O 2 and N 2 ) and one outlet connected with Hiden Analytical Quadrupole Mass Spectrometer HAL 201RC. The temperature of the center part of the reactor is controlled by a heating mantle.</p><p>The catalyst (porous Pt film) is prepared on yttria stabilized zirconia (YSZ) tube (Ortech, 2 mm thickness, 25 mm diameter and 245 mm length) by brush painting the organometallic Pt paste (Fuel Cell Materials) followed by a heat treatment at 900 °C for 2 h in air. The crystal structure and purity of the catalyst were determined by X-ray diffraction (XRD, Bruker, Cu Kα radiation, λ = 1.54056 Å) in the Bragg-Brentano configuration. Diffractograms were collected at a scan rate of 0.02°in the 2θ range of 20°-90°. Surface composition of the Pt film was investigated by X-ray photoelectron spectroscopy (XPS, ThermoFisher Scientific, K-alpha instrument). The surface morphology of the as prepared catalyst (Pt/YSZ) was characterized using a scanning electron microscope (FEI Quanta 3D FEG instrument) at an acceleration voltage of 3-5 keV (Fig. 1b and c). The Pt film has a porous structure and consists a network of percolated particles of the order of micron and thickness is around 14 µm. SEM micrographs before and after plasma experiments show no difference since the Pt catalyst is 15 cm far from the tail of active plasma area. This is in good agreement with the minimal temperature increase on the catalyst (i.e. 1-2 °C) upon plasma ignition. The catalyst geometrical area is 20 cm 2 , while the loading is 5 mg of Pt per cm 2 . The surface area of the catalyst was determined by hydrogen adsorption with potential deposition method 43 and it was found to be 230 nmol of Pt adsorption sites.</p><p>The experiments were performed by co-injecting and co-activating 5 mbar N 2 and O 2 (100 standard cubic centimeters per minute) by RF plasma source with plasma power as 80 W while maintaining 0 W reflected power through a tunable matching network. The calibrations to quantify the NO production and O 2 consumption were carried out by using 100 and 1000 ppm NO in He and 1% O 2 in He cylinders, respectively. In each case, the standard gas mixture was used without dilution and with He dilution in the levels of 25% and 50% keeping the flow rate constant. In all the cases, a linear relation between the signal level and amount of the gas in study, has been observed. The concentration of NO produced during plasma experiments is in good agreement (5%) with the oxygen level decrease. Experiments were repeated three times.</p><!><p>Thermal catalysis The reaction and activation energies of nitrogen oxidation were collected from literature, where the calculations were performed using DACAPO with core electrons described by Vanderbilt ultrasoft pseudopotentials and exchange and correlation effects described by the RPBE functional. 24 The standard entropies of gas molecules are from NIST-harmonic oscillator model. 45 The rate constants of adsorption, surface and desorption reactions were estimated with transition state theory. 34 The steady-state surface coverages and rates were solved with a mean-field microkinetic model, as detailed in Supplementary Methods. 46 The steady-state surface coverages described by ordinary differential equations(ODEs)</p><p>were firstly solved using a method that automatically switches between nonstiff (Adams) and stiff (BDF) solvers, as implemented in scipy.integrate.odeint in Python. The steady-state coverages were further converged with a system of algebraic equations using the coverages solved from ODE as initial guesses. Newton's method implemented in mpmath.findroot in Python was used.</p><!><p>To estimate the adsorption rate of vibrationally excited molecules, the dissociative adsorption rates were modified to be an explicit function of N 2 and O 2 vibrational states. 32 The non-Boltzmann population densities of vibrationally excited N 2 and O 2 states in the plasma were estimated using the Treanor formula at different vibrational temperatures. 47,48 The rate constant of each vibrational state are estimated individually and the overall rate constant is calculated with the summation of all vibrational states. The first ten vibrational excited states were included because of the depopulation of highly excited levels. See details in Supplementary Methods. The adsorption of N and O (Reaction 2a and 2b) were included in the plasma catalysis model. P O P O2 is set as 10 P N P N2 since O 2 dissociation fraction is observed to be about one order of magnitude higher than N 2 in N 2 -O 2 plasmas. 30,31 Plasma reactions Experimentally measured forward rate constants for reactions in the Zeldovich mechanism were used (Supplementary Table 3). 15 The backward rate constants were calculated with standard free energies to enforce thermal consistency. 44 The calculated backward rate constants are in agreement with experimental measurements. 36,37 Vibrational excitation of N 2 and O 2 are included using the same methods in plasma catalysis. Integral reactor models to predict the NO yield of plasma reactions and coupled plasma and catalysts are detailed in Supplementary Methods. In these models, we assume the flow is well-mixed in the reactor. The ordinary equation of gas compositions and surface coverages were solved</p>

ChemRxiv

The Mce3R stress-resistance pathway is vulnerable to small-molecule targeting that improves tuberculosis drug activities

One-third of the world's population carries Mycobacterium tuberculosis (Mtb), the infectious agent that causes tuberculosis (TB), and every 17 seconds someone dies of TB. After infection, Mtb can live dormant within macrophages for decades in a granuloma structure arising from the host immune response; and cholesterol is important for this persistence of Mtb. Current treatments require long-duration drug regimens with many associated toxicities, which are compounded by the high doses required. We phenotypically screened 35 6-azasteroid analogues against Mtb and found that at low micromolar concentrations, a subset of the analogues sensitized Mtb to multiple TB drugs. Two analogues were selected for further study to characterize the bactericidal activity of bedaquiline and isoniazid under normoxic and low-oxygen conditions. These two 6-azasteroids showed strong synergy with bedaquiline (fractional inhibitory concentration index = 0.21, bedaquiline minimal inhibitory concentration = 16 nM at 1 µM 6-azasteroid). The rate at which spontaneous resistance to one of the 6-azasteroids arose in the presence of bedaquiline was approximately 10 −9 , and the 6-azasteroid-resistant mutants retained their isoniazid and bedaquiline sensitivity. Genes in the cholesterol-regulated Mce3R regulon were required for 6-azasteroid activity, whereas genes in the cholesterol catabolism pathway were not. Expression of a subset of Mce3R genes was down-regulated upon 6-azasteroid treatment. The Mce3R regulon is implicated in stress resistance and is absent in saprophytic mycobacteria. This regulon encodes a cholesterol-regulated stress-resistance pathway that we conclude is important for pathogenesis and contributes to drug tolerance, and that this pathway is vulnerable to small-molecule targeting in live mycobacteria.

the_mce3r_stress-resistance_pathway_is_vulnerable_to_small-molecule_targeting_that_improves_tubercul

INTRODUCTION<!>6-Azasteroids were prepared from intermediate 1.<!>6-Azasteroids inhibit Mtb growth in combination with isoniazid.<!>MIC INH+6-aza/ MIC INH<!>6-Azasteroids improve the activity of other TB drugs and drug candidates.<!>6-Azasteroids show synergism with INH or BDQ.<!>6-Azasteroids improve bactericidal activity.<!>6-Azasteroids are more potent under low-oxygen conditions than under normoxic conditions.<!>6-Azasteroids improve the bactericidal activity of BDQ under low-oxygen conditions.<!>6-Azasteroids retain potentiation activity in cholesterol catabolism mutants.<!>The Mce3R regulon is required for potentiation of INH by 6-azasteroids.<!>Spontaneous resistance to 6a is Mce3R-independent.<!>6-Azasteroids suppress the transcription of Mtb stress-response genes including the Mce3R regulon.<!>DISCUSSION<!>The precise biochemical functions of the proteins encoded in the Mce3R<!>Synthesis of 6-azasteroid precursor 1 and the corresponding 17β-carboxylic acid.<!>Determination of MICs.<!>Aerobic time-kill assay.<!>Frequency of resistance.<!>RNA-seq transcriptional profiling and qRT-PCR.

<p>Cholesterol metabolism plays an important role in the persistence, virulence, and intracellular survival of Mycobacterium tuberculosis (Mtb). [1][2][3][4][5] Mtb is able to survive and replicate inside macrophages 6 by utilizing host-derived nutrients, including cholesterol. [1][2][3]7 Mtb catabolism of cholesterol provides a source of acetyl-coenzyme A (CoA), pyruvate, and propionyl-CoA, which can be utilized for energy production and as lipid precursors. 8 Catabolism proceeds through b-oxidation of the side chain, and oxidative cleavage of the sterol rings. 9 Comparison of transcriptional profiles of Mtb cultured with and without cholesterol identified over 200 genes that are regulated by cholesterol. 2 At least 52 cholesterol-regulated genes are clustered within the Mtb genome and encode the enzymes needed for catabolism. 2,8 Two TetR family regulators, KstR1 and KstR2, control transcription of the majority of these catabolism genes 10,11 and are de-repressed by CoA metabolites of the catabolism pathway. 12,13 Owing to the large number of potential targets in the Mtb cholesterol metabolism pathway, the protein target that would be the most vulnerable to inhibition is not clear. 14 Therefore, we undertook whole-cell phenotypic screening to identify inhibitors of Mtb growth. We reasoned that use of a steroid scaffold would bias the screen to target cholesterol metabolism. We further sought a scaffold with pharmacokinetic properties that would be advantageous for future in vivo experimentation.</p><p>For this purpose, we chose the 6-azasteroid scaffold (Figure 1). The 6-azasteroids were developed by Glaxo-Wellcome as part of a 5α-reductase inhibitor program for treatment of benign prostatic hyperplasia, but they were subsequently abandoned in favor of 4-azasteroids. [15][16][17][18] Several 6-azasteroids were found to be orally bioavailable in rats, dogs, or both and to have low in vivo toxicity. 16 Because of the pharmacokinetic properties of 6-azasteroids and their structural similarity to cholesterol, we chose to screen them as possible inhibitors of cholesterol metabolism in Mtb.</p><p>Previously, we demonstrated that 6-azasteroids with a large, hydrophobic R 1 side chain and an unsubstituted N6 atom (e.g., 4a, Figure 1) are competitive inhibitors of 3β-hydroxysteroid dehydrogenase. 19 However, because this enzyme is not essential for Mtb survival in the mouse or guinea pig models of infection, its relevance as a drug target was questionable. 20 We reasoned that other enzymes in the Mtb cholesterol metabolism pathway might also be susceptible to inhibition by compounds with the 6-azasteroid scaffold. In the present study, we identified several 6-azasteroids with anti-mycobacterial activity that improve the activity of existing anti-TB drugs, and we established a relationship between activity and side-chain structure. By investigating the mechanism of action and the target of two of the active 6-azasteroids in Mtb, we identified a connection between stress resistance and cholesterol-regulated genes that reside outside the cholesterol catabolic pathway. The inhibitors described herein offer a strategy for combating innate drug tolerance in Mtb and highlight the complex role of cholesterolregulated genes in Mtb infection. [16][17][18] and subjected to further elaboration as shown.</p><!><p>Following a previously established synthetic route, [15][16][17][18] we prepared a library of 6-azasteroids from key intermediate 1 (Figure 1), which was prepared on a 500 g scale. Briefly, hydrolysis of the methyl ester group of 1 followed by activation of the resulting carboxylic acid and substitution reactions with various primary amines provided amides 3 with more than 20 different substituents at R 1 . Subsequent deprotection of the ring nitrogen provided compounds 4a-4v. Alkylation of 4a or 4b at N6 provided compounds 5a-7a and 6b. Treatment of 4a with an excess of sodium hydride and iodomethane yielded 8a, which has a methyl group both on the ring nitrogen and at</p><!><p>We initially screened the synthesized 6azasteroids (Figure 1) to determine their minimal inhibitory concentrations (MICs) in a Mtb growth assay. As a carbon source, we used cholesterol solubilized in tyloxapol detergent micelles to ensure that we could detect activity against the cholesterol catabolism pathway. The 6-azasteroid concentrations were varied from 0.15 to 40 µM. We were unable to obtain accurate MICs, but we estimated the values to exceed 40 µM for the compounds that showed at least some activity. A clear structure-activity relationship was observed for the 6-azasteroids. Among the compounds with R 2 = H (4), an aniline side chain at R 1 was required. Compounds with side chains derived from aliphatic amines (4n, 4o, 4p, 4r, 4s) or aliphatic amines linked to aromatic groups (4m, 4q, 4t) were inactive. Furthermore, compounds with aniline side chains bearing a single substituent (4e, 4f, 4g), compounds</p><!><p>with aniline side chains bearing polar substituents (4b, 4c, 4h, 4i, 4j, 4k, 4l), and compounds with pyridyl-based side chains (4u and 4v) were inactive. Only 4a and 4d, the (2,5-di-t-butyl)phenyl and the (3,5-di-t-butyl)phenyl compounds, improved INH activity. The activity of parent compound 4a was retained upon installation of an alkyl R 2 substituent on the ring nitrogen (5a, 6a, 7a). Interestingly, moderate potentiation was observed upon conversion of the inactive 2-t-butyl-5-trifluoromethylanilinederived compound 4b to N-propyl derivative 6b. Likewise, compound 3c, which has a tbutyloxycarbamate-protected ring nitrogen, was active, whereas the corresponding unprotected compound (4c) was not. Thus, derivatization of the aza ring nitrogen was well tolerated and, in some cases, could convert inactive compounds into potentiators of INH activity. 6-Azasteroid activity is not specific for cholesterol catabolism. Next, we assessed the specificity of the 6azasteroid scaffold for cholesterol catabolism. We found that when glycerol was used as the carbon source, the seven active 6-azasteroids retained their ability to improve the efficacy of INH. As a negative control, we tested compound 4g, which was inactive with cholesterol as the carbon source, and found it remained inactive in glycerol (Figure S1). Thus, the activity of 6-azasteroids was not carbon-sourcespecific.</p><p>Two of the active compounds, 4a and 6a, were selected for further study. Their effect on the rates of CDC1551 Mtb growth was monitored by optical density (OD). As expected from our initial MIC screen, 10 µM 4a or 6a did not inhibit the growth of Mtb on glycerol as a carbon source over the course of 15 days (Figure 3A).</p><!><p>Compounds 4a and 6a were tested as potentiators of the following additional TB drugs and drug candidates: rifampin, pretomanid, bedaquiline (BDQ), clofazimine, pyrazinamide, moxifloxacin, linezolid, and ethionamide. We found that these two 6-azasteroids improved the activity of at least five of the TB drugs in vitro: rifampin, pretomanid, BDQ, pyrazinamide, and ethionamide (Tables 1 and S1A). The 6-azasteroids did not appear to improve the activity of clofazimine, linezolid, or moxifloxacin. BDQ, in addition to INH, was selected for further study because it is used for the treatment of multidrug-resistant Mtb infection.</p><!><p>Using the H37Rv(mlux) strain, we carried out a checkerboard growth-inhibition assay at various 4a or 6a and ΙΝΗ or BDQ concentrations. As expected, neither 10 nor 20 µM 4a inhibited the growth of H37Rv(mlux) (Figure 3B). A sub-MIC concentration of INH (0.032 µM) also failed to inhibit growth. However, when Mtb was treated simultaneously with 20 µM 4a and 0.032 µM INH, growth was inhibited. Similarly, at a low concentration (4 µM), BDQ alone inhibited growth by approximately 50%. Importantly, when BDQ was used in combination with 4a, complete inhibition was observed. Likewise, combinations of 6a and 0.032 µM INH or 4 µM BDQ resulted in complete inhibition (Figure 3C). Growth readings at day 9 were used to prepare an isobolographic plot for INH and 6-azasteroid (Figure S2). Although an accurate 6-azasteroid MIC could not be determined from the plots, the combinations of 4a or 6a and INH were clearly synergistic.</p><!><p>Combining 6a with BDQ (1.8 µM) improved the bactericidal activity of BDQ. Specifically, the combination of 10 µM 6a and BDQ resulted in a 2 log 10 CFU improvement in bacterial kill at day 7 of treatment. However, the rate of bacterial kill by BDQ was not increased, as evidence by the 2-day time points (Figures 3D and S3A). The combination of 4a and INH was also bactericidal (Figure S3B).</p><!><p>We also conducted a checkerboard low-oxygen recovery assay (LORA) of H37Rv(mlux) growth inhibition by 4a or 6a and ΙΝΗ or BDQ at various concentrations (Figure 4). Both 4a and 6a were active in the LORA, showing MICs of 11 and 13 µM, respectively. Although INH was not active under low-oxygen conditions (MIC » 400 µM), treatment with INH plus 4a or 6a reduced the MIC at least 2-fold (Figure 4A). The LORA fractional inhibitory concentration (FIC) indices for INH/4a and INH/6a were 0.75 and 0.69, respectively. BDQ had an MIC of 160 nM in the LORA. Compound 4a or 6a improved BDQ activity in the LORA (Figure 4B). At 4.4 µM, 4a reduced the BDQ MIC approximately 50-fold (to 3 nM), and 6a reduced the BDQ MIC to 23 nM. The FIC index for both BDQ/4a and BDQ/6a was 0.21, indicating strong synergy.</p><!><p>The bactericidal activities of 4a and 6a were determined under low-oxygen conditions. The minimum bactericidal concentration of 4a was 80 µM, and that of 6a was higher than 80 µM (80% kill). To assess the bactericidal activity of combinations of 6-azasteroids and BDQ, we performed a time-kill assay with 10 µM 4a or 6a in combination with BDQ at concentrations below the minimum bactericidal concentration (Figure 4C). At day 10 of treatment, the combination of 10 µM 4a and 0.15 µM BDQ resulted in a 3.2 log 10 CFU reduction in bacterial kill relative to that in a no-drug control group. In addition, this same combination (10 µM 4a plus 0.15 µM BDQ) resulted in a 1.4 log 10 CFU reduction in bacterial kill at day 10 relative to that in the BDQ-only group at a slightly higher BDQ concentration (0.25 µM); that is 10 µM 4a plus 0.15 µM BDQ was more potent than 1-0.25 µM BDQ alone (Figure 4C).</p><!><p>Because 6-azasteroids are cholesterol analogues, we determined whether the cholesterol catabolism pathway was required for azasteroid activity. For this purpose, we tested the activity of 6-azasteroids in combination with INH in mutant Mtb strains with disrupted cholesterol catabolism genes. Specifically, we selected strains with mutations in fadA5, chsE4 (fadE26), fadE31, and fadE33 (the transcription of which is repressed by the two main transcriptional regulators of cholesterol catabolism, KstR1 and KstR2) 10,11 and in hsd (which encodes the first enzyme in the pathway). 21 KstR1 regulates the transcription of genes encoding side- chain-and A/B-ring-degrading enzymes, whereas KstR2 regulates the transcription of genes encoding C/D-ring-degrading enzymes. These two regulators are de-repressed by CoA metabolites in the pathway. 12,13 We found that none of these five genes were required for 4a to improve INH activity against Mtb grown with either cholesterol or glycerol as the carbon source (Tables 2 and S1B).</p><!><p>Previously, we discovered that the Mce3R regulon encodes FadE17-FadE18, an unusual heterotetrameric acyl-CoA dehydrogenase that is associated with cholesterol metabolism. 22 Moreover, Mce3R regulates the mel2 operon, which is implicated in Mtb persistence in macrophages 23 and in resistance to oxidative stress 24 that is associated with lipid metabolism. 25,26 The regulon is de-repressed upon treatment with cholesterol, most likely by a cholesterol metabolite, 2 and expression of the regulon is up-regulated during hypoxia and in the stationary phase. 27 We found that azasteroid 4a no longer improved INH activity when tested against Mtb Mce3R regulon mutants fadE18, melF, and melH (Table 2), and this loss of activity was independent of carbon source.</p><p>The cholesterol catabolism pathway regulated by KstR1 and KstR2 is found in both saprophytic and pathogenic mycobacteria. 28 In contrast, the Mce3R regulon is conserved in M. marinum and Mtb but not in other pathogens (e.g., M. avium) or in saprophytic mycobacteria (e.g., M. smegmatis). In results consistent with our mutant potentiation data, we found that 6-azasteroid 4a did not improve INH or BDQ activity when tested against M. smegmatis or M. avium grown with glycerol as a carbon source (Table 2). However, inhibition of M. marinum growth by BDQ and INH was increased by addition of 4a.</p><!><p>We examined the frequency with which spontaneous resistance to one of the azasteroids arose (Table 3). Specifically, resistant mutants were raised against 6a alone and in combination with BDQ. The reported frequency of in vitro resistance to 20 µM BDQ ranges from 9 × 10 -9 to 5 × 10 −7 mutations per cell division, and mutations arise predominately in the atpE gene. 29 We found similar frequencies of BDQ resistance in this study. The frequencies of resistant mutant formation upon treatment with 6a in combination with BDQ were comparable to the frequencies of BDQ resistance. Treatment with 6a alone resulted in an approximately 5-10-fold higher frequency of resistance. All the resistant mutants that formed upon treatment of 6a alone or in combination with BDQ had significant alterations in colony morphology. The wild-type colonies were thick, rough, irregular, and yellow, whereas the mutant colonies were thin, small, round, and almost transparent (Figure S4). Twenty mutants were selected, and six of these grew in liquid culture. The mutant colonies unable to grow in liquid culture were restreaked on agar plates, and one more mutant grew. All the resistant mutants grew slowly in liquid culture or on agar plates. The 6a, BDQ, and INH MICs against these seven mutants were determined (Figure S1C). Against three of the mutants (ASR1, ASR4, and ASR5), the MIC of 6a increased more than 4-fold with respect to the MIC against WT-CDC1551, whereas 2-4 fold increase was observed for the remainder of the mutants. The INH and BDQ MICs against all the 6a-selected mutants were unchanged (1-2 fold increase) compared to their MICs against WT-CDC1551.</p><p>DNA from six of the resistant mutants was sequenced. Only three mutants (ASR1, ASR2, and ASR3) produced high-quality data; the DNA from the other three mutants was highly fragmented. Twenty-nine nonsynonymous single nucleotide polymorphisms (SNPs), occurring in 25 genes, were identified and confirmed by at least 10 Illumina reads (Figure S5). Of the 25 genes, 6 belong to the PE/PPE family and 3 encode enzymes involved in PDIM biosynthesis.</p><p>Because mutant ASR2 showed no significant increase in the 6a MIC, we focused on analyzing mutants ASR1 and ASR3, in which we identified, respectively, 21 and 11 nonsynonymous SNPs, with 3 SNPs that were identical in both mutants. A stop gain occurred at S154 in Rv2940c, which encodes a mycocerosic acid synthase involved in PDIM biosynthesis. In addition, we identified a frameshift insertion at P188 in Rv3467, a non-essential gene, and a recombination of PPE19. No Mce3R gene mutations were identified in the resistant mutants.</p><p>As for individual nonsynonymous mutations, we determined that ASR1 carried a nonsynonymous SNP in guaB3 (Rv3410c), which is an essential gene annotated as an inosine-5'-monophosphate dehydrogenase (Figure S5). However, GuaB3 does not perform this dehydrogenase enzymatic function. 30 The guaB3 gene resides in a three-gene operon Rv3409c-guaB3-guaB2, and Rv3409c is required for bacterial survival and growth in macrophages. 31 We also identified frameshift deletions in Rv0678, Rv2552c, and Rv3267 from ASR1. These genes are involved in the regulation of a cell efflux system, aromatic amino acid synthesis, and cell-wall synthesis, respectively. A frameshift deletion was also identified in Rv0204c from ASR3. The disruptions in genes involved in mycobacterial cell-wall biosynthesis are consistent with the phenotypic changes in colony morphology that we observed.</p><!><p>RNA-seq gene expression profiles of the wild-type CDC1551 Mtb and the resistant mutants ASR1 and ASR2 were generated after 6 h of exposure to 4a or 6a. Upon treatment of the wild-type with 4a or 6a, genes involved in lipid metabolism, cell-wall synthesis and cell processes, PE/PPE, and information pathways (TubercuList functional classes) were generally down-regulated (>2-fold). The global expression profiles for 4a and 6a treatment were like the profile for BDQ treatment, with >80% of the genes that were down-regulated by BDQ (>1.5-fold) also being down-regulated (>1.5-fold) by 4a. 32 In the absence of any of the test compounds, resistant mutants ASR1 and ASR2 had a common set of 49 upregulated (>2-fold) genes, including stress-response genes regulated by DosR, PhoP/R, MprA, Crp, and WhiB3. 33 In addition to the observed general response to drug treatment, six sets of genes that appeared to be specific to the 6-azasteroid mechanism of action were up-or down-regulated by treatment of wild-type CDC1551 Mtb with 4a or 6a (Figures 5 and S6). The magnitude of differential expression of these specific genes in response to treatment with 4a or 6a was much lower in ASR1 than in the wild-type, a result that is consistent with the increased drug MIC. In the wild-type CDC1551 Mtb, the KstR1 and KstR2 regulon genes that encode cholesterol catabolism enzymes were either up-regulated or unaffected (Figure 5). In contrast, KstR1 genes not directly related to cholesterol catabolism were greatly down-regulated (up to 25-fold, not shown). Additionally, KstR1regulated genes in the mce4 operon, which encodes the cholesterol transport system, were down-regulated by 4a or 6a. Genes in the mce1R operon, which is associated with mycolic acid transport, and in the mel1 operon, which encodes putative membrane protein and transglutaminases thought to be important for infection, 23 were greatly down-regulated (>5-fold). In the wild-type, the Mce3R-regulated echA13-fadE17-fadE18 operon was down-regulated by treatment with 4a or 6a, whereas the other two Mce3Rregulated operons were unaffected. In addition, guaB2, which is in the Rv3409c-guaB3-guaB2 operon, was up-regulated by 4a or 6a (Table S4). Gene expression of representative genes from Mce3R, KstR1 and KstR2 regulons and mel1 operon in Mtb CDC1551 exposed to 4a, was confirmed by qRT-PCR (Figure S6).</p><!><p>We tested 37 azasteroids in MIC assays and identified a set of eight 6-azasteroids that clearly sensitized Mtb to existing TB drugs (Figure 2). We observed a clear structure-activity relationship, indicating that the compounds acted on a specific target or targets. The TB agents that exhibited synergy with these 6azasteroids utilize a wide array of killing mechanisms, disrupting processes ranging from cell-wall biosynthesis to ATP biosynthesis (Table 1).</p><p>We found that genes involved in cholesterol catabolism were not required for the activity of 6-azasteroid 4a (Table 2), despite our previous work demonstrating that 6-azasteroids, including 4a, inhibit the first enzyme in the catabolic pathway, 3β-hydroxysteroid dehydrogenase (hsd), 19 and a body of evidence indicating that cholesterol catabolism is important for Mtb persistence and survival in a mouse model of infection. 1,2 Moreover, the 6-azasteroids were active against Mtb grown on a sugar carbon source, glycerol, as well as against Mtb grown on a cholesterol carbon source (Figure 2). Importantly, 6azasteroids 4a and 6a were active under low-oxygen growth conditions. Taken together, the evidence suggests that 6-azasteroids act on a target that contributes to drug tolerance regardless of carbon source or oxygen level.</p><p>Transcriptional, 2 phenotypic, 34 and biochemical profiling 22 have identified genes outside the Mtb cholesterol catabolism gene cluster that are regulated by cholesterol, that are important for growth on cholesterol in vitro, or that have cholesterol-specific structural motifs. Several of these genes are found within a TetR-like transcriptional regulon controlled by the Mce3R repressor. 26,27 Transposon disruption of mce3R significantly inhibits growth of Mtb on cholesterol. 34 However, the genes within the Mce3R regulon are not required for catabolism of cholesterol, 8,9 suggesting that these genes have an alternate role in Mtb survival. Therefore, we investigated the importance of four genes in the Mce3R regulon (fadE18, echA13, melF, and melH) in 6-azasteroid activity. Intriguingly, fadE18, melF, and melH were required for 6-azasteroid potentiation of INH activity. Although echA13, annotated to encode an enoyl-CoA hydratase, was not required, the Mtb genome encodes many other enoyl-CoA hydratases that might compensate for disruption of this gene.</p><!><p>regulon have yet to be established. However, bioinformatics and phenotypic assays allow for tentative assignment of their function. The fadE17, fadE18, and mel2 genes are important contributors to Mtb resistance to various cellular stresses. The entire Mce3R locus is present in M. marinum and provides resistance to reactive oxygen species and reactive nitrogen species. 24 Significantly, the Mce3R regulon is absent in saprophytic mycobacteria, 28 which suggests that the primary role of the regulon is mediating host-derived stresses, such as those encountered by Mtb in activated macrophages. Consistent with this observation, the mel2 operon is required for persistence and dissemination of Mtb infection in mice. 35 The fadE17 and fadE18 genes encode an acyl-CoA dehydrogenase with an unusual α 2 β 2 heterotetrameric that is characteristic of acyl-CoA dehydrogenases that oxidize cholesterol-derived substrates. 22 The melF and melH genes reside in a single operon (Rv1936-Rv1941) of the mel2 locus, which encodes homologs of the Lux luminescence system. This system catalyzes formation of a fatty acid aldehyde by means of an ATP-dependent process.</p><p>The mel2 operon is thought to provide resistance to oxidative stress, which is consistent with the idea that the operon encodes catalytic machinery to generate a fatty acid aldehyde that can remove oxidizing species from the cellular milieu. 24 When Mtb production of ergothioneine, which is a redox couple in Mtb, is disrupted, fadE18 is up-regulated. 36 When Mtb is exposed to a lysosomal soluble fraction prepared from activated macrophages, both fadE17 and fadE18 are induced. 37 In addition, a recent study has shown that cysteine in combination with INH or rifampin can enhance oxygen consumption by Mtb, thereby leading to increased production of reactive oxygen species and sterilization of Mtb cultures. 38 Upon addition of cysteine to INH-treated Mtb cultures, the transcription level of fadE18 increases nearly 4-fold. Treatment of Mtb with BDQ for 48-96 h results in up-regulation of mel2 genes, especially melF, which is upregulated about 8-fold. 32 These responses indicate that the Mce3R regulon plays a critical role in the Mtb response to oxidative, host-induced, or drug-induced stress.</p><p>In contrast to treatment with INH, rifampin, or BDQ, treatment with 6-azasteroid 4a or 6a for 6 h resulted in 3-to 4-fold down-regulation of the Mce3R-regulated echA13-fadE17-fadE18 operon. Although the transcription levels of mel2 genes were unchanged or only marginally changed after 6 h of 6-azasteroid treatment, the response of the mel2 gene to BDQ treatment is likely to be indirect, given the long post-BDQ treatment time (48-96 h) at which up-regulation is observed. 32 Together with our mutant experiment in which fadE18 was required for 6-azasteroid potentiation, the evidence supports a mechanism in which the 6-azasteroids suppress fadE18/Mce3R-dependent activation of a drug-induced stress-resistance pathway.</p><p>Mtb spontaneous resistant mutants that were raised against 6a showed a thinner, almost transparent cellwall phenotype and were difficult to culture. Importantly, we found that Mce3R genes were not directly involved in spontaneous resistance to the 6-azasteroid. Among the genes where nonsynonymous mutations occurred in the resistant mutants, PE/PPE genes were not investigated further as candidates for involvement in the azasteroid-resistance phenotype, because of their high variability in Mtb. In addition, it is unlikely that genes involved in PDIM biosynthesis are directly involved in the azasteroid-resistance phenotype, because the loss of PDIM production is commonly observed during in vitro propagation of Mtb cultures. 39 The remainder of the mutations identified were in genes involved in mycobacterial cellwall synthesis and efflux systems, which is consistent with a mechanism of resistance that decreases 6azasteroid uptake and simultaneously the growth fitness of Mtb is reduced.</p><p>In addition to suppressing Mce3R genes, 6-azasteroids also suppressed the expression of genes in the DosR, 40 PhoP/R, 33 and SigB regulons 41 and in the icl1 operon. 42 All these genes are involved in Mtb stress resistance and establishment of persistence. There is evidence that inhibitors of DosR genes also increase the activity of INH, 43 although the increase is not as dramatic as that observed for 6-azasteroids. This enhancement of INH activity by the 6-azasteroids indicates that they inhibit a network of Mtb stress resistance and that the disruption of this network leads to the potentiation of TB drugs.</p><p>Co-drug sensitization is emerging as a useful strategy for treating TB, including multidrug-resistant TB. [44][45][46][47] Variations on this strategy include targeting specific drug-activation pathways, 46 as well as more general drug-desensitization targets such as the DosRST regulon 43 and thiol stress. 47 Thus, these drugsensitization strategies are a powerful tool for killing resilient and/or nonreplicating Mtb.</p><p>In summary, our work further supports the idea that Mce3R-regulated genes are important for managing the Mtb cellular response to drug-induced stress (Figure 6). Moreover, our findings suggest that Mtb depends on cholesterol or its metabolites for activation of stress resistance through the Mce3R regulon (Figure 6). Our discoveries that 6-azasteroid suppression of drug tolerance depends on Mce3R-regulated genes and occurs under low-oxygen conditions present an intriguing avenue for further development of co-drugs that can improve the efficacy of existing TB drugs in vivo.</p><!><p>17β-Carbomethoxy-6t-butoxycarbonyl-6-azaandrost-4-en-3-one (1) was synthesized as reported previously [16][17][18] and then converted to 17β-carboxy-6-t-butoxycarbonyl-6-azaandrost-4-en-3-one as described in the literature. 17 Both compounds were assessed to be greater than 95% pure by 1 H NMR spectroscopy.</p><p>General procedure for coupling amines to the 17β-carboxylic acid and subsequent deprotection of the ring nitrogen. 17 17β-Carboxy-6-t-butoxycarbonyl-6-azaandrost-4-en-3-one (30 mg) was suspended in 1 mL of toluene. One drop of DMF and 15 µL of pyridine were added to the solution, which was then cooled in an ice bath, treated with 10 µL of thionyl chloride, and stirred for 1 h. The solution was filtered, and the filtrate was concentrated. The resulting acid chloride residue was dissolved in DCM (5 mL) and treated with the desired amine to give amide 3. Amide 3 was dissolved in DCM (1 mL) and treated with 2 mL of TFA at 20 °C to remove the BOC protecting group from the ring nitrogen. After 2 h, the reaction mixture was poured into saturated NaHCO 3 solution, the layers were separated, and the organic layer was washed with brine and dried over Na 2 SO 4 . Chromatography using 5% (v/v) MeOH in DCM provided deprotected amide 4.</p><p>Procedure for N6 alkylation of deprotected amides 4a and 4b. 8,18 Amides 4a and 4b (30 mg) were dissolved separately in THF (5 mL) and treated with 1.2 equiv of NaH. After stirring for 30 min at 20 °C, the reaction mixture was treated with the desired iodoalkane (1 equiv) for a further 30 min. When the reaction was complete as judged by TLC (approx. 5 h), 30 mL of EtOAc was added, and the resulting solution was washed with water and brine (3× each) and dried over Na 2 SO 4 . Chromatography using 5% (v/v) MeOH in DCM provided N6-alkylated amides 5a, 6a, 6b, and 7a. For the synthesis of 8a, amide 4a (30 mg) was dissolved in DMF and treated with 3 equiv of NaH. After 30 min at 20 °C, 5 equiv of iodomethane was added into the reaction, followed by stirring at 20 °C for 2 h to give compound 8a.</p><p>Active azasteroids 4a, 5a, 6a, 8a, and 3c were resynthesized for verification and were determined to be greater than 99% pure by LC-MS.</p><p>Bacterial strains and culture conditions. Mutant and complemented strains used in this study and their sources are listed in Table S2. Mtb strains CDC1551, H37Rv, and H37Rv(mlux), also known as H37Rv(pFCA-luxABCDE), were used as wild-type strains for this study. Mtb strains were cultured at 37 °C either in Middlebrook 7H9 medium (broth) supplemented with 0.2% glycerol or 0.1 mM cholesterol, 0.5% BSA, 0.08% NaCl, and 0.05% (v/v) tyloxapol (7H9/glycerol or 7H9/cholesterol medium) or on 7H11 medium (agar) supplemented with 10% oleate-albumin-dextrose-NaCl (OADC), 0.5% glycerol, and 0.05% Tween 80.</p><!><p>MICs were determined by means of a broth microdilution assay. 48 Briefly, cells were grown to mid-log phase and then diluted 1000-fold in defined media. Cell suspension was added to a 96-well plate containing compound dilutions to obtain a final volume of 100 µL. Plates were incubated at 37 °C for 14 days, and MICs were determined as the lowest concentration that resulted in complete inhibition of growth by visual inspection or luminescence readouts (mlux). For isobolograms, 5-fold serial dilutions of INH or BDQ were used so that the concentration range in a single plate was sufficient. If the 90% reduction in RLU could not be determined directly, a linear interpolation between two RLU values was used to determine the concentration at 90% inhibition.</p><p>LORA. 49 A low-oxygen-adapted culture of H37Rv(mlux) Mtb expressing a luciferase was used for the LORA. H37Rv(mlux) Mtb stored at -80 °C (1 × 10 5 CFU/mL) was thawed and exposed to 2-fold or 3fold serial dilutions of test compound diluted in 7H12 broth in black 96-well plates, which were incubated for 10 days at 37 °C in an anaerobic jar under hypoxic conditions created with an Anoxomat system (MART Microbiology). Luminescence readouts were obtained after 28 h of normoxic recovery in 5% CO 2 . To calculate the MIC, the dose response curve was plotted as percentage growth and fitted to the Gompertz model. The MIC was defined as the lowest concentration inhibiting recovery of the luminescence signal by 90% relative to bacteria-only controls. FIC is defined as the ratio of the MIC of the inhibitor when used in combination to the MIC of the inhibitor alone. FIC index is the sum of the FICs for the two drugs used. The minimum concentrations that worked in combination were used. The determination of drug MICs for rifampin, INH, linezolid, pretomanid, and BDQ were conducted as positive controls.</p><p>LORA time-kill assay. The low-oxygen-adapted Mtb strain H37Rv(mlux) was cultured as described above for the LORA. Compounds 4a and 6a were tested at 10 µM in combination with BDQ at 0.5, 0.25, 0.15, 0.0625, and 0.05 µM. At 0, 7, and 10 days, cultures were recovered in an aerobic environment (5% CO 2 ) for 28 h and then were serially diluted in 7H9 broth and plated on 7H11/OADC agar plates. Plates were incubated at 37 °C, and CFUs were counted after 3-4 weeks. Statistical analysis was done using an unpaired Student's t-test (GraphPad Prism 4).</p><!><p>Mtb was grown at 37 °C to mid-log phase and then diluted in fresh medium to 5 × 10 5 CFU/mL. Test compounds were added at defined concentrations. Aliquots of cultures were withdrawn at specified time points, and either OD (600 nm) (CDC1551) or luminescence (mlux) was recorded. At 0, 2, and 7 days, aliquots of cultures were serially diluted in 7H9 broth and plated on 7H11/OADC agar plates. Plates were incubated at 37 °C, and CFUs were counted after 3-4 weeks.</p><!><p>Mtb mutants resistant to 6a were isolated by means of the procedure reported by Ioerger et al. 50 H37Rv Mtb bacteria were grown at 37 °C to mid-log phase and then diluted in fresh Middlebrook 7H9 medium containing ADC-Tween 80 to 5 × 10 8 CFU/mL. Middlebrook 7H11/OADC agar plates with 6a at 200 or 400 µM with or without 10 or 20 µM BDQ were inoculated with 10 8 , 10 7 , 10 6 , and 10 5 CFU/plate, and the plates were incubated at 37 °C for 3-4 weeks. Resistance was tested by measuring the MICs of 6a, BDQ, and INH. The frequency of the appearance of resistant mutants was calculated.</p><p>Whole genome sequencing. Mtb strains CDC1551, ASR1, ASR2, and ASR3 were grown to log phase, and their genomes were extracted with cetyltrimethylammonium bromide and lysozyme as described in the literature. 51 Whole genome preparation, sequencing, assembly, and pairwise analysis were performed as previously described. 52 Briefly, DNA was sheared into ~20,000 bp fragments using Covaris G-tube spin columns and was end-repaired before being ligated to SMRTbell adapters (Pacific Biosciences). The resulting library was treated with an exonuclease cocktail to remove unligated DNA fragments and was size-selected on a Sage Science BluePippin system to obtain fragments of ≥7000 bp. The P5-C3 sequencing enzyme and chemistry were used to sequence the resulting libraries on the Pacific Biosciences (PacBio) RS II platform. Resulting PacBio sequencing data were assembled using HGAP3 (ver. 2.2.0). For Illumina sequencing, genomic DNA (1 µg) was sheared to achieve ~200 bp fragments using a Bioruptor Pico sonicator (Diagenode). Library preparation was performed using the end repair, A-tailing, and adaptor ligation NEBNext DNA library prep modules for Illumina from New England Biolabs, according to the manufacturer's protocol. The resulting libraries were sequenced on the Illumina HiSeq 2500 platform. Illumina reads were then mapped to the curated PacBio assemblies using samtools mpileup 53 to correct PacBio sequencing errors. Genome circularization, curation, and annotation were performed with a custom postassembly pipeline (https://github.com/powerpak/pathogendb-pipeline). 52 Finally, NUCmer (ver. 3.1) 54 was used for aligning mutant genome strains to the PS00103 reference genome to identify genetic variants. Whole genome sequencing data are available in the NCBI BioProject database under accession number PRJNA482894.</p><!><p>Mtb strains CDC1551, ASR1, and ASR2 were grown to OD ~0.6 and treated with 30 µM 4a or 6a or with vehicle control for 6 h in Middlebrook 7H9 medium supplemented with 0.2% glycerol, 0.5% BSA, 0.08% NaCl, and 0.05% (v/v) tyloxapol. Total RNA was extracted using TriZol, with chloroform back extraction and 70% ethanol precipitation. RNA was purified with an RNeasy kit (Qiagen) with DNAse treatment. Total RNA was processed for ribosomal reduction library construction. Libraries were sequenced as single-end 75 bp reads on an Illumina NextSeq500 sequencer following the manufacturer's protocols (Cofactor Genomics, Inc.). The sequence reads were aligned to the CDC1551 Mtb complete genome (The National Center for Biotechnology Information Database) using STAR (https://github.com/alexdobin/STAR). For each transcript or patch, the reads per kilobase million (RPKM) expression value was calculated for each sample. For each replicate group, the mean and coefficient of variation for each transcript or patch were calculated across the expression values for the samples in that group. These means were considered to be the expression values for the replicate group. P-values were calculated for comparisons between the means of each pair of replicate groups using a Welch's t-test corrected for false discovery rate by means of the Benjamini-Hochberg procedure. For each comparison, differentially expressed genes were identified as genes with an average normalized count of >100, differential gene expression of >2-fold, and a P-value of <0.05. Three biological replicates were performed. RNA-seq data are available in the NCBI GEO database under accession number GSE118482. Total RNA was reverse-transcribed using PrimeScript RT Master Mix (TaKaRa Bio). The resulting cDNA was used for PCR amplification using iTaq Universal SYBR Green Supermix (Bio-Rad).</p><p>Relative level of gene expression was calculated by the DDCq method with 16s rRNA as an internal control. Primers are listed in Table S3.</p>

ChemRxiv

Modular and Versatile Hybrid Coordination Motifs on \xce\xb1-Helical Protein Surfaces

We report here the construction of phenanthroline (Phen) and terpyridine (Terpy) based hybrid coordination motifs (HCMs), which were installed on the surface of the four-helical bundle hemeprotein, cytochrome cb562. The resulting constructs, termed HPhen1, HPhen2, HPhen3 and HTerpy1, feature HCMs that are composed of a histidine ligand and a Phen or Terpy functionality located two helix turns away, yielding stable tri- or tetradentate coordination platforms. Our characterization of the tridentate HCMs indicates that they accommodate many divalent metal ions (Co2+, Ni2+, Cu2+, Zn2+) with nanomolar to femtomolar affinities, lead to significant stabilization of the \xce\xb1-helical protein scaffold through metal-mediated crosslinking, assert tight control over protein dimerization, and provide stable and high-affinity binding sites for substitution-inert metal probes. Our analyses suggest that such tridentate HCMs may be used modularly on any \xce\xb1-helical protein surface in a sequence-independent fashion.

modular_and_versatile_hybrid_coordination_motifs_on_\xce\xb1-helical_protein_surfaces

Introduction<!>Construction of cyt cb562 variants with Quin, Phen and Terpy-bearing HCMs<!>Metal Binding Properties of Phen- and Terpy-based HCMs<!>Metal-mediated protein stabilization through Phen- and Terpy-based HCMs<!>Metal-Dependent Self-Assembly Properties of HPhen1<!>Using Phen-bearing HCMs for protein functionalization<!>Effects of Intervening Residues in i/i+7 HCMs<!>Conclusion<!>Materials and Methods<!>Mass Spectrometry<!>Site Directed Mutagenesis<!>General Protein Expression and Purification Protocol<!>Synthesis of Iodoacetic anhydride<!>Synthesis of 5-Iodoaceamido-1,10-phenanthroline (IA-Phen)<!>Synthesis of Iodoacetamido-8-hydroxyquinoline (IA-Quin)<!>Synthesis of 4-Iodoacetamido-2,2\xe2\x80\xb2:6\xe2\x80\xb2,2\xe2\x80\xb3-terpyridine (IA-Terpy)<!>General protocol for functionalization of cyt cb562 variants with Phen and Terpy chelates<!>Metal Binding Titrations<!>Chemical Unfolding Studies<!>Thermal Denaturation<!>Sedimentation Velocity Experiments<!>DFT Calculations<!>

<p>The primary biological roles of metals including catalysis, electron transfer and structural stabilization are generally established once they are firmly placed within a protein scaffold.1 Owing to the stability of the resulting complexes, the interactions between metals and the interiors of proteins are readily characterized and have justifiably formed the focus of Bioinorganic Chemistry. One could argue, on the other hand, that metals spend a good majority of their time interacting with protein surfaces, and that such transient, harder-to-characterize interactions carry in vivo and in vitro consequences that rival those of metal-protein interior interactions. The prevalence of metal-protein surface interactions become especially clear when picturing the behavior of metal ions and complexes within the crowded cellular environment, for example, as they are being passed on from one specific protein (e.g. a metallochaperone)2 to another, or as they crosslink together multiple proteins whose aggregation may have dire consequences.3 Similarly, outside the cellular realm, metal-protein surface interactions form the basis of immobilized metal ion affinity chromatography (IMAC)4 as well as the functionalization of protein surfaces with metal complexes that have served as invaluable spectroscopic and functional probes.5, 6 Given such broad importance and utility of metal-protein surface interactions, we envisioned a need for metal coordinating motifs that would enable a better control of inorganic chemistry on protein surfaces.</p><p>Our original interest in metal-protein surface interactions stems from our desire to use metal coordination chemistry to direct protein-protein interactions (PPIs)7 and protein self-assembly more predictably and readily than computational design approaches. One caveat to utilizing metal coordination to control PPIs is the presence of numerous metal binding sidechain functionalities on any given protein surface, which bring about the challenge of controlling metal localization. In one strategy to circumvent this challenge, we introduced a Cys-specific bidentate non-natural chelate (5-iodoacetamido-1,10 phenanthroline, IA-Phen) onto the surface of a four-helix-bundle protein, cytochrome cb562, which led to the Ni2+-driven formation of an unusual triangular protein architecture.8 More recently, to exert more control over metal localization as well as metal-directed protein self-assembly, we used another bidentate chelate (5-iodoacetamido-8-hydroxyquinoline, IA-Quin) attached to a Cys (C70) in combination with a His (H63) located two helix turns away on the cyt cb562 surface, yielding the construct HQuin1 (Figure 1a).9 The resulting i/i+7 hybrid coordination motif (HCM) was shown to coordinate various divalent metal ions in a tridentate fashion, which led to: 1) high affinity divalent metal binding with dissociation constants (Kd's) ranging from nM to fM, 2) stabilization of the protein scaffold via metal-mediated crosslinking of a two-helix turn segment, and 3) tight control over protein dimerization via an octahedral metal coordination geometry. Several potential applications arise from these advantages including site-selective labeling of proteins with metal probes, improved protein separation with IMAC, stabilization of small helical peptides for pharmaceutical purposes, and manipulation of cellular pathways that depend on protein dimerization.</p><p>Given such possibilities and the ease of constructing an HCM via iodoacetamide (IA)-Cys coupling, we have sought to examine in the present study whether the advantageous properties of the HQuin1 HCM are generalizable, i.e., whether the i/i+7 HCMs that consist of a His and a non-natural chelating ligand can be utilized in a modular fashion on any α-helical surface. Towards this end, we have created a series of additional cyt cb562-based constructs (Figure 1b-e), which have been functionalized with various non-natural chelates (Figure 2): a) HPhen1, the phenanthroline(Phen)-derivatized counterpart of HQuin1, was constructed to probe the generality of the non-natural component; b) HPhen2, which features the opposite placement of His and the Cys-Phen group as in HPhen1 (Cys63 in the i and His70 in the i+7 position), was constructed to test the sensitivity of the i/i+7 HCM to the relative placement of the natural and non-natural ligands, c) HPhen3, which has the HCM motif located elsewhere on the cyt cb562 surface (His70, Cys77), was generated to test the sensitivity of the HCM to location, and, d) HTerpy1, the terpyridine-derivatized counterpart of HQuin1 and HPhen1, as a tetradentate HCM motif. We present here the characterization of these constructs in terms of their metal binding affinities, metal-dependent stabilization and metal-dependent oligomerization properties (Figure 3). Our results suggest that the i/i+7 HCMs may be modularly utilized on any α-helical protein surface towards a number of applications.</p><!><p>One requisite for expanding the biological and chemical utility of HCMs is to demonstrate the modularity of the non-natural metal chelator within the HCM system and the ease of its incorporation. To this end, we site-specifically labeled cyt cb562 variants bearing a single surface Cys residue with IA-derivatized versions of the ubiquitous metal chelators Phen, Terpy and Quin to create the HCM variants shown in Figure 1. These non-natural ligands were chosen because they are commercially available as – or easily converted to – amino-functionalized precursors, their coordination chemistry has been extensively studied, and they represent a small but diverse set if ligands with variations in denticity and overall charge.</p><p>The amino-precursors of Phen, Quin and Terpy were converted in a one-pot reaction with iodoacetic anhydride or iodoacetyl chloride into IA-derivatives with 60–75% yield. Although IA-Phen, IA-Quin and IA-Terpy are sparingly soluble in water, they are easily introduced into cyt cb562 solutions after being solubilized in DMF and DMSO; we have found no adverse effects of these organic solvents on cyt cb562 up to a final volume fraction of 50% versus H2O. Cys functionalization reactions proceed rapidly and specifically (provided that the solution pH is kept below 8 to prevent Lys labeling), with overall yields of modification ranging from 60% for IA-Terpy to 95% for IA-Phen after purification. In the case of cyt cb562, the functionalized products are facilely separated from non-functionalized protein using anion-exchange chromatography (Figure S1.1).</p><!><p>We had previously examined the divalent metal-binding properties of HQuin1 and confirmed that the i/i+7 His-Quin HCM was able to coordinate metals in a facial, tridentate geometry.9 Here, we have performed similar metal-binding titrations for HPhen1, HPhen2 and HPhen3 using late first-row transition metals (Co2+, Ni2+, Cu2+, Zn2+) to probe if the Phen functionality behaves similarly to Quin in the context of an HCM. It is important to note that the relative positions of the coordinating atoms to the point of protein attachment in the Phen derivative are equivalent to those in the Quin derivative (Figure 2). Metal binding by the His-Phen HCMs were monitored by the distinct 10-nm red-shift in the π-π* absorption band for Phen (metal-free λmax= 272 nm; metal-bound λmax= 282 nm) upon metal binding (Figure 4). It was confirmed through CD spectroscopy that the α-helical fold is not significantly perturbed by metal binding to the HCMs (Figure S1.19).</p><p>As in the case of HQuin1, it was quickly established that Phen-based HCMs bind all tested divalent metals very tightly, which required all titrations to be performed in the presence of ethylene glycol tetraacetic acid (EGTA) as a competing ligand. Due to the inherent ability of Phen-based HCMs to undergo metal-mediated dimerization (Figure 3b), protein concentrations were kept sufficiently low (< 5 μM) to minimize dimer formation. In all cases (HPhen1-3 and all metals), the metal binding isotherms were satisfactorily described by a 1:1 binding model (Figures 5, S1.2 and S1.3).</p><p>An analysis of the determined dissociation constants (Table 1) reveals that all Phen based HCMs display a significant increase in affinity over free Phen for the late first row transition metals, which strongly suggests the participation of the His component of the HCMs in metal binding. Moreover, the affinities of HPhen1, 2 and 3 for divalent metals are similar and vary at most by six-fold, indicating that metal binding ability is not very sensitive to helix location or relative orientation of the HCM (see below for a discussion on the possible effects of intervening residues).</p><p>In addition to the HPhen variants, we also investigated whether HTerpy1 (Figure 1e) can engage both Terpy and His in a tetradentate coordination motif. Our metal binding titrations and sedimentation velocity (SV) experiments reveal that HTerpy1 almost exclusively forms a stable dimer with a saturation point reached upon addition of half an equivalent of M2+ (Figure S1.4), which has precluded the determination of the HTerpy1 metal binding affinities. While protein unfolding studies (see below) show evidence for metal coordination by both His and Terpy, the unstrained, facial coordination geometry observed in HQuin1 and HPhen variants cannot be accommodated by the large Terpy group, leading to the formation of the thermodynamically and kinetically stable bis-Terpy adduct involving two proteins.</p><!><p>We next sought to determine if the Phen- and Terpy-based HCMs would have any stabilizing effect on the protein scaffold. Since HCMs crosslink a ~7-Å long, two-helix-turn segment of cyt cb562 through metal coordination, an increase in the global stability of the protein should be expected. Metal crosslinking of both natural and non-natural residues at i/i+4 positions has extensively been shown to induce α-helicity in peptides and significantly stabilize helical protein structures.11, 12 Likewise, covalent cross-linking of sidechain functionalities in i/i+4, i/i+7 or i/i+11 positions can lock small peptides in α-helical conformations,13–15 which in turn have proven to be promising pharmaceutical agents that effectively disrupt protein-protein interactions and exhibit increased resistance to proteases in vivo.16</p><p>In order to investigate the cross-linking ability of HCMs in the presence of metals, chemical and thermal unfolding studies were undertaken. In a typical chemical unfolding experiment, a solution of folded HPhen or HTerpy variant was titrated with increasing amounts of unfolded protein solution prepared in 8 M guanidine hydrochloride (GuHCl). The folding/unfolding transition was followed by CD spectroscopy, monitoring changes in ellipticity at 222 nm. Thermal unfolding measurements spanning 298 to 373 K were similarly monitored at 222 nm; because of the high stability of all variants, 1.5 M GuHCl was included in each sample to ensure that complete unfolding took place before 373 K. In both chemical and thermal unfolding experiments, metals were present in large excess over protein to ensure full occupancy of HCMs, thereby preventing metal-induced dimerization.</p><p>The stability of all HPhen and HTerpy variants tested was found to increase in the presence of divalent metal ions. Figure 6 shows representative unfolding titrations of the variants, each of which display a particularly enhanced stability in the presence of Ni2+ (for other metals and thermal titrations, see Figures S1.5–9); a complete set of results is given in Table 2. At least in the case of the HPhen variants, we attribute the superior stabilizing effect of Ni2+ over other metals to the formation of an unstrained, facial coordination geometry by the His-Phen HCM, which was previously shown to be the case for the His-Quin HCM.</p><p>In order to establish that the observed protein stabilization is due to metal-mediated, intrahelical crosslinking, we carried out the unfolding titrations of HPhen1 and HTerpy 1 at pH 5.5, where the His component of the HCM should be partially protonated and unable to fully coordinate metals (Figures 7a and 7b). Additional unfolding titrations were performed for variants of HPhen1 and HTerpy1, where either the Phen or the Terpy moiety is replaced by a carboxymethyl group (CM-G70C cyt cb562, Figure S1.10) or the His63 residue is mutated to Ala (APhen1 and ATerpy1, Figures 7c, 7d and S1.11). The results indicate that, in all cases, Ni-induced stabilization is significantly diminished, confirming the involvement of both His and Phen (or Terpy) in metal coordination.</p><p>It is tempting to link the thermodynamics of metal binding by the HCMs (Table 1) to that of metal-induced protein stabilization (Table 2). Nevertheless, such a correlation is complicated by the fact that net protein stabilization is a function of metal binding not only to the folded but also to the unfolded state, which may display multiple modes of metal coordination (thus deviating from a two-state system). We therefore have avoided presenting free energies of unfolding – which assumes a two-state process – for our variants in the presence of metals. A good case in point is Cu2+, which displays by far the highest affinity for any HCM, yet leads to the smallest extent of stabilization (Table 2). Regardless, the protein unfolding titrations clearly indicate that: 1) all Phen-based HCMs lead to a measurable metal-induced increase in protein stability, 2) this stabilization is not specific to a particular HCM location or orientation, and finally, 3) the Terpy-His HCM displays a diminished stabilizing effect due likely to an unfavorable metal coordination geometry.</p><!><p>The ability to control protein self-assembly, both temporally and spatially, is an intensely pursued goal that is complicated by the necessity to design extensive molecular surfaces.17, 18 Particularly challenging is to direct the self-assembly of proteins into discrete shapes that can recognize biological targets. One of the most exciting findings about HQuin1 was its ability to specifically dimerize upon Ni2+ binding into a rigid architecture that was shaped appropriately to bind major grooves of a double-helical DNA.9 Since HPhen1 is the closest in composition and chemical behavior to HQuin1 which we had already structurally characterized, we chose to explore its metal-dependent self-assembly properties as a representative of all Phen-bearing variants.</p><p>Sedimentation velocity (SV) experiments reveal that HPhen1 readily dimerizes in the presence of half an equivalent of Ni2+ with a sedimentation coefficient of 2.6 S similar to that of the Ni:HQuin12 complex (Figure 8a).9 The dissociation constant for the Ni:HPhen12 dimer (Kd (2mer-1mer)) was determined by sedimentation equilibrium (SE) experiments to be ~ 9 μM, which is lower than the Kd (2mer-1mer) of 42 μM for Ni:HQuin12 (Figure 8b Figure S1.12).9 Significantly, the dimeric stability of Ni:Phen12 now closely approximates that of bZIP family transcription factors which use peripheral leucine zippers domains for dimerization, with Kd's in the low micromolar range.19</p><p>Complete structural characterization of the Ni:HPhen12 dimer has remained elusive to this point. However, we project – based on the similarities between HPhen1 and HQuin1 and the fact that dimerization in both cases is entirely dictated by metal coordination – that the structure of Ni:HPhen12 should closely resemble that of the Ni:HQuin12. In the case of Ni:HQuin12, it was determined through density functional theory (DFT) calculations that the most favored inner-sphere coordination arrangement would pose the Quin groups cis to one another in Λ configuration,20 whereby the two phenolate Quin oxygens would lie trans to each other, which is also the crystallographically observed configuration.9</p><p>We performed similar DFT (BP86 and OLYP) calculations on Ni:HPhen12, where we investigated the relative energies of two possible inner-sphere arrangements for the His-Phen HCM: one that presents Phen ligands cis to one another (cis-Phen) and one that presents Phen ligands in a trans configuration (trans-Phen) (Figure 9). These calculations suggest that the cis-Phen arrangement of cis- Phen arrangment is ~5.2 kcal/mol more stable than the trans-Phen arrangement. In the case of Ni2+:HQuin12, the higher stability of cis-Quin isomer was attributed to the trans-directing effect of the imine ligands which would render a mutual trans orientation of the weaker-field phenolate ligands the least destabilized configuration. With the N, O groups of Quin now replaced with the N, N groups in the Phen ligands, this argument cannot be made to explain the higher stability of the cis-Phen arrangement. Instead, a close inspection of the calculated structures reveals that in the trans-Phen arrangement, there would be considerable steric clashes between the Phen hydrogens that lie on the Ni equatorial plane, which would be relieved in the cis-Phen arrangement.</p><p>Taking together the DFT results and solution studies, we conclude that the i/i+7 His-Phen ICM would yield a Ni2+-induced V-shaped dimer that is equivalent to the crystallographically characterized Ni:HQuin12 architecture (Figure 10).9 Although the reason for the specific formation of this V-shaped structure is different for His-Quin and His-Phen HCMs, both examples demonstrate that self-assembly of proteins can be programmed through a simple consideration of inner-sphere metal coordination, which is far more facile than designing extensive protein interfaces to the same end.</p><!><p>Metal complexes site-specifically attached to protein surfaces have proven to be invaluable functional reporters. Among these, Ru-, Os- and Re-polypyridyl derivatives have been widely used due to their photophysical and photochemical properties.21, 22 Similarly, bifunctional, As-based fluorescent reporters have been designed to specifically bind bis-Cys patterns on proteins and are finding increasing use as target selective in vivo reporters.6 Given the high affinity of Phen-based HCMs for divalent metals and their two-point attachment to the protein scaffold, we envisioned that they could provide stable and specific target sites for functional metal-based probes on α-helical proteins. Moreover, we surmised that if such probes are based on substitution-inert metals, they could result in the improved and irreversible stabilization of α-helical proteins/peptides and may be of value in terms of constructing helical peptide-based pharmaceutical agents. To investigate such possibilities, we explored the interactions of HPhen1 with a p-cymene-capped Ru(II) compound. We chose this particular piano-stool complex as a test case, because the Ru center is capped with an arene group (p-cymene) which should prevent protein dimerization and accommodate facial binding by the His-Phen HCMs. Additionally, it is commercially available in a dimeric, chloro-substituted form (1), and weakly luminescent when bound to a polypyridines.23</p><p> </p><p>In a proof-of-principle study, a solution of HPhen1 was treated with 5-fold molar excess of compound 1 (i.e., 10-fold excess Ru) dissolved in DMSO and stirred at room temperature for ~ 4 days. Reactions were quenched by removing unreacted 1 via gel filtration and subsequently purified by ion exchange chromatography. The FPLC chromatogram and corresponding mass spectra indicate that the only major product of the reaction is HPhen1 bound to a single Ru(p-cymene) adduct (Figure 11), with no discernible unlabeled or multiply labeled species. The absorbance spectra of the isolated Ru(p-cymene)(HPhen1) complex features the expected shift to ~286 nm in the Phen π-π* band due to metal binding as well as a new band at 326 nm (Ru(II) → π* arene MLCT) contributed by Ru-adduct (Figure S1.13 and S1.14). When excited at 326 nm, Ru(p-cymene)- HPhen1 displays a weak emission band centered at 442 nm. Both the absorbance (Figure S1.14) and the emission (Figure S1.15) features of Ru(p-cymene)-HPhen1 are similar to those of a highly analogous model complex, [(p-cymene)Ru(phen)(1-(4-cyanophenyl)imidazole)], in support of the intended mode of Ru coordination to the His-Phen HCM.23</p><p>We then examined the chemical unfolding behavior of the Ru(p-cymene)-HPhen1 complex to study the effects of HCM capping by a substitution-inert metal complex on protein stability. As shown in Figure 12, binding of Ru(p-cymene) to HPhen1 leads to a significantly higher stabilization compared to substitution-labile divalent metals (Figure 6 and Table 2), with a corresponding shift in the unfolding midpoint of ~1.5-M GuHCl. Under the reasonable assumption that Ru(p-cymene) is still bound to the His-Phen HCM upon denaturation (which is not necessarily the case for labile metals), the unfolding of Ru(p-cymene)-HPhen1 can now be treated as a two-state process, allowing the determination of the free energy of stabilization (ΔΔGfolding) by Ru(p-cymene) binding to be 4.1 kcal/mol.24 The finding that global protein stability can be raised to such an extent by the metal-mediated crosslinking of a local fragment is particularly significant given that the free energy of unfolding for natural proteins typically ranges from 5 to 15 kcal/mol.25</p><!><p>To probe if tridentate i/i+7 HCMs may be used on any helical protein surface regardless of the amino acid content, we took a closer look at the structural features of our variants with particular focus on the residues that lie between the coordinating His and the functionalized Cys. Figure 13a shows the Ni coordination mode of the His-Quin HCM in the previously determined Ni:HQuin12 structure, and Figure 13b shows the proposed conformation for HPhen1 modeled after the same structure. These structures clearly indicate that the only intervening residues of interest are at the i+3 and i+4 positions, regardless of the relative positions of His and Cys on the helix (i.e., i/i+7 or i/i-7). Importantly, for both i+3 and i+4 positions (Asp66 and Ile67 for HPhen1 and HQuin1), the Cbackbone-Cα vectors that largely dictate the orientation of the sidechains are directed away from the coordinating groups. It could then be expected that the i/i+7 His-Phen or His-Quin HCMs may be universally installed on any regular α-helical surface to coordinate metals without significant interference by the intervening amino acids. This expectation is supported by the finding that HPhen1, HPhen2 and HPhen3 display more or less similar binding affinities for several divalent metal ions (Table 1) despite different sets of intervening residues: Asp66/Ile67 for HPhen1, Ile67/Asp66 for HPhen2 (inverse of HPhen1) and Asp73/Asp74 for HPhen3 (Figures 13a and b).</p><p>At the same time, a close inspection of the HQuin1 structure and the HPhen1 model (Figure 13c) shows that the side chain of Ile67 forms van der Waals contacts (d ~ 3.0 Å) with the Quin (or Phen) aromatic ring. These favorable interactions would be absent in the case of HPhen2 or HPhen3, which would respectively present Asp66 or Asp74 near the vicinity of Quin or Phen. Such differential interactions are likely culprits for the lack of any obvious trend in the metal binding affinities of HPhen1, HPhen2 and HPhen3 (Table 1). Nevertheless, we envision that the i+3 and i+4 positions within i/i+7 HCMs may be exploited as an additional handles to fine-tune metal coordination by HCMs.</p><!><p>Cumulatively, our studies establish that i/i+7 HCMs that include a single His and a non-natural bidentate ligand like Phen and Quin can form tridentate chelating platforms on α-helices, extending the scope of coordination chemistry on protein surfaces. Such tridentate HCMs not only provide unprecedented metal binding affinities, but are also able to stabilize α-helical structures, lead to the formation of discrete oligomers and provide high-affinity attachment sites for metal-based probes. Our findings and analyses suggest that these HCMs may be utilized as modular units on any α-helical protein surface in a sequence independent fashion.</p><!><p>Unless otherwise noted, all solvents, buffers were purchased from Fisher Scientific or VWR and used without further purification. ACS reagent grade metal salts (CoCl2, NiSO4, CuSO4 and ZnCl2) were purchased from Sigma-Aldrich and used without further purification.</p><!><p>Protein mass spectrometry was carried out at Biomolecular/Proteomics Mass Spectrometry Facility at UCSD using a Voyager DE-STR MALDI-TOF mass spectrometer. Protein samples (100 μL) were first washed with 3× with 400 μL of nano-pure water (Millipore) using a centrifugal spin column (Millipore) equipped with a 10 KDa cutoff filter. In a typical experiment, 5 μL of a protein sample was mixed in a 1:1 ratio with sinapinic acid (Aligent) as a matrix. 1 μL of the resulting protein/matrix samples was plated on a standard 100 well plate and dried completely before use.</p><p>Mass spectrometry (MS) of small molecules was carried out at the Molecular Mass Spectrometry Facility at UCSD using either electrospray ionization (ESI) or an atmospheric pressure chemical ionization (APCI) source on a ThermoFinnigan LCQDECA mass spectrometer equipped with a quadrupole ion trap mass analyzer and Xcalibur data system. The MS detector was operated under both positive and negative ion modes with a mass resolution range of 100 ppm.</p><!><p>Site directed mutagenesis was performed on the pETc-b562 plasmid (denoted as wild-type)26 using the QuikChange kit (Stratagene) and employing primers obtained from Integrated DNA Technologies. The mutant plasmids were transformed into XL-1 Blue E. coli cells and purified using the QIAprep Spin Miniprep kit (Qiagen). Point mutations were introduced to obtain the following cyt cb562 variants: G70C-cyt cb562, G70H/H63C-cyt cb562, K77C/G70H/H63A/W59H-cyt cb562, G70C/H63A-cyt cb562. Sequencing of all mutant plasmids was carried out by Retrogen Inc. (San Diego, CA).</p><!><p>The mutant plasmids isolated from XL-1 blue cells was transformed into BL21(DE3) E. coli cells along with the ccm heme maturation gene cassette plasmid, pEC86.27 Cells were plated on LB agar, containing 100 μg/mL ampicillin and 34 μg/mL chloramphenicol, and grown overnight. From these colonies LB medium was then inoculated and allowed to incubate for 16 h at 37°C, with rotary shaking at 250 rpm. No induction was necessary.</p><p>Mutant-expressing cells were sonicated, brought to pH 5 with the addition of HCl, and centrifuged at 16,000 g, 4° C, for 1 hr. The protein was then purified by ion-exchange chromatography on a CM-Sepharose matrix (Amersham Biosciences) using a NaCl gradient in sodium acetate buffer (pH 5). After exchange into sodium phosphate buffer (pH 8) using 6–8 kDa cutoff dialysis tubing (Fisher), the protein was further purified using an Uno-Q (BioRad) anion exchange column on a DuoFlow chromatography workstation (BioRad) using a NaCl gradient. Protein purity was determined by SDS-PAGE gel electrophoresis. Verification of mutations was made through MALDI mass spectrometry:</p><!><p>As a precursor, iodoacetic acid anhydride was freshly prepared by adding 2.64 g (12.8 mmol) of DCC to a stirred solution of 5.0 g (26.8 mmol) iodoacetic acid in 75 mL of ethyl acetate. Dicyclohexylurea precipitates immediately, but the mixture was allowed to stir for 2 h in the dark. The dicyclohexylurea was removed by filtration and the resulting solution was evaporated to dryness and used immediately.</p><!><p>0.5 g (2.56 mmol) of 5-amino-1-10-phenanthroline (Polysciences) was dissolved in 90 mL of acetonitrile with slight heating. To this stirred solution, the freshly prepared iodoacetic acid anhydride dissolved in 10 mL of acetonitrile was added. The mixture was allowed to react in the dark overnight. The precipitated product was isolated by filtration and washed with cold 5% sodium bicarbonate, followed by water and dried in vacuo. Both the ESI MS (Figure S1.16) and NMR spectra correspond to previously reported literature values.22 (Yield: 75%)</p><!><p>0.5 g (2.14 mmol) of 5-amino-8-hydroxyquinoline dihydrochloride (Sigma) was dissolved in 30 mL of acetonitrile by refluxing overnight with 975 μL (7 mmol) of triethylamine. The resulting solution was filtered, and iodoacetic acid anhydride, dissolved in 5 mL of acetonitrile, was added. The mixture was allowed to react in the dark overnight. The product evaporated to dryness and washed extensively with cold 5% sodium bicarbonate followed by water and dried in vacuo (Yield: 75%). Synthesis of IA-Quin was verified by mass spectrometry (ESI-MS, positive mode, Figure S1.17). Measured MW = 328.96 m/z (exp.: 328.9) (M + H+)</p><!><p>4-amino-2,2′:6′,2″-terpyridine (NH2-Terpy) was prepared as previously described.28 0.1 g (0.4 mmol) of NH2-Terpy was dissolved in 20 mL of dry dichloromethane. To this solution, 200 μL (1.4 mmol) of triethylamine was added and the reaction mixture was cooled to 0° C by stirring in an ice bath. Once the temperature equilibrated (approx. 20 min), 54 μL (0.6 mmol) of iodoacetyl chloride was added in a dropwise fashion. The mixture was left to react in the dark at 0° C for 30 min and then slowly brought up to room temperature. After a total period of 1 hour, the reaction volume was doubled with dichloromethane and washed extensively with cold 5% sodium bicarbonate followed by water. The organic fractions were evaporated dried in vacuo and used without further purification (Yield: ~60 %). In addition to IA-Terpy, small amounts of the amino precursor and a chloroacetamide terpyridine adduct were present as impurities. However, since only IA-Terpy can efficiently modify the protein under relevant labeling conditions, the product was used without further purification. Product formation was verified by mass spectrometry. (ESI-MS, positive mode, Figure S1.18). Measured MW = 417.05 m/z (exp.: 417.22 m/z) (IA-Terpy + H+); 325.29 m/z (exp.: 325.08 m/z) (ClA-Terpy + H+); 249.44 m/z (exp.: 249.11 m/z) (NH2-Terpy + H+)</p><!><p>A solution of 0.3 mM of cyt cb562 protein solution in degassed 0.1 M Tris buffer (pH 7.75) was treated with a 10-fold excess of dithiothreitol (DTT) (Sigma). The protein was allowed to reduce for a period of 30 min. The protein was then dialyzed against 2 × 1 L of degassed 0.1 M Tris buffer (pH 7.75) under an inert atmosphere to remove DTT. A 10-fold excess of iodoacetamide label was dissolved in 2 mL of degassed DMF and added dropwise to the protein solution over the course of 1 min. The mixture was allowed to react in the dark at 25° C overnight. The reaction mixture was then dialyzed again against 2 × 1 L of 10 mM sodium phosphate buffer (pH 8) and 1 mM EDTA. The crude labeled protein was subsequently purified on an Uno-Q anion-exchange column (BioRad) using a sodium chloride gradient. If further purification was necessary, the labeled fractions were combinded and dialyized against 2 × 1 L of 10 mM sodium acetate buffer (pH 5) and 1 mM EDTA. The protein mixture was then purified on an Uno-S (BioRad) cation-exchange column using a sodium chloride gradient. The final purity of the functionalized protein was determined to be greater than 95% by MALDI mass spectrometry and SDS-PAGE electrophoresis. (Labeling yield: 60–95%).</p><!><p>Unless otherwise stated, all metal (M2+) binding titrations were prepared by diluting a concentrated protein stock solution to a final volume of 2 mL with a final protein concentration ranging from 2 to 5 μM. All titrations were performed in 50 mM MOPS buffer (pH 7) previously treated with Chelex resin (BioRad). All pipette tips were rinsed 3× in an analytical grade 5% HNO3 (Fluka) solution before use. All further procedures followed to ensure a metal-free environment have been previously outline by Linse.29</p><p>Titration data obtained by monitoring changes in the Phen/Terpy absorption were fit using non-linear regression on Dynafit 3 (BioKin). All absorption spectra were obtained on an HP 8452A spectrophotometer. HPhen and HTerpy concentrations were determined based on the Soret absorption maximum for cyt at 415 nm (ε = 0.148 μM−1 cm−1). All data were baseline-corrected and adjusted cb562 for dilution. Due to the tendency of the HPhen and HTerpy variants to undergo metal-mediated dimerization, data were separately fit to both 1:1 and 1:1/1:2 models. The later model accounts for both metal binding and metal-mediated protein dimerization.</p><!><p>5 mL of an unfolded protein solution containing 5 μM of HPhen/HTerpy and 1 mM of M2+ or EDTA was freshly prepared in ~ 8 M guanidine HCl (GuHCl) in the appropriate buffer (either 100 mM Tris buffer (pH 7.5) or 100 mM sodium acetate (pH 5.5)). In parallel, 3 mL of a folded protein solution containing 5 μM of the same variant protein in the appropriate buffer and 1 mM M2+ or EDTA was prepared. The unfolded protein stock was titrated into the folded protein stock at 25°C using an autotitrator (Microlab 500 Series), keeping the sample volume constant at 2 mL. Protein unfolding was monitored by CD spectroscopy (222 nm) on an Aviv 215 spectrometer. For every titration point, the solution was allowed to stir for 30 seconds in order to reach equilibrium. This procedure was repeated for a minimum of 30 points covering a GuHCl range of 0.1–6.5 M. GuHCl concentrations were calculated using the refractive indices of the folded and unfolded protein stock solutions.30 Unfolding data were fit using Kaleidagraph (Synergy Software) with an expression that assumes a two-state folding/unfolding equilibrium as described by Pace (eq. 1):31</p><p> (1)FractionUnfolded=e(−m1×(m2−[GuHCl])RT1+e(−m1×(m2−[GuHCl])RT where m1 represents the slope of the unfolding transition and is defined as (∂ΔGH2O/∂ [GuHCl]), and m 2 represents the midpoint GuHCl concentration where 50% of the protein is unfolded. All HPhen and HTerpy variants remain in their monomeric form at under the conditions (5 μM of protein, 1 mM M2+) used in chemical denaturation experiments.</p><!><p>A 3 mL solution of 5 μM HPhen/HTerpy in 100 mM Tris buffer (pH 7.5) and 1.5 M GuHCl was prepared. Inclusion of 1.5 M GuHCl was necessary to ensure that the protein fully unfolds below 373 K. To the protein solutions either 1 mM Ni2+ or EDTA was added. The unfolding reaction was monitored over a range of 300–376 K by CD spectroscopy (222 nm). At each temperature point, the solution was allowed to stir for 30 seconds in order to reach equilibrium. Although the thermal unfolding of HCMs is not completely reversible, the curve was fit to a two-state model as described by John and Weeks32 to obtain an apparent ΔTmmetal. Unfolding data were fit using Kaleidagraph (Synergy Software).</p><!><p>SV experiments were performed in order to determine the solution-state oligomerization behavior of each HPhen/HTerpy variant. All SV samples were prepared in 20 mM Tris buffer (pH 7). Measurements were made on a Beckman XL-I Analytical Ultracentrifuge (Beckman-Coulter Instruments) using an An-60 Ti rotor at 41,000 rpm for a total of 250 scans per sample. The following wavelengths were used for detection: 418 nm (5 μM protein), 420 nm (10 μM protein), 425 nm (20 μM protein), 540 nm (40 μM protein), 545 nm (60μM protein) and 560 nm (100 μM protein).</p><p>All data were processed using SEDFIT.33 Buffer viscosity, buffer density, and protein partial specific volume values were calculated at 25° C with SEDNTERP (http://www.jphilo.mailway.com). Partial specific volume (Vbar) for each variant was calculated assuming a partial specific volume of heme of 0.82 mg/mL and 0.71 mg/mL, 0.75 mg/mL, 0.87 mg/mL for Phen, Quin and Terpy respectively. All data were processed using fixed values for buffer density (ρ) (0.99764 g/mL) and buffer viscosity (0.0089485 poise).</p><!><p>DFT calculations were performed with Amsterdam Density Functional (ADF) program suite,34, 35 version 2007.01,36 on a home-built 72-CPU (1 × 8 master, 8 × 8 slave) Rocks 4.3 Linux cluster featuring Intel Xeon E5335 Quad-Core 2.00GHz processors. Job control was implemented with the Sun Grid Engine v. 5.3. Crystallographic atomic coordinates were used as input where appropriate. Optimized geometries and molecular orbitals were visualized with the ADFView graphical routine of the ADF- GUI37 and the Gaussview 3 program.</p><p>In ADF program suite calculations, the triple-ζ Slater-type orbital TZ2P ADF basis set was utilized without frozen cores for all atoms. Relativistic effects were included by use of the zero-order regular approximation (ZORA).38 To ensure consistency over a range of exchange/correlation profiles, the molecular geometries and energies were evaluated with both the BP86 and OLYP functionals.</p><p>In BP86 calculations, the local density approximation (LDA) of Vosko et al.39 (VWN) was coupled with the generalized gradient approximation (GGA) corrections described by Becke40 and Perdew41, 42 for electron exchange and correlation, respectively. In OLYP calculations, the parameterized (X = 0.67) exchange-only LDA was coupled with the GGA corrections described by Handy and Cohen (OTPX)43 and Lee, Yang and Parr (LYP)44 for electron exchange and correlation, respectively.</p><!><p>Cartoon representations for various HCM-bearing cyt cb562 variants. Functionalities that comprise the HCMs and the heme groups are shown as sticks and highlighted in magenta.</p><p>Phenanthroline (Phen), terpyridine (Terpy) and hydroxyquinoline (Quin) derivatives used for the construction of HCMs in this study. The iodoacetamide moiety is attached to the chelating functionalities through the amide nitrogen.</p><p>(a) Cartoon depiction for the three coordinate facial binding mode of a His-Phen HCM. (b) Proposed mode of metal-dependent dimerization mediated by a i/i+7 His-Phen HCM.</p><p>Spectral changes that accompany Zn2+ binding to HPhen1 as monitored by UV-vis spectroscopy. Spectra show a typical ferric heme spectrum with a Soret band at 415 nm (ε = 0.148 μM−1 cm−1) along with a transition between metal-free (λmax= 272 nm) and metal-bound (λmax= 282 nm) Phen species. Inset: Close-up view of the UV region showing a clean isobestic point at 274 nm consistent with a 1:1 Zn2+:Phen binding model.</p><p>Metal-binding titration data and fits for HPhen1 as monitored by UV-visible spectroscopy. A typical titration sample contained 2–5 μM of HPhen1, 50 mM MOPS buffer (pH 7) and 20–100 μM of the competing ligand EGTA. All data were described satisfactorily by a 1:1 binding model. Binding isotherms for HPhen2, HPhen3 are shown in Figure S1.2 and S1.3. Dissociation constants (Kd) determined are listed in Table 1 and S2.1.</p><p>Chemical unfolding titrations of (a) HPhen1 (b) HPhen2 (c) HPhen3 and (d) HTerpy1 in the presence and absence of Ni2+ as monitored by CD spectroscopy at 222 nm.</p><p>Chemical unfolding titrations of (a) HPhen1 at pH 5.5, (b) HTerpy1 at pH 5.5, (c) APhen1 at pH 7.5, and (d) HTerpy1 at pH 7.5. The lack of significant protein stabilization in the presence of Ni2+ ions indicates that both His and Phen or Terpy moieties are involved in metal-mediated protein cross-linking.</p><p>(a) Sedimentation velocity profiles of 20 μM HPhen1 in the absence and presence of Ni2+. (b) Sedimentation equilibrium (SE) profiles for 20 μM HPhen1 in the presence of 10 μM Ni2+, with equilibrium speeds of 20,000 (green), 25,000 (blue), 30,000 (cyan), 35,000 (yellow) and 41,000 rpm (red). Data were fit to a monomer-dimer self-association model with a log K = 5.05 (2) or 8.9 (1) μM.</p><p>Energy minimized structures (BP86) for the proposed inner-sphere coordination geometry of the Ni:HPhen12 dimer.</p><p>(a) The proposed Ni:HPhen12 architecture modeled after the crystallographically determined Ni:HQuin12 structure. (b) The corresponding inner coordination sphere.</p><p>(a) Anion-exchange FPLC chromatogram of the crude Ru(p-cymene)-HPhen1 reaction mixture. Product eluted at 0.2–0.25 M NaCl using a linear NaCl gradient (0–0.5 M in 10 mM sodium phosphate, pH 8.0). (b) MALDI mass spectra of the major FPLC product identified as the Ru(p-cymene)-HPhen1 complex.</p><p>Chemical unfolding titrations (monitored by CD spectroscopy) showing the higher stability of the Ru(p-cymene)-HPhen1 complex (blue) with respect to HPhen1 in the absence of metals (red).</p><p>(a) Cartoon representation of the HQuin1 structure showing the Ni coordination mode by the i/i+7 His-Quin HCM. Other residues on Helix3 that are important either for the construction of HPhen3 HCM (positions 70 and 77) or those corresponding to the i+3 and i+4 positions for all variants are shown also as sticks. (b) The analogous representation of HPhen1 modeled after the HQuin1 structure/(c) Closeup view of the model for His-Phen HCM coordinated to a Ni2+.</p><p>Dissociation constants for HPhen-metal complexes compared to those for free 1,10-phenanthroline (Phen).</p><p>Dissociation constants determined by competition with EGTA in 50 mM MOPS (pH 7).</p><p>pH-adjusted values based on reported Kd's.10</p><p>Observed changes in the midpoint for the unfolding transition (Δ[GuHCl]m) for HPhen and HTerpy variants upon metal binding.</p>

PubMed Author Manuscript

Imbalance in chemical space: How to facilitate the identification of protein-protein interaction inhibitors

Protein-protein interactions (PPIs) play vital roles in life and provide new opportunities for therapeutic interventions. In this large data analysis, 3,300 inhibitors of PPIs (iPPIs) were compared to 17 reference datasets of collectively ~566,000 compounds (including natural compounds, existing drugs, active compounds on conventional targets, etc.) using a chemoinformatics approach. Using this procedure, we showed that comparable classes of PPI targets can be formed using either the similarity of their ligands or the shared properties of their binding cavities, constituting a proof-of-concept that not only can binding pockets be used to group PPI targets, but that these pockets certainly condition the properties of their corresponding ligands. These results demonstrate that matching regions in both chemical space and target space can be found. Such identified classes of targets could lead to the design of PPIclass-specific chemical libraries and therefore facilitate the development of iPPIs to the stage of drug candidates.Protein-protein interactions (PPIs) play essential roles in nearly all biological processes, and their deregulation is often associated with disease states. Therefore, there is a growing interest to target PPIs for therapeutic interventions not only with biologics but also using low-molecular-weight (LMW) compounds (< 1,000 g.mol −1 ). Still, targeting PPIs with LMW drugs remains one of the most difficult challenges in molecular medicine 1 . Although great innovations have been achieved to facilitate the identification of inhibitors of PPI targets (iPPIs) (e.g., fragment-based drug design, Nuclear Magnetic Resonance (NMR), X-Ray crystallography, etc.), experimental screening procedures for PPI targets still suffer from the unavailability of suitable fragments and chemical libraries 2-4 . Indeed, the molecular topography of most known PPIs, which are often described as shallow, large, and hydrophobic, makes them harder to tackle with small compounds. This situation has often been translated into designing larger, more hydrophobic and more aromatic compounds 2-4 . Such compounds dramatically diminish the likelihood of obtaining a safe and specific drug at the end of the development process [5][6][7] . In addition to such impeding properties, other studies have nevertheless highlighted specific physicochemical characteristics that may be necessary for iPPIs to bind PPI interfaces. These characteristics include the specific 3D shapes 8-10 of those compounds. Recently, our group has identified new 3D characteristics of inhibitors of PPI targets 11 . In this seminal analysis, four shape properties were shown to be specific to the structure of iPPI compounds, including the globularity (glob) and the Volsurf 12 properties EDmin3, CW2, and IW4 in a distribution of putative hydrophobic and hydrophilic interacting regions around the compound. Most noticeably, EDmin3, which describes the capacity of a compound to efficiently bind the hydrophobic patch that is often present at the core of a PPI interface, is an important structural feature for nearly all iPPI compounds regardless of the heterogeneity of the PPI target space. In contrast to the previously identified shape features 8,9 , such properties correlate with neither the size nor the hydrophobicity of the compounds and could therefore allow chemists to prioritize the selection of PPI-compliant LMW compounds without having to drive potency through molecular obesity 5 .The goal of this study was to capitalize on this cumulative knowledge to obtain further insight into the iPPI chemical space and to evaluate the heterogeneity of the known PPI target space. We aimed to establish a proof-of-concept that some classes of known PPI targets can be identified either by detecting shared properties

imbalance_in_chemical_space:_how_to_facilitate_the_identification_of_protein-protein_interaction_inh

<!>Results and Discussion<!>Assessment of conventional ADME/tox rules and iPPI prediction rules.<!>Confirmation of known iPPI specific descriptors.<!>Discrepancies among PPI targets.<!>Visualization of the iPPI chemical space.<!>Conclusion<!>Methods<!>Molecular descriptors.<!>Ligand analysis. PAINS/ADME-Tox filter. PAINS (pan assay interference compounds) (filters A, B and<!>Filter Explanation

<p>and chemotypes for the ligands meant to modulate them or by analysing the properties of the PPI interfaces' binding cavities. This identification would facilitate the design of PPI-class-specific chemical libraries and would boost the identification of active compounds on PPI targets.</p><p>To this end, we have combined iPPI compounds and pharmacological data from both iPPI-DB 13 and TIMBAL 14 . To the best of our knowledge, this study is the first of its kind, applying such a wide dataset of iPPI compounds (~3,250 representative compounds after preparation) across 29 PPI targets. We also selected a series of annotated libraries as reference chemical datasets that contain compounds that are not considered iPPIs (called hereafter non-iPPIs): natural compounds, allosteric compounds, advanced drug candidates, launched drugs, active compounds on conventional targets, and compounds from commercial libraries. This series collectively corresponds to 566,000 non-iPPI compounds. Furthermore, we selected particular group molecular descriptors known to be iPPI-selective, namely the octanol/water partition coefficient (AlogP), the molecular weight, the aromatic ratio (proportion of aromatic atoms) and the 4 shape descriptors cited above (glob, EDmin3, CW2, and IW4). These descriptors were used to define for these joined datasets the foundations of a chemical space in which iPPIs can be more suitably distinguished from non-iPPIs. These iPPI-selective descriptors were combined with conventional descriptors that are commonly used to depict chemical space. The resulting chemical space was then visualized and analysed using a principal component analysis (PCA). Further along this line, a new metric is described hereafter to quantify the overlap in chemical space between two populations of compounds. This metric uses the probability density functions for each dataset along with each principal component of a PCA calculated from joined sets of compound libraries and a given set of descriptors. The use of such an approach has confirmed several known trends about the iPPI chemical space compared to conventional molecules, such as their overall higher size, higher aromaticity, and higher hydrophobicity. This approach supports the selective character of the 4 shape descriptors described above (glob, EDmin3, CW2, and IW4). However, this approach also underpins novel observations, such as the important heterogeneity of the PPI target space. Indeed, this study allowed us to create some classes of PPI targets by comparing the properties of their corresponding ligands. The determination of such ligand-driven PPI classes was then compared to the classes that were obtained using a purely pocket-driven analysis of the PPI targets themselves. Most of the PPI target classes could thus be confirmed using this latter orthogonal approach. Therefore, as a proof of concept, we show trends that confirm some of the observations that were made using the iPPI compounds alone and provide a perspective with which to generalize such pocket-driven approaches PDB-wide 15 (Protein Data Bank). Indeed, it seems that not only can the properties of binding pockets be used to group PPI targets, but they can certainly condition the properties of their corresponding ligands in a predictable manner.</p><!><p>Description of the datasets. For the present analysis, we utilized a set of iPPI compounds and a set of non-iPPI compounds. For iPPIs, we combined the compounds and binding data of the iPPI-DB and TIMBAL databases that contain, after standardization, a total of 3,248 non-redundant iPPIs across 29 PPI targets: 1,650 from iPPI-DB across 13 targets, and 1,598 iPPI from TIMBAL across 16 other targets (Fig. S1). We chose to apply the standards of iPPI-DB to the compounds of TIMBAL in order to maximize the homogeneity of the data, including the availability of binding data in the form of a XC 50 (e.g., IC 50 , K d , EC 50 , K i ) inferior to 30 μM, the existence of a clearly identified PPI target (e.g., excluding TIMBAL's Integrins), and the exclusion of all peptides (defined as 3 sequential peptide bonds) and of macrocycles. The present procedure was used to limit the presence of non-specific binders and to focus on small molecules (molecular weight < 1,200 g.mol −1 ).</p><p>For the sets of non-iPPI compounds, we selected different databases. The goal was to better represent the diversity of compounds that can be considered negative data with respect to iPPIs. The list of those datasets is presented in Table 1 and refers to different types of small molecules: active compounds under development, actual drugs, allosteric modulators, natural compounds, or compounds from commercial chemical libraries. From MDDR (www.biovia.com), we gathered compounds that are presently under development, either in biological testing (early development), preclinical phases, and clinical phases or launched drugs. Along the same lines, we collected a subset of drugs that are orally bioavailable from the E_Drugs3D database 16 , called e_Drug hereafter. We also collected the allosteric modulators of the ASD database 17 or a selection of conventional inhibitors that are active in the top 100 most studied targets from BindingDB 18 (version 2012), which are all conventional targets, i.e., GPCRs (G-protein coupled receptors), enzymes, kinases, ion channels, or nuclear receptors. We also selected a chemically diverse set of compounds, BDM, from three different commercial chemical libraries: Asinex 19 (v2012), ChemDiv 20 (v2012) and Enamine 21 (v2012). Finally, we selected the natural compounds of the Nubbe database 22 . Together, this overall set of 566,208 non-iPPI compounds represents a negative dataset for iPPI compounds and provides a comparative analysis of the physicochemical properties of iPPIs with respect to the different types of annotated compounds. This dataset will facilitate us obtaining a global positioning system in chemical space for iPPI compounds. To treat equally positive (iPPIs) and negative (non-iPPI compounds) data, all of the datasets of iPPI compounds and non-iPPI compounds were standardized using the exact same procedure (see the standardization procedure in Methods).</p><!><p>We first estimated the drug-like profile of each dataset using in silico routines. To this end, we ran the FAF-Drugs3 web server 23 to measure and compare different physicochemical rules that are commonly used to predict the pharmacokinetic profiles of drug candidates. These rules include PAINS-containing substructures 24 ; rules for drug likeness and/or bioavailability, such as Lipinski's RO5 25 , Veber's 26 , Eli Lilly 27 , Gleeson's X4_400 28 , and Egan's 29 ; and some predictive models for iPPI compounds, such as PPI-HitProfiler 9 and the 2P2I RO4 30 .</p><p>The detection of PAINS substructures (Fig. S2) was performed for all of the compounds from all of the datasets by differentiating the three established filters: PAINS-A, PAINS-B and PAINS-C. Those filters were created by Baell et al. 24 to consider three levels of pan-assay-interfering substructures depending on the number of occurrences, such chemical moieties, that were observed in their experimental screenings over the years. More precisely, the most problematic substructures are listed in PAINS-A: they have been observed to be active in more than 150 screening campaigns. The substructures from PAINS-B have been observed in experimental assays between 15 and 150 times. Finally, the substructures from PAINS-C have been observed fewer than 15 times. The present PAINS analysis indicates that iPPIs tend to show on average more PAINS-containing compounds (20% for iPPI-DB, mostly from filter PAINS-C, and 15% for TIMBAL, mostly from filter PAINS-A) than active compounds on conventional targets (less than 10%, except for kinase with 15%). Nevertheless, this pattern is not the case for all of the PPI targets. Indeed, Bcl-2, MDM2, CD80, ITGAL and E1 are the PPI targets that have the highest proportions of PAINS-containing compounds, although not necessarily for the same filter (PAINS-A, PAINS-B, or PAINS-C). Interestingly among the detected PAINS substructures, for example, among the Bcl-2 inhibitors, one can find catechol from filter PAINS-B. However, this chemical moiety is also found in gossypol or apogossypol, which are currently in clinical trials as treatments for leukaemia. Thus, for conventional inhibitors, most of the iPPI compounds are not predicted as promiscuous binders or frequent hitters according to Baell's PAINS.</p><p>The level of compliance of all of the datasets with respect to all of the remaining chemistry rules cited above confirms that iPPIs exhibit different profiles compared to conventional drug-like compounds (Fig. S3). The use of such rules on orally bioavailable drugs, such as e_Drug compounds, highlights the importance of the Lipinski's RO5, Veber's, Gleeson's X4_400, Egan's and Eli Lilly's compounds to estimate the drug-likeness or the bioavailability of compounds. In contrast, both iPPI-DB and TIMBAL compounds have the worst profiles according these rules, underpinning the now known hydrophobic character and the larger size of iPPIs with respect to active compounds on conventional targets (GPCR, enzyme, etc.) or drug candidates, as most of these rules partially rely on molecular weight and logP (octanol/water partition coefficient). However, one concern is the compliance of all of these rules in the iPPI datasets when subdividing them according to the activity of the compounds (Fig. S4). Indeed, when considering two bins of activity, namely pXC50 < 7 or pXC50 > 7, the chemistry rules penalize the most potent iPPI compounds even more dramatically. Again, most of these rules partially rely on molecular weight and hydrophobicity (i.e., octanol/water partition coefficient, or logP), indicating that there is still a price to pay for iPPI in terms of size and hydrophobicity to reach a critical level of potency, impeding the development of drug candidates.</p><p>Finally, we also assessed known predictive models of iPPI, such as PPI-HitProfiler 9 and 2P2I RO4 30 , on the iPPI dataset. Those models confirm their good levels of predictability (from 70 to 92%) on the iPPI datasets and their satisfactory specificity with respect to non-iPPI datasets (Figs S3 and S4).</p><!><p>We then used the datasets of iPPI compounds (iPPI-DB + TIMBAL) and non-iPPI compounds (all remaining datasets) to evaluate the levels of specificity of already known discriminative descriptors towards iPPI and to further confirm the results of our recently reported work 11 . To this end, we tested the following descriptors: molecular weight (MW), hydrophobicity (AlogP), proportion of aromatic atoms (aromatic ratio), and the 4 shape descriptors cited above: globularity (glob), CW2, IW4, and EDmin3. To measure the levels of specificity of those descriptors with respect to non-iPPI compounds, we ran an ANOVA (Analysis Of Variance) followed by post hoc comparisons using a pairwise test (see Methods) between the iPPI datasets and all of the individual non-iPPI datasets (Fig. 1). As a global trend, iPPI compounds display a significantly higher molecular weight (MW iPPI-DB = 540 g mol −1 ± 145; MW Timbal = 521 g mol −1 ± 179) than that of all of the non-iPPI datasets. Compared to most of the non-PPI datasets, iPPIs present a significantly higher hydrophobicity (AlogP iPPI-DB = 4.61 ± 1.98; AlogP Timbal = 4.19 ± 2.39), except for ion channel inhibitors (AlogP ion channel = 4.64 ± 1.52). The aromatic ratio, i.e., the proportion of aromatic atoms in a molecule, is most often significantly higher for iPPI compounds (aromatic ratio iPPI-DB = 0.29 ± 0.10; aromatic ratio TIMBAL = 0.29 ± 0.12), except when compared to ion channel, BDM, ASD, and kinase compounds (aromatic ratio ion channel = 0.32 ± 0.10; aromatic ratio BDM = 0.29 ± 0.11; aromatic ratio ASD = 0.33 ± 0.11; aromatic ratio kinase = 0.38 ± 0.10), for which significantly higher values have been found. Furthermore, iPPIs tend to have significantly lower values of CW2 (CW2 iPPI-DB = 1.98 ± 0.17; CW2 Timbal = 1.97 ± 0.20) than do non-iPPI compounds, except for GPCR compounds (CW2 GPCR = 1.9 ± 0.18), which have even more significantly lower values. This property illustrates that the ratio between the hydrophilic regions is defined at − 0.5 kcal.mol −1 and the molecular surface of the molecule. Thus, it seems that iPPIs have on average smaller proportions of exposed soft polar regions than the compounds of non-iPPI datasets. In contrast, iPPIs tend to have significantly higher values of IW4 (IW4 iPPI-DB = 2.69 ± 1.00; IW4 TIMBAL = 2.78 ± 1.16) than do non-iPPI compounds, except for GPCRs (IW4 GPCR = 2.78 ± 1.15) and nuclear receptor compounds (IW4 Nuclear = 2.87 ± 1.31), which have significantly higher values of IW4 than do iPPIs. This property illustrates the distance between the centre of mass of the molecule and the barycentre of its polar regions defined at − 2 kcal.mol −1 . This property therefore expresses a higher degree of concentration of those soft polar regions at one extremity of the molecule, most often in the case of iPPIs, with the exception of GPCRs and nuclear receptor compounds. Globularity has also been confirmed as an important factor for iPPI. These compounds are indeed on average significantly more globular (Glob iPPI-DB = 0.16 ± 0.09; Glob Timbal = 0.15 ± 0.11) than are all of the non-iPPI datasets, except for existing drugs, i.e., MDDR/launched (Glob Launched = 0.15 ± 0.08), e_Drugs (Glob eDrugs = 0.15 ± 0.09), and natural compounds (Glob Nubbe = 0.15 ± 0.09). This result therefore confirms the importance of having chemical structures with a higher isotropic occupation of the 3D space. Interestingly, natural compounds, which are defined by a higher number of sp3 carbon atoms and a higher level of chirality and complexity, have a globularity that is not significantly distinguishable from those of iPPI compounds. Furthermore, existing drugs have on average a globularity that is in the same range as those of iPPI compounds. This pattern is not the case for drug candidates in Phases I, II and III. Indeed, these compounds have a significantly lower globularity than do iPPI compounds, confirming the importance of discerning suitable molecular shapes when designing drugs and the concept that particular molecular shape is definitely worth investigating as an alternative to driving potency through molecular obesity 5 . Finally and most importantly, the property EDmin3, which describes the capacity of a compound to efficiently bind a hydrophobic patch at a protein surface, has been confirmed for iPPI compounds as significantly lower in energy (i.e., more efficient) (EDmin3 iPPI-DB = − 2.85 kcal.mol −1 ± 0.19; EDmin3 Timbal = − 2.84 kcal.mol −1 ± 0.26) than all of the non-iPPI datasets. This property still appears to be key in the binding of such compounds to their respective PPI targets.</p><!><p>To address the heterogeneity of the PPI target space first using iPPI properties and to ensure that those properties are not specific to the PPI targets that have the highest numbers of iPPI compounds, we also compared the distributions of those descriptors for each PPI target individually (28 PPI families from TIMBAL and iPPI-DB that have more than 5 iPPI compounds) by running an ANOVA followed by post hoc comparisons using a pairwise test (see Methods). Indeed, the distribution of each descriptor for each PPI target was compared to that of each non-iPPI dataset. Each significant difference with respect to a non-iPPI dataset was counted and plotted (Fig. 2). The most frequent discriminative descriptors were again the molecular weight and EDmin3, which are significantly different with respect to the compounds of non-iPPI datasets for most of the PPI targets. To a lesser extent, the globularity, aromatic ratio and AlogP were also significantly different from those of the non-iPPI datasets for a significant number of PPI targets. The other properties, CW2 and IW4, display good discriminative power but only for some PPI targets, namely Bcl2, MDM2, ITGAL, Xiap, PSIP1, bromodomain, HIF-1A, and neuropilin. Interestingly, CW2, IW4, and globularity may constitute surrogate descriptors for EDmin3 for the rare PPI targets; this property does not fall into the expected PPI profile. Indeed, it seems that alternatively for timbal_Xiap, PSIP1, HIF-1A and cyclophilin, these descriptors are discriminative with regards to most of the non-iPPI datasets, while EDmin3 is not. Finally for only 6 PPI targets out of 28 (E1, timbal_CTNNB1, timbal_IL2, timbal_tak1, timbal_Rad51, and timbal_S100B), none of the identified 3D descriptors were significantly different from a substantial number of non-iPPI datasets. Thus, for most of the known iPPI compounds, the above-cited 3D descriptors were specific by themselves, such as EDmin3; as a combination; or as surrogate descriptors for EDmin3, such as globularity, IW4, and CW2. In contrast, when plotting the discriminative power of the descriptors with reference to the non-iPPI datasets, EDmin3 was again confirmed as a significantly specific descriptor with respect to PPI targets (Fig. S5), confirming the importance of that property despite the heterogeneity of the PPI target space.</p><p>To further confirm this trend, we plotted for each descriptor (globularity (glob) EDmin3, IW4, CW2, molecular weight (MW), AlogP, and aromatic ratio) the number of PPI targets as a function of the number of non-iPPI datasets for which they are significantly different (Fig. 3). For example, the compounds of 27 PPI targets had a significantly lower value of EDmin3 than that of 4 non-iPPI datasets, and the compounds of 21 PPI targets had a significantly higher molecular weight than that of 9 non-iPPI datasets. It is clear from Fig. 3 that the molecular weight of iPPI remained significantly higher for most of the PPI targets with respect to most of the non-iPPI datasets.</p><p>Nevertheless, globularity and, more importantly, EDmin3 are among the most significantly specific descriptors of iPPI compounds, even more so than the AlogP or aromatic ratio. Indeed, EDmin3 seems to better characterize the specificity of iPPI compounds. Finally, because CW2 and IW4 are significantly different only for a subset of PPI targets, their discriminative power as a global trend is not as important. These descriptors nevertheless remain extremely efficient when used to characterize the appropriate PPI targets and must therefore be kept and used to help to address the PPI target space heterogeneity. These 3D properties therefore provide us with a set of descriptors to separate iPPIs from non-iPPI compounds as efficiently as the hydrophobicity, aromaticity and molecular weight of the compounds.</p><!><p>Having identified discriminative descriptors to properly distinguish iPPIs from non-iPPI compounds, we then attempted to visualize the joined chemical space comprising all of the datasets. A common and convenient tool with which to visualize and analyse the chemical space of molecular datasets is the principal component analysis (PCA) with an appropriate set of molecular descriptors. Thus, the combination of the 7 discriminative descriptors cited above (molecular weight, AlogP, aromatic ratio and the 4 shape descriptors uncorrelated with neither the size nor the hydrophobicity) with 13 commonly used descriptors to depict chemical space (Table 2) provides us a way to depict the joined chemical space comprising all of the datasets. Moreover, in this chemical space, iPPI compounds can be characterized and distinguished from For example, the iPPIs of at most 27 PPI targets had a significantly different EDmin3 (lower in the case of EDmin3) than did the compounds of 3 non-iPPI datasets, and the iPPI compounds of at most 21 PPI targets had a significantly different molecular weight (higher in the case of molecular weight) than did the compounds of 9 non-iPPI datasets. non-iPPI compounds. We therefore performed a principal component analysis on all of the cumulated datasets with those 20 descriptors (Fig. 4), representing a total of 569,456 molecules and 20 descriptors (See Methods). From Fig. 4, both iPPI-DB and TIMBAL compounds clearly show a nice global overlap between the two iPPI datasets (coloured dots and coloured squares in Panel A). Even though the two iPPI datasets can have different PPI targets, their corresponding iPPI compounds seem to overlap nicely as a whole. This trend is particularly noted when comparing the PPI targets that have iPPI compounds in both of the PPI datasets (but no duplicates), such as Bcl2, MDM2, Xiap, Brd, and IL2. For example, Xiap compounds (green dots and green squares) from both iPPI-DB and TIMBAL overlapped well in the individual map of Fig. 4 (Panel A). However, within the chemical space region corresponding to iPPI compounds, the compounds of different PPI targets interestingly occupy different regions. It is visible that iPPIs of Xiap (green), MLL (yellow), K-RAS (grey), and IL2 (orange) occupy the top of the individual map in Panel A, which, according to the circle of correlation (Panel B), defines a higher number of chiral centres and sp3 carbon atoms. In contrast, iPPIs of CD80 (in red) and of CTNNB1 (in beige) are at the bottom right corner of the individual map (Panel A), illustrating their higher proportion of aromatic atoms according to the same circle of correlation. The example given in Fig. 4 also shows the inhibitors of GPCRs (black dots). It is clear from the individual map (Panel A) that iPPIs (all of the non-black dots and squares) and GPCR inhibitors (black dots) occupy different regions in chemical space. Interestingly the most important chemical space imbalance comes from the first component of the PCA, which is mostly associated with EDmin3, confirming that iPPIs clearly have lower values than do non-iPPI compounds for that property.</p><p>Evaluation of the imbalance in the iPPI chemical space. It is clear from the analysis of the PCA that interesting comments can be made to evaluate the qualitative imbalance in chemical space between datasets to link it to the principal components and therefore to the descriptors/properties that are responsible for this imbalance. Nevertheless, given the multiple dimensions of the PCA and the number of datasets (29 PPI targets when iPPI-DB and TIMBAL are combined and 17 non-iPPI datasets), it would be impossible to account for the imbalance between the datasets just by plotting their corresponding compounds into the successive components of the PCA, nor would it be sufficient to quantify it. To address this issue and provide ourselves with a procedure to actually quantify the imbalance in chemical space between the different datasets, we designed a simple metric relying on the principal components of a PCA. This procedure relies on the probability density functions of each dataset along each component of the PCA (see Methods). For each possible pair of datasets and each successive PCA component, the individual overlap δ (Equation 5 in Methods) between the two corresponding probability density functions was calculated. These one-component overlaps δ were then combined into an arithmetic mean Δ weighted by the variance associated with the corresponding component (see Equation 6in Methods). The idea is to put more weight for the component overlap δ corresponding to the components carrying most of the variance. Collectively, this metric, in the form of a normalized score between 0 and 1, with 1 being complete overlap, quantifies the imbalance in chemical space for each possible pair of datasets in accounting for all of the components of the PCA. The results of such a measure can then be displayed as an array of pairwise normalized overlap scores for all possible pairs of datasets and colour-coded according to the values of their overlap score, as shown in the surface map of Fig. 5.</p><p>To assess the pertinence of such a metric, several observations can be made. First, the observed common region of chemical space shared by the iPPI-DB and the TIMBAL datasets depicted in Fig. 4 was confirmed by an overlap score Δ of 0.88. Second, commercial chemical libraries are poorly adapted for identifying modulators of PPIs because they have been designed for the modulation of more conventional targets, such as GPCRs, enzymes, and, more recently, kinases, nuclear receptors, and ion channels. This trend was confirmed by the overlap metric Δ between BDM, which is a representative dataset of three chemical compound providers, and the subsets GPCR, enzyme, kinase, nuclear receptors, ion channels, or the full MDDR dataset. Indeed, overlap scores Δ between BDM and these datasets were most often close to 0.85, while they were equal to 0.74 and 0.76 for the pairs (BDM, iPPI-DB) and (BDM, TIMBAL), respectively.</p><p>Third, the actual drugs and compounds in current drug development, such as those of e_Drug, launched (MDDR), phase I/II/III (MDDR), preclinical (MDDR) and biological testing (MDDR), are optimized to fit as much as possible a certain physicochemical profile in terms of size, complexity and hydrophobicity. This profile maximizes the chances of eventual success and might avoid a series of pharmacokinetic issues during development. Moreover, the present study shows that the 4 shape properties EDmin3, IW4, glob and CW2 were significantly different on average between those drug datasets and the two iPPI datasets (iPPI-DB and TIMBAL). These properties were represented among the 20 descriptors that were used to calculate the PCA and therefore were accounted for in the overlapping scores. It is therefore not surprising to picture these datasets as close in chemical space. In fact, it is clearly visible from the surface map of Fig. 5 that the overlap scores Δ of these datasets correspond to the largest values of overlap (largest green square), with overlap scores Δ greater than 0.9. Not only do these datasets seem to be characterized by the above-mentioned physicochemical profile, but they also seem to be equally dissimilar in chemical space to iPPI compounds regarding their shape properties. These points collectively justify the global overlap between these datasets and the selection of those descriptors.</p><p>Finally, it is interesting to note that enzymes and GPCRs have a higher overlap score Δ with respect to advanced phases of development in MDDR (Phases I, II, and III) than do kinases, nuclear receptors, and ion channels. This result again confirms a well-established fact that most of the developed drugs still aim at GPCRs or enzymes.</p><p>Having confirmed a series of pertinent observations, we then attempted to use the overlap metric Δ to identify new trends in this chemical space. The most striking observation concerns the heterogeneity of the PPI targets. Indeed, a rapid overlook of all of the individual PPI targets from both of the iPPI datasets (right lower quarter of the figure) is sufficient to perceive that they do not share the same level of overlap as the GPCR, enzyme, kinase, ion channel, and nuclear receptor datasets. We have already identified such heterogeneity in a previous study using a rather different approach 31 , and the need to address this heterogeneity by considering subclasses of PPI targets has also been underlined by others 32 . Furthermore, we can confirm some of the observations that we made simply using the individual map of Fig. 4 (Panel A). Indeed, PPI targets that have iPPI compounds in both iPPI-DB and TIMBAL (but no duplicates) have clearly higher overlap values: (Bcl2_family, timbal_Bcl2_family, Δ = 0.80), (MDM2, timbal_MDM2, Δ = 0.77), (Xiap, timbal_Xiap, Δ = 0.80), (Brd, timbal_Brd, Δ = 0.80), and (IL2, timbal_IL2, Δ = 0.81). In contrast, the overlap scores also distinguish iPPI datasets belonging to different regions of chemical space. The values of the overlap scores are rather different when considering the pair Xiap, IL2 (Δ = 0.54), which belongs graphically to the same region of the individual map of Fig. 4, or when considering the pair Xiap, CD80 (Δ = 0.31) which belongs to rather different regions. Furthermore, the surface map of Fig. 5 also shows that PPI targets, such as HIF-1A, have a higher overlap score Δ with non-iPPI datasets than with other PPI Figure 5. Surface map of the pairwise imbalance in chemical space of all of the datasets. In the top left corner are the Nubbe, MDDR, BindingDB, ASD, BDM, TIMBAL, and iPPI-DB datasets. The rest of the map contains the sub-datasets corresponding to MDDR (biological testing, preclinical, Phase I/II/III, and launched), e_Drug, BindingDB (ion channel, nuclear receptors, kinases, GPCR, and enzyme), and the individual PPI targets from both TIMBAL and iPPI-DB. Surface peaks are coloured based on the pairwise overlap in chemical space for two given populations. Overlap scores are between 0 and 1, such that according to the colour code, green regions demonstrate higher overlap, and red or even transparent regions demonstrate low overlap, if any, and therefore greater imbalance. targets. Finally, PPI targets, such as CTNNB1, timbal_MLL, or E1, are clearly described as outliers with respect to the rest of the targets (PPI or not).</p><p>Defining homogenous regions of the iPPI chemical space. The surface map of Fig. 5 could highlight the global positioning of each PPI target with respect to the remaining PPI targets, and that classes of PPI targets could be identified using the same metric. To confirm those observations, we then used the metric described herein as a distance criterion and proceeded to an agglomerative hierarchical clustering of all of the datasets using the Ward method 33 . The clustering was combined with a heat map (see Methods) to visualize the different classes and gauge the internal homogeneity of the resulting clusters (Fig. 6). To prevent redundancy and any bias of the results, we only kept sub-datasets. Indeed, we kept all of the PPI targets individually but removed iPPI-DB and TIMBAL as a whole. We proceeded to the same operation for MDDR and BindingDB by just keeping the sub-datasets that comprised them. Except for the datasets that were removed to prevent redundancy, the same array of pairwise overlap scores Δ that was used to make Fig. 5 was also used to feed the agglomerative hierarchical clustering using the Ward method.</p><p>As expected, clustering gathers all of the subsets of MDDR (biological testing, preclinical, phase I/II/III, launched), e_Drug, Nubbe and enzyme active compounds. Indeed, the natural compounds of Nubbe appear to be close to MDDR compounds. Similarly, active compounds on enzymes are closer to MDDR compounds because they still account for the majority of the existing active compounds and drug candidates.</p><p>Clustering showed a strong overlap between nuclear receptors, GPCRs, ion channel, BDM, ASD, and tim-bal_HIF-1A, with an overlap score Δ greater than 0.7 on average. This result confirms the observation made on the surface map from Fig. 5 that this PPI target is closer to non-iPPI datasets than to other PPI targets. It is worth noting that these datasets have also an important overlap with MDDR subsets. Finally, these results confirm that commercial libraries, such as BDM (~40,000 from three providers), are closer to active compounds on conventional targets than to iPPI. Similarly, allosteric compounds from ASD are also closer to commercial databases and GPCR compounds, which account for a substantial number of allosteric compounds, than inhibitors of protein-protein interactions. However, most interestingly, the results show a clear separation between inhibitors of most PPI targets from non-iPPI datasets and, more convincingly, of the corresponding PPI targets among themselves. Indeed, the conclusive aspects of clustering are the creation of classes of PPI targets, whose iPPI compounds seem to share physicochemical profiles. One can note the proximities of the following PPI targets: (IL2 and Xiap), (CD80 and CTNNB1), (Brd, annexinA2, and GP120), (MDM2, ITGAL and, to a lesser extent, BCL2). This pattern also confirms that iPPIs of a given PPI target from both iPPI-DB and TIMBAL, such as (timbal_IL2, IL2), (timbal_XIAP, XIAP), (timbal_brd, brd), and (timbal_MDM2, MDM2), most often fall into a highly homogeneous class (Δ~0.8). Therefore, such an approach allows the definition of homogenous regions in the iPPI chemical space. There are also PPI targets that are close to non-iPPI datasets, such as PSIP1 with BindingDB subsets; S100B, MAX, and cyclophilin with kinase; and timbal_HIF-1A as mentioned above.</p><p>Pocket-driven evaluation of the PPI target space. To further measure the heterogeneity of the PPI target space and to confirm some of the groups of PPI targets mentioned above, we used a purely pocket-driven approach without relying on the physicochemical properties of the iPPI compounds. To this end, we proceeded to analyse the pocket properties of all of the PPI targets that were present in both TIMBAL and iPPI-DB and whose experimental structures (X-ray crystallography or NMR) were available in the PDB.</p><p>Using the programs VolSite 12 and MOE 34 , we detected the binding pockets of the PPI targets and calculated a series of 117 pocket descriptors: 89 from VolSite, 10 using a combination of VolSite descriptors and 18 calculated with MOE on the negative image of the binding pockets provided by VolSite (see Methods). These descriptors were then used to calculate the Euclidean distance between pockets and to carry out a clustering (Fig. S6) of all of the available crystal structures of the PPI target binding pockets from the datasets using an agglomerative hierarchical clustering with the Ward method (see Methods).</p><p>From the clustering, it is clear that most of the structures of a given PPI target are grouped together when considering 4 clusters, with the exception of Bcl-2, which falls into two distinct groups. This target has a very large pocket that binds the BH3 domain of its partners, consisting of an 18-amino-acid-long α -helix lining the groove of Bcl-2. The present pocket is highly flexible and accommodates according to the nature of the BH3-only partners that it binds 35 . This result may explain the separation of the BCL-2 structures into two groups. Moreover, all of the structures of a given PPI target fall into the same cluster regardless of whether the experimental structures were resolved with or without a ligand (holo-or apo-form of the protein). However, more importantly, classes of PPI targets could be identified, some confirming the results of the ligand-based clustering approach that was described in the previous paragraph. Indeed, clustering unambiguously provides similar results to that carried out with the iPPIs alone. Thus, when considering 4 clusters, Brd is grouped with GP120, MDM2 is grouped with ITGAL and partially with Bcl-2, Xiap is grouped with IL-2, and the rest of the Bcl-2 structures are grouped together. It is worth noting that PSIP1 is now grouped with MDM2 and ITGAL but mostly with non-iPPI datasets before in the ligand analysis.</p><p>In order to characterize each of these 4 groups of targets, we selected the corresponding structures that clustered within each of these groups and calculated an ANOVA followed by post hoc comparisons using a pairwise test between groups and for each pocket descriptor. This process has allowed us to determine which pocket descriptors were significantly different (p-value < 1.10 −4 ) for each group with respect to the 3 remaining groups and provides a physicochemical profile of a representative pocket for each of the 4 groups. Thus, the group of Brd and GP120 is characterized by more globular and more buried pockets with more hydrogen bond acceptors. MDM2 and ITGAL are more aromatic, less hydrophobic, moderately buried, and with fewer hydrogen bond acceptors and fewer hydrogen acceptors and donors combined. Xiap and IL-2 are smaller and more exposed (shallow), rod-like with more hydrogen bond donors and with more positive and negative charges. Bcl-2 pocket are larger, more hydrophobic with negative charges, less aromatic, less globular and with fewer hydrogen bond donors.</p><p>Crossing the PPI chemical and target spaces. The fact that some PPI targets fall into the same classes using two orthogonal methods, such as pocket-based and ligand-based approaches, led us to retrospectively inspect which regions of the chemical space these groups correspond to. Thus, we decided to closely inspect the regions of chemical space enclosing the iPPI compounds of GP120, Brd, MDM2, ITGAL, Xiap, IL-2 and Bcl-2. To do so, we plotted an individual map of the PCA, including all of the corresponding iPPI compounds, by colour coding the above-mentioned groups, i.e., Group 1 (Brd with GP120), Group 2 (MDM2 with ITGAL), Group 3 (Xiap with IL-2) and Group 4 (Bcl-2) (Fig. 7). From the PCA individual map, one can note that, even for the two first components (46% of the total variance), Groups 1, 2, and 3 can be easily separated. Group 4 partially overlaps with Group 2, confirming the results of the pocket-driven clustering that Bcl-2 and both MDM2 and ITGAL share a region for their chemical and target spaces. This figure nicely confirms that such clustering can be performed independently and cohesively using iPPI compounds or PPI pocket descriptors. Not only do some of the PPI pockets seem to share properties, but these properties seem to condition the physicochemical properties of their respective ligand. More precisely, the perspective to further identify homogenous regions of the PPI target space corresponding to homogenous regions of the iPPI chemical space is key to designing focused chemical libraries that are dedicated to classes of PPI binding cavities. This process would become particularly efficient when combined with properties such as EDmin3, which manages to differentiate more generally an iPPI compound from a conventional molecule regardless of the PPI target.</p><p>To inspect the properties of the shared regions of the chemical space and of the target space, we summarized the properties of both the iPPI compounds and the PPI targets corresponding to the PPI groups mentioned above (Fig. 8). This summary demonstrates that homogenous and cohesive regions of chemical and target space can be found and that these identified groups of targets could lead to the design of PPI-class-specific chemical libraries.</p><!><p>Fundamental processes in living cells are largely controlled by protein-protein interactions. The deregulation of these interactions plays a critical role in the pathogenesis of numerous diseases. Thus, PPIs represent attractive targets even though targeting them with LMW compounds still represents a major challenge. The aim of this analysis was to capitalize on existing knowledge of the iPPI chemical space and to gauge the heterogeneity of the PPI target space. During this analysis, we confirmed that iPPIs present a specific profile according to their size, hydrophobicity, aromaticity, capacity to bind hydrophobic patches (EDmin3), globularity (glob), percentage of exposed polar groups (CW2), and location of these polar groups (IW4). Interestingly, we could distinguish properties that seem to apply to nearly all of the PPI targets, such as a higher molecular weight or the shape property EDmin3, and properties that seem to be more specific to some PPI families, such as logP, aromatic ratio, globularity, CW2 and IW4. These results highlight that more effort should be put into designing or selecting compounds that have hydrophobic chemical moieties at the right location rather than being on average simply more hydrophobic. A proper layout can be easily evaluated with descriptors, such as EDmin3. These results also confirm some discrepancies among PPI targets. The metric Δ is presented to quantify the overlap in chemical space between two molecular datasets using PCA calculated with some of the iPPI-specific descriptors cited above. The use of this metric has not only confirmed a high discrepancy among the different PPI targets, but more importantly has highlighted classes of PPI targets according to the profile of their iPPI compounds. Most interestingly, when using an orthogonal analysis based on a purely pocket-driven approach, some of the most important classes of PPI targets could be confirmed, demonstrating that homogenous regions of the iPPI chemical space can be affixed to homogenous regions of the PPI target space. We anticipate new types of PPI pockets in the near future that will either merge or supplement those described herein. However, given the profiles that are presented in this study, we might speculate that any given PPI pocket characterized by similar pocket properties may require iPPIs with similar properties. Moreover, the fact that we managed to partially propose the independent and cohesive clustering of the PPI targets regardless of whether they are processed using iPPI compounds or pocket descriptors represents a proof-of-concept that homogenous regions of the PPI chemical space can be identified and be coherent with homogenous regions of the PPI target space. This proof of concept is therefore independent of the methods used herein, and we anticipate that new methodological developments will support this strategy because not only do the PPI pockets seem to share properties, but they seem to condition the physicochemical properties of their respective ligands. This characteristic will be a strong factor in further developing such approaches in the near future and promulgating PDB-wide pocket analysis. Such large data analyses could provide essential insights into the feasibility of a given project based on the proximity of the PPI target of interest with previously chemically probed interfaces. In turn, these results could help to prioritize the choice of PPI targets based not only on cellular mechanisms or the druggability of their interfaces, but also on the chemical risks that one may encounter by investigating it. This information will certainly assist in the preparation of dedicated compound libraries tuned for classes of homogenous PPI targets and will in turn facilitate the identification of protein-protein interaction inhibitors.</p><!><p>Dataset compilation. Inhibitors. A dataset of iPPIs was constructed from iPPI-DB (version 2013) 13 (1,650 compounds) and TIMBAL (version May 2015) 3 (8,107 compounds). In order to select the more accurate dataset, several criteria were applied. The information about the PPI had to be unambiguous. The modulation had to be inhibition and not stabilization. We selected only those compounds with the following measures of activity: K d , K i , IC 50 and EC 50 . Moreover, this activity had to be less than 30 μM. We excluded small metal-based compounds, peptides, macrocycles and molecules containing atoms different from C, N, O, S, P and halogens. Finally a set of 3,248 iPPIs was constructed: 1,650 from iPPI-DB and 1,598 from TIMBAL.</p><p>We also constructed different datasets of so-called non-iPPIs from different sources. In this study, non-iPPIs are molecules that do not inhibit protein-protein interfaces: inhibitors of conventional targets, such as GPCR or enzymes; known drugs; allosteric compounds; natural products; compounds in preclinical or clinical phases; or compounds from commercial databases. The description and number of compounds in each dataset are described in Table 1.</p><p>For all of the compounds (iPPIs and non-iPPIs), we removed redundant compounds, peptide molecules defined as compounds with more than 3 contiguous peptide bonds, salts, compounds with less than 10 heavy atoms, carbocations, and molecules containing atoms different from C, N, O, S, P and halogens. Only compounds with a molecular weight of less than 1,200 g.mol−1 were selected. All of the compounds were normalized and standardized with PipelinePilot (v9.0.2) using the same standardization protocol. Interfaces. A total of 62 protein-protein interfaces were collected from the PDB 36 or 2P2I-DB 30 and referred to the PPI complexes in iPPI-DB or TIMBAL. A total of 84 inhibitor-protein complexes (from PPI) were collected from the PDB and 2P2I-DB 37 . These complexes were found across 15 different PPI families.</p><!><p>Inhibitors. A set of 16 molecular 2D descriptors was calculated in this analysis. These descriptors were calculated with a java program using ChemAxon JChem library (version 5.10.1, https:// www.chemaxon.com/). These descriptors were a set of classical descriptors and are described in Table 2.</p><p>Another set of 3D descriptors (vsurf_EDmin3, vsurf_CW2, vsurf_IW4 and glob) was also calculated using Moe software (version 2012.10) using a previously described protocol 11 .</p><p>Interfaces. A set of 89 descriptors for the interface were calculated using VolSite software 38 (version 2014). The different parameters for all of the different interfaces were the minimal number of cubes required to consider it as a cavity equal to 20, the edge length of the main box equal to 20 Å, the minimal number of neighbours for buried cavity boxes equal to 7, and the edge length of each box equal to 1.5 Å, and hydrogen atoms were considered. For the PPI interfaces, the minimal threshold for buriedness was set to 50 and 60 for the non-PPI interfaces.</p><p>All of the different interfaces were manually inspected to ensure that the detected pocket was at the right location on the protein surface.</p><p>For all of the pockets that were divided into different sub-pockets, different steps were used to combine them and obtain descriptors for a global pocket. The cavity volume was obtained by adding the cavity volume of different sub-pockets (see Equation 1for two sub-pockets).</p><p>Equation 1: The calculation of the total number of points present in the combined pocket with nb tot1 as the total number of points present in sub-pocket 1 and with nb tot2 as the total number of points present in sub-pocket 2.</p><p>The number of points of each feature for each pocket (see Equation for an example: hydrophobic points) divided by the total number of points was calculated by to evaluate the global percentage (see Equation 3) for an example: hydrophobic points) of each feature (hydrophobic, aromatic, H-bond acceptor, negative ionizable, H-bond acceptor/donor, H-bond donor, positive ionizable and dummy atoms)</p><p>Equation 2: Calculation of the number of hydrophobic points present in sub-pocket 1, with %hyd1 as the percent of hydrophobic points in the sub-pocket. For all of the evaluated pockets, 10 descriptors were added for this analysis. These descriptors were the sum of the different features (hydrophobic, aromatic, H-bond acceptor, negative ionizable, H-bond acceptor/donor, H-bond donor, positive ionizable and dummy atoms) at each distance (Equation 6). Equation 6: Calculation of the combination of descriptors at each distance.</p><p>For each evaluated pocket, VolSurf provides a mol2 file with all of the different probes. This file was used as a template in Moe (version 2012.10), and the 18 descriptors that are present in Table 3 were calculated.</p><p>Finally, a set of 117 descriptors (89 from VolSite, 10 from a combination at each distance and 18 from Moe) was calculated on the different interfaces.</p><!><p>C) and ADME-Tox filter (Table 4) were calculated using the online program FAFDrugs 3 23 (http://fafdrugs3.mti. univ-paris-diderot.fr/). All of the parameters were set to default values. Molecular descriptors analysis. All of the multivariate or univariate analyses were performed using scripts from the software package R (version 2.15.1). Graphical analyses were performed using R (version 2.15.1) or Excel (Mac 2008). A PCA was performed using the FactoMineR library, which scaled the input data prior to the analysis. The probability density functions on the PCA were calculated based on the successive principal components with the R module Density set to default parameters. Heat maps were calculated using either with heat map packages using R or with the graphics type "surface" in Excel, and dendrograms on the heat maps were implemented with the Ward method 33 using Euclidean distances.</p><p>Analysis of the discrimination between two populations. For all of the different statistical tests, only families with five or more compounds were analysed 39 . Thus, out of the 39 PPI families (13 from iPPI-DB and 26 from TIMBAL), only 28 PPI families were analysed.</p><p>The two datasets were compared using Student's test if the two datasets followed a normal distribution according to the Shapiro test and had equivalent variances according to the Fisher-Student test.</p><p>If the previous conditions were not fulfilled, comparisons were made using the nonparametric Mann-Witney-Wilcoxon test. The population mean values were compared using a nonparametric Kruskal-Wallis ANOVA. Post hoc comparisons were carried out using a pairwise Mann-Whitney-Wilcoxon test 40 .</p><p>For all of the different comparisons, the statistical tests were two-tailed with an alpha equal to 0.05 and were considered significant when the p-value was less than 0.05. Determination of the overlap between two populations. First, a principal component analysis was calculated using R and the FactoMineR package on all of the different datasets (Table 1). Then, the coordinates from the PCA were extracted for the two considered datasets (see example in Fig. S7).</p><p>On each component of the PCA and for each of the two datasets, the probability density function was calculated using the R density module set to default parameters. The overlap on one component between two datasets was the area in common between the two densities (in violet in the Fig. S7). To calculate this overlap δ on one component between two datasets, we used Weitzman's Measure (Equation 7) 41 . ∫ δ = f x f x dx min( ( ), ( )) 1 2</p><!><p>Lipinski rule of five 25 An oral bioavailability evaluation Veber 42 An oral bioavailability evaluation</p><p>Egan 42 An oral bioavailability evaluation GSK's 4/400 28 A drug safety profiling Pfizer's 3/75 43 A drug safety profiling Lilly MedChem Rules 27 Identify compounds that may interfere with biology Table 4. ADME-tox filter used on FAF-Drug3, names and descriptions.</p><p>Equation 7: Weitzman's Measure 41 with f 1 as the density for the first dataset and f 2 as the density for the second dataset.</p><p>Then, to combine the overlap for all of the components, Equation 8was used. This combination is an arithmetic mean weighted by the variance of each component (to reach 100% of the cumulated variance). Finally, an overlap score Δ for each pair of datasets was obtained. This score was between 0 and 1, with 0 meaning completely dissimilar datasets and 1 meaning identical datasets. These operations were performed for all of the datasets versus all of the components to obtain an N× N array of Δ scores.</p><p>Equation 8: Weighted arithmetic mean with δ i is the overlap between the two densities for the axis i from the PCA, Vi is equal to the variance percentage for the axis i from the PCA and n is the number of dimensions to reach 100% of the variance.</p><p>Interface analysis. Analysis of Hclust. Clusters were calculated on the different binding pockets using Euclidian distances and the Ward method within the Hclust R package. A comparison test was performed on each different group of the clustering. The corresponding structures of each group were selected, and an ANOVA was calculated followed by post hoc comparisons using a pairwise test between groups. The population mean values were compared using a nonparametric Kruskal-Wallis ANOVA. Post hoc comparisons were carried out using a pairwise Mann-Whitney-Wilcoxon test 40 . For all of the different comparisons, statistical tests were two-tailed with an alpha equal to 1E-04 and were considered significant when the p-value was less than 1E-04. Therefore, only very significant discriminating descriptors were selected to describe each group.</p>

Scientific Reports - Nature

Decrypting a cryptic allosteric pocket in H. pylori glutamate racemase

One of our greatest challenges in drug design is targeting cryptic allosteric pockets in enzyme targets. Drug leads that do bind to these cryptic pockets are often discovered during HTS campaigns, and the mechanisms of action are rarely understood. Nevertheless, it is often the case that the allosteric pocket provides the best option for drug development against a given target. In the current studies we present a successful way forward in rationally exploiting the cryptic allosteric pocket of H. pylori glutamate racemase, an essential enzyme in this pathogen's life cycle. A wide range of computational and experimental methods are employed in a workflow leading to the discovery of a series of natural product allosteric inhibitors which occupy the allosteric pocket of this essential racemase. The confluence of these studies reveals a fascinating source of the allosteric inhibition, which centers on the abolition of essential monomer-monomer coupled motion networks.

decrypting_a_cryptic_allosteric_pocket_in_h._pylori_glutamate_racemase

H<!>Results<!>Occupancy of the allosteric site by either NP-020560 or<!>Determination of the source of the disruption in allosteric communication by small molecule inhibitors of H. pylori GR.<!>Discussion<!>Methods

<p>elicobacter pylori (H. pylori) is a spiral-shaped Gramnegative pathogenic bacterium estimated to infect more than half of the world's population 1,2 . H. pylori infections may persist for a lifetime if treated inappropriately and have been associated with numerous gastrointestinal diseases like chronic gastritis, gastric ulcers, and several forms of gastric cancer 3,4 . Current treatment options involve the use of relatively broad antibiotics, which do not specifically target H. pylori 5,6 . The current work focuses on structure-based design targeting an essential enzyme in the H. pylori life cycle, glutamate racemase (GR), which catalyzes the stereo-inversion of glutamate. D-glutamate (D-Glu) is an essential component of the peptidoglycan layer of bacterial cell walls, which protects the microbe from osmotic rupture and proteolytic damage 7 . Importantly, although glutamate racemases are ubiquitous in the bacterial world, the H. pylori GR possesses a unique cryptic allosteric pocket, which has been the subject of robust drug discovery efforts, from both the pharmaceutical industry [8][9][10][11][12] and academia [13][14][15][16][17][18] .</p><p>In one of the most successful studies to date, researchers at AstraZeneca performed a large high-throughput screen (of ~400 K compounds) and identified a single viable lead compound, a pyrazolopyrimidinedione (designated 'Compound A' -Fig. 1) as an uncompetitive allosteric inhibitor of H. pylori GR 8 . Lundqvist et al., reported that Compound A had an IC 50 value of 5 μM, and X-ray crystallography studies showed it to be binding in a cryptic allosteric pocket remote from the active site (Fig. 1). Despite extensive optimization cycles, no clinical candidate emerged, due to poor solubility and high protein binding for this class of compounds 9,10,12 . Thus, despite being a validated antibiotic drug target, there is no clinically available inhibitor of H. pylori GR. A compounding factor is that a mechanism of action, by which occupancy of the cryptic allosteric pocket remotely leads to inhibition has never been elucidated. Indeed, it is difficult to imagine gaining medicinal chemistry traction on this target in the absence of such knowledge. This is a trend that has plagued allosteric drug discovery in general 19,20 . In Lundquist et al., the authors hypothesized that uncompetitive inhibition occurs due to Compound A binding via limiting a hinge movement in each monomer, thus restricting substrate-product release 8 . Our group recently disproved the proposed uncompetitive kinetic mechanism of H. pylori GR, and proposed an alternative mechanism of action, based on computational and experimental data 15 . MD and QM/MM studies in Witkin et al., found that H. pylori GR goes through a large conformational change that facilitates the acidification of the substrate Cα-proton via a protonation of the α-carboxylate oxygen from the catalytic Cys185; the entire process is a key part of a pre-activation step that occurs before the racemization catalytic cycle (Supplementary Fig. 1) 15 . However, essentially nothing is known about how occupancy of the allosteric pocket by this inhibitor promotes dampening of the chemical activation process that makes the extraordinarily difficult chemistry of racemization possible.</p><p>An important scientific and pharmaceutical question is whether or not structure based discovery approaches can be successfully applied to a cryptic allosteric pocket, as in the case of H. pylori GR, especially given our limited understanding of the manifold forms that such a pocket may take; it is particularly problematic due to the global flexibility of GR and a lack of experimental structural data about the apo form of the enzyme 7 . Although allosteric drugs hold enormous promise, the truth is that only 19 of more than 3700 FDA approved drugs are allosteric, and only a single drug was derived using approaches centering on in silico/structural methods 20 . These are extremely challenging problems, which necessitate the use of methods that stress test our computational and structural approaches. In this study we report an MD/Docking workflow which is stress tested via the Receiver Operator Characteristic (ROC) statistical method, and results in the successful discovery of a new allosteric inhibitor of H. pylori GR, but with a completely different chemical space than Compound A, proving that the workflow presented herein is able to address the challenges presented by cryptic allosteric pockets. Importantly, the MD-based discovery process itself reveals a distinct allosteric communication in the native system (no inhibitor) which occurs between monomers of the dimeric enzyme via a dynamic C-terminal α-helix, which is dramatically dampened by complexation with the allosteric inhibitor. The dynamical data is strongly supported by the pattern in the crystallographic B-factors in the inhibited and uninhibited complexes of H. pylori GR, which has not been previously recognized. These findings constitute a definitive and novel type of Allosteric Structure Activity Relationship (ASAR).</p><!><p>Overall summary of structure-based discovery workflow for cryptic allosteric pocket. In the current study we report a virtual screening regime developed specifically for highly flexible enzyme targets like H. pylori GR and which contain a cryptic allosteric pocket (Fig. 2). Here we briefly describe the broad outlines and philosophy of this approach, while a more detailed description can be found in the Methods section. The central feature of this approach involves selection of the "best-performing" combination of receptor (from MD trajectories) and docking protocol/program for virtual screening. The workflow employs a hybrid Molecular Dynamics-Docking approach involving all atom classical MD simulations followed by global structural clustering of MD snapshots to generate different receptors forms. Known actives and decoys were docked into each form of the receptors using different docking protocols. The results were then analyzed using the statistical analysis technique of Receiver Operating Characteristic (ROC) curves [21][22][23][24] , to select the best performing MDreceptor/docking protocol pair (defined in the ROC approach as its ability to separate true positives from false positives given a user-defined selection threshold value (i.e., docking score), which will be elaborated on below) (Fig. 2). Due to the very low hit rates obtained in HTS screening against H. pylori GR using drug-like synthetic compounds observed by Lundqvist et al. 8 , and the solubility challenges with Compound A, 9,10,12 along with the aim to challenge our workflow with unconventional chemical matter, we chose to focus on in silico screening of natural product libraries. In silico screening of natural product libraries, which are commercially available in a purified form, have recently been used to great effect against difficult drug targets 24 . Importantly, natural products provide an unrivaled success rate in drug lead discovery [25][26][27] , but have been under-exploited in in silico approaches, often due to the difficulty and expense in obtaining purified compounds. Thus, the best performing MD-receptor/ docking protocol pair was employed to screen the AnalytiCon's MEGx natural products library, which resulted in 177 hits that scored better than the cut-off established by ROC analysis (as designated by our defined threshold docking score (Fig. 2)).</p><p>A post-docking optimization protocol, which employs an automated all atom MD simulated annealing energy minimization, with explicit solvation (referred to as SAEM 28 ) followed by local docking, which is described in detail in Hengel et al., 24 was used in affinity rank ordering the docked poses. The resulting top five virtual hits were identified, and these natural products were purchased in pure form (>95% purity) from AnalytiCon Discovery GmbH (Potsdam, Germany), and experimentally evaluated for binding to GR via surface plasmon resonance (SPR) which yields stoichiometric binding information, and is thus explicitly able to differentiate specific binders from nonspecific effectors, and will be discussed in more detail below. The utility of SPR is additionally important in the context of natural product screening, due to the scarcity and expense of hits. Compounds that were shown to successfully bind to H. pylori GR via SPR were then evaluated for their ability to inhibit GR's enzymatic activity in a coupled-enzyme assay. Four of the five compounds bound to the protein with micromolar affinity and inhibited H. pylori GR with varying degrees of potency and are described in detail below. The reader is directed to a detailed description of this workflow, centered around MD and ROCbased selection, in the Methods section.</p><p>Finally, we performed a 50 ns MD simulation on one of these novel active allosteric inhibitors (NP-020560, which was discovered from the workflow outlined above) bound to the H. pylori GR-D-Glu complex, as well as a similar study on Compound-A bound to the GR-D-Glu complex. Dynamic cross correlation matrix (DCCM) analyses of these systems clearly indicated specific suppression of inter-monomer correlated motions in the inhibited systems. The loss in inter-monomer correlated motions has never been studied in the case of GR and provides a much-needed mechanism of action for allosteric inhibition of H. pylori GR by Compound A and NP-020560.</p><p>Identification of the optimal receptor-docking pair via ROC analysis. The selection of the appropriate protein receptor structure is of paramount importance when employing structurebased drug discovery approaches. This problem is compounded when working with highly flexible enzymes 29,30 . The use of receptors derived directly from crystal structures raises questions about whether ligands that bind in the same pocket associate with similar protein confirmations in solution. Ultimately, what is most useful in the in silico screening process is identification of receptor forms that can best differentiate between actives and inactives. Using MD simulations to sample additional conformational space and refining receptor selection is a known approach for improving screening outcomes [31][32][33][34] . To implement this, we performed all atom classical MD simulations on two available H. pylori GR dimer-D-Glu-allosteric inhibitor complexes (PBD ID: 2JFZ 8 and 4B1F 9 ) followed by global structural clustering of MD snapshots to generate different receptors forms (Fig. 3a highlights the allosteric site residues used for clustering). Previous work by Jeremy Smith et al. on a wide range of MDderived receptors found that there was no a priori characteristics that enabled them to know which snapshots of an ensemble would yield the best receptor for identifying true positives 35 . To address this and to identify a receptor-docking pair capable of enriching for true positives while minimizing false positives and negatives, we utilized a statistical method employing the 'Receiver Operating Characteristic' or ROC curves approach [21][22][23]36 , which is fully described in the Methods section, which relies on the use of decoy compounds generated with the DUD-E application 37 for use as negatives. Briefly, the compounds were docked, and the final docking scores (kcal/mol) were then used for generating ROC curves which were analyzed based on their area under the curve (AUC) (Fig. 3b, c shows ROC curves of the best performing MOE and FlexX data sets; Table 1 shows the tabulated ROC Fig. 2 Biphasic workflow for natural product inhibitor discovery targeting the allosteric cryptic pocket of H. pylori GR. The graph is divided into two phases, depicted on the left and the right, respectively. On the left half of the chart, we describe the selection of the best receptor and docking protocol combination; snapshots from MD simulations of ~800 ns were clustered and centroids were examined for their ROC performance (ability to discriminate between decoys and known hits). The best performing pair was "2JFZclstr3" as the receptor (derived from the PDB structure 2JFZ, as described in the Methods section) and docking was performed with FlexX (BiosolveIT), as described in the Methods section. The flowchart on the right-hand side is the actual virtual screening protocol that employs the validated 2JFZclstr3-FlexX (receptor and docking protocol) pair, including the experimental biophysical hit validation using SPR. The screening library employed was AnalytiCon's MEGx natural products library. data). An ideal curve would reach the upper left corner of the graph and have AUC of 1.0 (Fig. 3c green trace). The analysis also enables the determination of the desired docking score cutoff to use for highest enrichment for true positives while avoiding false positives and negatives (Table 1), which is one of the great advantages of the ROC approach [21][22][23]36 . In general, all forms of 2JFZ outperformed 4B1F while FlexX outperformed MOE (Table 1). MD cluster 3 for 2JFZ (2JFZclstr3) when used for docking using FlexX performed the best with AUC 0.9671 and score cutoff of −20.09 kcal/mol and thus was used for performing the virtual screening (Fig. 3c). Importantly, superimposition of Compound A's position in the cryptic pocket using this approach was nearly identical to the co-crystallized structure (Supplementary Fig. 2) with an RMSD of 0.883 Å (lower than the crystal structure resolution of 1.86 Å), validating the protocol. Additionally, we could not find any simple pocket metric (such as volume, polarity, or druggability metrics 38,39 ), which would predict the ROC performance of any given receptor. In other words, the elucidation of the superior performance of 2JFZclstr3 in lead selection (Table 1) compared to other receptor forms must be directly determined by ROC plots. Nevertheless, simple pocket scoring metrics did correctly identify that MD clustered forms of the protein had better predictability (in terms of ROC performance) than crystal structure forms (Table 1), which is highly encouraging.</p><p>In silico screening using the ROC-validated receptor-docking protocol. The AnalytiCon MEGx Natural Products Library was Fig. 3 Use of the ROC method for identifying the best docking-receptor pair. Clustering MD snapshots based on allosteric pocket geometries and use of these receptors in the Receiver Operator Characteristic (ROC) statistical procedure for identifying the best combination of receptor and docking protocol for the resolution of true hits and decoys. a MD snapshots were clustered based on the structure of the residues in the allosteric pocket (shown in magenta; see Computational Methods for details). 2JFZ provided 6 clusters and 4B1F provided 4 clusters in addition to average and low energy forms for each. b, c The ROC distributions for the best performing MD cluster for the MOE (b) and FlexX (c) docking protocols, respectively. The curve measures the ability of the docking procedure to resolve true positives and true negatives. The fraction of true positives that score above a user defined threshold (sensitivity) is plotted on the y-axis and the fraction of true negatives (specificity) that score above the threshold is shown on the x-axis. The theoretically perfect curve is shown in green, while the average score for both the MOE and FlexX protocols with cluster '2JFZclstr1' and '2JFZclstr3' is shown in red, respectively. The dotted lines show the standard deviation for the fit. screened using the validated approach for FlexX described above (fully described in Methods). This yielded an initial hit rate of 4.7% (not unusually high for in vitro natural product screens) or 177 compounds that scored above the optimum cutoff value of −20.859 kcal/mol. Since docking programs are optimized to best predict the optimum binding pose for a ligand in complex with a macromolecule, we employed an optimized version of our previously developed computationally expensive Simulated Annealing Energy Minimization (SAEM) docking approach as a secondary refinement screen to rank-order the virtual hits based on predicted affinity (fully described in Methods) 28 . Visual inspection of the top high scoring compounds indicated good shape complementarity between ligand and the allosteric site, π-stacking and/or hydrophobic interactions with Trp-252 (major interaction for Compound A), presence of multiple hydrogen bonding and other hydrophobic interactions with allosteric site residues and lack of steric constraints in the docked poses. Natural product compound selection for experimental testing was based on the SAEM binding score, compound cost (which varies widely in natural product libraries) and availability. Five compounds were selected for experimental evaluation (Table 2 lists these top five compounds with their structures scores).</p><p>Experimental validation of natural product inhibitor hits from in silico screen and ROC analysis. We first evaluated the binding affinity of these compounds to H. pylori GR by Surface Plasmon Resonance (SPR). A combination of amine coupling and His-Capture was used to immobilize the protein on the surface of the SPR biosensor. Compounds were injected onto the sensor in increasing concentrations using a dilution series ranging from 500 µM to 7.81 µM (7 concentrations). Using a Langmuir 1:1 binding model, the K d for the virtual hits was obtained as listed in respectively (Fig. 4a, b, Table 2). The binding curves for the evaluated hits and control compound represent stoichiometric 1:1 binding with reasonable on and off rates and overall binding response (RU), and do not indicate any non-specific binding to the protein or chip. The binding affinity of the rest of the three compounds is as listed in Table 2 and the SPR graphs are as shown in Supplementary Fig. 3a-c. These binding experiments indicate that compound A and our hits from virtual screening bind to H. pylori GR in the absence of the substrate (D-Glu), which is contrary to the uncompetitive nature of Compound A originally proposed by Lundquist et al. 8 . After establishing that the virtual hits bind to H. pylori GR, we next evaluated the ability of these hits to inhibit the enzymatic activity of the enzyme employing a previously established coupled-enzyme assay 8,40 .</p><p>Four of the five tested compounds inhibited GR to different degrees, with NP-020560 being the most potent, having an IC 50 of 6.6 ± 3.1 µM (Table 2 and Fig. 4c), while the other compounds had IC 50 values in the high micromolar range (Table 2 and Supplementary Fig. 3d-f). NP-020560 has an IC 50 value within statistical error of Compound A, but clearly binds more strongly than Compound A when measured by SPR (Table 2). Additionally, the difference in K d (54.7 ± 0.3 μM) and IC 50 (6.6 ± 3.1 µM) for NP-020560 indicates that it binds tighter to H. pylori GR in presence of its substrate (D-Glu), and this trend is similar to one observed for compound A as reported in the literature, in which changes in intrinsic fluorescence were used to determine K d values with and without D-Glu. 15 The source of these disparities is due to the enormous global conformational changes in GR associated with glutamate binding. 15 .</p><p>The other natural product leads all have about an order of magnitude weaker IC 50 values than Compound A.</p><p>The consensus pharmacophore from the structures of bound natural product hits. Next, we examined the binding mode of NP-020560 in the allosteric pocket in more detail. As shown in Fig. 4d, e & Supplementary Fig. 4a, NP-020560 occupies the allosteric pocket making several polar and hydrophobic interactions. One of the anthracenedione ring of the biantracene ring system in NP-020560 is inserted into the inner pocket making the key π-stacking interaction with Trp-252 along with forming several hydrophobic interactions with the side chain of Leu-186. The second anthracenedione ring forms an edge to face π-stacking interaction with Trp-244. The hydroxyl and ketone moieties on NP-020560 form several hydrogen bonding interactions with several pocket residues including Ser14, Ile-149, Ser-152, and Gln-248. As shown in Supplementary Fig. 4 the remaining four virtual hits also make similar binding contacts with π-stacking interaction with Trp-252 and several hydrogen bonding interactions including with Gln-248 being the key contacts.</p><p>The salient feature of allosteric inhibition is dampening of a dynamic C-terminal α-helix. A superimposition of structures 2JFX (H. pylori GR-D-Glu) and 2JFZ (H. pylori GR-D-Glu-Compound A) shows very few distinct differences, with a notable exception around the C-terminal helix, which can be seen to significantly move in order to accommodate the allosteric inhibitor. However, Fig. 5a also shows that there is a very large increase in the B-factors for the C-terminal region, which does not appear to be connected with crowding from the neighboring units. These observations suggests that there may be useful dynamical information contained within the changes in the normalized B-factors between these two data sets, which has heretofore not been addressed.</p><p>Having identified the above natural product allosteric inhibitors of H. pylori GR provided us with a unique opportunity to further probe this mechanism of allosteric inhibition, in addition to that of complexation with Compound A. An all-atom MD simulation was performed on 2JFX (uninhibited GR, bound to D-Glu), 2JFZclstr3 (GR inhibited allosterically by Compound A, bound to D-Glu) and GR allosterically inhibited by NP-020560 (from 2JFZclstr3, as described above, also bound to D-Glu). The simulations were performed using the AMBER14 force field in YASARA Biosciences software package for 50 ns with snapshots being collected every 100 ps (see Methods section for simulation details) [41][42][43] . In Fig. 5b, c we compare the ΔRMSF per residue (inhibited minus uninhibited, using NP-020560), obtained from the MD simulations juxtaposed with the Δnormalized B-factors per residue (inhibited minus uninhibited, 2JFZ-2JFX). The juxtaposition between the two plots is striking; there is a clear similarity in the major peaks and troughs between the MD and X-ray crystallography data. Notably, there is a large trough in the C-terminal α-helical region, signifying a loss of motion in this region in both the MD and crystallography data upon binding the allosteric inhibitor. Additionally, comparison of panels b and c, shows large peaks centered around residue 200 and 171, for both MD and X-ray crystallography data, which indicates two small loop regions that are positioned at the opposite end of α-helices 156-168 and 187-199, which point into the allosteric pocket. Extraordinarily, we see that changes in B-factors, due to binding of an allosteric inhibitor can be directly traced to dynamical changes in the enzyme as calculated from MD simulations of these systems. The ΔRMSF per residue was also calculated from the MD simulations with Compound A (Supplementary Fig. 5a), which, not surprisingly, yields the same pattern as with NP-020560. Additionally, the ΔRMSF per residue (for both NP-020560, Supplementary Fig. 5b, as well as Compound A, Supplementary Fig. 5c) and Δnormalized B-factors per residue (for Compound A, Supplementary Fig. 5d) for the B-monomer exhibits the same pattern, albeit with a lower signal-to-noise ratio, which is an asymmetry in the dimer that is more fully evaluated in the DCCM analysis discussed below. Overall, these data provide an extraordinary structural nexus between atomistic simulations and experimental data, due to the action of allosteric enzyme inhibition, which is the foundation for the deeper insights from observed changes in correlated motion, vide infra.</p><!><p>Compound A yields non-optimal Cα proton transfer geometries. Glutamate Racemase performs an exotic cofactorindependent "two base" (Cys181 and Cys70) racemization reaction to catalyze the stereo-inversion of glutamate 15,44,45 . Computational studies using MD/QM/MM have found that there is a Cys181-mediated "pre-activation" step, which enables racemization of D-Glu by GR, in the absence of any cofactors such as chelating metals or PLP (a remarkable chemical feat) 15 . Cys181 (one of the two flanking bases) donates a proton to α-carboxylate oxygen of the D-Glu substrate; model studies show this change in substrate protonation state lowers the pKa of its α-carboxylate carbon from 22.8 (carboxylate form) to 14.4 (carboxylic acid form) (Fig. 6a), dramatically reducing the overall barrier for Cα proton transfer (the difficult and rate determining step in amino acid racemization). This chemical pre-activation is linked to larger global allosteric motions in the H. pylori GR dimer, which remain, heretofore, undefined. The use of MD simulations with both the newly discovered natural product inhibitor, as well as Compound A, have shed significant light on how such an allosteric path is occurring, which is defined below. 2). The binding affinity of the rest of the three compounds is as listed in Table 2 and the SPR graphs are as shown in Supplementary Fig. 3. These binding experiments indicate that compound A and our hits from virtual screening bind to H. pylori GR stoichiometrically in the absence of the substrate (D-Glu). c Evaluation of inhibitory activity of NP-020560 against H. pylori GR employing a previously established coupled-enzyme assay 40 . Four of the five tested compounds inhibited the enzyme to different extents with NP-020560 being the most potent, having an IC 50 of 6.6 ± 3.1 µM (which is roughly equivalent to the IC 50 for compound A, see Table 2), while the other compounds had IC 50 values in the high micromolar range (Table 2). Values represent mean ± SE, n = 3. d, e NP-020560 fits into the allosteric pocket making several polar and hydrophobic interactions. One of the anthracenedione ring of the biantracene ring system in NP-020560 is inserted into the inner pocket making the key π-stacking interaction with Trp-252 (a key recognition feature of Compound A as well) along with forming several hydrophobic interactions with the side chain of Leu-186. The second anthracenedione ring forms an edge to face π-stacking interaction with Trp-244. The hydroxyl and ketone moieties on NP-020560 form several hydrogen bonding interactions with several pocket residues including Ser14, Ile-149, Ser-152, and Gln-248. the light purple cartoon structure represents the neighboring unit of 2JFZ and the teal structure is the neighboring unit of 2JFX. An analysis of these neighboring asymmetric units shows no clashes that could account for these selective large differences in B-factors in the C-terminal region, suggesting that dynamics may be an important factor. b, c A juxtaposition between computational (MD simulations) and experimental (X-ray crystallography) data. In panel b, the changes in RMSF (Å) between the inhibitor-bound system (GR-D-Glu-NP-020560) and the inhibitor-free system (GR-D-Glu) are shown for MD simulations as a function of the residue number. In c, the changes in normalized B-factors between the inhibitor-bound structure (2JFZ: GR-D-Glu-Compound-A) and the inhibitor-free system (2JFX: GR-D-Glu) are plotted as a function of residue number. In both cases, b and c, the data are derived from subtracting the uninhibited data from the inhibited data (Supplementary Fig. 6). In terms of the effects of these dynamics on active site catalytic chemistry, the orientation and distance of Cys181 from the αcarboxylate oxygen of D-Glu is important for the for the preactivation step (i.e., protonation of the Cα-carboxylate oxygen). Only after the proton transfer to the α-carboxylate oxygen does the pKa of C-α hydrogen reduce enough to facilitate its abstraction by Cys70 15 . During MD simulation of H. pylori GR inhibited by NP-020560, Cys181 rotates frequently between conformations facing towards and away from D-Glu (Fig. 6b), relative to the uninhibited simulation. More importantly, the distance between Cys181 and α-carboxylate of D-Glu is considerably higher in the presence of NP-020560 as compared to uninhibited state of GR (Fig. 6c, d), hindering pre-activation chemistry. Overall, what we observe is that when NP-020560 is bound in the allosteric pocket, it alters the distance and orientation of Cys181 (part of the active site) and predisposes the enzyme to a conformation that is unfavorable to perform the racemization.</p><!><p>To further understand the effect of occupation of the allosteric pocket on the overall dynamics of GR, a dynamic crosscorrelation matrix (DCCM) analysis was performed on various complexes of H. pylori GR. DCCM provides information on dynamic relationships between all residues of the protein shedding light on how each residue's motion correlates with every other residue during a MD simulation. (A full description of this technique is located in the Methods section, and a comprehensive application to GR can be found in Dean et al. 46 ). To explore the effect of NP-020560 on the entire GR dimer, the DCCM for GR-NP-020560-D-Glu was subtracted from the DCCM of native GR-D-Glu. This provided a difference DCCM (ΔDCCM) as shown in Fig. 7a and it highlights the coupled motions that are lost (or gained) in the presence of NP-020560. The color bar of the topographical map in Fig. 7a indicates the intensity of the loss in coupled motion due to the binding of the NP-020560 inhibitor, with red being the greatest loss and blue being a gain in coupled motion. As we can see, most of the topography is very small and close to zero, which brings the important structural dynamics into stark relief. Indeed, the ΔDCCM is particularly insightful for our purposes of trying to define a set of allosteric structure activity relationships (ASAR) for this difficult to understand inhibition mechanism. The salient loss of coupled motion is represented in the striking "L-shaped" pattern, which corresponds to an interaction between monomers, which can be pinpointed to the C-terminal α-helix (Fig. 7a). Specifically, the C-terminal residue of monomer A (residues 235-255) loses coupled motion with several key helices and sheets in monomer B (Fig. 7a, b). Again, the C-terminal α-helix composes nearly half of the residues making up the allosteric site with Trp252 being the key residue that forms π-stacking interactions with both NP-020560 and Compound A. Importantly, this same pattern also appears in the ΔDCCM for Compound A (Supplementary Fig. S7).</p><p>Based on the previous and current observations, it is clear that the pre-activation step (protonation of the Cα carboxylate) is key to the ability of GR to carryout stereo-inversion of D-Glu and is dependent on the highly flexible and dynamic nature of GR. Occupancy of the cryptic allosteric pocket by a small molecule dampens coupled motions between two monomers of GR (and this is the operative mechanism for Compound A as well). This reveals the existence of a deeper monomer-monomer interaction in GR in the uninhibited state, the inhibition of which is a novel means of allosteric control by a small molecule, and for the first time explains why GR enzymes in general are almost always found as active dimers. Indeed, this supports the possibility that H. pylori GR must be a viable dimer to have catalytic activity. This fundamental knowledge about how H. pylori GR is harnessing catalytic power via monomer-monomer cross-talk is a very unexpected and important result from this allosteric drug discovery campaign.</p><!><p>Helicobacter pylori glutamate racemase is an essential protein for microbial survival and is an attractive drug target. Due to its high inherit flexibility and poor druggability at the active site, application of classic structure-based design approaches has been Fig. 7 Topology map of the difference between the DCCM for the uninhibited system (GR-D-Glu) and the inhibited system (GR-D-Glu-NP-020560). ΔDCCM = DCCM-Uninhibited -DCCM Inhibited . The original DCCM plots are shown in Supplementary Fig. 8. Panel a shows that most coupled motions are not different between the two systems, except for a strong region of positive changes in coupled motion are seen at precisely the interaction of subunit A's C-terminal α-helix (residues 235-255) and a number of structural elements of subunit B; this salient loss of coupled motion is represented in the striking "L-shaped" pattern, which corresponds to an interaction between monomers, which is surprising and unprecedented in the GR enzyme family. The affected areas include large portions of the active site, which encompasses the catalytic Cys181, which is responsible for the substrate acidification. Panel b shows this key coupled motion, which is lost upon complexation with the allosteric inhibitor, mapped onto the dimeric structure, in which the C-terminal α-helix is shown in green and the regions it is coupled to are shown in red.</p><p>challenging. These challenges were overcome by developing a hybrid molecular dynamics-based docking workflow, which targeted a cryptic allosteric pocket, taking into account both global dynamics in the enzyme as well as the fitness of the receptor for selecting true positives in in silico screening. The key steps in the protocol include an all-atom MD simulation on the starting protein to evaluate additional conformations followed by using a decoy library to challenge the workflow and identify the best performing receptor-docking pair for virtual screening. The power of the developed workflow was demonstrated by screening and identifying natural product inhibitors of H. pylori GR.</p><p>Additionally, molecular dynamics studies with and without the allosteric inhibitor yielded a striking comparison to changes in the normalized B-factors from the crystallographic data, strongly pointing to dynamics as the source of these differences. Analyses of these MD simulations uncovered the existence of monomer-monomer interactions that are dampened by the occupancy of NP-020560 (or Compound A) in the cryptic allosteric pocket. This further reveals a novel mechanism of allosteric control which works by dampening the coupled motions of several key helices within and between monomers. The outcome is a reduction in the flexibility of GR which in turn also orients Cys181 in an unfavorable orientation to carry out the key pre-activation step of Cα-carboxylate protonation, thus inhibiting the ability of GR to stereo-invert D-Glu. Importantly, these studies suggest that the native (i.e., uninhibited) dimeric GR relies on a complex monomer-monomer allosteric crosstalk for its catalytic activity. These studies provide a foundation for an allosteric-centered structure activity relationship for targeting H. pylori GR, and a rationale for why GR enzymes largely operate as obligate dimers.</p><!><p>Clustering of receptors. We performed all atom classical MD simulations on two available H. pylori GR dimer-D-Glu-allosteric inhibitor complexes (PDB ID: 2JFZ 8 and 4B1F 9 ) using the YAMBER3 knowledge-based force field in YASARA biosciences [41][42][43] . We chose the two protein complexes because of differences in resolution (4B1F = 2.05 Å; 2JFZ = 1.86 Å) and the presence of structurally analogous inhibitors with different affinities for GR. 2JFZ lacked a loop which was build using the homology model functionality in YASARA Structure package from YASARA biosciences 42 . Before starting the MD simulations, the starting structures were energy minimized followed by generation of periodic simulation cell with explicit solvent extending 10 Å from the surface of complexes. The simulation cell was then neutralized with NaCl (0.9% by mass) as described previously 47 . The MD runs were performed in YASARA following methods as described in Dean et al. 46 , with minor modifications. The simulations were performed using YAMBER03 force field 41 that uses long-range Coulomb electrostatic potentials calculated using Particle Mesh Ewald 48 method with a van der walls cutoff of 7.86 Å. The calculations were ran using NVT ensemble at a temperature of 25 °C and pH 7.4 for a total of 700 ns for 2JFZ and 120 ns for 4B1F with snapshots collected every 100 ps. The root-mean-square deviation (RMSD) of the structures was performed using YASARA MUSTANG8 from YASARA Biosciences, while the MD trajectories were analyzed using MDAnalysis toolkit 42 . Lastly, the MD snapshots produced by YASARA were clustered using the Ensemble Cluster tool of the UCSF Chimera package 49 which applies the methodology of Kelley et al. 50 (developed by the Resources for Biocomputing, Visualization, and Informatics at the University of California, San Francisco and supported by NIGMS P41-GM103311). The ensembles achieved global RMSD convergence at 60 ns for 2JFZ (Supplementary Fig. 9a) and at 15 ns for 4B1F (Supplementary Fig. 9b) and trajectories from the equilibrated portion of the MD run were clustered based on cryptic allosteric site residues: Val10, Gly11, Phe13, Ser14, Lys17, Ile149, Glu150, Ser152, Leu154, His183, Leu186, Trp244, Gln248, Trp252, and Leu253. These residues were all within 5 Å from compound A in 2JFZ (Fig. 3a). The resulting centroids from clusters representing 75% of MD trajectories (Table 1) along with the time-averaged, low energy and PDB deposited structures of both receptors were used for further studies.</p><p>Generation and preparation of known actives and decoys 'test' compound library. A library of known inhibitors (actives) and decoys was generated to test the ability of the workflow to correctly identify true positives and minimize false positives and false negatives. A library of 65 active pyrazolopyridiminedione analogs was compiled from published AstraZenaca studies and respective inactive/ decoy compounds were generated using the Database of Useful Decoys -Enhanced (DUD-E) website (Supplementary Table 1) 37 . In short, decoys are propertymatched to true actives using physicochemical properties like molecular weight, logP, number of hydrogen bond donors and acceptors, etc. but are distinct in chemical topology to true binders. On average, one decoy was retained for each active compound to give a library of 144 compounds (Supplementary Table 1) that was preprocessed in Molecular Operating Environment (MOE) program by adding hydrogens, adjusting partial charges and energy minimizing to give the final library used for the study 38 .</p><p>Docking known actives and decoys 'test' compound library. Two different docking programs where then used since each takes a different approach for ligand placement and for searching minimum energy confirmations: FlexX (part of Lea-dIT available from BioSolveIT GmbH) 51,52 and MOE 2016 38 . The 144-compound test library was docked into the cryptic allosteric pocket of each of the 16 protein structures (9 for 2JFZ and 7 for 4B1F) (Table 1).</p><p>For MOE, the site for docking was defined by selecting the protein residues within 5 Å from co-crystallized ligand. Docking parameters were set as Placement: Triangle matcher; Scoring function: London dG; Retain Poses: 30; Refinement: MMFF94x force field based refinement; Re-scoring function: MM/GBSA dG (this includes an implicit solvation energy calculation and captures changes in the solvent exposed surface area of the pose, and is a highly parameterized version of the popular MM/PBSA and MM/GBSA methodologies 53,54 ; Retain poses: 1.</p><p>FlexX is a module available within the LeadIT software package and it predicts protein-ligand complexes by fragmenting a ligand at rotatable bonds, determining and docking a base fragment, and incrementally building up the ligand 51,52 . For docking, individual protein were loaded into FlexX and the binding site was defined by selecting the protein residues within 5 Å from co-crystallized ligand (Compound A). Docking parameters were set at default with base placement employing the hybrid approach and retention of 1 pose per compound. The program performed well with the best pose for Compound A having rmsd of 0.883 Å Receiver Operator Characteristic (ROC) using decoys. To identify the best docking protocol capable of enriching for true positives and reduce false positives and false negatives we used a statistical method employing ROC curves 21 . For generation of the curves, all 'actives' were assigned the value of 1, while all decoys were named as 0. The corresponding docking score for each compound for a receptor-software pair was noted. ROC curves were then created using MedCalc program for each of the receptor-software pair by scoring ranks of actives vs inactive poses 55 . The generated plot represents 'False positive rate' on the x-axis and 'True positive rate' on the y-axis for a wide range of cut-off scores, and helps examine how well a classifier (i.e., the cutoff threshold chosen for a particular scoring function) is capable of correctly selecting actives and discarding inactives (Fig. 3b, c, Table 1). The curves were analyzed using the metric of the area under the curves (AUC) 22 . An ideal curve would reach the upper left corner of the graph, while a random classifier would cross the diagonal of the graph area (Fig. 3c). The best performing receptor-software pair had curves reaching close to ideal curve with an AUC close to 1.</p><p>Rationale for library selection. Natural product libraries are often the best place to begin a screening campaign, having a storied history in drug discovery. However, the expense of isolation, purification, and structural validation have often outweighed their advantages. In silico screening from commercially available collections of natural products opens up new possibilities for structure-based drug discovery. Although employing a natural product library can be about an order of magnitude more expensive than traditional libraries, the higher hit rate and distinct chemical space of natural products may prove advantageous for highly benchmarked workflows such as the one presented here. To challenge our protocol, we screened a library of natural products from AnalytiCon Gmbh (Potsdam, Germany) (MEGx Purified Natural Product Screening Compounds), which is a collection of 3734 compounds.</p><p>In silico screening. An in silico version of the AnalytiCon Discovery MEGx Natural Products Screen Library along with Compound A as control was prepared for screening as described above and docked into the cryptic allosteric pocket of 2JFZclstr3 (best performing receptor) following the procedure detailed above using FlexX (best performing docking program). The top scoring pose for each of the 177 compounds scoring above the score cutoff of −20.859 kcal/mol were exported as individual pdb files to be used for SEAM-docking.</p><p>All atom-simulated annealing energy minimization (SAEM)docking protocol. With FlexX being optimized to identify true binders, we decided to employ the SAEM approach 23,28 as a secondary screen, which is designed to provide reliable rank-ordering of the hits. The all atom simulated annealing energy minimization with YAMBER03 knowledge-based force field followed by rescoring in AutoDock Vina 56 is a customized program completely automated with a script in the Python-based Yanaconda scripting language and run in the YASARA software package 41,42 . We employed a slightly modified version of the method explained in detail in our previous reports 28,47 . In short, each chosen complex (output from FlexX docking) was placed in a simulation cell, solvated, charge-neutralized followed by optimization of solvent and hydrogen bonding network and then phased simulated annealing was performed without any restraints as described in Whalen et al. 28 . All the protein and ligand atoms were kept free. The convergence criteria for the run was reaching the energy minimum with a maximum of five failures allowed. Once the convergence was reached, the affinity of the ligand in this optimized complex along with binding energy was determined by employing the docking utility of AutoDock Vina 56 and scoring with the same. A key step in this process is the retention of interstitial water molecules during the AutoDock Vina docking and scoring.</p><p>Procedure for DCCM and ΔDCCM. The DCCM between residues i and j was calculated by dividing the dot product of any two residue displacements relative to an average structure, as described by Equation 1:</p><p>The value of d is the displacement of an atomic position from the ensemble average position, and the brackets represent averaging over the snapshots from the MD simulation of the H. pylori GR-ligand complexes. The ΔDCCM values were calculated using the Pandas package within Jupyter Notebook Release 6.1.4</p><p>Expression and purification of H. pylori glutamate racemase. E. coli BL21 (DE3) pLysS cells were transfected with 6XHis-tagged H. pylori glutamate racemase (GR) and GroEL/ES chaperone proteins inserted in the pET-15b and pCH1 vectors, respectively. Cells were cultured overnight at 37 °C with rotation in 5 ml of Terrific Broth (TB), supplemented with 50 μg/mL ampicillin, 30 μg/mL chloramphenicol, and 100 μg/mL kanamycin. The 5 mL starter culture was back-diluted into 750 mL of TB medium containing antibiotics, and grown at 37 °C with shaking until the OD600 reached 0.8-1.0. Protein expression was induced upon addition of IPTG at a final concentration of 0.1 mM. Following induction, protein was expressed for 16-18 h at 20 °C with shaking. Cells were harvested by centrifugation at 5000 × g at 4 °C for 20 min. Supernatant was discarded and cell pellets were resuspended in buffer A (100 mM Tris, 100 mM NaCl, 10 mM imidazole, 1 mM TCEP, pH 8.0). An Emulsiflex microfluidizer was used to lyse the cells. Insoluble materials were pelleted by centrifugation at 30,000 × g for 75 min at 4 °C and the supernatant was passed through a 0.22-μm filter. Clarified lysate containing 6XHis-tagged protein was then loaded onto a 1 mL HisTrap IMAC HP (GE Healthcare) cobalt resin column equilibrated with buffer A. Cobalt-bound GR was washed with buffer A and eluted over several fractions via a linear gradient from buffer A to buffer B (100 mM Tris, 100 mM NaCl, 250 mM imidazole, 1 mM TCEP, pH 8.0). The purity of select fractions were analyzed using 12% SDS-PAGE gels. Pooled fractions were concentrated utilizing a 10,000 MWCO Amicon centrifugal filter device. Protein destined for crystallography purposes was dialyzed into thrombin cleavage buffer (50 units thrombin, 20 mM Tris, 150 mM NaCl, 25 mM CaCl 2 , pH 8.4) overnight to remove the 6XHis-tag. GR was further purified by Size-exclusion chromatography using a HiLoad 16/200 Superdex column (GE Healthcare) equilibrated with protein storage buffer (50 mM Tris, 100 mM NaCl, 0.2 mM DTT, pH 8.0). Protein stocks were stored at a final concentration of 5-7 mg/mL with 20% v/v glycerol at −20 °C.</p><p>Coupled-enzyme assay for H. pylori glutamate racemase. The D-to L-glutamate racemization activities of H. pylori GR was assayed through a previously established coupled-enzyme method, utilizing L-glutamate dehydrogenase and diaphorase 40 . The H. pylori GR required for this assay was purified as detailed above. The assay consists of 1.25 mM D-Glu, 2.51 mM ADP, 5 mM NAD, 0.506 mM INT, 5 Units of L-GDH, 0.4 units of diaphorase and 2 µM GR in 50 mM TRIS, pH = 8. Compound stocks solutions of varying concentrations (100 µM-50 mM) were made in DMSO and introduced in the assay vials. The final DMSO concentration after addition of compounds in the assay was 5% v/v. The assays were performed as three independent repeats for Compound A and NP-020560, while other hits being weak inhibitors and restricted by amounts available, were evaluated in a single repeat. Absorption for the reduced iodonitrotetrazolium was collected by measuring absorbance at 500 nm at 1 min interval for 60 min on Cary 300 UV-VIS Spectrophotometer from Varian (Palo Alto, CA). Absorbance for each compound concentration at every time point were normalized to absorbance for blank, followed by calculating slopes for each concentration. The calculated slopes with the respective log (inhibitor concertation) were fitted to log(inhibitor) vs response (three parameters) model within 'Nonlinear Regression-Dose Response Inhibition' in GraphPad Prism version 8.4.3. The program provides output as IC 50 and LogIC 50 along with respective standard errors, and for ease of expression, we report inhibition data as IC 50 (mean ± SE).</p><p>Surface Plasmon Resonance (SPR) GR immobilization. Experiments were performed using a SensiQ Pioneer SPR. The biosensor chip was first equilibrated with SPR running buffer (50 mM Tris-HCl, 100 mM NaCl, TCEP, 1% v/v TWEEN, pH 8.0), then the surface of the biosensor chip was seasoned by injecting 200 μL 50 mM NaOH at 40 μL/min, 200 μL 0.1% v/v SDS at 40 μL/min, and 200 μL 10 mM HCl at 40 μL/min in the test and reference biosensor chip flow channels. This process of seasoning was repeated twice. All buffers and solutions were filter sterilized with 0.22-μm buffer filters before using on the SPR instrument. After seasoning, the lines were primed three times with SPR buffer. The biosensor chip was then prepped for protein immobilization by washing the HisCap chip with 200 μL of 400 mM EDTA (pH = 8) injected at 20 μL/min followed by a 15 s dissociation phase. Next 500 μL of 100 mM NiCl 2 was injected at 40 μL/min followed by a 15 s dissociation phase, then 200 μL of 700 mM imidazole was injected at 40 μL/min followed by a 15 s dissociation phase. To use amine coupling in addition to HisCapture, 500 μL of 0.4 M EDC/0.1 M NHS was injected at 40 μL/min, 325 μL of 180 nM GR is injected at 10 μL/min. Protein was injected across the nickel activated channel until it reached the desired response units (RU) of immobilized protein, approximately 8000-12,000 RU. This density of protein on the biosensor surface is necessary for measuring binding interactions of small molecules. Once achieved, the remaining protein was washed out with SPR buffer and buffer was continuously pumped until the baseline stabilized with a drift of no more than 3 RU min −1 . This was followed by a manual injection of SPR running buffer at 10 μL/min.</p><p>Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/ licenses/by/4.0/.</p>

Nature Communications Chemistry

Trivalent PROTACs enhance protein degradation via combined avidity and cooperativity

Bivalent PROTACs work drive protein degradation by simultaneously binding a target protein and an E3 ligase and forming a productive ternary complex. We hypothesized that increasing binding valency within a PROTAC could enhanced degradation. Here, we designed trivalent PROTACs consisting of a bivalent BET inhibitor and an E3 ligand, tethered via a branched linker. We identified VHL-based SIM1 as a low picomolar BET degrader, with preference for BRD2. Compared to bivalent PROTACs, SIM1 showed more sustained and higher degradation efficacy, which led to more potent anti-cancer activity. Mechanistically, SIM1 simultaneously engages with high avidity both BET bromodomains in a cis intramolecular fashion and forms a 1:1:1 ternary complex with VHL exhibiting positive cooperativity and high cellular stability with prolonged residence time. Collectively, our data along with favorable in vivo pharmacokinetics demonstrate that augmenting the binding valency of proximity-induced modalities can be an enabling strategy for advancing functional outcomes.

trivalent_protacs_enhance_protein_degradation_via_combined_avidity_and_cooperativity

Introduction<!>Structure-guided design and synthesis of trivalent PROTACs<!>Supplementary Information).<!>Trivalent PROTAC SIM1 is a highly potent BET degrader<!>SIM1 is a preferential BRD2 degrader<!>SIM1 is more efficacious than bivalent PROTACs or inhibitors<!>SIM1 forms a 1:1:1 complex with VHL and BET BD1 and BD2<!>BRD2:SIM1:VHL ternary complex shows avidity, cooperativity, and enhanced residence time<!>Supplementary Information).<!>Discussion<!>LC-MS/MS analysis.<!>Monitoring cMyc Loss and Cell Viability in MV4<!>Size exclusion chromatography (SEC).

<p>Bispecific molecular agents that induce proximity between two proteins are an emerging paradigm of pharmaceutical intervention into biology and medicine 1 . Targeted degradation compounds, classified as PROTACs or molecular glues, have shown great promise as a new class of chemical probes to study biology and therapeutics for treatment of disease [2][3][4] .</p><p>PROTACs are traditionally conceived as bifunctional, i.e. composed of two ligands joined by a linker, forming a ternary complex consisting of Target Protein:PROTAC:E3 ligase, resulting in ubiquitination and subsequent degradation of the target protein via the proteasome [5][6][7] . PROTACs have successfully been applied to degrade a wide spectrum of protein targets including nuclear [8][9][10][11][12] , cytoplasmic 6,13,14 , membrane-bound 15 , and multi-pass transmembrane proteins 16 , most commonly by recruiting either the von Hippel-Lindau (VHL) or cereblon (CRBN) E3 ligases.</p><p>PROTACs have shown unexpected advantages compared to the inhibitors of which they are composed. PROTACs can discriminate amongst highly homologous targets [17][18][19][20] , and can exhibit much greater potencies than expected, due to a catalytic mechanism of action, which can compensate for low binary binding affinities or poor cellular permeability, and allow for use of weak, non-functional ligands 12,18,21,22 . Unlike inhibitors, degraders must function beyond simple binary engagement. Instead they must work throughout a cascade of events, not only inducing proximity between two proteins which do not natively interact, but also yielding a productive ternary complex which structurally positions the target protein for efficient ubiquitination by the E3 ligase [23][24][25] . Recent X-ray crystallographic structures and allied biophysical studies of PROTAC ternary complexes have demonstrated that some PROTAC-mediated ternary complexes are, like molecular glues, capable of cooperative binding, most notably shown for BRD4 BD2 -MZ1-VHL 17 . This and subsequent studies have shown how in order to drive productive target ubiquitination and profound degradation at catalytic low concentrations, degraders need to form complexes of sufficient stability and residence time, which can be enhanced by cooperativity (defined as the ratio between the dissociation constant of a binary complex and that of the corresponding binding step in the ternary complex) and favourable intracomplex interactions 12,17,18,23,25 . Such optimal "glueing" within the ternary complex can be challenging to realize with conventional PROTAC degraders that are by definition monovalent at the target of interest. Indeed, unfavourable PROTAC ternary complexes are often observed, which even if permissive to downstream protein ubiquitination and degradation, can lead to pronounced hook effect at higher concentration and/or result in slow and incomplete target degradation 13,26 . We hypothesized that the molecular recognition process could be aided by multivalency and avidity, i.e. the accumulated strength of multiple affinities of individual binding interactions 27,28 .</p><p>Here, we present the design, synthesis, and mechanistic characterization of trivalent PROTACs as a strategy to enhance targeted protein degradation. We elected the Bromo and Extra Terminal (BET) domain family member proteins BRD2, BRD3 and BRD4 as ideal model systems for this study because of their therapeutic relevance in many diseases including cancer 29,30 . Several bivalent BET inhibitors and PROTACs, including the afore-mentioned MZ1, have been developed from parent monovalent inhibitors 5,31,32 . Altogether, these compounds provided suitable benchmark reference for our study, where we set out to synergize the effects of a bivalent target ligand with E3 ligase recruitment to produce a trivalent PROTAC with enhanced target degradation. Our trivalent PROTACs were further characterized in a series of biochemical, biophysical and cellular assays to understand potency, kinetics, and specificity of degradation as compared to bivalent molecules, as well as for functional outcomes and underlying mechanism of action.</p><!><p>Insights to design a trivalent PROTAC came from analysis of the crystal structure of BRD4 BD2 -MZ1 (1)-VHL ternary complex 17 , which revealed a central portion of the PEG3 linker as solvent exposed, suggesting it as a branching point to link to a second BET ligand (Fig. 1a).</p><p>Similarly, a site for E3 ligase ligand attachment came from the co-crystal structure of the bivalent BET inhibitor MT1 (2) bound to two monomers of BRD4 BD2 33 (Fig. 1b). In contrast, the bivalent inhibitor Bi-BET is fully buried in its co-crystal structure 34 (Extended Data Fig. 1a), therefore we elected MZ1 (1) and MT1 (2) as the progenitor bifunctional molecules in our design strategy.</p><p>We next envisaged that a trifunctional PROTAC could be assembled around a 'core scaffold' connected to each of its three ligands via variable linkers (Fig. 1c). We reasoned that 1,1,1-tris(hydroxymethyl)ethane, also known as trimethylolethane (TME), could provide fit-forpurpose scaffold because it features three primary alcohol groups in a neopentyl core structure, thus acting as branched, close bioisosteric replacements of PEG units. The achirality of the central quaternary carbon of the TME core could be readily achieved by keeping the chemical linkers to each BET ligand moiety identical. To allow flexibility in exploring the relative constraints between the different monomeric ligands, while at the same time keeping the overall chemical structure as close as possible to those of MZ1 and MT1, we designed three branched PROTACs (SIM1 (3), SIM2 (4) and SIM3 (5)) bearing PEG3 or PEG4 at each linker to the BET ligand (n=3,4), and PEG0 or PEG1 towards the VHL ligand VH032-amine 35 (m=0,1) (Fig. 1c).</p><p>We also designed analogous compounds SIM4 (6), SIM5 (7) and SIM6 (8) composed of the CRBN ligand pomalidomide 4'-alkylC2-amine 11 in place of VHL ligand (Fig. 1c). We synthesized trivalent PROTACs in nine overall steps from TME, with installation of either VHL or CRBN ligand followed in the final step by the 2:1 coupling of the BET ligand (see</p><!><p>To evaluate the ability of SIM1-6 trivalent compounds to induce intracellular degradation of BET proteins, we first treated human HEK293 cells for 4h at 1µM and assessed protein levels by western blot. Profound degradation across BET proteins was observed with VHL-based SIM1-SIM3, while minimal to partial degradation was observed with CRBN-based SIM4-SIM6 (Fig. 1d). To confirm the results we next used live cell continual luminescent monitoring of CRISPR/Cas9 endogenously tagged HiBiT-BRD4 in HEK293 cells over 24h 23 . Faster rates of BRD4 degradation accompanied by greater maximal degradation levels were seen with SIM1-SIM3 as compared to the slower and partial loss observed with SIM4-6 (Extended Data Fig. 1b).</p><!><p>To identify the best degrader, concentration-dependent profiling at 4h treatments using immunoblots evidenced much lower DC50 values of 0.7 -9.5nM for SIM1-SIM3 compared to MZ1 (DC50 values of 25 -920nM) across all the BET proteins (Fig. 1e and Extended Data Fig. 1c). To confirm greater potency of VHL-based degraders, we evaluated growth inhibition profiles of BET-sensitive cancer cell lines MV4;11 (Fig. 1f), A549, and HL-60 (Extended Data Fig. 1d) with SIM1-6 treatments. SIM1-SIM3 consistently showed more potent activity relative to SIM4-SIM6 and bivalent molecules MZ1 and MT1 (Fig. 1f and Extended Data Fig. 1d). In both these experiments, SIM1 emerged as the most potent of the three VHL-based degraders.</p><p>To determine whether the increase in target binding valency of SIM1 improved the degradation activity, we synthesized two diastereomeric analogues of SIM1 as negative controls:</p><p>(R,S)-SIM1 (3a), which has inverted stereochemistry at one of the two BET ligands making it inactive at binding one of the two BET bromodomains (Fig. 1g) 36 , and the non-degrading negative control isomer cis-SIM1 (3b) (Fig. 1g and Supplementary Information) 5,7 . While the synthesis of cis-SIM1 was straightforward as with its trans diastereomer, initial attempts to synthesize (R,S)-SIM1 compound by using half equivalent of BET ligand at the final coupling step failed to produce the desired 1:1 coupling with an acceptable yield. Therefore, we revised the synthetic route to allow subsequent, independent coupling steps with (+)-JQ1 (9) 36 first, followed by (-)-JQ1 (9a) 36 (Supplementary Information). Upon successful synthesis, (R,S)-SIM1 was tested for degradation activity and found to behave similarly to MZ1 and less potently than SIM1 for degradation of all BET family members (Fig. 1h).</p><p>To assess whether the trivalent PROTAC induced also a more sustained degradation of BET proteins in cells compared to MZ1 or (R,S)-SIM1, degradation washout experiments were performed. CRISPR HiBiT-BET HEK293 cells were treated with equivalent concentration (100nM) of SIM1, (R,S)-SIM1 and MZ1 compounds for 3.5h, then media was removed and replaced with media lacking compounds. We also tested SIM1 at 10x lower concentration (10nM) to account for its higher potency of degradation across BET proteins. The HiBiT-BET protein levels were continuously monitored from the initial addition of the compounds and immediately after the wash for a total time of 50h. At the 100nM treatment, degradation of all BET family members by SIM1 remained at constant low levels after washout over the time course, while at the 10nM SIM1 treatment partial recovery was observed for BRD2, BRD3, and BRD4 after washout (Fig. 1i). Recovery of all BET family members following washout for cells treated with 100nM (R,S)-SIM1 or 100nM MZ1 was greater and occurred faster than those observed with SIM1 (Fig. 1i).</p><!><p>To further characterize degrader activity, we performed quantitative analysis of live-cell kinetic degradation of the CRISPR HiBiT-BET family members treated with SIM1 titrations over more than 6-log-order concentration range (10pM-30µM) (Fig. 2a and Extended Data Fig. 2a). All BET family members showed complete degradation across a 4 log concentration range (3nM-30µM). A slight slowing of the initial rate of degradation (due to the hook effect) was manifested only at the highest concentrations (3-30µM) with BRD3 and BRD4, and to a much lesser extent BRD2 (Extended Data Fig. 2a). From the kinetic analyses, degradation rates and Dmax values were calculated and plotted versus concentration to obtain λmax and Dmax50 (Fig. 2b). SIM1 exhibited Dmax50 values of 60-400pM and BET family degradation preference of BRD2>BRD4>BRD3 on both λmax and Dmax50 (Fig. 2b). This differs from MZ1 which has preference for BRD4 (Fig. 1d and 1e) 5,17,23,25 .</p><p>We next sought to further understand the SIM1 degradation preference for BRD2 compared to other PROTACs. Live-cell kinetic degradation dose response profiles were performed with the bivalent BET degrader ARV-771 (10) 31 using the CRISPR HiBiT-BET cell lines (Extended Data Fig. 2b) and degradation parameters were quantitated and compared with those previously determined for MZ1 23 and SIM1 (Fig. 2b). Shown in Fig. 2c increased degradation potency for BRD2 as compared to ARV-771 and MZ1, respectively (Fig. 2c). As enhanced degradation rates tend to correlate with enhanced ubiquitination, cellular studies were performed to monitor the kinetics of ubiquitination of BET proteins using bioluminescence resonance energy transfer (NanoBRET) assays consisting of the HiBiT-BET CRISPR cell lines expressing fluorescently labelled HaloTag-Ubiquitin 23 . Shown in Fig 2d , kinetic increases in cellular ubiquitination were greater for BRD2 as compared to BRD3 and BRD4 after a 10nM SIM1 treatment. These same trends were observed at 100nM SIM1 concentration and comparison to MZ1 (Fig. 2d), revealing that SIM1 led to higher levels of ubiquitination of all BET family members (Extended Data Fig. 2d), with the greatest difference observed for BRD2.</p><p>To assess the cellular selectivity of SIM1 for BET proteins in a BET-relevant cellular background and also determine if BRD2 preference was observed in this context, multiplexed tandem mass tag (TMT) labeling mass spectrometry proteomic experiments were performed to monitor protein levels in a quantitative and unbiased fashion. Acute myeloid leukemia (AML) MV4;11 cells were treated in triplicate with DMSO, 10nM SIM1, or 10nM cis-SIM1 for 4h.</p><p>Among the 5,232 proteins quantified, BRD2 was found as most significantly degraded by SIM1, followed by BRD3 and BRD4 (Fig. 2e and Extended Data Fig. 2f). No significant changes in BET protein abundance were observed in cells treated with cis-SIM1 (Extended Data Fig. 2f).</p><p>Together, the data support SIM1 as a degrader with unusual preference for BRD2.</p><!><p>To quantitate time-dependent loss of cMyc, a known downstream effect of BET-induced degradation, cMyc was endogenously tagged with HiBiT in MV4;11 and protein levels were monitored in cell lysates at varying times with different concentrations of SIM1, cis-SIM1, and MT1 (Fig. 3a, left and Extended Data Fig. 3a). Rapid and complete loss of cMyc was observed with 1nM SIM1 concentration (Fig. 3a, left) whereas similar levels of cMyc loss with MT1 or cis-SIM1 were not achieved until 50-100nM treatments (Extended Data Fig. 3a). Cell viability assays revealed that, at 1nM treatment, only SIM1 (but not cis-SIM1 or MT1) resulted in measurable cellular death after 6h (Fig. 3a, right). Similarly, at higher concentrations SIM1 was found to be significantly more cytotoxic than control compounds (Extended Data Fig. 3a)</p><p>We next moved to study compounds in a BET-sensitive cell line, the prostate cancer line 22Rv1. Treatment of 22Rv1 cells with varying concentrations of compounds at 4h confirmed the enhancement in BET degradation potency and cMyc level suppression activity of SIM1 compared to MZ1 and ARV-771, as well as non-degrading controls MT1 and cis-SIM1 (Fig. 3b and Extended Data Fig. 3b). The expected mechanistic dependency of SIM1-induced degradation on functional E3 ligase and proteasomal activity was confirmed with co-treatment with VHL inhibitor VH298 (11) 37 and proteasome inhibitor MG132 (Extended Data Fig. 3c).</p><p>The superior activity of SIM1 on the viability of 22Rv1 cells was evidenced in a colonyformation assay, where only SIM1-treated cells showed significant cytotoxicity as compared to vehicle control or cells treated with bivalent counterparts at the same concentration of 10nM (Fig. 3c). Substantial cell death was observed after 24h treatment with 10nM SIM1, as shown by PARP cleavage assays (Fig. 3d and Extended Data Fig. 4a). In contrast, 10nM MT1 or MZ1 did not cause observable PARP cleavage even up to 48h, and 1µM concentration was needed to induce levels of PARP cleavage similar to 10nM SIM1 (Extended Data Fig. 4b). Cells died of apoptosis as indicated by prevented cleavage of PARP upon co-treatment with QVD-OPh, a pancaspase inhibitor, and not of necroptosis, as shown by co-treatment with Necrostatin-1 (Fig. 3d and Extended Data Fig. 4b). Caspase-Glo assays confirmed the more potent activity of SIM1 (EC50 2nM) compared to MZ1 or ARV-771 (EC50 150 and 90nM, respectively), with much greater maximal signal than non-degraders cis-SIM1 and MT1 (Fig. 3e), and apoptosis was blocked by co-treatment with the VH298 and QVD-OPh (Extended Data Fig. 4c).</p><p>Our proteomics analysis (Fig. 2e) showed SIM1 induced decrease in protein levels of HMOX1 (heme oxygenase 1) suggesting early initiation of apoptosis 38 . Early and late apoptotic induction between compound treatment was compared and notably, SIM1 induced a much greater degree of both early and late apoptosis at 1nM compared to all bivalent counterparts tested, even when compared to tenfold higher concentration of (R,S)-SIM1, MZ1, or MT1 (Fig. 3f and Extended Data Fig. 5). Together, the biological data supports more potent degradation and more substantial downstream functional activity of the trivalent degrader SIM1 compared to parent bivalent degraders or inhibitors.</p><!><p>As bivalent BET inhibitors simultaneously engage BD1 and BD2 bromodomains 33,34 , we hypothesized that trivalent SIM1 would also display a cis intramolecular engagement of a BET protein. To explore this, we first employed biophysical binding assays with recombinant proteins using size-exclusion chromatography (SEC) 39 with varying tandem BD1-BD2 constructs from BRD4. These included wild-type (WT), capable of cis intramolecular binding, or point mutations in either BD1 or BD2 at a conserved asparagine residue in the ligand-binding pocket (N140F in BRD4 BD1 , or N433F in BRD4 BD2 ), to abrogate cis binding 34 . SIM1 and MT1 both shifted the SEC profile of BRD4 WT BD1-BD2 tandem construct to a higher elution volume, compared to free or MZ-1-bound BRD4, consistent with the formation of a more compact intramolecular 1:1 complex (Fig. 4a). In contrast, when either N140F or N433F mutant tandems were used, we observed a significant shift to lower elution volumes with both SIM1 and MT1, consistent with the formation of higher-molecular weight 2:1 species in solution (Fig. 4a and Extended Data Fig. 6a). Having established that SIM1 engages BD1 and BD2 in a cis fashion, we next asked whether SIM1 could form a 1:1:1 complex between the BD1-BD2 tandem domain and VHL. Indeed, a sample containing 1:1:1 equivalents of SIM1, BD1-BD2, and VHL-Elongin B-Elongin C complex (VCB) ran at lower elution volumes compared to either of the two peaks observed from a sample containing the same equivalent ratio of cis-SIM1, BD1-BD2 tandem and VCB, where only the 1:1 cis-SIM1:BD1-BD2 complex and unbound VCB are formed (Fig. 4a).</p><p>To determine if SIM1 could induce a conformational change of BRD4, known to occur with bivalent BET inhibitors 34 , we utilized a NanoBRET biosensor containing the BD1-BD2 tandem domains of WT BRD4 or mutant N433F, flanked respectively by a NanoLuc donor and HaloTag acceptor (Fig. 4b). With the BD1-BD2 tandem WT biosensor, compounds SIM1, cis-SIM1 and MT1 all showed a change in BRD4 conformation, manifested by an increase in NanoBRET signal which reached and maintained a plateau, as expected for an intramolecular engagement (Fig. 4b). As control, the BD1-BD2 N433F mutant sensor showed no response indicating that simultaneous binding of BD1 and BD2 is required for the conformation change (Fig 4b).</p><p>Interestingly cis-SIM1 and SIM1 showed higher EC50 values for BRD4 binding than MT1 (Fig. 4b). To determine if this is due to a reduced binding affinity of BRD4 and/or reduced permeability, NanoBRET target engagement assays were performed measuring displacement of a fluorescent BET tracer molecule bound to HiBiT-BRD4 34 . In permeabilized cells, we observed binding of SIM1, cis-SIM1, and MT1 to endogenous HiBiT-BRD4 with near-identical binding affinities and IC50 values (Extended Data Fig. 6b). However, in live cells, SIM1 showed reduced binding affinity to BRD4 compared to MT1 (Extended Data Fig. 6b), suggesting the EC50 shift observed in the conformational sensor assay reflects reduced permeability of the trivalent PROTAC relative to MT1. Reduced permeability was also observed relative to VH298, using NanoBRET target engagement with NanoLuc-VHL, however cellular permeability of SIM1 was within 2-fold lower than MZ1 (Fig. 4c) 23,40 .</p><p>To further characterize ternary complex binding thermodynamics, we next used isothermal titration calorimetry (ITC) by performing reverse titrations 17 . First, in titrations of BRD4 N433F or N140F tandems (competent for BD1 or BD2 binding alone, respectively) into preformed SIM1:VCB complex we observed 2:1 stoichiometry, molar binding enthalpy of ΔH = -11.6 and -9.1 kcal/mol, and Kd = 1.2 and 0.12 µM, respectively (Fig. 4d). In contrast, titration of BRD4 WT BD1-BD2 under identical conditions displayed stoichiometry of 1:1, and a large negative binding enthalpy ( H = -20 kcal/mol) corresponding to the sum of BD1 and BD2 binding (Fig. 4d). To study ternary complex formation in a cellular context, we interrogated VHL binding to full-length BRD4 WT, N140F, or N433F mutations using kinetic NanoBRET assays 23 . In these experiments, the ternary complex was rapidly formed with SIM1, but not with cis-SIM1 or MT1 (Fig. 4e, right). Ternary complex formation however was markedly reduced with BRD4 N140F (Fig. 4e, middle) and almost abolished with N433F (Fig. 4e, left). These results confirm SIM1 utilizes both BD1 and BD2 for cellular ternary complex formation and suggest preferential binding to BD2 over BD1, consistent with the ITC results and what was found previously with MZ1 17,25 . Additional experiments showed a more robust and sustained ternary complex for BRD2 and BRD4 with VHL induced by SIM1 as compared to MZ1, while that with BRD3 did not appear to be as prolonged or stable (Extended Data Fig. 6c). Together, these data evidence SIM1 intramolecularly engages BD2 and BD1 to form a 1:1:1 ternary complex with VHL and BRD4.</p><!><p>We next asked to what extent SIM1 might exhibit both avidity, i.e. enhanced binding affinity for BET proteins due to intramolecular BD1 and BD2 binding, and cooperativity within the ternary complex, i.e. enhanced affinity of forming ternary complex relative to the corresponding binary complex. We used established phage-based bromodomain displacement assays to quantitatively measure compound binding with tandem bromodomain constructs (Supplementary Information). Bidentate SIM1 showed picomolar affinity to tandem bromodomain constructs BRD2(1,2), BRD3(1,2), BRD4(1,2) and full-length BRD4, with 50-90X increase in affinity compared to monodentate (R,S)-SIM1 evidencing its avidity (Fig 5a and</p><!><p>In forming ternary complexes with VHL and BET proteins, SIM1 exhibited a positive cooperativity α value of 3.5 as shown in competitive AlphaLISA assays (Fig. 5b), measuring binding of SIM1 alone (IC50 = 205nM) or SIM1:VCB binary complex (IC50 = 58nM) via the displacement of biotin-JQ1 41 from BRD4. As a cross-validation, we evaluated cooperativity in a competitive FP assay measuring binding at the VHL end of the PROTAC molecule via the displacement of a fluorescent HIF-1α peptide probe 37 . In this experiment too SIM1 exhibited positive cooperativity from enhanced affinity of its competitive displacement once pre-engaged with BRD2 or BRD4 tandem ( = 4.4 or 7.3, respectively) compared to SIM1 alone (Fig. 5c).</p><p>To assess dissociation kinetics of ternary complexes of SIM1, we used an SPR binding assay 25 . We immobilized biotinylated VCB onto the surface chip and measured binding parameters for SIM1 alone (binary) or SIM1 pre-incubated in excess of BRD2 or BRD4 tandem BD1-BD2 proteins (ternary) in single and multicycle kinetic modes (Fig. 5d and Extended Data Fig. 7a), which gave comparable results (Supplementary Table 1). SIM1:BET-tandems bound to VCB with 1:1 stoichiometry, as the %Rmax value for the expected 1:1 binding model (47-77%) was comparable to that observed from reference titration of SIM1 alone (52%) (Supplementary Table 1). SIM1 formed high-affinity, stable and long-lived complexes with VCB and BET tandem proteins (for BRD2(1,2): t½ = 36s, Kd = 45nM, = 13.8; for BRD4(1,2): t½ = 29s, Kd = 100nM, = 6.4) (Fig. 5d and Supplementary Table 1). In comparison to the binary complex formation, the VHL:SIM1:BRD2/4(1,2) ternary complexes showed faster association rate and slower dissociation kinetics, resulting in the lower dissociation constants and positive cooperativity (Supplementary Table 1).</p><p>Having established avidity, cooperativity and stability of biophysical ternary complex recognition with recombinant proteins, we moved to interrogate complexes kinetics in live cells.</p><p>To monitor residence time, termed the complex half-life t½, the HiBiT-BRD2 and BRD4 CRISPR cells were first incubated with saturating concentrations of SIM1, cis-SIM1 or parent compounds, followed by a competitive BET fluorescent tracer. The NanoBRET signal produced from this displacement can be monitored kinetically, the rate and intensity of which directly correlates to the residence time of the initial compound-bound complex. As controls, JQ1 had a short residence time similar to the monovalent tracer alone, while MT1 showed a longer residence time, both in terms of calculated rate, Kobs (h -1 ) and complex half-life, t½ (h) (Fig. 5e).</p><p>SIM1 showed significantly slower rates and longer half-lives, indicting prolonged residence time, particularly for BRD2, and to a lesser extent also for BRD4 (Fig. 5e). These trends matched well to ubiquitination and degradation potency for each compounds and BET protein.</p><p>Interestingly cis-SIM1 tracked nearly identically with MT1 in both BRD2 and BRD4 traces (Fig. 5e) indicating that cooperative engagement of VHL improves residence time of SIM1 vs cis-SIM1. Further experiments which were performed with (R,S)-SIM1, MZ1, and cis-MZ1 revealed enhanced residence times of MZ1 and (R,S)-SIM1, though neither as significant as for SIM1 (Extended Data Fig. 7b). The increase observed from cis-MZ1 to MZ1 (Extended Data Fig. 7b) again supports that cooperative ternary complexes 17 can increase residence time, yet not to as great of extent as SIM1 as it is lacking the added avidity. Together, the results indicate that SIM1 favorability for BRD2 is facilitated by intramolecular binding that results in both a structural change as well as extended ternary complex half-life.</p><p>Finally, we evaluated the pharmacokinetics (PK) of SIM1 following intravenous and subcutaneous administration in mice (Fig. 5f). SIM1 exhibited highly favourable bioavailability and stability, including high AUCs, low clearance and long half-lives, comparing positively to those of the more canonical small-molecule components i.e. monovalent JQ1 36 and bivalent MZ1 (data available on OpnMe.com) (Fig. 5f and Supplementary Table 2). Such favourable PK profile is remarkable given its large size (molecular weight 1,619 Da) and qualifies SIM1 as chemical probe appropriate for in vivo use.</p><!><p>The study presented here finds inspiration from an often-used strategy for improvement of compound efficacy, i.e. increasing binding valency, and has leveraged this for improved PROTAC-mediated protein degradation. Our work qualifies the novel trivalent PROTAC SIM1 as a profoundly potent and fast degrader of BET proteins. Our biological and mechanistic investigation with SIM1 provides proof-of-concept for augmenting the valency of degraders as an advantageous strategy to enhance their mode of action by positively impacting the ternary complex. SIM1 bound intramolecularly the BET protein in a cis-fashion to both BD1 and BD2, inducing a conformational change, to then form a 1:1:1 complex (Fig. 6a). Further investigation with BD1 or BD2 mutants suggested there is preferred BD2 binding of SIM1 with BRD4 (Fig. 6a). Interestingly BRD2 was found in a series of orthogonal assays to show the most favorable ternary complex formation and prolonged residence time, the most robust level of ubiquitination, and correspondingly the fastest and highest level of degradation amongst the family members with SIM1. This is unprecedented preference amongst known BET PROTAC degraders. We cannot exclude the possibility that the structural change induced by SIM1 better positions BRD2 in a more favourable state for more productive ubiquitination, as compared to BRD3 or BRD4.</p><p>Structural, thermodynamic and kinetic favorability of ternary complex formation are critical for efficient PROTAC mode of action 17,26,42 . These factors are determined by neointeractions within the ternary complex, including between E3 ligase and target, similar to the mechanism of monovalent molecular glues. This results in stable, cooperative, and long-lived complexes which drive efficient catalytic ubiquitination 17,23,25,43 . Such ternary complexdependent outcomes can be optimized through rational design 12,44 . For compounds which do not have this, their window of degradation efficacy will be limited by the hook effect as nonproductive binary complexes with either the target or E3 ligase will more readily compete the ternary complex at higher PROTAC concentration (Fig. 6b) 13,45 . In our studies with the trivalent PROTAC, we find that optimization of structural, energetic, and kinetic ternary complex parameters occurred from combined binding avidity and cooperativity in the process, resulting in stable biophysical recognition and prolonged cellular residence times (Fig. 6b). Our data suggests that it is this combination which increased most significantly the degradation activity, as the trivalent CRBN-based PROTACs, which we predict would also have avidity, were much less active. Indeed, more in-depth mechanistic experiments with representative CRBN-based trivalent PROTAC SIM4 showed that this compound, despite its ability to efficiently form ternary complexes, showed very low levels of BET family ubiquitination, and as a result induced only partial protein degradation (Extended Data Fig. 8). For the VHL-based SIM1, an improvement in all the parameters with the trivalent PROTAC resulted in a vast expansion of the degradation window, from rapid rates of BET family loss at very low concentration to maximal degradation with minimal hook effect observed at concentration up to 30µM, which is 500,000-fold above the Dmax50. Some of these characteristics have been observed previously with bivalent PROTACs such as cooperative complexes with MZ1 5,17 and potent degradation with dBET-6 32 and ARV-771 31 , but the combination of these to achieve maximal favorability in the ternary complex to enhance degradation had not yet been shown.</p><p>The transition from bivalent to trivalent degraders might not immediately seem like an obvious approach for improvement of degradation outcomes, particularly given the chemical synthesis challenges and presupposition that increasing molecular weight of degraders would be accompanied by lack of cellular permeability or poor pharmacokinetics. However, with the trivalent PROTAC studied here, we have demonstrated that this is not the case. While indeed SIM1 has slightly reduced permeability compared to parent bivalent inhibitor and PROTACs, it outperformed those in relevant cellular assays used for assessment of BET compound potential for therapeutic use. Furthermore, the remarkably favourable PK profile of SIM1 suggests trivalent PROTACs will be appropriate for in vivo use despite their increased molecular weight.</p><p>The enhanced potency and increased sites of binding of trivalent PROTACs might potentially allow to alleviate or circumvent some of the emerging cancer resistance mechanisms with monovalent and bivalent degraders, such as missense mutations on the target protein 46 .</p><p>While the chemical design and synthesis of a trivalent PROTAC is more involved than for bivalent ones, the increased effort showed significant benefits and allowed to overcome these perceived challenges, affording a much-improved degrader. To achieve this, outlined is a new linker design strategy for generation of a branched trifunctional scaffold to which both target and E3 binders could be conjugated, which provides highly modular design and opens to numerous future applications (Fig. 6c). Though we have shown here increased valency to address two repeat domains within a target, trivalent compounds could be directed towards two distinct domains on the same protein. A multivalent design concept could be applied to any three protein targets if binding ligands for each are known, wherein all three warheads engage if structural constraints allow (Fig. 6c). Multivalent avidity may allow leveraging weak intramolecular binding at the ternary complex interface, and best exploiting weak binding ligands which can be readily found but can be challenging to optimize as part of mono-or bivalent agents 47 . One might imagine two subunits within a multi-protein complex, or even two distinct proteins (albeit at the expense of avidity), could be recruited either simultaneously or independently to the E3 ligase. While this manuscript was in advanced stages of revision, a dual-target PROTAC compound consisting of three warheads was reported, 48 however not shown to engage all three warheads simultaneously as we report here.</p><p>Initial work with multi-specific drug modalities has been transformative for drug discovery and greatly expanded the therapeutic landscape 1 . This trivalent compound concept is not limited to E3 ligase recruitment, and could find utility beyond PROTACs for emerging approaches for small-molecule induced proximity [49][50] . We thus anticipate broad applicability of the approach to improve performance of a wide range of multi-specific agents and modalities for chemical biology and pharmaceutical development. CRISPR HiBiT-BRD2 or HiBiT-BRD4 cells were incubated with each of the indicated compounds at their pre-determined EC80 values (listed in Methods) followed by addition of a competitive fluorescentlylabeled BET tracer. NanoBRET was measured and is plotted as fractional occupancy over time. From these graphs, residence time rates (Kobs (h -1 ) and half-life (t½ (h)) were determined. Data are presented as mean values with error bars representing the SD of technical triplicates. f) SIM1 exhibits excellent availability and pharmacokinetics exposure in vivo. Mean plasma concentration-time profiles of SIM1 after single intravenous (IV) or subcutaneous (SC) administration (5 mg/kg) to male C57BL/6 mice (n =3) are shown. Further details are in the associated Supplementary Table 2 and Supplementary Data Set 2. Figure 6. Models of trivalent ternary complex formations and advantages over monovalent and bivalent compounds. a) Proposed mechanism for the formation of a 1:1:1 ternary complex between trivalent PROTAC, VHL and BET protein. Preferential initial binding of the PROTAC to BD2 of BRD4 is followed by conformational change and bidentate binding to BD1. Avidity and cooperativity contribute to formation of a highly stable ternary complex with enhanced residence time at extraordinarily low concentrations of SIM1. b) Shown are different types of degrader-induced ternary complexes, depicted at their varying extents as a function of degrader concentration. A trivalent complex combining avidity with cooperativity shows the highest and most sustained levels of ternary complex formation, with a minimized hook effect. A cooperative bivalent PROTAC complex is next, followed by a non-cooperative bivalent complex. Lastly, the ternary complex induced by molecular glue compounds is shown, which reaches a plateau and unlike PROTACs are not predicted to experience the competitive hook effect at higher concentrations. c) A general model for trifunctional compound-induced ternary complex utilizing a compound with three different warheads (or ligands) to recruit together three distinct protein.</p><p>absorbance at 562nm measured by spectrophotometry (NanoDrop ND1000) or on a plate reader (BMG Labtech Pherastar). Samples were run on SDS-PAGE using NuPAGE Novex 4-12% Bis-Tris gels (Invitrogen) with 20-40µg total protein/well, transferred to 0.2μm pore nitrocellulose membrane (Amersham) by wet transfer and blocked with 3% w/v BSA (Sigma) in 0.1% TBST. Blots were incubated in anti-BRD2 (1:2000, abcam #ab139690), anti-BRD3 (1:500, abcam #ab50818), anti-BRD4 (1:1000, abcam #ab128874), anti-c-myc (1:1000, abcam #32072), anti-PARP (1:1000, CST #9542S), anti-cleaved PARP (1:1000, BD Pharmingen #51-9000017), anti-caspase-3 (1:1000, CST #9662S), anti-tubulin (1:3,000, Bio-Rad #12004165) or anti-β-actin (1:2500, CTS #4970S) antibody overnight at 4°C with rotation. Blots were then incubated in goat anti-mouse or donkey anti-rabbit IRDye 800CW secondary antibodies (1:10,000, LICOR #925-32210 and #926-32213) for 1h at room temperature with rotation. Bands were detected using a ChemiDoc MP imaging system (BioRad) and quantified (Image Studio Lite, version 5.2) with normalisation β-actin and the DMSO control per time point. Data are the average of two biological repeats unless indicated otherwise. Degradation data were plotted and fitted by nonlinear regression using a single-phase exponential decay model using GraphPad Prism.</p><p>Cell Viability Assay. MV4;11 cells were incubated with compounds at the desired concentration for 48h on a clear-bottom 384-well plate. MV4;11 cells were kept in RPMI medium supplemented with 10% FBS, L-glutamine. Initial cell density was 3 × 10 5 per mL. The cells were treated with various concentrations of compound or 0.05% DMSO in triplicates for each concentration point. After treatment, cell viability was measured with Promega CellTiter-Glo luminescent cell viability assay kit according to the manufacturer instructions. Signal was recorded on a BMG Labtech PHERAstar luminescence plate reader with recommended settings. Data were analyzed with GraphPad Prism software to obtain EC50 values of each test compound. Kinetic Degradation, Quantitation, and compound washout experiments. HEK293 cells (ATCC) stably expressing LgBiT (Promega) were edited using CRISPR/Cas9 to endogenously HiBiT tag the Nterminal genomic loci of BRD2, BRD3, or BRD4 23 . For kinetic degradation assays, cells were plated in quadruplicate in white 96-well tissue culture plates at a density of 2 × 10 4 cells per well in 100µL of growth medium and incubated overnight at 37°C, 5% CO2. The following day, medium was replaced with CO2-independent medium (Gibco) containing a 1:100 dilution of Endurazine (Promega) and were incubated at 37°C in 5% CO2 for 2.5h before addition of a 3-fold serial dilution of the indicated concentrations of SIM1, SIM2, SIM3, or ARV-771 (MedChemExpress). Plates retaining lids were placed into the GloMax Discover Microplate Reader (Promega) set to 37°C, and continuous luminescent measurements with readings every 5-15min were made over a 22-24h period. Degradation rate (λ), degradation rate plateau (λmax), and degradation plateau (Dmax) were calculated from above determined kinetic degradation profiles. Briefly, the degradation portion of each kinetic concentration curve was fitted to a single exponential equation where ƛ = degradation rate in units of h -1 . The degraded fraction, Dmax, was calculated as 1plateau. For each curve, the data points before onset of degradation were excluded from the fits. The Dmax was then plotted against concentration to determine Dmax50 values. For compound washout assays, the CRISPR HiBiT BRD2, BRD3, and BRD4 HEK293 cells were plated as described above for kinetic assays and treated with SIM1 (100nM and 10nM), (R,S)-SIM1 and MZ1 (both at 100nM), and equivalent volume of DMSO for a period of 3.5h, with continual luminescence monitoring. The concentrations were chosen to allow achieving >70% Dmax of all three BET proteins at 3.5h with all three compounds, with the caveat that they reflect different relative ratios with DC50 values for each protein-compound combination. At 3.5h, media was removed and replaced with CO2-independent medium containing Endurazine. The plates were placed back in the luminometer for continued monitoring of protein levels for a further 46.5h.</p><p>Mass spectrometry proteomics. Sample preparation. MV4;11 cells in RPMI (Invitrogen) were seeded at 5 × 10 6 cells on a 100mm plate 24 h before treatment. Cells were treated in triplicate by addition of test compound. After 4 h, the cells were centrifuged at 250g for 5 min and washed twice with 12mL of cold PBS. Cells were lyzed in 500µL of 100mM TRIS pH 8.0, 4% (w/v) SDS supplemented with protease inhibitor cocktail (Roche). The lysate was pulse sonicated briefly and then centrifuged at 15,000g for 10 min at 4°C. Samples were quantified using a micro BCA protein assay kit (Thermo Fisher Scientific) and 200µg of each sample was processed and digested using the filter aided sample preparation method followed by alkylation with iodoacetamide and with trypsin as previously described 17 . The samples were then desalted using a 7mm, 3mL C18 SPE cartridge column (Empore, 3M) and labeled with TMT 10-plex Isobaric Label Reagent Set (Thermo Fisher Scientific) as per the manufacturer's instructions. After labeling, the peptides from the nine samples were pooled together in equal proportion. The pooled sample was fractionated using high pH reverse-phase chromatography on an XBridge peptide BEH column (130Å, 3.5µm, 2.1 × 150mm, Waters) on an Ultimate 3000 HPLC system (Thermo Scientific/Dionex). Buffers A (10mM ammonium formate in water, pH 9) and B (10mM ammonium formate in 90% acetonitrile, pH 9) were used over a linear gradient of 5 to 60% buffer B over 60 min at a flow rate of 200µL min −1 . Then, 80 fractions were collected before concatenation into 20 fractions on the basis of the ultraviolet signal of each fraction. All the fractions were dried in a Genevac EZ-2 concentrator and resuspended in 1% formic acid for mass spectrometry analysis.</p><!><p>The fractions were analyzed sequentially on a Q Exactive HF Hybrid Quadrupole-Orbitrap Mass Spectrometer (Thermo Scientific) coupled to an UltiMate 3000 RSLCnano ultra HPLC system (Thermo Scientific) and EasySpray column (75µm × 50 cm, PepMap RSLC C18 column, 2µm, 100Å, Thermo Scientific). Buffers A (0.1% formic acid in water) and B (0.08% formic acid in 80% acetonitrile) were used over a linear gradient from 5 to 35% buffer B over 125 min at 300nL min −1 . The column temperature was 50°C. The mass spectrometer was operated in data dependent mode with a single mass spectrometry survey scan from 335-1,600 m/z followed by 15 sequential m/z dependent MS2 scans. The 15 most intense precursor ions were sequentially fragmented by higher energy collision dissociation. The MS1 isolation window was set to 0.7 m/z and the resolution set at 120,000. MS2 resolution was set at 60,000. The automatic gain control (AGC) targets for MS1 and MS2 were set at 3 × 10 6 ions and 1 × 10 5 ions, respectively. The normalized collision energy was set at 32%. The maximum ion injection times for MS1 and MS2 were set at 50 and 200 ms, respectively.</p><p>Peptide and protein identification. The raw mass spectrometry data files for all 20 fractions were merged and searched against the Uniprot-sprot-Human-Canonical database by MaxQuant software v.1.6.0.16 for protein identification and TMT reporter ion quantitation. The MaxQuant parameters were set as follows: enzyme used trypsin/P; maximum number of missed cleavages equal to two; precursor mass tolerance equal to 10 ppm; fragment mass tolerance equal to 20 ppm; variable modifications: oxidation (M), acetyl (N-term), deamidation (NQ), Gln→pyro-Glu (Q N-term); fixed modifications: carbamidomethyl (C). The data was filtered by applying a 1% false discovery rate followed by exclusion of proteins with fewer than two unique peptides. Quantified proteins were filtered if the absolute fold-change difference between the three DMSO replicates was ≥1.5.</p><!><p>;11 Cells. CRISPR cMyc-HiBiT MV4;11 cells (Promega) were plated at a density of 5 × 10 4 cells per well in solid, white 96-well tissue culture plates (Corning Costar #3917). Following an overnight incubation, they were treated with 1-100nM of the indicated compounds and at the plotted time points, and cMyc levels were determined using luminescent measurement with NanoGlo HiBiT lytic reagent (Promega). Replicate plates of all compound treatments were prepared and at identical timepoints as the protein level measurements, and cell viability was measured using Cell-Titer Glo (Promega). Plates were shaken on an orbital shaker for 10-20 min before reading luminescence on a GloMax Discover Microplate Reader (Promega).</p><p>Caspase-Glo® 3/7 assays. 22Rv1 cells were seeded at 10,000 cells/well of white 96 well plates (Corning #3917) 12-24h before treatment with test compounds with and without inhibitors or an equivalent volume of DMSO for 24h. 100µL/well of Caspase-Glo 3/7 Reagent (Promega) was added and the plate shaken at 500 rpm for 30 s. The plate was incubated for 2h and luminescence measured using a PHERAstar FS plate reader (BMG Labtech). Clonogenic assay. 22Rv1 cells were treated with 10nM SIM1, cis-SIM1, MT1, MZ1 and ARV-711 for 24h. The next day, cells were trypsinised and counted. 500 cells were re-plated and allowed to grow at 37ºC and 5% CO2 for 20 days. After 20 days incubation, the cells were fixed with ice-cold 100% (v/v) methanol for 30 min at 4ºC. Afterwards, methanol was removed, and the cells were stained with 500µl 0.1% crystal-violet dye (in MeOH) for 30 min at room temperature. Following incubation, the cells were washed with dH2O and left to dry overnight. Plates were scanned on an Epson Perfection V800 Photo scanner. And image analysis was done in ImageJ software v. 1.52n. Plating efficiency (PE) was calculated by counting colonies for each treatment condition and dividing the average by number of cells plated. Survival fraction was determined by diving PE of treated cells by PE of untreated cells, multiplied by 100 52 . Bar graphs were generated using GraphPad Prism software. Two independent experiments were performed. Flow Cytometry. MV4;11 cells were counted on a Countess 3 Automated Cell Counter (Thermo Fisher, UK) with the addition of trypan blue. Cells (1x10 6 ) were aliquoted, spun down and resuspended in RPMI media containing test compounds at indicated concentrations. Additionally, QVD OPh (Sigma-Aldrich, UK) and Necrostatin-1 (Sigma-Aldrich) were added to the SIM1 treatment at 20µM final concentration. Treated cells were left to incubate for 24h at 37°C and 5% CO2. On the following day, the cells were collected in a Falcon tube and spun down at 500g for 5 min. Supernatant was aspirated and cells were washed once in 1mL FACS buffer (PBS, 5% FBS, 0.05 % NaN) and afterwards resuspended in 100µL of the same buffer containing Apotracker Green (Biolegend, UK) and DAPI (Sigma-Aldrich) at final concentration of 400nM and 1µg/mL, respectively. Cells were incubated for 20 min on ice and afterwards washed in 1mL of FACS buffer and finally resuspended in 500µL of the same buffer. Measurements were done on BD FACS Canto II flow cytometer (Flow Cytometry and Cell sorting facility, University of Dundee, UK) using blue (ex: 488 nm; em: 530±30 nm) and violet (ex: 405 nm; em: 450±50 nm) laser for detection of FITC and DAPI, respectively. Data were analysed on FlowJo™ 10.7.1. Software and GraphPad Prism. Gating strategy is detailed in Supplementary Figure 1.</p><p>Protein expression and purification. For expression of BRD4 tandem construct, N-terminally His6-SUMO-tagged BRD4 (1-463) or similar mutants were expressed in Escherichia coli BL21(DE3) at 18°C for 16h using 0.4mM isopropyl β-D-1-thiogalactopyranoside (IPTG). N-terminally His6-SUMO-tagged BRD2 tandem (73-455) was induced to express in E. coli BL21(DE3) with 0.3 mM IPTG at 18°C for 16h. E. coli cells were lysed using a pressure cell homogenizer (Stansted Fluid Power) or a CF1 Cell Disruptor (Constant Systems Ltd) and lysate clarified by centrifugation. Proteins were purified on a HisTrap HP 5 mL affinity column (GE Healthcare) by elution with an imidazole gradient. The proteins were dialyzed overnight into low imidazole concentration buffer in dialysis bags (14.5 kDa MWCO) with either TEV protease for BRD4 or SENP1 protease for BRD2 to remove the His6-SUMO tags. The cleaved proteins were then flowed through the HisTrap HP column a second time, allowing impurities to bind, as the recombinant proteins eluted without binding. The proteins were then additionally purified by cation exchange and size-exclusion chromatography using HiTrap SP HP 5 mL and Superdex-200 16/600 columns (GE Healthcare), respectively. The final purified proteins were stored in 20mM HEPES, pH 7.5, 100mM sodium chloride and 1mM TCEP. The VCB complex was expressed and purified as described previously 17 . Briefly, N-terminally His6-tagged VHL (54-213), ElonginC (17-112) and ElonginB (1-104) were co-expressed and the complex was isolated by Ni-affinity chromatography, the His6 tag was removed using TEV protease, and the complex further purified by anion exchange and size-exclusion chromatography. The BET protein BDs were expressed and purified as described previously 17 . Briefly, N-terminally His6-tagged BRD2-BD1 (71-194), BRD2-BD2 (344-455), BRD3-BD1 (24-146), BRD3-BD2 (306-416), BRD4-BD1 (44-178) and BRD4-BD2 (333-460) were expressed and isolated by Niaffinity chromatography and size-exclusion chromatography.</p><!><p>SEC experiments were carried out in a ÄKTA pure system (GE Healthcare) at room temperature. The oligomeric state of the BRD4 BD1-BD2 tandem protein in solution was analyzed by gel filtration in a buffer containing 20mM HEPES (pH 7.5), 100mM NaCl and 1mM TCEP using a Superdex 200 Increase 10/300 GL column (GE Healthcare) calibrated with globular proteins of known molecular weight (GE Healthcare, 28-4038-41/42). BRD4 tandem (25 µM) was incubated with SIM1 (25µM), MZ1 (25µM), MT1 (25µM) or DMSO (0.5 %) for 30 min at room temperature prior to injection. Sample volume for each injection was 200µl, and the flow rate was 0.8 ml/min. Peak elution was monitored using ultraviolet absorbance at 280nm. ITC. Titrations were performed as reverse titration on an ITC200 micro-calorimeter (Malvern). SIM1 was not soluble enough to be loaded at the required concentrations in the syringe (normal direct titration), therefore reverse titrations were performed. The titrations consisted of 19 injections of 2µl tandem BRD4 BD1-BD2 construct (WT or N140F or N433F) solution in 20mM Bis-Tris propane, 100mM NaCl, 1mM TCEP, 1.6% DMSO, pH 7.5, at a rate of 0.5µl/s at 120 s time intervals. An initial injection of protein (0.4µl) was made and discarded during data analysis. All experiments were performed at 25°C, whilst stirring at 750 r.p.m. SIM1 from 10mM DMSO stock solution and VCB were diluted in buffer containing 20mM Bis-Tris propane, 100mM NaCl, 1mM TCEP, pH 7.5. The final DMSO concentration was 1.6% v/v. BRD4 protein (200µM, in the syringe) was titrated into the SIM1-VCB complex (SIM1 16µM, VCB 32µM, in the cell). Data were fitted to a single-binding site model for each BRD4 mutant to obtain the stoichiometry (n), the dissociation constant (Kd) and the enthalpy of binding (ΔH). Data for WT BRD4 was fitted to a two sets of sites model to account for the reverse titration set-up whereby a two-site protein is titrated into a bivalent ligand (see Malvern MicroCal ITC analysis software using Origin™ User Manual, pg. 102). Data fitting was performed using Microcal LLC ITC200 Origin software provided by the manufacturer.</p><p>AlphaLISA assays. Ligands were titrated against 4nM His-tagged BRD4 BD2 and 10nM biotinylated JQ1. All reagents were diluted in 50mM HEPES, 100mM NaCl, 0.1% BSA, 0.02% CHAPS, pH7.5 (final concentration). On VCB premixed condition, the buffer also included 12.5µM VCB. Ligands were tested over an 11-point 3-fold serial dilution in duplicates for each concentration point, starting at 100µM without VCB or starting at 10µM with 20µM VCB, and giving a final DMSO concentration of 1%. Binding was detected using anti-His6 antibody-conjugated AlphaLISA acceptor beads and streptavidincoated donor beads (PerkinElmer), with a final concentration of 10µg/ml for each bead). Titrations were prepared in white 384-well Alphaplates (PerkinElmer), and read on a Pherastar FS plate reader (BMG) equipped with an AlphaLISA excitation/emission module. Data was analyzed and dose-response curves generated using GraphPad Prism. Each assay well had a final volume of 25µl. First 10µl of 2.5X ligand or 2.5X ligand with VCB was mixed with 5µl of a 5X mix of bromodomain and biotinylated JQ1 and incubated for 1h at room temperature. The assay plate was then moved to a dark room and 5µl of 5X acceptor beads were added and incubated for 1h. Then (still in darkness) 5µl of 5X donor beads were added, the plate was incubated for 1h before being read.</p><p>Fluorescence polarization assay. Fluorescence polarization (FP) competitive binding assays were run as described previously 25 with a final volume of 15µL, with each well solution containing 15 nM VCB protein, 10nM FAM-labeled HIF-1α peptide (FAM-DEALAHypYIPMDDDFQLRSF, "JC9") and decreasing concentrations of SIM1 (15-point 2-fold serial dilution starting from 10µM) or SIM1:BET tandem bromodomain protein (15-point 2-fold serial dilutions starting from 10µM SIM1:20µM BET protein). Assays were prepared in triplicate on 384-well plates (Corning 3575) and all measurements taken using a PHERAstar FS (BMG LABTECH) with fluorescence excitation and emission wavelengths (λ) of 485 and 520nm, respectively. Components were dissolved from stock solutions using 100mM Bis-Tris propane, 100mM NaCl, 1mM TCEP, pH 7.0, and DMSO was added as appropriate to ensure a final concentration of 1%. Control wells containing VCB and JC9 with no compound (zero displacement), or JC9 in the absence of protein (maximum displacement) were also included to allow for normalization. Normalized (%) displacement values were plotted against log[SIM1] and curves were fitted by nonlinear regression using GraphPad Prism to determine the IC50 values for each titration. Ki values were backcalculated from the Kd for JC9 (~2nM, determined from direct binding) and fitted IC50 values, as described previously 53 . Cooperativity (α) values were calculated from the ratio of binary Ki and ternary Ki values determined for JC9 displacement by SIM1 alone or SIM1+BET protein, respectively. SPR binding studies. SPR experiments were performed on Biacore T200 instruments (GE Healthcare) as described previously 25 . Immobilization of biotinylated VCB was carried out at 22 °C on a Series S SA chip in an immobilisation buffer containing 20 mM HEPES, 150 mM NaCl, 1 mM TCEP, 0.05 % TWEEN 20, pH 7.0. For binary studies (binding of SIM1 only) the final surface density of biotinylated VHL was approximately 1600-1800 RU; for ternary studies (binding of pre-formed SIM1:target protein complex) multiple lower surface densities of biotinylated VHL used (30-50 RU) to minimize mass transfer effects. Biotinylated VCB was prepared as previously described 54 . All interaction experiments were performed at 22 °C in a running buffer containing 20mM TRIS, 250mM NaCl, 0.2% (w/v) PEG 3350, 0.2% (w/v) BSA, 1mM TCEP, 0.05% TWEEN 20, 1% DMSO; pH 7.5. For binary interaction experiments, SIM1 were initially prepared at 1.5µM solution in the running buffer containing 1% DMSO. This stock solution was then serially diluted in the running buffer containing 1% DMSO (two-fold serial dilution). Solutions were injected individually (duplicate wells) in multi-cycle kinetic format (contact time 60s, flow rate 50 μL/min, dissociation time 150s) using a stabilization period of 30s and syringe wash (50 % DMSO) between injections. For ternary interaction experiments, SIM1 were initially prepared at 500nM in the running buffer containing 2% DMSO. This solution was mixed 1:1 with a solution of 1µM of the BET tandem protein in the running buffer without DMSO, to prepare a final solution of 250nM SIM1 and 500nM BET in running buffer containing 1% DMSO. This complex was then serially diluted in the running buffer containing 500nM BET and 1% DMSO (5-point three-fold serial dilution). All ternary experiments were run in both single-cycle and multi-cycle kinetic modes (two replicate series per experimental repeat, contact time 60s, flow rate 100µL/min, dissociation time 600s) using a stabilization period of 30s and syringe wash (50% DMSO) between injections. Two series of blank injections were performed for all experiments. Sensorgrams from reference surfaces and blank injections were subtracted from the raw data before data analysis using Biacore Insight Evaluation Software. To calculate the association rate (kon), dissociation rate (koff), and dissociation constant (KD), experiments were fitted using a 1:1 Langmuir interaction model, with a term for mass-transport included.</p><p>NanoBRET Ubiquitination, Ternary Complex, and Biosensor Experiments. For endogenous live cell BET:Ubiquitin,BET:VHL, and BET:CRBN assays, CRISPR HiBiT-BRD2, HiBiT-BRD3, and HiBiT-BRD4 HEK293 cells stably expressing LgBiT were transfected with 2µg of HaloTag-UBB,HaloTag-VHL, or HaloTag-CRBN vectors in 6-well plates using FuGENE HD (Promega). For full transient NanoBRET experiments with NanoLuc-BRD4 WT, N433F, or N140F mutants, HEK293 cells (8 ×10 5 ) were co-transfected with 0.02µg NanoLuc-BRD4 and 2µg of HaloTag-VHL vectors. For transient NanoBRET experiments with the BRD4 NL-BD1-BD2-HT biosensor containing either the WT tandem BD1-BD2 domain the N433F mutation, HEK293 cells (8 ×10 5 ) were transfected with 0.02µg biosensor plasmid and 2µg carrier DNA. The following day, transfected cells (2 × 10 4 ) were replated in quadruplicate into white 96-well tissue culture plates in the presence or absence of HaloTag NanoBRET 618 Ligand (Promega) and incubated overnight at 37°C, 5% CO2. For kinetic experiments, medium was replaced with Opti-MEM (Gibco) containing a 1:100 dilution of Vivazine (Promega), and plates were incubated at 37 °C, 5% CO2, for 1h before addition of DMSO or 10nM-1µM final concentration of the indicated compounds. Continual BRET measurements were then made every 3 min up to 5h on a CLARIOstar equipped with an atmospheric control unit (BMG Labtech) set to 37 °C and 5% CO2. For the biosensor experiments, the cells were treated with a 3-fold serial titration of 10µM of the indicated compounds. NanoBRET NanoGlo (Promega) substrate was added and BRET was measured two</p>

ChemRxiv

Enzyme repurposing of a hydrolase as an emergent peroxidase upon metal binding

As an alternative to Darwinian evolution relying on catalytic promiscuity, a protein may acquire auxiliary function upon metal binding, thus providing it with a novel catalytic machinery. Here we show that addition of cupric ions to a 6-phosphogluconolactonase 6-PGLac bearing a putative metal binding site leads to the emergence of peroxidase activity (k cat 7.8 Â 10 À2 s À1 , K M 1.1 Â 10 À5 M). Both X-ray crystallographic and EPR data of the copper-loaded enzyme Cu$6-PGLac reveal a bis-histidine coordination site, located within a shallow binding pocket capable of accommodating the o-dianisidine substrate.

enzyme_repurposing_of_a_hydrolase_as_an_emergent_peroxidase_upon_metal_binding

Introduction<!>Results and discussion<!>Conclusions

<p>Metal ions are present in nearly half of the characterized proteome, 1,2 and metalloenzymes catalyze some of nature's most challenging reactions. 3 This versatility has inspired a number of different strategies to create articial metalloenzymes in the past twenty years. [4][5][6] Guided either by computation or intuition, [4][5][6][7] articial metalloenzyme design by-and-large relies on the de novo introduction of a metal-binding site within a protein scaffold or peptide. [8][9][10][11][12][13] For this purpose, various strategies have been pursued including covalent-, supramolecular-and dative anchoring. [4][5][6] The latter approach may rely on incorporating either natural-or non-natural aminoacids as ligands (i.e. bipyridinylalanine etc.). [14][15][16] Alternatively, metal-substitution within a metalloenzyme can lead to novel catalytic activity. [17][18][19][20][21] Eventually, the nascent catalytic activity may be further improved by directed evolution strategies. [22][23][24][25] To complement these efforts and in an "enzyme tinkering spirit", 26,27 we hypothesized that non-metal containing proteins may, through the course of random mutations, evolve a putative metal binding site. Upon acquisition of a metal, nascent catalytic activity may arise, Fig. 1. As with enzyme promiscuity, [28][29][30] if the newly acquired asset provides a competitive advantage to the cell, 31,32 it may evolve to a highly efficient metalloenzyme.</p><p>To test this hypothesis we employed the "Search for Three dimensional Atom Motifs in Protein Structure" (STAMPS) algorithm, 33 which allows to identify proteins by a systematic search in the protein databank (PDB) for structures possessing a given motif in a topology similar to that adopted by the functional motif in a reference protein. Using this approach, we recently reported on the in silico identication of non-metallated twohistidine one-carboxylate metal binding motifs within the structurally characterized proteins of the PDB. 34 Herein we report our efforts to valorise the in silico study by creating an articial metallo-peroxidase upon addition of a metal cofactor to a protein harboring a latent mononuclear metal binding site, Fig. 1.</p><!><p>The proteins bearing a latent non-metallated two-histidine onecarboxylate motif (HHD/E, hereaer) were identied previously using the STAMPS algorithm. For this study, we selected seven scaffolds bearing an HHD/E motif located within a binding pocket predisposed to bind metals. 33,34 The search was expanded to include HHN/Q motifs, revealing six additional potential metal binders following a single point mutation. In total, six of the thirteen cloned proteins could be overexpressed in E. coli and puried using a Strep-tag II (Fig. S1-S4 †).</p><p>These wild-type proteins or their single mutant isoforms (pdb code: 3D53, 2F99, 1JSY, 1FHI (bearing a Q83E mutation), 1MEJ (bearing a N106D mutation), 3OC6 (bearing an N131D), see Table S1 †) were tested for their peroxidase activity in the presence of various transition metal salts including VOSO 4 , MnCl 2 , FeCl 2 , CoSO 4 , NiCl 2 and CuSO 4 . For this purpose, odianisidine (0.75 mM) and hydrogen peroxide (1.5 mM) were added to a buffered solution containing the protein (20 mM) and the metal salt (25 mM). The appearance of the quinone-diimine oxidation product was monitored at 460 nm. 35 Gratifyingly, this screen revealed one hit: in the presence of CuSO 4 , 6-phosphogluconolactonase bearing an N131D mutation (Cu$6-PGLac hereaer) catalyzes the oxidation of o-dianisidine (Fig. 2A, S5 and S6 †).</p><p>Inspection of the X-ray structure of the putative 6-phosphogluconolactonase (6-PGLac, apo form) at 1.81 Å reveals a Rossmann fold annotated to the glucosamine-6-phosphate isomerases/6-phosphogluconolactonase family (PF01182) in PFAM database, with the hydrolytic Glu149-His151 dyad and the latent metal binding site 26 Å apart, (Fig. 3A and Table S4 †). The alleged 6-phosphogluconolactonase activity 36 was conrmed by 31 P NMR, both in the absence and in the presence of cupric ions, emphasizing the moonlighting nature of this emergent peroxidase activity, Fig. S7. † 29 Among the PF01182 family proteins, the triad residues (His67, His104, and Asn(Asp) 131) are not conserved, thus suggesting that this potential metal binding motif is not functionally relevant (Fig. S8 †).</p><p>No peroxidase activity was observed upon substitution of hydrogen peroxide by the dioxygen/ascorbate couple. However, the enzyme cascade consisting of glucose oxidase and Cu$6-PGLac restores catalytic activity in the presence of glucose, dioxygen and o-dianisidine (Fig. 2A and S6B †). Signicant rate enhancement was achieved using t-butyl hydroperoxide (t-BuOOH) instead of H 2 O 2 as oxidant (Fig. 2A). The peroxidation of catechol and guaiacol by Cu$6-PGLac was investigated (Fig. S5 †). No peroxidation product was detected for guaiacol with either H 2 O 2 or t-BuOOH as oxidant. Cu$6-PGLac catalyzed the peroxidation of catechol. However, the background peroxidation with CuSO 4 alone was signicant, which contrasts to both o-dianisidine and guaiacol.</p><p>To gain structural insight into the metalloenzyme activity, 6-PGLac crystals (vide supra) were soaked with a mother liquor containing excess CuSO 4 . This procedure however did not yield suitable diffraction data. Thus, 6-PGLac was co-crystallized with 3 mM CuSO 4 followed by soaking with 12 mM CuSO 4 (Cu$6-PGLac). The X-ray structure was rened to a resolution of 1.39 Å (Fig. 3B, Table S4 and Fig. S9 †). The overall structure is nearly identical to that of 6-PGLac, apo form (Root Mean Square Deviation (RMSD) ¼ 0.32 over 243 Ca atoms, Fig. 3A and B and S9A †). The structure contains three fully occupied copper ion binding sites with strong anomalous signals (Table S6 †): two copper ions are bound to surface histidines and aspartates crosslinking two protein monomers in the crystal (Cu2 is bound to His9 and Asp3 0 and Cu3 is bound to His95 and Asp158 0 and Asp220 0 respectively, see Fig. S9B †). The third copper (Cu1) is located in the putative metal binding site and displays a Tshaped [2 + 1] coordination geometry. The copper is bound to His67 and His104 (at a distance of 2.00 Å and 1.92 Å, respectively) and displays a weak contact with a water molecule (WATxx 2.80 Å), which is held in place via a hydrogen bond to the carbonyl oxygen of Val101 (O/O 2.81 Å). This structure also contains further unspecically bound copper ions in the lactonase active site as well as on the surface (Fig. S9C and Table S6 †). However, these binding sites are characterized by low occupancy, as reected by the height of anomalous difference density peaks, and thus presumably have a far lower binding affinity for copper ions. The coordination geometry around Cu1 is T-shaped with two trans-histidines and a water ligand. Although reminiscent of the Cu-coordination recently reported for the lytic polysaccharide monooxygenase from Aspergillus oryzae, it lacks the terminal amine ligand, characteristic of the "histidine brace". 37 While two-and three coordinate copper geometries are frequently encountered in metalloenzymes and coordination complexes, earlier transition metals oen prefer higher coordination numbers. 3 We speculate that the lack of peroxidase activity observed with all other metal salts tested in the presence of 6-PGLac may be caused by the ill-suited low coordination geometry imposed by the putative active site.</p><p>The affinity of 6-PGLac for Cu(II) was determined using tryptophan-uorescence quenching. Two dissociation constants could be extracted from the titration prole: K d1 ¼ 0.83 AE 0.11 mM, K d2 ¼ 130 AE 3.3 mM; conrming the presence of one tight and weaker copper binding sites (Fig. S13 †).</p><p>The potentially coordinating oxygen of Asp131 identied by the in silico search and of the Tyr69 are located 6.55 Å and 4.25</p><p>Å from Cu1 respectively. The low coordination number of Cu1, coupled with the possibility of photoreduction by the X-ray dose, led us to explore whether cupric state may have additional ligands.</p><p>In order to investigate the involvement of neighboring amino acids in the coordination of catalytically competent cupric ions, the kinetic saturation proles of Cu$6-PGLac and mutants (Fig. S12 †) thereof were determined. The results are summarized in Table 1, Fig. 2B and S12A. † Cu$6-PGLac bearing the HHD triad displays Michaelis-Menten behavior with an efficiency of k cat /K M ¼ 6.9 Â 10 3 M À1 s À1 (Table 1, entry 1). Interestingly, the K M value of Cu$6-PGLac is comparable to that of the naturally occurring peroxidase (For horseradish peroxidase (HRP), 13 mM). 38 The k cat /K M however is lower than that of HRP (7.1 Â 10 7 M À1 s À1 ), 38 whereas it's equal to or greater than those of the articial peroxidases (ferric porphyrin-binding antibody (3.4 Â 10 4 M À1 s À1 ) 39 and ferric porphyrin-binding xylanase (3.7 Â 10 2 M À1 s À1 ) 40 ). Deletion of the surface histidines conrms that both Cu2 and Cu3 (bound to His9 and His95 respectively) are not catalytically competent. Indeed both Cu$6-PGLac H9A and Cu$6-PGLac H95A display nearly identical kinetic behavior when compared to Cu$6-PGLac (Table 1, entries 2 and 3). Unfortunately, all attempts to mutate two surface histidine residues invariably led to inclusion bodies which could not be renatured. In stark contrast, mutation of either His67 or His104 binding residues into non-coordinating amino acids completely shuts off catalytic activity, (Table 1, entries 4 and 5), conrming that the catalytic metal is indeed located in the cavity anked with residues His67, His104, Asp131 and Tyr69. Unexpectedly, mutation of Asp131 to either a hydrophobic, a coordinating or a polar amino acid does not affect the catalytic performance. This suggests that this residue does not bind either to cupric-or cuprous ions (Table 1, entries 6-9). Position 69 in contrast has a larger impact on the catalytic performance: Cu$6-PGLac Y69L displays a moderately improved k cat , albeit at the cost of K M (Table 1, entries 10-12). Based on this limited mutagenesis study, we are condent that the catalytic performance could be improved signicantly with a larger mutagenesis campaign.</p><p>The involvement of Tyr69 was assessed by performing a docking simulation between Cu$6-PGLac and o-dianisidine. It revealed one major low-energy conformation with the polar edge of the biaryl-substrate pointing towards the copper active site, Fig. 4. One methoxy oxygen atom of o-dianisidine is hydrogen bonded to the oxygen atom of Tyr69 (H 3 CO/O Y 2.8 Å) and interacts with Cu1 (H 3 CO/Cu 2.6 Å). The other hydrogen bonds were observed between nitrogen atom of o-dianisidine and the oxygen atom of Tyr69 at a distance of 3.0 Å, and the carboxyl oxygen atom of Asp131 at 2.6 Å.</p><p>The articial peroxidase was scrutinized by X-band EPR analysis at 77 K (Fig. 5). The presence of multiple Cu 2+ species in Cu$6-PGLac results in a complicated spectrum in the parallel region (Fig. 5 and S14A †), highlighting the presence of multiple Cu 2+ binding sites as observed in the crystal structure (Cu1; His67 and His104, Cu2; His9, Cu3; His95). Analysis of histidine-deleted variants revealed that Cu$6-PGLac H9R exhibits an axial EPR spectrum, typical of a single Cu 2+ species (Fig. 5 and S14A †), whereas Cu$6-PGLac H95F shows an almost identical spectrum to that of Cu$6-PGLac. This suggests that His9 binds to Cu 2+ more tightly than His95. In contrast, the H67F and H104F single mutants resulted in markedly different spectra (Fig. 5), suggesting that both His67 and His104 are involved in copper coordination in Cu$6-PGLac. A reasonable simulation of the spectrum of Cu$6-PGLac H9R provided detailed EPR parameters (see Fig. S14C † for further details). Again here, a more detailed analysis was hampered by the formation of inclusion bodies upon mutation of multiple surface histidine residues. Given the parameters of the parallel region (g z ¼ 2.230 (5.9 AE 0.9) Â 10 3</p><p>a The kinetic measurements were determined in triplicate and tted using the Michaelis-Menten equation. The minimal background reaction caused by free copper was subtracted from the raw data for tting purposes. b The catalytic activity was too small to be determined (Fig. S12B). and A z ¼ 17.7 mT), Peisach-Blumberg analysis 41 suggests that this Cu 2+ species is a type 2, reminiscent of the type 2 copper in the multicopper oxidase (g z $ 2.24 and A z ¼ 13-19 mT). 42 This copper center bears two nitrogens and one or two water ligand(s) in the equatorial plane in the reported crystal structure. 43,44 Superhyperne coupling pattern due to the ligating nitrogen atoms is apparent in the perpendicular region (Fig. S14 †). Thus, these results suggest that the 6-PGLac provides the Cu1 ion with a coordination environment suitable for coordination of both the peroxide and the substrate to yield a transient end-on (alkyl-)peroxide copper(II) species. 45,46 Although, no known natural peroxidase contains a mononuclear copper active site to our knowledge, mononuclear cupric coordination complexes display peroxidase activity and have been shown to operate via concerted H atom abstraction with O-O bond scission and subsequent attack at a substrate by a Cu(II)-oxyl radical (i.e. L n Cu II -Oc). [45][46][47]</p><!><p>The enzyme repurposing strategy contrasts with other articial metalloenzyme approaches. Indeed, as we demonstrate herein, the latent metal binding site is present but not metallated in the wild-type 6-PGLac. Simple metal salt addition thus suffices to endow a protein with novel catalytic activity, very different from the native activity. Furthermore, this strategy relying upon metal acquisition can be viewed as complementary to enzyme promiscuity, offering a novel means to swily acquire enzymatic catalytic activity, vastly different from the native activity. Although the STAMPS algorithm suggested a possible facial triad coordination, a T-shaped Cu coordination was observed upon Cu(II) supplementation. We believe that this may be due to the size of the latent metal binding site which allows for signicant side chain exibility: a feature not taken into consideration in the STAMPS algorithm. Gratifyingly, this lowcoordination bis-histidine copper geometry combined with the additional surrounding Tyr69 and Asp131 residues provides unique opportunities to ne-tune or alter the emergent peroxidase activity. The compatibility of the articial metalloenzyme with other enzymes (e.g. glucose oxidase) as well as cell lysates suggests that directed evolution strategies may be used to further optimize its performance.</p>

Royal Society of Chemistry (RSC)

RGD conjugated cell uptake off to on responsive NIR-AZA fluorophores: applications toward intraoperative fluorescence guided surgery

The use of NIR-fluorescence imaging to demarcate tumour boundaries for real-time guidance of their surgical resection has a huge untapped potential. However, fluorescence imaging using molecular fluorophores, even with a targeting biomolecule attached, has a major shortcoming of signal interference from non-specific background fluorescence outside the region of interest. This poor selectivity necessitates prolonged time delays to allow clearance of background fluorophore and retention within the tumour prior to image acquisition. In this report, an innovative approach to overcome this issue is described in which cancer targeted off to on bio-responsive NIR-fluorophores are utilised to switch-on first within the tumour. Bio-responsive cRGD, iRGD and PEG conjugates have been synthesised using activated ester/amine or maleimide/thiol couplings to link targeting and fluorophore components. Their off to on emission responses were measured and compared with an always-on nonresponsive control with each bio-responsive derivative showing large fluorescence enhancement values. Live cell imaging experiments using metastatic breast cancer cells confirmed in vitro bio-responsive capabilities. An in vivo assessment of MDA-MB 231 tumour imaging performance for bio-responsive and always-on fluorophores was conducted with monitoring of fluorescence distributions over 96 h. As anticipated, the always-on fluorophore gave an immediate, non-specific and very strong emission throughout whereas the bio-responsive derivatives initially displayed very low fluorescence. All three bio-responsive derivatives switched on within tumours at time points consistent with their conjugated targeting groups. cRGD and iRGD conjugates both had effective tumour turn-on in the first hour, though the cRGD derivative had superior specificity for tumour over the iRGD conjugate. The pegylated derivative had similar switch-on characteristics but over a much longer period, taking 9 h before a significant emission was observable from the tumour. Evidence for in vivo active tumour targeting was obtained for the best performing cRGD bio-responsive NIR-AZA derivative from competitive binding studies. Overall, this cRGD-conjugate has the potential to overcome the inherent drawback of targeted always-on fluorophores requiring prolonged clearance times and shows excellent potential for clinical translation for intraoperative use in fluorescence guided tumour resections.

rgd_conjugated_cell_uptake_off_to_on_responsive_nir-aza_fluorophores:_applications_toward_intraopera

Introduction<!>Results and discussion<!>Synthesis and characterisation<!>Photophysical properties<!>In vitro live MDA-MB 231 cell imaging<!>In vivo tumour imaging<!>Conclusion<!>Experimental section

<p>Intraoperative uorescence imaging to guide surgical resections in real-time has huge untapped potential. Advantages lie in its ease of use, enhanced safety prole over radiolabelling and the ability to acquire image data in real-time during surgical procedures. 1 Currently indocyanine green (ICG) is the sole clinically approved near infrared red (NIR) uorophore. 2 Clinical uses include vascularisation assessments during reconstructive 3 and bowel anastomoses 4 surgeries and lymph node mapping in digestive tract, 5 cervical 6 and breast 7 tissues. Due to its non-specicity and very short in vivo half-life, its use as an agent to demarcate tumour boundaries for surgical resection is limited to hepatocellular carcinoma of the liver. 8,9 As a result, new classes of NIR-uorophore have recently emerged, several of which are bio-conjugated to enhance their affinity for specic cancer types. 10 However, a remaining complexity for in vivo uorescence imaging using molecular uorophores exists. Following intravenous administration, uorophore distributes to all vascularised regions within seconds, resulting in a strong non-specic uorescence. This necessitates an unpredictable time delay to allow background uorophore clearance, following which imaging is achievable if sufficient uorophore is retained in the region of interest (ROI) (Fig. 1a). This limitation is irrespective of whether the uorophore alone is used or if conjugated to a targeting group (e.g. antibody), as an initial broad distribution will still occur. The time between administration and imaging depends on several parameters such as rates of accumulation and clearance from both the ROI and surrounding tissues and elimination from the body via metabolic and excretion pathways. Each of these factors can be inuenced by the structure of uorophore itself and by groups conjugated to it, but a time lag before imaging is unavoidable. To provide sufficient contrast for imaging, it is necessary to identify an optimal time point at which a maximum quantity of uorophore is retained in the ROI with a minimum remaining in the surrounding tissues. For example, antibody conjugated uorophores have been adopted in recent clinical trials for visualising breast and colorectal cancers utilising labelled bevacizumab and carcinoembryonic antigen (CEA) respectively. 11,12 Yet in spite of using these expensive cancer specic antibody technologies, uorescence images could only be acquired between two and four days post administration. The prolonged waiting period to achieve sufficient tissue contrast is due to the very long biological half-lives of antibody labelled agents. This time delay adds signicant uncertainty to their practical use and raises doubts as to whether all of the cancer would then be detectable by the low levels of remaining uorophore. In effect, what makes large molecular weight antibodies attractive for sustained drug delivery, can work against them when used for the delivery of contrast agents (Fig. 1a).</p><p>Thus, for rapid and accurate intraoperative imaging innovative alternative approaches are needed to enhance target-tobackground signal ratio at early stages following uorophore introduction. One plausible solution is to exploit a mechanism of selective uorescence quenching in the background areas, whilst rst establishing the emitting potential of the uorophore in the ROI (Fig. 1b). This overcomes the issue of waiting for background clearance and allows observation of dynamic tissue accumulation in real-time during the course of the surgical procedure. While beyond the scope of this work, it may become feasible that dynamic images are recorded continuously, with tissue classications determined using in-line soware image analysis.</p><p>In our previous work, we have shown that bio-responsive NIR-AZA uorophore 1 performs as an excellent probe capable of real-time continuous imaging of fundamental cellular processes such as endocytosis, lysosomal trafficking and efflux (Fig. 2). 13a Specically, the highly photostable NIR uorescent probe 1 has off/on uorescence switching controlled by a reversible phenol/phenolate interconversion (Fig. 2). Emission from the probe was shown to be highly selective for cellular lysosomes and, as the off/on switching mechanism is reversible, it is capable of real-time continuous imaging of lysosomal trafficking in 3D or 4D over prolonged time periods without perturbing normal cellular function. 14 Preliminary in vivo imaging in a mouse tumour xenogra model showed good tumour discrimination 24 h post i.v. injection of 1 with no observable toxicity. 13a These positive in vitro and in vivo features are good indicators that bio-responsive NIR-AZA uorophores warrant further investigation for translation towards clinical use in uorescence-guided surgery. In recent preclinical tests, always-on NIR-AZA uorophores have shown their potential for lymph node mapping and ureter identication using clinical instrumentation. 15 In this report, we describe the synthesis, photophysical characterisation, in vitro and in vivo imaging assessment of bio-responsive NIR-AZA uorophores conjugated to cyclic-RGD peptide sequences and polyethylene glycol polymer acting as active or passive targeting agents respectively. Breast cancer is a key health concern for women, with over two million new cases diagnosed worldwide annually. Screening programs have resulted in most breast cancers being identied in the early stages with over 80% of breast cancer patients undergoing surgery as part of their treatment. Numerous trials have shown that for patients with between zero and three node metastases, breast-conserving surgery has similar or superior outcomes to mastectomy. 16 As tumour-free surgical margins are critical to the success of breast-conserving surgery, utilising uorescence guidance to improve surgical outcomes could have signicant patient benet.</p><p>Integrins are membrane bound cell adhesion receptors important for cell-cell and cell-extracellular matrix (ECM) interactions. They act as transmembrane linkers between extracellular ligands such as ECM proteins, growth factors, matrix degrading proteins and the cytoskeleton, which serves to modulate various essential signalling pathways in most cells. 17 Integrins such as avb3 and avb5 (among others) are known to play a key role in tumour angiogenesis and are associated with the metastasis of solid tumours. 18 Of the integrins, avb3 is one of the most studied as it is the most prevalent integrin involved in the regulation of angiogenesis and is widely expressed on tumour blood vessels. 19 Over-expression of avb3 integrin has been associated with increased tumour growth in breast cancer and it has been shown that the activation of avb3 is a contributing factor for metastasis in breast cancer models. 20 The tripeptide arginine-glycine-aspartic acid (RGD) sequence can recognise and bind anb3 and avb5 integrins and promote cellular internalisation with conjugates of the more stable cyclic variant c(RGDfK) being widely investigated as a selectivity enhancer for tumour therapies and diagnostics. 21 The related iRGD peptide sequence (cCRGDKGPDC) has been reported to provide both specic integrin targeting and increased tumour uptake and penetration. It contains the RGD motif, which mediates binding to the endothelial cell membrane expressing the av integrins but upon proteolytic cleavage a second neuropilin-1 (NRP-1) binding motif (CRGDK) is revealed to promote internalisation. 22 Both RGD and iRGD conjugates have been investigated to improve drug selectivity with chemotherapeutic conjugates such as RGD-doxorubicin and RGDpaclitaxel showing promising preclinical results in breast carcinoma mouse models. 23 RGD conjugates for uorescent imaging using always-on probes has been explored in several preclinical models including breast cancers. 24 For this study we have chosen one cRGD (c[RGDfK(PEG-PEG)]) and one iRGD (cCRGDKGPDC) peptide sequence for conjugation to the bio-responsive NIR-AZA imaging platform. It was hoped that these low molecular weight peptides would promote rapid uptake and switch-on of emission preferentially within tumours allowing a high tumour to background ratio (TBR) to be established without waiting for prolonged clearance times. In practice, it is envisaged that they would be administered and visualised intraoperatively, thereby not impeding the normal surgical or hospital workow.</p><!><p>Incomplete tumour removal during surgical resection is closely related to cancer reoccurrence and patient survival rates. A major challenge in achieving cancer free margins is to fully distinguish between all of the cancerous growth and normal tissue during surgery. While high denition images obtained by PET, CT or MRI scans identify and diagnose tumour growths prior to surgery, such images are not overly useful to guide surgical resection during the operation. Currently, tumour margins are typically assessed by visual assessment and palpation of the tumour intraoperatively. However, the possibility of micro-invasion of the surrounding tissues can make it difficult to determine an adequate tumour-free excision margin. In this report, we have developed synthetic routes to RGD conjugated bio-responsive uorophores, examined their photophysical and in vitro cellular emission proles and tested their in vivo tumour imaging performance using a human breast tumour model in mice.</p><!><p>Cell uptake responsive probes 2 and 3 were selected for synthesis using activated ester/amine coupling to conjugate the cRGD sequence and cysteine to maleimide addition for the covalent linkage of the iRGD peptide sequence (Fig. 3). Two bioconjugation approaches were adopted to conrm synthetic exibility of NIR-AZA bio-uorophores to functionalisation with targeting moieties.</p><p>Synthesis of 2 required uorochrome 4 which has been previously reported, though only in reaction with an aminopegylated polymer to produce 1 (Scheme 1). 13a For this study, the amino-pegylated substituted cRGD substrate 5 (cyclo[Arg-Gly-Asp-D-Phe-Lys(PEG-PEG-NH 2 )]) was selected as it is known to be an ideal construct for housing the integrin recognising tripeptide sequence (Scheme 1a). The reaction of 4 and 5 in DMSO at rt was followed by HPLC and 1 H NMR which showed a clean conversion to conjugate 2 in 4 h. Transformation of activated ester 4 into conjugate 2 was clearly distinguishable by the shi of the key methylene 1 H NMR peak from 5.53 to 4.61 ppm (Scheme 1b). Purication of product was achieved using preparative reverse phase HPLC and the structure conrmed by high-resolution MS and NMR methods.</p><p>The generation of iRGD conjugate 3 rst required the synthesis of the corresponding maleimide-substituted uorochrome 7 (Scheme 2). This was readily achievable starting from the previously reported derivative 6 which was subjected to the nitration conditions of KHSO 4 /KNO 3 at reux in CH 3 CN/H 2 O to yield the o-nitro phenol substituted substrate 7. 25 Synthesis of iRGD peptide 8 followed literature procedures to produce (cCRGDKGPDC) which was coupled with N-acetyl protected cysteine, through the amine of the lysine residue, to provide the nal thiol substituted peptide. 26 Bio-conjugation via cysteine to maleimide addition was efficiently achieved by reaction of 7 with iRGD 8 in DMSO at rt for 30 min to produce the required derivative 3 (Scheme 2). Product purication utilised preparative reverse phase HPLC with the structure conrmed by usual analytical methods (ESI †).</p><p>To allow comparisons be made between bio-responsive and non-responsive conjugates the iRGD substituted always-on control 10 was synthesised. Conjugate 10 is similar in structure to 3 but has the o-nitro phenol uorescence switching substituent replaced by a water solubilizing alkylsulfonic acid group (Scheme 3). The synthetic route adopted to make this control utilised the reaction of known uorochrome 9 with peptide 8 to produce 10. 25 The cysteine/maleimide coupling proceeded smoothly in phosphate buffered saline (PBS) at pH 7.2 with iRGD conjugated 10 obtained in good yield (Scheme 3, ESI †).</p><p>To test the extent of advantage gained from utilising integrin targeting RGD peptide conjugates such as 2 and 3 versus a passive accumulating agent such as a PEG group, the pegylated bio-responsive 1 was also included for testing in the study as a comparative control (Fig. 2). This we envisaged would allow a direct imaging performance evaluation between bioresponsive uorophores using either active targeting peptides or the passive enhanced permeability and retention (EPR) effect of a PEG group.</p><!><p>The photophysical properties of the bio-responsive uorophores 2 and 3 and the always-on control 10 were studied in solutions of Dulbecco's modied eagle's cell medium (DMEM) containing 10% fetal bovine serum (FBS). Absorption and emission wavelengths are listed in Table 1 and are, as would be expected for the NIR-AZA class, in the 690-730 nm range. At pH 7.4 uorescence intensity of always-on 10 was 17-and 15-fold greater than 2 and 3 respectively (Table 1). This illustrates both the emissive potential of the peptide conjugates and the ability to quench the bio-responsive derivatives in relevant biological media (Table 1, spectra and inset).</p><p>In order to display the responsive nature and full emissive potential of 2 and 3, their DMEM solutions were sequentially acidied (Fig. 4a and b). This caused a successive increase in emission intensity as acidity increased, with a maximum intensity reached at approximately pH 4. Plotting the measured data revealed pK a values of 4.9 for both 2 and 3 which is consistent with the previously reported value of 4.6 for 1 (Fig. 4 insets, Fig. S1-S3 †). 13a This shows that the conjugating group does not overly inuence the important p-nitro phenol emission-controlling feature. While it is recognized that the extracellular matrix of a solid tumour can be more acidic than normal tissue, intracellular organelles such as late endosomes and lysosomes are also acidic ranging between pH 4.5 and 5.5. 27 Encouragingly, the measured uorescence enhancement factor (FEF) for 2 and 3 between pH 7.2 and that of 4.5 (as found in lysosomes) was 23 and 18 respectively (Fig. 4c). As such, a switch-on of emission could be expected to occur both in the localised extracellular tumour microenvironment and upon cancer cell uptake. As emission quenching in the off states of 2 and 3 at pH 7.2 is highly effective, good background to noise differentials could be anticipated. In contrast, control uorophore 10 showed no absorption or emission spectral changes between pH 7.2 and 4.5 (Fig. S4 †).</p><p>The next stage involved testing 2, 3 and control 10 in live cell imaging using the epithelial human breast cancer cell line MDA-MB 231 that is known to express membrane integrins including avb3 and avb5. MDA-MB 231 is a highly aggressive triple-negative cell line, with its invasiveness mediated by proteolytic degradation of the extracellular matrix. 28 The metastatic invasive nature of MDA-MB 231 cells is closely associated with large acidic vesicles (LAV) in which endocytosed extracellular matrix can be digested by activated lysosomal proteinases such as cathepsin. 29 As these intracellular LAVs have a pH of approximately 4, they also could activate the bio-responsive uorophores upon cancer cell uptake in addition to lysosomes.</p><!><p>With the responsive nature of 2 and 3 established, the potential for translation of these constructs to real-time live cell imaging was investigated. In order to illustrate the imaging effect of the bio-responsive characteristics of 2 and 3 the always-on 10 was also imaged as a positive control. For live MDA-MB 231 cell experiments, chamber slide seeded cells were placed in a wideeld microscope surrounded by an incubator to maintain the temperature at 37 C and CO 2 at 5%, following which an imaging eld of view containing viable cells was chosen. The cells were treated with either 2, 3 or 10 (1-5 mM) and time-lapse NIR-uorescence and differential interference contrast (DIC) images were acquired over 120 min. Image data showed that for always-on 10, a uorescence specic to the plasma membrane was rapidly observed within 15 min, which could be attributed to its strong association with the cell membrane (Fig. 5a). As expected, the endothelial-like morphology of the cell line is distinguishable by its membrane lopodia projections which is characteristic of its metastatic invasive phenotype (Movie S1 †). Aer 60 min incubation, both cell membrane and intracellular vesicle staining of the cytoplasm could be observed, both of which persisted at 120 min (Fig. 5b and c). A wider eld of view showing a larger number of cells and Z-stack images can be seen in Fig. S5, S6 and Movie S2. † Revealingly, the bio-responsive NIR-AZAs 2 showed no cell membrane staining in the rst 15 min of incubation and only following this time point could intracellular regions of uorescence be observed (Fig. 6a). The intracellular punctate staining pattern is consistent with those previously observed for 1 and are due to a selective bio-responsive switch-on of emission within the acidic vesicles of the cytoplasm. 13a,14 The images revealed two distinct vesicle sizes, the smaller of which are attributable to cellular lysosomes and the bigger LAVs specic to the metastatic nature of MDA-MB 231 cells (Movie S3 and S4 †). At 60 and 120 min the intracellular uorescence intensity increased, but at no point was membrane uorescence observed (Fig. 6b, c and Movie S5 †). This lack of plasma membrane uorescence shows that 2 can translocate across the membrane without uorescence being activated. A wider eld of view showing a larger number of cells can be seen in Fig. S7. † Z-Stack analysis of cells conrmed that regions of uorescence were within the cytoplasm (Fig. S8 and Movie S6 †). Similar results were obtained from imaging experiments using bio-responsive iRGD NIR-AZA 3 which can be seen in Fig. S9, S10 and Movie S7. † In addition, similar results were obtained from imaging experiments with HeLa Kyoto cells using bio-responsive 2 which can be seen in Movie S8. †</p><p>The different cell staining patterns between 10 and 2, 3 over time shows the delity of the uorescence switching and the potential signal to background contrast advantage of the bioresponsive NIR-AZA probes. The next challenge of this work was to examine if a preferential in vivo switch-on of bioresponsive NIR-AZAs in cancerous tumour could allow both early (due to initial switch-on) and later (due to retention of switched on probe) stage discrimination of tumour from background.</p><!><p>For this study, the human breast cancer cell line MDA-MB 231 was selected for its relevance to clinical forms of aggressive breast cancers for which the rst line of treatment is oen surgical resection. The ability of bio-responsive conjugates 1, 2 and 3 were tested using subcutaneous tumours grown in nude mice. Fluorophore 10 was also included as a positive control in the study to demonstrate the advantage of using off to on responsive uorescence over a constant emission. It was anticipated that 1, 2 and 3 would remain predominately uorescent silent until tumour uptake occurred causing a uorescence signal modulation to on. Experimental measurement of changes in tumor-to-background ratio (TBR) over time was the preferred method to quantify differences between bioresponsive and always-on uorophores. Pegylated bioresponsive 1 was included to compare the turn-on time differences between passive PEG and the active integrin targeting of 2 and 3. The expectation being that the larger EPR dependent PEG conjugate would be slower.</p><p>Each uorophore was subjected to in vivo analysis using a standard dosing set at 2 mg kg À1 delivered by i.v. tail vein injection. Post injection, images were acquired initially at regular intervals between 10 min and 9 h and thereaer less frequently at 24, 48 and 96 h. The method used for image analysis was consistent across all experiments with TBR values calculated by measuring tumour ROI uorescence against an average of three equally sized background ROI regions, two of which were close to and one distant from the tumour (Fig. S11 †). In previously reported preclinical studies, when imaging through the skin, a TBR ratio of two was shown to be a clinically relevant threshold. 24a,30 As such, we adopted this value as a point of reference to compare result from different uorophores and different time points for individual uorophores.</p><p>For always-on iRGD control 10, from 10 to 60 min post i.v. injection an immediate strong and non-specic uorescence was observable throughout the animals, with no discernible bias for tumour as demonstrated by the measured TBRs of below 1.3 (Fig. 7a). The TBR value marginally improved over the following 2 h with a TBR value of 1.5 achieved at 3 h post administration. While it is likely that 10 has begun to accumulate at the tumour site, the cancerous ROI is not readily distinguishable from the background uorescence (Fig. 7c, see Fig. S12 in ESI † for additional time point images). By the 6 h time point the TBR had improved further to 1.9, though, it was not until 24 h post administration that a TBR above 2 was obtained. By this time, the overall uorescence intensity had dropped approximately 75% fold from its peak (Fig. 7b). The TBR value of 2 was maintained out to 48 h as the emission intensity further decreased (90% of peak), and by 96 h it had fallen below the threshold. This sequence of TBR values comes about due to an initial distribution through normal and cancerous tissues followed by a faster clearance of 10 from normal tissue with retention within cancerous tissue. The sequence of images shown in Fig. 7c illustrates the general challenge facing always-on uorophores, regardless of whether they are substituted with cancer specic targeting agents or not.</p><p>As the process of accumulation and clearance of uorophore from normal and cancerous tissues are both dynamic processes, the success or failure of an always-on probe relies on identifying the time point at which uptake and clearance for the different tissue types are most divergent from each other. This poses signicant challenges for their use in surgical oncological practice with respect to patient-to-patient variances and complex hospital scheduling.</p><p>Analysis of the image timelines for bio-responsive RGD NIR-AZA 2 showed remarkable differences from control 10 (Fig. 8). As 2 is administered in solution at pH 7.2, it is non-uorescent and remained virtually uorescent silent within the vasculature immediately post injection ((Fig. 8b), see Fig. S13 in ESI † for additional time point images). At 60 min post injection, 2 begun to accumulate at the tumour region and the uorescent signal had turned on giving a measured TBR of 1.4, with some background also observed from the adjacent liver (Fig. 8a and c). By 3 h a signicant 4.1-fold increase in tumour uorescence intensity had occurred along with a jump in TBR, surpassing the threshold of 2. The TBR value (2.5) reached a maximum at 6 h and maintained this level until 24 h. Importantly, emission intensity from the tumour reached a maximum within 3 h and there was no reduction in intensity between 3 and 6 h coinciding with when the TBR was at its maximum (Fig. 8b and inset). Encouragingly, even at 9 h only a $20% intensity reduction had occurred which provides a wide time frame in which tumour visualisation could be achieved (Fig. 8a). The ability of 2 to effectively tumour stain shortly aer administration can be judged by the sequence of images shown in Fig. 8c. This we view very positively as not only is the threshold reached quickly, it is maintained for a prolonged time. This ts well with a clinical surgical workow whereby the contrast agent could be administered at the start of surgery with intraoperative tumour visualization possible.</p><p>Similarly, responsive iRGD NIR-AZA 3 also had emission suppressed at the start of imaging with very low uorescence at 20 min and tumour accumulation evident at 60 min with a TBR of 1.4 (Fig. 9c, see Fig. S14 in ESI † for additional time point images). Between 20 min and 3 h there was a 4.7-fold increase in tumour uorescence intensity (Fig. 9b and inset). Yet, in comparison to 2 the extent of background signal was larger, though not brighter than the tumour itself and there was a longer delay until 6 h before the TBR threshold reached 1.7 (Fig. 9a and c). Disappointingly, the TBR never exceeded the threshold value of 2 with the best value of 1.8 recorded 48 h post injection.</p><p>Finally, the pegylated bio-responsive 1 was studied to determine the inuence of the conjugating group on the time taken to reach maximum tumour uorescence and TBR. Again, emission remained off in the beginning with a similar, but considerably time delayed, prole to that of 2. TBRs remained low ($1.2) for the rst 60 min then rose to 1.6 and 1.8 at 6 and 9 h respectively. It took 9 h to reach maximum tumour uorescence intensity with this level remaining relatively unchanged at 24 h. Pleasingly, the TBR threshold of 2 was reached at 24 h and this was maintained for a further 48 h. Comparing the changing TBRs of pegylated 1 and RGD 2 over time illustrates that substituting with the peptide sequence provides a considerably faster tumour accumulation. If clinically adopted, the slower time frame of 1 would most likely require its administration 24 h before surgery, though its long retention within the tumour may provide a prolonged window in which it could be imaged.</p><p>To complete this study, further in vivo tests were performed on the most promising derivative RGD NIR-AZA 2. To gain insight into the initial rate of tumour uorescence turn on, images were acquired every 10 min for 3 h immediately aer introduction of 2. To establish that the RGD substituent was inuencing the rate of tumour uptake, competitive binding or blocking experiments were also carried out. Experimentally this was achieved by rst administering an i.v. tail injection of the RGD peptide 5 (6.8 fold equivalence excess) and following a short time period (5 min) 2 was then administrated. It would be expected that the rst administration of RGD 5 would result in the integrin receptors being bound by the free peptide such that when NIR-AZA 2 was next introduced there would be a reduced uorophore uptake and as such a lower switch-on of uorescence. In order to make direct comparisons, pairs of mice with closely matching sized tumours were selected for each experiment (n ¼ 4 pairs). One animal was rst given RGD 5 then both animals were administered with 2, following which uorescence images of both animals were taken every 10 min for the following 3 h. Averaged results from four experiment showed a 1.75 fold reduction in the total tumour uorescence intensity aer three hours for the mice which were rst treated with the peptide 5 prior to receiving the NIR-AZA conjugate 2 versus the mice which only received 2 (Fig. 11a).</p><p>Additionally, the rate of uorescence turn on in the rst 80 min within the tumour that was not exposed to free RGD peptide was 1.9-fold higher than that exposed to the competing peptide (Fig. 11b). These results conrm that the peptide substituent of the RGD-uorophore 2 is positively inuencing tumour accumulation immediately following administration, which facilitates early time point imaging. Upon completion of imaging, tumours were resected from an animal pair with quantication of their uorescence intensities showing a 3.9 fold suppression of emission intensity from the RGD peptide pre-treated animal (Fig. 11c and S15 †). Encouragingly, dissection and imaging of the tumours showed that the uorescence intensity was highest at the outer boundary of the tumour, which would be most benecial for operative identication of the full extent of tumour margins (Fig. 11c). Finally, analysis of the uorescence turn on prole of 2 alone showed that intensity had reached a near maximum at 80 min (Fig. 11a and b red traces). This indicates that 2 could be administered at the start of a surgical procedure with intraoperative tumour visualization taking place without signicantly impeding the normal surgical workow.</p><!><p>In summary, three bio-responsive NIR-AZA uorophore constructs have been synthesised conjugated to either active (RGD) or passive (PEG) tumour targeting groups and their photochemical, cellular and in vivo properties compared with an always-on uorescent control. Each bio-responsive derivative showed excellent off to on uorescence switching characteristics with large enhancement values. In vitro live MDA-MB 231 cell imaging experiments showed internal acidic organelle cell staining with the responsive probes 2 and 3, contrasting with the always-on derivative 10 which rst showed plasma membrane before internal organelle staining. This result proves that the delity of uorescence switching is maintained in cellular experiments and is independent of the conjugating group.</p><p>A comprehensive in vivo assessment of tumour imaging performance for bio-responsive probes 1, 2, 3 and always-on derivative 10 was conducted with monitoring of the uorescence distributions over 96 h following administration. As anticipated, the always-on 10 gave an immediate, non-specic and very strong emission prole throughout animals whereas the bio-responsive 1, 2 and 3 displayed relatively very low initial uorescence. In the case of 10, clearance from normal tissue with accumulation and retention in tumour, allowed for a TBR above 2 to be reached between 9 and 24 h. All three bioresponsive derivatives switched on within tumours at time points consistent with their conjugated targeting groups. cRGD 2 and iRGD 3 both had effective switch-on in the rst hour though 2 had superior specicity for tumour than 3. Probe 2 achieved the threshold TBR value of above 2 within 3 h and this was maintained for a further 24 h. Relatively, the PEGylated 1 had slower similar turn on characteristics taking 9 h to reach maximum uorescence from the tumour. Despite the slower accumulation, its retention was biased to the tumour tissue with the threshold TBR value being reached at 24 h and maintained out to 96 h. The side-by-side imaging comparison of 1 and 2 is an important and unique illustration of the dynamic differences between passive EPR and active targeting in action.</p><p>Overall, the cRGD-conjugate 2 has been identied as showing excellent potential for clinical translation for intraoperative uorescence guided tumour margin identication. Its bio-responsive nature with early accumulation at the tumour periphery may overcome the inherent drawback of always-on uorophores requiring prolonged clearance times. The PEGylated derivative 1 does also offer potential for clinical translation though its slow switch-on rate may ultimately limit its clinical scope.</p><p>The next ongoing stage of this research is to record continuous NIR-uorescence video of the bio-responsive turn on at tumour margins to gather more kinetic data on the tissue dependent rates of emission increase over the rst 90 min. This real-time data will be utilised in conjunction with specically developed algorithms for dynamic image analysis that could provide the surgical team with an augmented reality (AR) heat map representation of the tissue to be excised during the operation. An intraoperative use of dynamic uorescence tissue imagery combined with AI analysis and a clinical AR interface has the potential to transform surgical practice.</p><!><p>Detailed experimental procedures and characterisations are provided in the ESI. † uorophores (EP2493898 and US8907107) in which he has a nancial interest.</p>

Royal Society of Chemistry (RSC)

Machine Learning 3D-resolved prediction of electrolyte infiltration in battery porous electrodes

Electrolyte infiltration is one of the critical steps of the manufacturing process of lithium ion batteries (LIB). Along with being the most time-consuming step in manufacturing, electrolyte wetting directly impacts the cell energy density, power density and cycle life. We present here an innovative machine learning (ML) model, based on deep neural networks (DNN), to fast and accurately predict fluid flow in three dimensions, as well as wetting degree and time for LIB electrodes. The ML model is trained on a database generated using a 3D-resolved physical model based on the Lattice Boltzmann Method (LBM). We demonstrate the ML model with a NMC electrode mesostructure obtained by X-ray micro-computer tomography. The extracted pore network from tomography data was also used to train our ML neural network. The results show that the ML model is able to predict the electrode filling process, with ultralow computational cost (few seconds) and with high accuracy when compared with the original data generated with the physical model. Also, systematic sensitivity analysis was carried out to unravel the spatial relationship between electrode mesostructure parameters and predicted infiltration process characteristics, such as saturation dynamics, filling time among others.The ML model is able to speed up the infiltration predictions by several orders of magnitude compared to the LBM model which usually requires several days of calculation. This paves the way towards massive computational screening of electrode mesostructures/electrolyte pairs to unravel their impact on the cell wetting and optimize the electrolyte infiltration conditions.

machine_learning_3d-resolved_prediction_of_electrolyte_infiltration_in_battery_porous_electrodes

Introduction<!>Computational procedures<!>Machine learning prediction<!>Comparison of LBM simulation and prediction based on ML<!>Parameters influencing the saturation and their physical interpretation<!>Conclusions

<p>Lithium ion batteries (LIBs) can provide high energy and power densities with long cycle life, 1 constituting the technology of choice nowadays for electronic gadgets and electric vehicles. 2 Therefore, the demand for LIB increases rapidly and its cost becomes one of the critical issues to overcome. Generally, the price depends on the battery's cell chemistry and manufacturing process. 3 And the electrolyte infiltration in the battery cell is one of the bottlenecks in the manufacturing process. 4 It is crucial to optimize the electrolyte infiltration as it takes a relatively long time compared to the other manufacturing steps. 5 Moreover, it can also impact the electrochemical performance of the cell. Indeed, a poor electrolyte impregnation decreases the active surface area (active material/electrolyte interface), and creates an inhomogeneous SEI layer in LIB negative electrodes, which may lead to low energy and power densities, and shorter cycle life. 6,7,8,9,10 Despite its importance, it is experimentally challenging to analyze electrolyte flow through the porous electrodes. Several attempts were made to capture the dynamic path of the infiltrating electrolyte by using 2D in-plane imbibition, transmission neutron and X-ray imaging. 11,12,13 Nevertheless, these studies lack appropriate resolution and detailed information due to the limitations of the techniques. In addition, the experimental results constitute average values, making very challenging the differentiation of the effect of various manufacturing conditions on electrolyte wetting. Moreover, performing high throughput experimental characterizations to unravel parameters interdependencies in the infiltration process is not a trivial task, 13,14 since the experimental techniques reported in the literature are costly and require sophisticated tools. 15 On the other side, a recent increase in computational power enables performing three dimensional (3D) fluid flow computational simulations to quantify the permeability of complex porous materials and electrolyte penetration, which can be carried out in electrode images obtained by micro-computer tomography (CT). One of the most prominent tools to evaluate the permeability of 3D mesostructures is the Lattice-Boltzmann Method (LBM). 7 The strength of this method comes from its mesoscopic nature based on the discrete kinetic theory, which straightforwardly includes biphasic interface dynamics. Thus, the LBM is an accurate numerical method to simulate physical phenomena in realistic electrode mesostructures. Typically, LBM simulations are performed in representative elementary volumes (REVs), where relatively small sub-volumes of the bigger mesostructure are selected, such that the global mesostructural properties are preserved. 16,17 For the first time, we recently reported this approach to simulate electrolyte infiltration into LIB electrode mesostructures in 3D. 7 Results arising from LBM simulations are generally accurate, reliable and allow deep physical interpretation of the infiltration process.</p><p>Nevertheless, performing routine calculations with LBM remains computationally expensive and time-consuming: typically, 48 to 120 hours are needed for simulating electrolyte infiltration in one electrode, running the code in a supercomputer. Still, LBM constitutes a great tool to produce big data for further analysis (100-300 Gigabytes per electrode), something which is not possible with current experimental tools. Still, the bottleneck of the LBM model is the inability to quickly screen a massive amount of electrode architectures and electrolyte types. Consequently, it remains crucial to speed up the simulation of the electrolyte infiltration process to pave the way towards the computational screening of the impact of electrolyte and electrode properties on the electrolyte infiltration dynamics and therefore envisage autonomous algorithms able to optimize the electrolyte infiltration for low required times.</p><p>Meanwhile, Artificial Intelligence (AI) has seen a tremendous rise in the last decade, becoming essential for modern industry and finance, among many other fields. 18 In LIBs, machine learning (ML) techniques have enabled tools that significantly reduce the slow time frames related to trial-and-error approaches or physics-based simulations for faster and more efficient data assessment. [19][20][21][22] Deep Neural Networks (DNNs) are the most popular technique in the AI field due to the good performances they show for modeling complex data structures with many non-linear relationships. [23][24] Particularly, Convolutional Neural Networks (CNNs), a type of DNN, are a perfect example, having outstanding performances in different applications involving many types of data such as images-to-images translations, image classification, or autonomous driving. [25][26][27] Such techniques have also been applied to datasets produced from LBM calculations in the geology domain through images-based prediction to obtain fluid flow properties in porous media. 16,28,29 They generally consist of supervised regression ML models, which reduce computational costs and predict relevant physical properties such as porosity or permeability from tomography (micro-CT) X-ray images. [30][31][32] In the field of energy storage, DNN architectures have also become very popular to accelerate physical-based simulations and reduce trial-and-error efforts to optimize LIBs. 33 . Therefore, our aim in this study is to report, for the first time to our knowledge, a ML model based on a multi layers perceptron model (MLP) that can describe the dynamics of the electrolyte infiltration process in 3D, given a particular mesostructure of a LIB electrode and its associated pore-network, while accounting for different external infiltration pressure conditions. The ML model was trained with the data coming from LBM simulations due to the lack of big data from the experimental side, but keeping in mind that the results of the LBM simulations are based on experimentally measured input parameters. Still, it is essential to notice that one more strength of the reported ML model is that it can be adapted for different sources of data (pure experimental or hybrid between experimental and simulated) as far as this concerns the same type of data. The manuscript is divided into three sections: first we present the different processing steps in obtaining the REVs using micro-CT X-ray tomography data coming from our previous work; 7 then we present the LBM simulation details to generate the data and the adopted MLP architecture; lastly, we present electrolyte infiltration prediction results from our ML model along with detailed sensitivity analysis on the effects of the pore-network properties and LBM simulation conditions on the predicted electrolyte infiltration dynamics. Finally, we discuss why this approach has the potential to pave the way towards fast computational screening of electrode architectures/electrolyte pairs for the accelerated optimization of electrolyte infiltration and LIB manufacturing process as a whole.</p><!><p>Extraction of REVs: From the full tomography dataset of the NMC 94% -CBD 6% electrode, eleven 100•100•75 µm 3 sub-volumes of similar porosities were extracted with a maximal relative error of 5 %. The carbon binder domain (CBD) location in the REVs was resolved using an in house stochastic algorithm. 22 Individual pore identification: An accurate reconstruction of the three-dimensional pore spaces and the subsequent identification of individual pores was done by the PoroDict library within the GeoDict ® software using the watershed algorithm. It is known that the surface roughness of 2D images (or 3D voxel data in our case) generally induces over-segmentation in watershed-based methods and many approaches exist to solve it. 34,35 GeoDict ® handles this by reconnecting overly segmented pore-fragments back into a single pore only when the shared interface between the pore fragments is larger than a chosen value. For consistency, this interface threshold value is kept constant at 10 % for each of the representative sub-volumes that were extracted and analyzed.</p><p>Then, their volume, surface area, and surface area of contact with other pores were calculated based on a six neighbors approach. 36 Compared to other pore-network modelling approaches, [37][38][39] where the pores are approximated as spheres and cylindrical throats, the watershed algorithm identifies individual pores by labeling every voxel in the pore phase. This is especially useful when setting a one-to-one correspondence between the pore-wise labeled volumes and other voxel-based volumetric data coming from LBM simulations.</p><p>LBM simulations: Simulations were carried out using the open-source Palabos library version 1.0. 40 The model simulates the streaming and collision of particles on a grid. All the simulations are in the laminar flow regime. The Navier-Stokes macroscopic kinetic theory was applied to describe fluid in the bulk flow at the mesoscopic level. Further details of the model and its description can be found in our previous LBM publication. 7 All the input parameters such as the density, the fluid contact angle with the solid phase, the viscosity, the surface tension and the sizes of simulation boxes are given in Table 1. After, the outputs from the Palabos library were further treated using NumPy 41 with the PyVista library 42 in order to obtain the individual pore-resolved saturation curves. While no specific rule exists for selecting DNN hyperparameters, 43 this architecture was sufficient to fit the training data correctly and to obtain trustable predictions as shown in the results section.</p><p>The LBM model simulations were performed using an Intel ® Xeon ® E5-4627 Cache @ 3.30 GHz with 264 GB of RAM. Each simulation took approximately between 48 and 120 hours, depending on the input pressure parameter. The NN models' training took around 10 minutes by using 48 processors Intel ® Xeon ® Silver 4116 CPU @ 2.10 GHz with 64 GB of RAM. from LBM simulations with voxels labeled as liquid (orange), void (blue), and solid (white). (D) (right) Individual pore-resolved saturation curves. For clarity, two pores (blue and red) were highlighted, spanning the same spatial regions between the watershed and LBM voxel data. The respective pore-resolved saturation curves are also shown, alongside their relevant features (S , Tf0, and Tf1). (E) Structure of the data used for the training of the NN, columns enumerate each pore, while rows are divided into the inputs (Xi)(i≤6) (blue shaded region) and outputs (Yi)(i≤7) (red shaded region). (F) Architecture of the neural network used for training, nodes in blue (red) represent the input (output) layers, respectively. Nodes in yellow represent the hidden layer nodes Sj,k where k is the layer index, and j is the node index. Nodes in green represent the bias that is applied to each hidden layer.</p><!><p>The evaluation of the electrolyte infiltration dynamics was done by extracting the relevant features of pore resolved saturation curves as shown in Figure 1D. Specifically, we extracted the values of the times at which the pore filling starts (Tf0) and stops (Tf1), as well as the saturation values associated to ten in-between evenly spaced time steps, as the outputs of the MLP. The set of saturation points are defined as S = {S1, S2, S3, S4, S5, S6, S7, S8, S9, S10}.</p><!><p>As mentioned above, our trained MLP can accurately predict the saturation, initial electrolyte entering and fully wetting time at an individual pore in the structure for the test dataset. In order to further compare and contrast our model, an additional REV was used, whose pore-network was not part of neither the training nor testing datasets. After inputting the parameters of the brandnew pore-network in our MLP, the overall saturation curves where reconstructed from the obtained outputs, as shown in Figure 3. The obtained results closely match the saturation curves obtained with LBM, which gives a hint of the ability our ML model to perform well in a variety of electrode mesostructures.</p><p>Five different applied pressures were also used as input parameters to study and predict their effect on electrolyte penetration. Again, Figure 3 shows the overall saturation curves simulated by LBM, and predicted by our MLP for different applied pressures. Generally, all saturation curves for both real and predicted cases show an asymptotic growth rate where the saturation curve increases steeply and slows down while it reaches the convergence point. Also, the wetting time for the electrode increases as applied pressure decreases for both real and predicted cases. Furthermore, the lower the applied pressure, the lower the overall saturation will be and the longer it will take to reach the convergence point. The saturation curves for real (LBM simulated) cases under the applied pressures p8, p4 and p2 tend to rise monotonically and reach complete wetted conditions.</p><p>For the lower pressure values (p1 and p0.5), the electrode wetting degrees are only 60% and 25%, respectively. In addition, the penetration rate, i.e. the rate at which the saturation will reach its convergence point, is slower for p1 and p0.5 compared to higher applied pressures. The predicted (MLP based) saturation curves agree with those from LBM simulated results, especially at high applied pressures where the predictions are extremely precise. In essence, the vital part is that the MLP can closely match the general simulation trends and is also able to precisely predict converging points, where the saturation degrees predicted by our MLP closely match those obtained by LBM simulations for high applied pressure inputs. Our model also allows following the electrolyte wetting process in 3D, since its outputs depend on spatially resolved pore-networks. Figure 4 shows the temporal evolution of the saturation degree of individual pores in the electrode mesostructure. It is known that electrolyte flows through the porous electrode due to the pressure difference between the electrolyte and air phases, known as capillary forces, while local resistance forces drive the electrolytes path within the porous electrode. Usually, the electrolyte is always directed towards larger pores, as shown in our previous the LBM simulations. 7 Figure 4 shows an excellent agreement between the MLP prediction and physical-based LBM model, in the path that the electrolyte takes within the porous electrode. There is a slight deviation at time step 1x10 4 lu, but the difference disappears at the converging point. The corresponding videos of the filling process as predicted by LBM and ML are provided in supplemental information.</p><!><p>Performing a sensitivity analysis of computational models is a clear and straightforward way to assess how the outputs of a model vary as a function of their inputs. In this work the Sobol indexes were extracted in order to evince the individual and total impact that the input parameters of the MLP model have on the seven outputs that we aim to predict, i.e. Tf0, Tf1 and the saturation values S . Additionally, the Sobol indexes corresponding to the initial (Sf0 = {S1, S2}) and final (Sf1 ={S8, S9, S10}) saturation values are averaged out, which allows condensing the results in order to facilitate their physical interpretation (Figure 5). Let the decomposition of the total variance for one specific output (named Y) according to the Sobol decomposition be expressed as</p><p>-/&,'/. . &0- (1) where Vi is the conditional variance for Y over knowing Xi, and Vi,...,n are the rest of conditional variances knowing different input parameters.</p><p>The Sobol indexes are the partial variances normalized by the total variance of the output, they are noted as (Sobi)(i≤n) and they are comprised between 0 and 1 (Eq. 11). They represent the individual effects (1 st order Sobol indexes) of each input parameter: 4 [8] (2)</p><p>In this study, the Sobol index depends on the chosen output (Yj)(j≤7), meaning that the equation above can be simplified as</p><p>where</p><p>is the conditional expectation of Yj giving Xi.</p><p>By successively adding higher order terms to the indexes, we obtain the total effect for each input parameter</p><p>where the different Sobol indexes 123 &'B are related to the combination of various input parameters. Such indexes are calculated according to the Saltelli sampling method. 44,45 The latter global approach is very popular to obtain convergences when the number of samples varies significantly and represents the entire space of input parameters, whereas random sampling approaches may perform meaningless evaluations of sensitive indexes. Therefore, the Saltelli method extends the size of input parameters combinations for the resulting calculations, which are uniformly distributed over the full input parameters space. The variances calculated from Eq. 10 to Eq. 12 can be described as projections along the different ranges of input parameters. In this study, a batch of 14000 samples is used to evaluate the Sobol indexes for each output of the MLP.</p><p>Figure 5 shows the 1 st order Sobol indexes regarding all possible combinations between inputs (geometrical properties of the pores and applied pressures) and outputs (Tf0, Tf1, Sf0 and Sf1) in this study. We can see that the initial pore filling time (Tf0) highly depends on the pore volume with a Sobol index of 0.7 followed by the pore total surface area and its location with Sobol indexes of 0.3. The total wetting time of the pore (Tf1) is also influenced mainly by these three parameters.</p><p>The pore volume has a major effect with a Sobol index of 0.6 and the second biggest factor is the pore surface area with Sobol index 0.4. It is intuitive that the bigger the pore size, the easier it is for the electrolyte to occupy its volume. Also, other geometrical attributes such as pore location and pore surface area play a significant role in the pore filling start time (Tf0). Additionally, it is important to mention that all the input parameters have almost the same effect on the saturation values at the end of the pore filling (Sf1) with Sobol indexes about 0.45. The onset of saturation Sf0 is, on the contrary, strongly influenced by the pore location with Sobol index around 0.7.</p><p>To sum up, the total electrode's wetting degree strongly depends on the pore network properties and applied pressure. The full wetting time (Tf1) is strongly affected by pore size distribution. Thus, to have optimal electrode mesostructures to reach complete wetting at the shortest possible time, pore size distribution and interconnectivity of the pores must be well designed during the manufacturing. This can be a challenging task with traditional manufacturing process, still it is achievable with alternative techniques such as SPS sintering and 3D printing. 46,47 The zoom region (black circle) illustrates the pore volume (grey zone), pore surface area (blue outline), pore contact surface area (red outline) and pore location (dotted turquoise arrow) of a given identified pore inside the REV.</p><!><p>In</p>

ChemRxiv

High Melting Point of Linear, Spiral Polyethylene Nanofibers and Polyethylene Microspheres Obtained Through Confined Polymerization by a PPM‐Supported Ziegler‐Natta Catalyst

In this work, different types of polyethylene (linear, spiral nanofibers and microspheres) were obtained via confined polymerization by a PPM-supported Ziegler-Natta catalyst. Firstly, the Ziegler-Natta catalyst was chemical bonded inside the porous polymer microspheres (PPMs) supports with different pore diameter and supports size through chemical reaction. Then slightly and highly confined polymerization occurred in the PPM-supported Ziegler-Natta catalysts. SEM results illustrated that the slightly confined polymerization was easy to obtain linear and spiral nanofibers, and the nanofibers were observed in polyethylene catalyzed by PPMs-1#/cat and PPMs-2#/cat with low pore diameter (about 23 nm). Furthermore, the highly confined polymerization produced polyethylene microspheres, which obtained through other PPM-supported Ziegler-Natta catalysts with high pore diameter. In addition, high second melting point (T m2 : up to 143.3 °C) is a unique property of the polyethylene obtained by the PPM-supported Ziegler-Natta catalyst after removing the residue through physical treatment. The high T m2 was ascribed to low surface free energy (σ e ), which was owing to the entanglement of polyethylene polymerized in the PPMs supports with interconnected multimodal pore structure.

high_melting_point_of_linear,_spiral_polyethylene_nanofibers_and_polyethylene_microspheres_obtained_

Introduction<!>Structure Data of the PPMs Supports and Ti Contents of the PPM-Supported Catalyst<!>Ethylene Confined Polymerization<!>Micromorphology of Obtained Polyethylene<!>Different T m2 of Initial Obtained Polyethylene<!>Further analysis of polyethylene with high T m2 after physical treatment<!>Conclusion<!>Experimental Section Materials<!>Preparation of the PPM-Supported Ziegler-Natta Catalyst<!>Ethylene Polymerization<!>Physical Treatment of Initial Obtained Polyethylene<!>Characterization

<p>Polyolefin, as one of the most important polymer materials, has been widely used in many industries and human life, resulting in greatly improvement of people's livelihood and life quality. [1] In order to break the barriers between laboratory and industry, and promote the large-scale industrial production of polyolefin, polyolefin catalyst needs to be supported on the carriers. There are several advantages for the supporting catalysts: i) the catalytic activity could be improved effectively; ii) it could meet the requirements of existing industrial facilities well; iii) the morphology and apparent density of polyolefin products could be improved; iv) the use of co-catalysts could be greatly reduced, thereby reducing production costs. [2] With further research on catalyst supports, the field of the catalyst supports has been expanded form original inorganic MgCl 2 to multiple types and scales, such as molecular sieve, carbon nanotube, polystyrene, etc. [3] In recent decades, with the rapid development of nanotechnology, it is found that the structure of polymer materials changed at the nanometer scale, and this phenomenon has drawn wide attention in the academic field. [4] According to the researches, the molecular chain structure, condensed matter structure, phase structure and stability of polymers will change at nanoscale, leading to quite different properties of polymers compared with the bulk state. [4e,5] Therefore, the concept of polymerization in a nano-confined space (defined as confined polymerization) has been proposed by scientists. Here the active center is loaded inside the supports, and then polymerization occurs in the nano space, hoping to obtain the product with different structure and performance compared with corresponding bulk product. After that, the goal, controlling the structure of polymer at different scales, could be achieved. [4c,5f,6] Aida and co-workers [4a] chose mesoporous silicon fiber (MSF) as a carrier to support Cp 2 TiCl 2 catalyst, then ethylene was polymerized in the MSF support. Moreover, the pore diameter of MSF was 2.7 nm with uniform pore structure. During the polymerization process, polyethylene molecular chains did not fold but to grow along the axis parallel to the MSF pores, because the 2.7 nm pore diameter was much smaller than the thickness of polyethylene crystal lamellar. Hence, polyethylene nanofibers with a straight chain structure and high melting point and ultrahigh molecular weight were finally obtained. According to this study, the concept of nanoextruder and extrusion polymerization in nanopores were proposed. After that, a number of attempts to polymerize olefin in the nano-confined space have been made. Choi et al. [7] used anodic aluminium oxide (AAO) film with the diameter of 60 nm as a template to carry Cp*Ti(OCH 3 ) 3 for styrene polymerization. After confined polymerization, polystyrene fibers with high molecular weight (M w = 928.000 g/mol) and high melting point were synthesized. According to the study of Liu et al., [8] the preparation of polyethylene nanofibers requires a durable confined space, and the weakening or disappearance of confined space is not conducive to generate fibers. Although there are many researches on the confined polymerization of inorganic supports, some defects are still exist: i) the active center could be deactivated by the inorganic supports; ii) the supports will remain as inorganic ash in the product; iii) the polymerization by the inorganic supported catalyst can't provide an environment that is similar to homogeneous polymerization. Therefore, developing a kind of organic support for confined polymerization is urgent.</p><p>Roscoe et al. [2a,c] synthesized polystyrene microspheres with nanopores through copolymerization, and then the polystyrene microspheres were used as supports for olefin polymerization. The results demonstrated that polymerization occurred independently in each polystyrene microspheres. Since then, more and more porous polymer supports were obtained for olefin confined polymerization. Uemura et al. [4b,9] studied the effects of porous polymer supports channels on polymerization activity, molecular weight and structure. It was found that chain termination was effectively suppressed when the polymerization was carried out inside the pores. The obtained polystyrene represented a low molecular weight distribution and high isotacticity. In our previous work, three catalysts, Cp 2 ZrCl2, Cp 2 TiCl 2 , and Ziegler-Natta, were supported into porous polymer microspheres (PPMs) (pore diameter: 9.0 nm), which was designed and synthesized all by ourselves, for ethylene confined polymerization. [10] The results indicated that three different confined polymerization (highly confined polymerization, slightly confined polymerization, and non-confined polymerization) occurred in the confined space of PPMs. Here highly confined polymerization was defined as the supports were gradually expanded rather than broken up during polymerization process, and the supports could provide continuous confined space for the whole polymerization. Slightly confined polymerization was defined as the supports were broken up gradually, and the supports could provide confined space part of polymerization process. Non-confined polymerization means that the supports could not provide any confined space for the polymerization process. However, most of the studies, including our previous work, were concentrated on the effect of porous polymer supports with a particular pore diameter on the confined polymerization. Based on the literature, [8] polymerizing in inorganic supports with different nanopores could obtain various products. Hence, changing pore diameter of the porous polymer supports with interconnected pores and better toughness might result in more complicated phenomena compared with that of inorganic supports.</p><p>In this article, two types of PPMs with different pore diameter and supports size were synthesized. After that, Ziegler-Natta catalysts were supported inside PPMs followed by ethylene polymerization. The aims of this context are those: i) Firstly, studying the effect of PPMs with different structures on confined polymerization activity; ii) Secondly, figuring out the variation of the structure and properties of confined polymerization products under the influence of PPMs supports structure changes.</p><!><p>The pore structure data of PPMs supports and corresponding Ti loading amounts are illustrated in Table 1. As shown, the PPMs supports can be divided into two types according to the variation of pore diameter and supports size: i) one is the similar pore diameter (around 22-23 nm) and different supports size (10.30 and 7.68 μm), i. e., PPMs-1# and PPMs-2#; ii) the other is the increasing pore diameter (from 23.3 to 86.3 nm) and similar size of PPMs supports (approximate 7 μm), i. e., PPMs-2#, PPMs-3# and PPMs-4#. The Ti contents of PPM-supported catalysts were measured by ICP-OES, and the loading amounts of Ti are around 6 wt % (Table 1). The SEM images of PPMs before and after Ziegler-Natta catalyst loading are listed in Figure 1. It is found that spherical structure of PPMs supports is maintained after Ziegler-Natta catalyst loading, and the size of PPMs changes little.</p><p>Then, the pore structure spectra of PPM-supported Ziegler-Natta catalysts are illustrated in Figure 2. Obviously, the interconnected tri-modal pore structure is still existed inside PPM-supported Ziegler-Natta catalysts. Figure 2a-d shows that the pore size distribution of the tri-modal pore structure is mainly between 3-10 nm and 20-70 nm. The unique tri-modal pore structure is caused by PPMs supports, which were formed [a] The BET specific surface area (S BET ), average pore diameter (d p ), and specific pore volume (V p ), were measured from BJH adsorption data, and average size of PPMs (S p ) was obtained from SEM images. through the copolymerization of acrylonitrile, polystyrene and 1,4-divinylbenzene.</p><p>FTIR and XPS were used to study the connection between Ziegler-Natta catalysts and PPMs supports. As shown in Figure 3a, reduced intensity of cyano groups and increasing intensity of imine groups indicate that chemical reaction occurred between activating agents (CH 3 MgCl) and cyano groups. It is the chemical reaction that transformed the cyano groups into imine groups. The chemical reaction was also confirmed by the increasing characteristic peak of N 1s binding energy (from 399.0 eV to 399.6 eV) after adding the activating agents (CH 3 MgCl) in Figure 3b. [11] Furthermore, Figure 3c indicates that the binding energies of Ti 2p3/2 and Ti 2p1/2 for the PPM-supported Ziegler-Natta catalyst are 458.5 and 464.3 eV, respectively. The two values of Ti 2p3/2 and Ti 2p1/2 are higher than that of homogeneous TiCl 4 catalyst in the literature, [12] indicating the formation of cationic active species through chemical reaction. Therefore, the results of FTIR and XPS show that Ziegler-Natta catalyst and PPMs supports are connected by chemical bonding. Hence, the preparation mechanism of PPMsupported Ziegler-Natta catalysts could be deduced, as shown in Figure 3d.</p><p>As mentioned above, Ziegler-Natta catalysts are supported into the two types PPMs supports through chemical bonding. In addition, the spherical structure with interconnected multi-modal nanopore is still maintained after Ziegler-Natta catalysts loading in the PPMs, which could provide a confined space for ethylene polymerization.</p><!><p>Then ethylene polymerization was catalyzed by the PPMsupported Ziegler-Natta catalysts with different structure. The mechanism of the confined polymerization and corresponding polymerization results are shown in the following Figure 4 and Table 2, respectively.</p><p>As shown in Table 2, the PPM-supported Ziegler-Natta catalysts with larger support size (PPMs-1#) has higher polymerization activity, when using PPM-supported Ziegler-Natta catalysts with similar pore diameter and different supports size, i. e., PPMs-1# and PPMs-2#. For instance, the activity of PE-6# is 567.8 kg PE/(mol of Ti h MPa), which is higher than that of PE-13# (276.5 kg PE/(mol of Ti h MPa)). This phenomenon is reasonable because the S BET , average size and V p of PPMs-1# and PPMs-2# are different, even though they own similar average pore diameter. The larger S BET , average size and V p of area, the easier contact between ethylene monomer and Ti active center, leading to an increasing polymerization activity.</p><p>It can be seen that the PPMs-3# supported Ziegler-Natta catalyst with the pore diameter of 48.0 nm has the highest polymerization activity among the three kinds of PPMs supports (PPMs-2#, PPMs-3# and PPMs-4#), when the PPMs supports size are similar and the pore diameter increases from 23.3 to 86.3 nm. For example, the activity of PE-17# is 930.4 kg PE/(mol of Ti h MPa), which is higher than PE-15# (865.9 kg PE/(mol of Ti h MPa)) and PE-23# (778.7 kg PE/(mol of Ti h MPa)).This result is attributed to three reasons: i) the polyethylene produced in nanopores will block the pores and hinder subsequent polymerization when the pore diameter is too small, leading to a low polymerization activity; ii) the specific surface area is small if the pore diameter is too large, and then the Ti active centers can't fully contact with ethylene monomer, resulting in the reduction of the polymerization activity; iii) the concentration of co- catalyst Triethyl Aluminium (TEA) was high in the large pore diameter, also increasing the probability of Ti active centers that transferred to the TEA chain, which could also reduce the polymerization activity; iV) What is more, PPMs-3# samples have the highest Ti concentration than others, leading to the highest activity than other PPM-supported catalyst. Therefore, we can deduce that only the proper pore diameter of PPMs supports is conducive to the release of activity, and PPMs-3# supported catalyst is the best among the above three supported catalysts (PPMs-2#, 3# and 4#).</p><!><p>Figure 5 illustrates the micromorphology of polyethylene obtained by confined polymerization of PPM-supported Ziegler-Natta catalysts. A large number of polyethylene microspheres are clearly observed in Figure 5a, and the particle sizes of these microspheres are between 60-100 μm. Moreover, the microspheres structure can be observed in all the products prepared by the four kind PPM-supported Ziegler-Natta catalysts. The surface of these microspheres is magnified to further analyze their structure. As shown in Figure 5a'-a'', the microspheres are composed of multiple nanofibers with the diameter of 100 nm, which are similar to the morphology of polyethylene obtained in our previous study. [10] The same morphology of polyethylene indicates that confined polymerization also can be occurred in the nanopores with the diameter of 23.3 to 86.3 nm.</p><p>It is an interesting phenomenon that is also appeared in SEM images, as illustrated in Figure 5b-d. Individual linear nanofibers (Figure 5b-c) and spiral nanofibers (Figure 5d-e) are observed in the polyethylene catalysed by PPMs-1#/cat and PPMs-2#/cat, in addition to the typical polyethylene microspheres composed of nanofibers (Figure 5a-a''). The Linear and spiral polyethylene nanofibers are not observed in the highly confined polymerization in our previous research. [10] This phenomenon is not surprising after further analysis, and it is also reasonable. On the basis of our previous research, confined polymerization could be divided into slightly confined (defined as the supports were broken up gradually, and the supports could provide confined space part of polymerization process) and highly confined (defined as the supports were gradually expanded rather than broken up during polymerization process, and the supports could provide continuous confined space for the whole polymerization). The former resulted in polyethylene nanofibers, while the latter led to polyethylene microspheres. Different morphology of the obtained polyethylene was caused by different initial polymerization activity of Cp 2 TiCl 2 and Ziegler-Natta catalysts. The initial polymerization activity of Cp 2 TiCl 2 was high, and the generated polyethylene would be broken into nanofibers gradually. While the activity of Ziegler-Natta catalysts was gentle, the obtained polyethylene would be polymerized in PPMs and the spherical structure was maintained during the process. However, the gentle activity of Ziegler-Natta does not mean that all of the PPM-supported catalysts were not broken up during the polymerization process. In this work, the PPMs with the lowest pore diameter (PPMs-1# and PPMs-2#: about 23 nm) exhibit more possibility to be broken up gradually. The reason of this phenomenon is that the low pore diameter makes the PPMs supports suffer from more pressure during polymerization process than the PPMs with high pore diameter. Thus, the PPMs-1#/cat and PPMs-2#/ cat may not afford the expansibility of polymerization, then the polyethylene would be broken and generated nanofibers gradually (Figure 5b-e), which is similar to PPM-supported Cp 2 TiCl 2 catalysts.</p><p>As for the formation of linear and spiral nanofibers, this is attributed to the unique nanopore structure of PPMs support (Figure 2). The structure of our supports is the interconnected multi-modal pore, which is different from traditional inorganic supports (such as SBA-15 and carbon nanotube). Here the pore is formed from the crosslinking of styrene and 1,4-divinylbenzene, leading to the ordered nanopores, as well as some disordered channels interconnected with each other. Therefore, the linear nanofibers will be produced through the ordered nanopores, and the spiral nanofibers are generated from the disordered channels.</p><!><p>There are also significant differences in the melting point of the initial obtained polyethylene (especially the secondary melting point, T m2 ), besides various surface morphology. Most of the primary melting point of initial obtained polyethylene is higher than 140 °C, which is the characteristic of confined polymerization. [4a,7] However, the T m2 of the initial obtained polyethylene have changed after eliminating thermal history. The polyethylene with linear and spiral nanofibers are still higher than 138 °C, which was polymerized by PPMs-1#/cat and PPMs-2#/cat (low pore diameter).The T m2 of polyethylene with microspheres structure are reduced to135 °C, and they were obtained through PPM-supported catalysts with high pore diameter. This phenomenon has not been observed before, and the thermal properties we tested were the polyethylene that removing the residue of catalyst and the matrix through physical method in the previous research. [10] Two kinds of initial obtained polyethylene with different T m2 (PE-15#: 136.9 °C and PE-5#: 140.2 °C) were chosen with the purpose of studying the difference between them. First of all, their structures are analyzed, and the obtained data are demonstrated in Figure 6. It can be seen from Figure 6a that the initial obtained polyethylene with different T m2 are orthorhombic crystal, and the characteristic peaks are located at � 21°and 24°. In Figure 6b, there is only the characteristic peak of methylene at 30 ppm, indicating that the obtained polyethylene are linear polyethylene. That is to say, there is no significant difference in the structure of the polyethylene with different T m2 . (PE-15#: 136.9 °C and PE-5#: 140.2 °C).</p><p>As we all know, melting point (T m ) is affected by the thickness of lamellar (d c ) and surface free energy (σ e ), and the relationship between them could be analyzed by the Gibbs-Thompson equation. As a result, SAXS was used to study the crystallization of polyethylene with different T m2 , and onedimensional SAXS profiles and corresponding data are illustrated in Figure 7 and Table 3, respectively. Here the amorphous layers thicknesses (d a ) and the thickness of lamellar (d c ) of polyethylene were calculated from DSC and SAXS results.</p><p>In equation ( 1) the equilibrium melting point is represented by T 0 m , and the enthalpy of fusion per unit volume is symbolized by DH 0 f . According to our calculation, there is a difference (about 1.2 nm) in the d c of the two kinds of polyethylene. The thicker of d c is, the higher value of T m2 is. Therefore, it can be concluded that the difference in d c of polyethylene results in various T m2 .</p><!><p>According to the research in section 2.4, it is found that polythene obtained through highly confined polymerization exhibits different values of T m2 (from 135 °C to 140 °C), and this result is different from our previous results. [10] The T m2 of the polythene we obtained before was high (up to 143.8 °C), [10] and the high T m2 polythene (143.8 °C) was treated through physical method before DSC test, in order to eliminate the effect of residue (catalyst and the matrix) on thermal performance. The same physical treatment of polythene obtained in this work was chosen to further analyze the difference between the previous and present work, and the corresponding treatment process is illustrated in Experimental section. Here the two samples we chosen show low T m2 (138.3 and 137.5 °C), and polythene prepared by SBA-15-supported Ziegler-Natta catalyst (T m2 :135.8 °C) is also selected for comparison.</p><p>Table 4 and Figure 8 demonstrate the variation of the thermal performance of polyethylene before and after removing residue through physical method. As shown in Table 4, the T m2 of polyethylene prepared by PPM-supported catalyst increases significantly, from 137.5 to 143.3 °C, while that by SBA-15-supported catalyst only increases 1.8 °C (T m2 : 137.6 °C). This changing indicates that the real thermal performance of the as-prepared polyethylene is inhibited by the residue, which existed in the initial obtained polyethylene. The characteristic of highly confined polymerization, high T m2 , will appear after removing the residue, and the high T m2 is consistent with our previous work.</p><p>SAXS test for the polyethylene was also conducted and the result was analyzed. As can be seen from Figure 9 and Table 5, the value of q decreases by about 0.01 after removing residue through physical treatment (from 0.17 to 0.16), after that the values of d c and d a are calculated based on DSC result. In Table 5, the d c of both two polyethylene increase to about 23.4 nm, while there is a little decrease in d a from 15.6 to 15.0 nm, indicating that the increasing T m2 is attributed to the thicker d c after removing residue.</p><p>It is found that the products exhibit the high T m2 (up to 143.3 °C) after removing residue through physical treatment, which are not observed in other researches. [4a,5f,6d] According to the literatures, [13] the constant values of T 0 m and DH 0 f are 145.5</p><p>°C and 280 J/cm 3 , respectively, and theoretical value of σ e is 90 × 10 À 7 J/cm 2 . After that, the value of d c could be calculated based on these values (Equation 1), and the results are illustrated in Table 6. In comparison with the d c obtained by SAXS result through our polymerization, the d c calculated based on equation 1 is much higher, which is a very interesting phenomenon. The huge difference in the two values of d c by different methods indicate that one of the values in equation 1 is not constant, and it is the uncertain value that causes the big deviation of d c . Coming back to equation 1, there is only σ e , which is just a theoretical value. Therefore, the key factor of high T m2 is the reduction of σ e rather than the increasing d c . As described in Section 2.1, the supports (PPMs) exhibit unique structure with interconnected multi-modal nanopores, leading to more entanglement of polyethylene prepared by confined polymerization compared with traditional supports (SBA-15, CNTs), and then generate fewer chain ends. [10] In the process of heating, a large number of entangled chains and fewer chain ends can reduce the value of σ e , resulting in high T m2. [14] Results showed that the high T m2 is the inherent property of polyethylene prepared by PPM-supported Ziegler-Natta catalyst, which is attributed to the unique interconnected multi-modal pore structure of the PPMs supports. [10,15]</p><!><p>In this work, linear, spiral polyethylene nanofibers and polyethylene microspheres were prepared via slightly and highly confined polymerization by PPM-supported Ziegler-Natta catalyst, respectively. DSC data of the initial obtained polyethylene showed that the residue (catalyst and the matrix) had an adverse effect on the thermal properties of polyethylene (especially the T m2 ), and the high T m2 (up to 143.3 °C) emerged after removing the residue through physical treatment. After the physical treatment, the thickness of lamellar (d c ) increased, but this is not the main reason for increasing T m2 (from 137.5 °C to 143.3 °C). Based on our study, the key factor to high T m2 is the low σ e , which is ascribed to the entanglement of polyethylene synthesized in the PPMs supports with interconnected multi-modal pore structure. Furthermore, the high T m2 was the unique feature for the polyethylene confined polymerization by PPM-supported Ziegler-Natta catalyst.</p><!><p>The materials used in this work and syntheses of porous polymer microspheres (PPMs) supports with different pore structure are illustrated in the Supporting Information.</p><!><p>All experiments that sensitive to air and moisture were using the standard Schlenk technique in nitrogen condition.</p><p>First of all, PPMs supports were activated to remove moisture and air in a vacuum oven for 12 h at 60 °C. CH 3 MgCl and a certain weight of PPMs were mixing in 60 mL toluene, then the mixture were stirred at 50 °C for 4 h. After that, the obtained product was filtered, washing the obtained solid with toluene more than three times. Then, adding 60 mL toluene and TiCl 4 in the washed solid, the mixture was stirred at 50 °C for 4 h followed by filtered again. At last, the obtained solid was washed five times with toluene, and then the solid was vacuum dried at room temperature.</p><!><p>Ethylene slurry polymerization by PPM-supported Ziegler-Natta catalysts was polymerized in a 0.1 L autoclave stainless steel reactor. The steel reactor contains mechanical stirring and inlets for adding catalyst and ethylene monomer. First of all, the steel reactor was heated to 80 °C and then cooled down to setting temperature under vacuum. Then TEA was added into the steel reactor under inert gas. After stirring a few minutes the PPM-supported Ziegler-Natta catalyst solution was added into the steel reactor. Then the setting pressure of ethylene monomer was injected to start ethylene confined polymerization. After certain minutes, the polymerization was terminated by acidified ethanol. Then polyethylene product was obtained and then vacuum dried.</p><!><p>The purpose of physical treatment is to eliminate the effect of residue (catalyst and the matrix) on thermal performance. Firstly, initial obtained polyethylene was wrapped by filter paper. Then put them into the inner device with pores in its wall. Toluene and a little antioxidant were added into the assembled device (Figure S1 in the Supporting Information). Subsequently, this device was</p><!><p>The surface areas (S BET ), average pore diameter (d p ), pore volume, and specific pore volume (V p ) of porous polymer microspheres (PPMs) support and corresponding PPM-supported Ziegler-Natta catalysts were measured by a specific surface area physisorption apparatus, and the model was NOVA-1000, USA. The Ti active center contents of the PPM-supported catalysts were obtained through an ICP-OES, and the model was Thermo iCAP 6000, USA. SEM pictures of PPMs supports, corresponding PPM-supported catalysts and polyethylene were observed by a FESEM, and the model was Philips XL30 ESEM, Netherlands. The structure of polyethylene was tested by an XRD, and the model was Brucker D8, Germany, a SAXS (France, Xenocs) and 13 C-NMR (Germany, Bruker DPX-300). Thermal properties of polyethylene before and after physical treatment were measured by a DSC (Switzerland, Mettler Toledo) at 10 °C/min from 30 °C to 190 °C. Molecular weight (M w ) and molecular weight distribution (PDI) of polyethylene were measured by a gel permeation chromatography (GPC, USA, PL-GPC 220).</p>

Chemistry Open

IL-1\xce\xb2 ENHANCES NUCLEOTIDE-INDUCED AND \xce\xb1-SECRETASE-DEPENDENT APP PROCESSING IN RAT PRIMARY CORTICAL NEURONS VIA UPREGULATION OF THE P2Y2 RECEPTOR

The heterologous expression and activation of the human P2Y2 nucleotide receptor (P2Y2R) in human 1321N1 astrocytoma cells stimulates \xce\xb1-secretase-dependent cleavage of the amyloid precursor protein (APP), causing extracellular release of the non-amyloidogenic protein sAPP\xce\xb1. To determine whether a similar response occurs in a neuronal cell, we analyzed whether P2Y2R-mediated production of sAPP\xce\xb1 occurs in rat primary cortical neurons (rPCNs). In rPCNs, P2Y2R mRNA and receptor activity were virtually absent in quiescent cells, whereas overnight treatment with the pro-inflammatory cytokine interleukin-1\xce\xb2 (IL-1\xce\xb2) upregulated both P2Y2R mRNA expression and receptor activity by four-fold. The upregulation of the P2Y2R was abrogated by pre-incubation with Bay 11-7085, an I\xce\xbaB-\xce\xb1 phosphorylation inhibitor, which suggests that P2Y2R mRNA transcript levels are regulated through NF-\xce\xbaB signaling. Furthermore, the P2Y2R agonist UTP enhanced the release of sAPP\xce\xb1 in rPCNs treated with IL-1\xce\xb2 or transfected with P2Y2R cDNA. UTP-induced release of sAPP\xce\xb1 from rPCNs was completely inhibited by pretreatment of the cells with the metalloproteinase inhibitor TAPI-2 or the phosphatidylinositol 3-kinase (PI3K) inhibitor LY294002, and was partially inhibited by the MAPK/ERK kinase (MEK) inhibitor U0126 and the protein kinase C (PKC) inhibitor GF109203. These data suggest that P2Y2R-mediated release of sAPP\xce\xb1 from cortical neurons is directly dependent on ADAM10/17 and PI3K activity, whereas ERK1/2 and PI3K activity may indirectly regulate APP processing. These results demonstrate that elevated levels of pro-inflammatory cytokines associated with neurodegenerative diseases, such as IL-1\xce\xb2, can enhance non-amyloidogenic APP processing through upregulation of the P2Y2R in neurons.

il-1\xce\xb2_enhances_nucleotide-induced_and_\xce\xb1-secretase-dependent_app_processing_in_rat_prim

INTRODUCTION<!>Materials<!>Primary cell culture of cortical neurons<!>Real-Time and Reverse Transcription-PCR analysis of P2Y2R mRNA expression<!>Single cell calcium assay<!>Overexpression of P2Y2 receptors in rPCNs<!>Western blot analysis<!>Statistical analysis<!>Upregulation of functional P2Y2Rs in rat primary cortical neurons (rPCNs) by IL-1\xce\xb2<!>P2Y2R-mediated release of sAPP\xce\xb1 from rPCNs treated with IL-1\xce\xb2<!>P2Y2R-mediated release of sAPP\xce\xb1 is dependent on ADAM10/17<!>P2Y2R-mediated sAPP\xce\xb1 release is partially dependent on PKC<!>P2Y2R-mediated sAPP\xce\xb1 release is dependent on PI3K<!>P2Y2R-mediated release of sAPP\xce\xb1 is partially dependent on PKC<!>DISCUSSION

<p>Inflammation is a central component of several chronic neurological diseases including Alzheimer's disease (Skaper, 2007). Various inflammatory mediators, such as cytokines, chemokines, proteases and protease inhibitors, have been detected in the brains of patients with Alzheimer's disease (Parihar and Hemnani, 2004). Interleukin-1β (IL-1β) and tumor necrosis factor-α (TNF-α) are rapidly upregulated in a variety of tissues in response to pro-inflammatory stimuli and amplify downstream signaling cascades that mediate inflammation (Stoll et al., 2002). In Alzheimer's disease, IL-1β regulates the increased expression and processing of amyloid precursor protein (APP) (Griffin et al. 1994), the phosphorylation of tau (Sheng et al., 2000), the induction of secretory phospholipase in astrocytes (Moses et al., 2006) and the increased production of acetylcholinesterase (AChE) in neurons (Li et al., 2000; Sue and Griffin, 2006). These inflammatory processes result in neuronal stress and injury, which in turn promotes microglial cell activation and IL-1β overexpression in a self-propagating cycle (reviewed by Standridge, 2006). Recent studies, however, indicate that certain aspects of the inflammatory response, specifically constitutive IL-1β production, may have therapeutic potential for alleviating amyloid pathologies (Wyss-Coray et al., 2001; Shaftel et al., 2007).</p><p>Nucleotides co-released with neurotransmitters can activate a widely distributed family of nucleotide receptors in the central nervous system (CNS). P2 nucleotide receptors (i.e., ionotropic P2X receptors and G protein-coupled P2Y receptors) have been implicated in the regulation of numerous functions in the mammalian nervous system, including neurotransmission, glial cell migration, cell proliferation, ion transport, and neuronal cell apoptosis and survival (Burnstock, 2000; Ciccarelli et al., 2001; Weisman et al., 2005). The P2Y2 nucleotide receptor (P2Y2R) is distinguished among other human P2 receptors by its ability to be activated equipotently by ATP and UTP. The P2Y2R is upregulated in response to stress or injury in various cell types, including activated thymocytes, salivary gland epithelial cells, and vascular smooth muscle and endothelial cells (Koshiba et al., 1997; Turner et al., 1997; Seye et al., 1997 & 2002).</p><p>Previously, we demonstrated that the heterologously expressed G protein-coupled P2Y2R mediates the α-secretase-dependent release of sAPPα from human 1321N1 astrocytoma cells that are devoid of endogenous P2Y receptors (Camden et al., 2005). Cleavage of APP by α-secretase results in the shedding of nearly the entire ectodomain of APP to yield a large soluble protein called sAPPα, and essentially precludes the generation of neurotoxic Aβ peptide from the same APP molecule (Kojro and Fahrenholz, 2005). Several zinc metalloproteinases, including a disintegrin and metalloproteinase (ADAM) 9 and 10, tumor necrosis factor-alpha convertase (TACE/ADAM17), and the aspartyl protease β-secretase beta-site APP cleaving enzyme (BACE2) can cleave APP at the α-secretase site (Allinson et al., 2003). While the proteolytic cleavage of APP is constitutive, it can also be increased through activation of protein kinase C (PKC) or other signaling pathways by substances such as G protein-coupled receptor (GPCR) agonists (Nitsch et al., 1997; Kirazov et al., 1997). It has been postulated that the proteolytic products of APP can modulate neuronal excitability, synaptic plasticity, neurite outgrowth, synaptogenesis, and cell survival (Mattson, 1997).</p><p>The aim of this study was to determine whether cytokines such as IL-1β regulate functional expression of the P2Y2R in rat primary cortical neurons and whether P2Y2R activation can modulate APP processing in these cells.</p><!><p>Fetal bovine serum (FBS) was obtained from Hyclone (Logan, UT). Dulbecco's modified Eagle's medium (DMEM), minimum essential medium (MEM), Neurobasal medium, penicillin (100 units/ml), streptomycin (100 units/ml) and B27-AO Neurobasal medium (B27 medium without cortex antioxidants) were obtained from Gibco-BRL (Carlsbad, CA). Rabbit anti-rat extracellular signal-regulated kinase 1 and 2 (ERK1/2), horseradish peroxidase (HRP)-conjugated goat anti-rabbit IgG, and HRP-conjugated goat anti-mouse IgG antibodies were obtained from Santa Cruz Biotechnology (Santa Cruz, CA). Rabbit anti-rat sAPPα antibody was obtained from Signet Laboratories (Dedham, MA). All other antibodies were obtained from Cell Signaling Technology (Beverly, MA). The Dual Color Protein Standards and nitrocellulose membranes (0.45 mm) were obtained from Bio-Rad (Hercules, CA). LumiGLO chemiluminescent substrates were obtained from New England Biolabs (Beverly, MA). The RNeasy Plus Mini Kit was obtained from Qiagen (Chatsworth, CA). The First Strand cDNA Synthesis Kit was obtained from Roche (Indianapolis, IN). Real-time polymerase chain reaction (PCR) was performed on an Applied Biosystems 7500 Real-Time PCR machine with TaqMan Gene Expression assay probes (P2Y2R NM_017255.1 and glyceraldehyde 3-phosphate dehydrogenase (GAPDH), NM_017008.3) and TaqMan Universal PCR Master Mix (2X) (Foster City, CA). The GenCarrier™2 Transfection Reagent Kit was obtained from Epoch Biolabs (Sugar Land, TX). Nucleotides and all other biochemical reagents were obtained from Sigma Chemical Co. (St. Louis, MO).</p><!><p>Experimental procedures for cell cultures of 95% pure rat primary cortical neurons (rPCNs) were carried out as previously described (Kong et al., 2005). Briefly, cerebral cortices from 18-day-old embryos of Sprague-Dawley rats were removed and the meninges were discarded. The brain tissue was mechanically dissociated in HGDMEM comprised of DMEM, 10% (v/v) FBS, 2 mM glutamine, 100 IU/ml penicillin, 100 mg/ml streptomycin, and 7.5 mg/ml fungizone. The tissue clumps were dispersed with a 10 ml pipette and suspended in 6 ml of 0.25% (w/v) trypsin at 37 °C for 10 min. Then, 2 ml of heat-inactivated horse serum were added to neutralize trypsin activity. The cell suspension was centrifuged at 900 g for 5 min and the cell pellet was suspended in HGDMEM. The resulting cell suspension was filtered through a sterilized 75 mm cell strainer (Becton Dickinson, Franklin Lakes, NJ) and the cells were seeded on plastic culture plates precoated with poly-D-lysine (0.1 µg/ml). After 16–18 h and every 3 days thereafter, the medium was replaced with B27-AO Neurobasal medium (2 mM glutamine, 100 IU/ml penicillin, 100 mg/ml streptomycin, 7.5 mg/ml fungizone, 10 ml of B27-AO Neurobasal medium, and Neurobasal medium to 500 ml) and 15 µg/ml 5-fluoro-2'-deoxyuridine was added to retard glial cell proliferation. The neurons were used for experiments after 7 days in culture (DIV7). Unless otherwise stated, DIV7 cells were treated for 24 h at 37 °C in serum-free DMEM with or without IL-1β at the indicated concentration. The next day, cells were either harvested or the medium was replaced with fresh serum-free DMEM containing UTP, PMA, or other compounds, as indicated in the figure legends.</p><!><p>Total RNA was isolated from rPCNs using the RNeasy Plus Mini Kit (Qiagen). cDNA was synthesized from 500 ng purified RNA using the First Strand cDNA Synthesis Kit for RT-PCR (AMV; Roche). Specific oligonucleotide primers were designed to selectively amplify cDNA for P2Y2R and GAPDH, as previously described (Wang et al., 2005). The amplification was performed using 35 cycles of denatuation for P2Y2R and 30 cycles for GAPDH at 95 °C for 30 sec, with annealing at 60 °C for 30 sec and extension at 72 °C for 1 min. The resulting PCR products were resolved on a 1% (w/v) agarose gel containing 10 mg/ml ethidium bromide and photographed under UV illumination. Ten percent of the synthesized cDNA was used as a template in a 50 µl real-time PCR. For TaqMan quantitative real-time PCR analysis, the probes were labeled on the 5' end with 6-carboxy-fluorescein phosphoramidite (FAM) (for P2Y2R) or VIC (for GAPDH), and at the 3' end with minor groove binder (MGB) dye as the quencher. The GAPDH gene was used as a stable endogenous control. The samples were run in quadruplicate for the P2Y2R target and the endogenous GAPDH control. After computing the relative amounts of target gene and endogenous control for each sample, the final amount of target gene in the sample was calculated as a ratio of the amounts of P2Y2R to GAPDH, using Applied Biosystems software. Relative mRNA levels of the control were normalized to 1.</p><!><p>The intracellular free calcium concentration, [Ca2+]i, was quantified in single cells using the Ca2+-sensitive fluorescent dye fura-2, and the InCyt Dual-Wavelength Fluorescence Imaging System (Intracellular Imaging, Cincinnati, OH). Briefly, rat PCNs cultured on poly-D-lysine-coated coverslips were incubated with 2.5 mM fura-2-acetoxymethylester at 37 °C for 30 min in physiological salt solution (PSS) containing (mM) NaCl 138, KCl 5, CaCl2 2, MgCl2 1, HEPES 10, glucose 10, pH 7.4, and washed with PSS. The coverslips with fura-2-loaded cells were positioned on the stage of an inverted epifluorescence microscope (Nikon; model TMD) and stimulated with agonist at 37 °C, as described in the figure legends. The percentage of cells that responded to agonist was determined, as previously described (Kong et al., 2005). Viable cells were identified by responsiveness to carbachol and non-responding cells were eliminated.</p><!><p>Rat primary cortical neurons at DIV7 expressing undetectable levels of endogenous P2Y2R were transiently transfected with 1 µg of human P2Y2 receptor cDNA using the GenCarrier™2 Transfection Reagent Kit (Epoch Biolabs), according to the manufacturer's instructions. The transiently-transfected cells were cultured for an additional 24–48 h in Neurobasal medium plus B27-AO Neurobasal medium and used to detect sAPPα release.</p><!><p>rPCNs were plated on 6-well plates and grown until DIV7. Cells were serum-starved and incubated in 1 ml of serum-free DMEM with or without various compounds for specified time periods, as indicated in the figure legends. At the end of the incubation period, the medium was collected and centrifuged at 12,000 × g for 5 min to remove cellular debris. The supernatant, representing conditioned medium, was stored at −80 °C until further use. To measure the level of secreted APPα, 200 µl of supernatant were diluted 4:1 with 50 µl of 5X Laemmli sample buffer (187.5 mM Tris-HCl, pH 6.8, 6% (w/v) SDS, 1.8% (v/v) β-mercaptoethanol and 0.003% (w/v) bromophenol blue). The cells were washed twice with ice-cold PBS, and lysed in 2X Laemmli sample buffer (200 µl/well).</p><p>Cell lysate and conditioned medium were heated for 5 min at 65 °C, subjected to 7.5% (w/v) SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to nitrocellulose membranes for protein immunoblotting. After overnight blocking at 4 °C with 5% (w/v) fat-free milk in TBS-T (10 mM Tris-HCl, pH 7.4, 120 mM NaCl, and 0.1% (v/v) Tween-20), membranes were incubated with either 1:1000 dilution of anti-ERK1/2, anti-phospho-ERK1/2, or anti-rat actin antibodies for 2 h at room temperature, or 1:1000 dilution of anti-sAPPα antibodies overnight at 4 °C followed by incubation with HRP-conjugated anti-rabbit or anti-mouse IgG antibodies (1:1000 dilution in TBS-T containing 5% (w/v) fat-free milk) for 1 h at room temperature. Protein immunoreactivity was visualized on autoradiographic film using the LumiGlo Chemiluminescence System (New England BioLabs), according to the manufacturer's instructions. The protein bands detected on X-ray film were quantified using a computer-driven scanner and Quantity One software (Bio-Rad, Hercules, CA). The activation levels of kinases or sAPPα release were expressed as a percentage of controls (i.e., total kinase or actin, respectively).</p><!><p>Results are expressed as the means ± S.E.M. of data obtained from at least 3 experiments. Statistical analysis of data was performed using Graph Pad Prism version 5.0. Statistical significance was determined by one-way ANOVA followed by Newman-Keuls multiple comparison tests or 2-way ANOVA followed by Bonferroni post-tests. Differences were considered to be statistically significant when p < 0.05 (*) and not significant when p > 0.05 (n.s.).</p><!><p>Several subtypes of P2 nucleotide receptor mRNA are expressed in rPCNs, including P2Y1, P2Y4, P2Y6, P2X3, P2X5, P2X6 and P2X7 (Kong et al., 2005; Weisman et al., 2005). However, P2Y2R mRNA and receptor activity are barely detectable in these cells (Weisman et al., 2005, Kong et al., 2005). Since previous studies demonstrated that IL-1β causes an increase in P2Y2R mRNA expression in mouse primary striatal astrocytes (Stella et al., 1997), we used real-time PCR analysis to examine whether this cytokine could also increase P2Y2R mRNA expression in rPCNs. After overnight incubation with IL-1β (0–100 ng/ml), P2Y2R mRNA expression was upregulated in a dose-dependent manner (Figures 1A and B). IL-1β did not significantly increase the expression of P2Y4R or P2Y6R mRNA (data not shown). Furthermore, P2Y2R mRNA upregulation was abrogated in a concentration-dependent manner by a 30 min pre-incubation with Bay 11–7082 (Figure 1B), an IκB-α phosphorylation inhibitor (Pierce et al., 1997), which suggests that IκB/NF-κB signaling pathway regulates transcriptional activation of the P2Y2R. Stimulation of IL-1β-treated rPCNs with the P2Y2R agonist UTP caused an increase in [Ca2+]i (Figure 1C) and activation of the mitogen-activated protein kinases (MAPKs) ERK1/2 (Figures 2A and B), indicating that functional expression of the P2Y2R is also upregulated by IL-1β in rPCNs. Co-incubation of rPCNs with TNF-α and IL-1β did not further enhance P2Y2R-mediated calcium mobilization (data not shown) or ERK1/2 phosphorylation as compared to IL-1β alone (Figures 2C and D).</p><!><p>Overnight treatment of rPCNs with IL-1β (1–100 ng/ml) caused a dose-dependent increase in sAPPα release when Il-1β treated neurons were stimulated with UTP (100 µM) (Figure 3). UTP also increased the release of sAPPα from rPCNs transfected with P2Y2R cDNA, as compared to untransfected rPCNs (Figure 4), confirming the hypothesis that UTP enhances sAPPα release from rPCNs through activation of the P2Y2R subtype.</p><!><p>Currently, several zinc metalloproteinases have been identified as potential α-secretases, including TACE/ADAM17 that catalyzes the shedding of the ectodomain of APP and other transmembrane proteins (Black et al., 1997; Moss et al., 1997), and ADAM10 and MDC9 that also catalyze sAPPα production (Lammich et al., 1999; Koike et al., 1999). In IL-1β-treated rPCNs, UTP-induced release of sAPPα was inhibited by the selective metalloproteinase inhibitor TAPI-2 (Figure 5), suggesting that P2Y2R-mediated APP processing in rPCNs is likely mediated by ADAM10/17.</p><!><p>P2Y2R activation causes the phospholipase C-dependent stimulation of PKC, and the activation of several other GPCRs has been reported to induce sAPPα release through PKC-dependent and -independent pathways (Nitsch et al., 1994 & 1995; Checler, 1995; LeBlanc et al., 1998). Therefore, we determined whether P2Y2R-mediated sAPPα release in rPCNs is dependent on activation of PKC. The general PKC inhibitor, GF109203, partially inhibited UTP-induced sAPPα release in IL-1β-treated neurons (Figure 6). Similarly, GF109203 significantly inhibited PMA-induced sAPPα release (Figure 6). These results indicate that P2Y2R-mediated sAPPα release is partially dependent on PKC.</p><!><p>The release of sAPPα induced by insulin requires activation of the PI3K/Akt pathway (Solano et al., 2000). Since P2Y2Rs activate PI3K in HeLa cells, endothelial cells and rat primary astrocytes (Muscella et al., 2003; Kaczmarek et al., 2005; Wang et al., 2005), we determined whether PI3K/Akt regulates P2Y2R-mediated sAPPα release in IL-1β-treated rPCNs. As shown in Figure 7, incubation of IL-1β-treated rPCNs with the selective PI3K inhibitor LY294002 completely inhibited UTP-induced sAPPα release, but had no effect on constitutive release of sAPPα in IL-1β-treated rPCNs. These results suggest that UTP-induced sAPPα release in rPCNs is mediated through the PI3K/Akt signaling pathway.</p><!><p>Previous studies in our laboratory have shown that inhibition of MAPK (i.e., ERK1/2) partially inhibits UTP-stimulated sAPPα release in human 1321N1 astrocytoma cells expressing human P2Y2R cDNA (Camden et al., 2005). Therefore, we investigated the role of the MAPK pathway in UTP-stimulated sAPPα release in IL-1β-treated rPCNs. U0126 (1 µM), an inhibitor of the MAPK/ERK kinase (MEK), partially inhibited UTP-induced sAPPα release, but not constitutive release of sAPPα in IL-1β-treated rPCNs (Figure 8). These results suggest that P2Y2R-mediated sAPPα release in neurons is partially dependent on ERK1/2 activation.</p><!><p>In the present study, we demonstrated that the pro-inflammatory cytokine IL-1β upregulated expression of the P2Y2R via the IκB-α/NF-κB signaling pathway in rat primary cortical neurons. Furthermore, we showed that activation of the P2Y2R in IL-1β-treated neurons enhanced the release of sAPPα, a non-amyloidogenic cleavage product of APP (Li et al., 1997; Luo et al., 2001; Stein et al., 2004), by stimulating α-secretase activity mediated by the metalloproteinases ADAM10/17.</p><p>Previous studies have demonstrated that P2Y2R mRNA expression and function are upregulated in stressed or injured tissue in salivary gland epithelium (Turner et al., 1998, Ahn et al., 2000, Schrader et al., 2005), in intimal lesions of rat aorta (Seye et al., 1997) and in endothelial and smooth muscle cells of collared rabbit carotid arteries (Seye et al., 2002). P2Y2R mRNA expression is also upregulated in smooth muscle cells by IL-1β (Hou et al., 2000), a cytokine that mediates CNS inflammation in a mouse model of Alzheimer's disease (Brugg et al., 1995; Hauss-Wegrzyniak et al., 1998). Furthermore, activation of the P2Y2R expressed in human 132N1 astrocytoma cells induces sAPPα release (Camden et al., 2005). The present study demonstrates for the first time that IL-1β-mediated P2Y2R upregulation and subsequent P2Y2R-mediated sAPPα release occurs in primary neurons. These results are consistent with reports that chronic inflammatory responses can enhance microglial cell-mediated amyloid-beta clearance and reduce plaque burden in murine models of AD (Wyss-Coray et al., 2001; Shaftel et al., 2007).</p><p>APP is a transmembrane glycoprotein that can be detected in most tissues but is abundantly expressed in the brain (Anderson et al., 1989). Post-translational processing of APP occurs by the action of at least three proteases called α-, β- and γ-secretases (Esler and Wolfe, 2001). Cleavage of APP by β- and γ-secretases results in the release of several protein fragments, including amyloid β (Aβ), the main component of senile plaques found in the brains of patients with Alzheimer's disease (Martins et al., 1991). Cleavage of APP by α- and γ-secretases causes the release of sAPPα, thereby preventing secretion of amyloidogenic Aβ (Furukawa et al. 1996; Mattson, 1997; Kojro and Fahrenholz, 2005). Although sAPPα is constitutively secreted, release of this protein is modulated by neural activity and can be increased by agonists of various receptors (Robert et al., 2001; Canet-Aviles et al., 2002; Camden et al., 2005). The studies presented here show that both IL-1β-induced upregulation of P2Y2Rs in primary neurons (Figure 3) and transfection of neurons with P2Y2R cDNA (Figure 4) significantly increase UTP-stimulated sAPPα release, suggesting an important physiological role for the P2Y2R in Alzheimer's disease pathology. IL-1β alone also significantly enhances sAPPα release and reduces sAPPβ and Aβ40/42 release in human neuroblastoma SK-N-SH cells (Tachida et al., 2008), although IL-1β alone did not enhance sAPPα release in rPCNs (Figure 3).</p><p>Responses to extracellular UTP in mammalian tissues can potentially be mediated by 3 P2Y receptor subtypes, P2Y2, P2Y4, and/or P2Y6 (Brunschweiger and Müller, 2006). P2Y4 receptors are endogenously expressed in rat neurons and have an agonist potency order of ITP = ATP = UTP = ATPγS = 2-MeSATP = Ap4A (Bogdanov et al., 1998). Therefore, it is reasonable to suggest that P2Y4Rs could contribute to the enhanced release of sAPPα evoked by UTP. In primary neurons without IL-1β treatment, P2Y4R mRNA is expressed at higher levels than P2Y2R mRNA (data not shown). However, there have been no studies to suggest that the P2Y4R can activate metalloproteinases, whereas P2Y2R activation has been shown to stimulate ADAM10/17 activity (Camden et al., 2005). The small percentage of rPCNs responding to UTP with an increase in [Ca2+]i in the absence of IL-1β stimulation (Figure 1C) may be due to P2Y4R activation. Nonetheless, in the absence of IL-1β pre-treatment, UTP did not significantly increase sAPPα release from rPCNs (Figure 3), suggesting that the P2Y4R does not contribute to APP processing. Moreover, since IL-1β does not increase P2Y4 or P2Y6 receptor mRNA expression in rPCNs (data not shown), it is most likely that the P2Y2R mediates UTP-induced sAPPα release in rPCNs treated with IL-1β.</p><p>There are three major isoforms of amyloid precursor protein (APP770, APP751 and APP695; Kitaguchi et al., 1988) resulting from alternative splicing of APP mRNA. APP695 is the predominant isoform in the brain (Tanaka et al., 1989), and has received the most attention in research on Alzheimer's disease. The ratio of APP770:APP751:APP695 mRNA is 1:10:20 in the cerebral cortex of the brain (Tanaka et al., 1989). Compared with neurons that express high levels of APP695, astrocytes synthesize relatively higher levels of APP770 and APP751, as expected for these Kunitz protease inhibitor (KPI) containing isoforms. In our study, soluble APPα released into the UTP-containing medium of P2Y2R-expressing rPCNs was detected as a doublet of 100–120 kDa (data not shown), as has been described for proteolytic degradation of APP 695 (Johnson et al., 1988). These results support the notion that soluble APPα released from rPCNs is not due to glial cell contamination in the neuronal preparation (LeBlanc et al., 1991 & 1996 & 1997; Lahiri et al., 1994; Parvathy et al., 1998; Solano et al., 2000; Ma et al., 2005).</p><p>The data presented here also suggest that UTP induces APP processing in IL-1β-treated rPCNs through activation of the metalloproteases ADAM10/17 (Figure 5). Another enzyme that is inhibited by TAPI-2, TNF-convertase, is not involved in the regulated release of sAPPα from neuronal cells (Parkin et al., 2002).</p><p>Stimulation of GPCRs has been shown to regulate APP processing by protein kinase C (PKC)-dependent signaling pathways (Nitsch et al., 1992). However, APP processing can also be regulated by increases in [Ca2+]i, and activation of phospholipase A2 (PLA2), protein kinase A and tyrosine kinases (reviewed in Mills and Reiner, 1999). In human 1321N1 astrocytoma cells expressing human P2Y2R cDNA, UTP-induced sAPPα release was independent of P2Y2R-mediated activation of PKC (Camden et al., 2005). Our results with pharmacological inhibitors of PI3K, PKC and ERK1/2 indicated that P2Y2R-mediated sAPPα release from IL-1β-treated neurons requires PI3K activity (Figure 7) and is partially dependent on PKC and ERK1/2 activation (Figure 6 and Figure 8). The MAPK/ERK1/2 signaling pathway has been implicated in the regulation of both PKC-dependent and PKC-independent APP catabolism in PC12 cells, human embryonic kidney cells, and cortical neurons (Mills et al., 1997; Desdouits-Magnen et al., 1998). PI3K has also been reported to regulate α-secretase-dependent APP processing (Solano et al., 2000; Fu et al., 2002), and IL-1β increased the release of sAPPα via ERK1/2-dependent activation of α-secretase cleavage in neuroglioma U251 cells independent of PI3K (Ma et al., 2005). In SH-SY5Y cells, sAPPα release is induced by muscarinic receptor activation and regulated by arachidonic acid generation via PLA2 activation (Cho et al., 2006). In HEK239 cells, a direct role for Gαq/11 protein in the regulation of muscarinic M3 receptor-mediated sAPPα release was shown (Kim and Kim, 2005). Thus, there is significant divergence in agonist-induced signaling pathways that regulate sAPPα release in different cell types.</p><p>The results of this study strongly suggest that IL-1β via activation of the IκB/NF-κB pathway induces functional upregulation of the P2Y2R in primary cortical neurons and subsequent activation of the P2Y2R signaling cascade by UTP results in a significant increase in ADAM10/17-mediated sAPPα release which is dependent on PI3K and partially dependent on PKC and ERK1/2 activities. Together, these lines of evidence suggest that under inflammatory conditions in the brain where IL-1β levels are increased (Rothwell and Luheshi, 2000; Allan and Rothwell, 2001; Caquevel et al., 2004; Colangelo et al., 2003), functional P2Y2R expression in neuronal cells may serve a neuroprotective role by promoting non-amyloidogenic APP processing. Since P2Y2R expression in glial cells has been postulated to regulate chronic inflammatory responses seen in Alzheimer's disease (Weisman et al., 2005), future studies are required to evaluate how glial cell-mediated chronic inflammation can be used to promote neuroprotective sAPPα release from neurons in response to P2Y2R activation during the progression of Alzheimer's disease.</p>

PubMed Author Manuscript

Nanoparticle Shape Determines Dynamics of Targeting Nanoconstructs on Cell Membranes

Nanoparticle carriers are effective drug delivery vehicles. Along with other design parameters including size, composition, and surface charge, particle shape strongly influences cellular uptake. How nanoparticle geometry affects targeted delivery under physiologically relevant conditions, however, is inconclusive. Here, we demonstrate that nanoconstruct core shape influences the dynamics of targeting ligand-receptor interactions on cancer cell membranes. By single-particle tracking of translational and rotational motion, we compared DNA aptamer AS1411 conjugated gold nanostars (AS1411-AuNS) and 50-nm gold spheres (AS1411-50NPs) on cells with and without targeted nucleolin membrane receptors. On nucleolin-expressing cells, AS1411-AuNS exhibited faster velocities under directed diffusion and translated over larger areas during restricted diffusion compared to AS1411-50NPs, despite their similar protein corona profiles. On nucleolin-inhibited cells, AS1411-AuNS showed faster rotation dynamics over smaller translational areas, while AS1411-50NPs did not display significant changes in translation. These differences in translational and rotational motions indicate that nanoparticle shape affects how targeting nanoconstructs bind to cell-membrane receptors.

nanoparticle_shape_determines_dynamics_of_targeting_nanoconstructs_on_cell_membranes

<p>Nanoparticle (NP) shape is an important parameter that influences cellular uptake and intracellular trafficking.1–2 Nanoconstructs—NP cores surrounded by ligand shells—with different particle geometries can promote different interactions with cell membranes based on surface ligand presentation,3 including altered binding kinetics to receptors.3–5 Efforts to prepare nanoconstructs with different shapes have focused on various core materials, including noble metal NPs,6–7 metal organic frameworks,8–9 DNA nanostructures,10 and polymer NPs.11 In cellular environments, the non-specific adsorption of proteins on NP surfaces forms a so-called protein corona that can screen the engineered targeting moieties as well as change the physicochemical properties of NPs.12–13 Although the composition of the protein corona has been shown to depend on NP geometry14–15 and conformation of ligand shell,16 the impact on the subcellular fate of targeting nanoconstructs is unknown.</p><p>In situ monitoring of nanoconstruct-cell interactions using NPs as probes can deconvolute the effects of NP shape and protein corona in physiologically relevant conditions. However, because of the intrinsic low optical contrast compared to cells, ligand-functionalized organic NPs and DNA origamis cannot function as probes without conjugation to dyes or inorganic nanoparticles.17–19 In contrast, gold nanoparticles (AuNPs) can be tracked via optical methods because of their strong scattering properties20 and also have advantages as probes because they are biocompatible, can be synthesized into various shapes,21–22 and can be covalently functionalized with diverse ligands.23–25 Anisotropic AuNPs such as gold nanostars (AuNS) show angle-dependent patterns in differential interference contrast (DIC) microscopy26–28 with potential for 3D orientation tracking,29 which can provide information on ligand-receptor binding and endocytosis at the molecular level.</p><p>DNA aptamer AS1411, the first aptamer to enter clinical oncology trials,30 is a widely used tumor-targeting ligand that binds to nucleolin (NCL), an overexpressed and ubiquitous protein on the surface of various cancer cell types.31 Previous work showed that AS1411-conjugated AuNS (AS1411-AuNS) are internalized after binding to NCL and shuttled to the peri-nuclear region to induce cell apoptosis with in vitro efficacy higher than free AS1411 aptamer.23, 32 Recently, we found that AuNS nanoconstructs functionalized with targeting versus non-targeting ligands exhibit distinct translational and rotational behavior even with similar protein corona profiles;26 hence, the ligand shell properties are preserved on anisotropic NPs and are critical for NP-cell membrane interactions.24, 26, 32–33 The influence of nanoconstruct shape on the presentation of targeting ligands to receptors on cell membranes has received little attention.</p><p>Here we demonstrate that NP shape determines whether engineered nanoconstructs maintain their targeting abilities on cell membranes. By real-time single-particle tracking using DIC-epifluorescence imaging, we compared dynamics of AS1411-conjugated AuNS (AS1411-AuNS) and AuNP spheres (AS1411-50NPs) on NCL-expressing (NCL+) and NCL-inhibited (NCL−) MCF-7 cells. We confirmed AS1411-NCL binding by tracking AS1411-AuNS rotation on NCL+ and NCL− cells; limited angular rotation was observed on NCL+ cells but free, fast rotation on NCL− cells. We then compared translational dynamics of AS1411-AuNS and AS1411-50NPs. Despite similar protein corona profiles, AS1411-AuNS exhibited faster velocities under directed diffusion and translated over larger areas during restricted diffusion compared to AS1411-50NPs. On NCL− cells, AS1411-AuNS showed much shorter translational trajectories, while AS1411-50NPs showed translation similar to that on NCL+ cells, which demonstrates that the two nanoconstructs have different targeting specificities for NCL. Our results suggest that NP core shape should be considered an important design parameter to maintain the targeting properties of nanoconstructs, especially for drug-delivery carriers.</p><p>Scheme 1 depicts representative single-particle dynamics of targeting constructs on cancer cell membranes. We prepared AS1411-AuNS and AS1411-50NPs with the same surface ligand density (Table S1) to investigate effects of particle geometry on AS1411-NCL interactions. We compared the single-particle dynamics of the two nanoconstructs on both NCL+ and NCL− MCF-7 cells to confirm the specificity of AS1411-nanoconstructs towards NCL. NCL− cells were prepared by pre-incubating MCF-7 cells with a NCL inhibitor to determine how blocking the NCL receptor would change nanoconstruct motion (SI, Methods). 30-s video streams were acquired to track individual AS1411-AuNS and AS1411-50NPs motion on cell membranes with either expressing or blocked NCL.</p><p>Figure 1 illustrates how the single-particle dynamics of AS1411-AuNS and AS1411-50NPs on cancer cell membranes were characterized via multi-channel and multi-wavelength imaging. Figures 1a–b depict a series of DIC images of AS1411-nanoconstructs on a glass substrate under rotation over 180° in increments of 45°; AS1411-AuNS showed angle-dependent DIC image patterns because of their anisotropic structure. AS1411-50NPs did not show significant changes at different angles because of symmetry, and moreover, were difficult to distinguish from spherical cellular vesicles by DIC microscopy alone (Figure S1).</p><p>We carried out single-channel DIC imaging of AS1411-AuNS at 700 nm, near the localized surface plasmon resonance of the AuNS, which showed high signal-to-noise ratios for both cellular features and nanoconstruct probes (Figure 1c). We conjugated Cy5-labeled AS141134 to the 50-nm spheres and tracked particle motions with multi-channel DIC-epifluorescence imaging (Figure 1d). NP motion was tracked in the epifluorescence channel (λ = 689 nm) by the Cy5 signal; the DIC channel (λ = 543 nm)20 was used to locate the particles with respect to the cell membrane. We identified the image plane of the cell membrane by optical sectioning (Figure S2, S3).</p><p>We first evaluated rotation of the nanoconstructs because of their relevance to ligand/receptor binding and clustering.35 Figure 2 shows that AS1411-AuNS on NCL+ cells and on NCL− cells have distinct rotational behaviors, indicating that targeting specificity of AS1411-AuNS to NCL was maintained in vitro. AS1411-AuNS exhibited a constant DIC pattern over several seconds on NCL+ cells (Figure 2a and S4). In contrast, AS1411-AuNS on NCL− cells fluctuated rapidly over a similar time period (Figure 2b), indicating a higher degree of rotational freedom for AS1411-AuNS when NCL was unavailable for binding. We calculated the contrast for AuNS in each frame, which we define as the average intensity of the circular AuNS pattern normalized against the background after subtracting the background intensity, (I − Ibkg)/Ibkg (SI Methods).27–28 On NCL+ cells, the contrast remained mostly constant, either positive (bright) or negative (dark); on NCL− cells, fast changes from negative to positive were observed (Figure 2c–d).</p><p>To distinguish between large and small angular changes during construct rotation, we defined the maximum contrast duration instead of the previous rotational frequency analysis method.26 We quantified the longest time when the nanoconstruct contrast remained the same sign over the 30-s stream; on average, the maximum duration for AS1411-AuNS was ~1.8 times longer on NCL+ cells compared to NCL− cells (Figure 2e). We also confirmed that rotational dynamic variations between AuNS/NCL+ and AuNS/NCL− were not from cell behavior changes after NCL blocking by comparing rotations of non-targeting control aptamer-conjugated AuNS on NCL+ and NCL− cells; no statistical difference was observed between the two cases (p > 0.05) (Figure S5). This result supports that the faster rotation of AS1411-AuNS on NCL− cells may be attributed to the unavailability of NCL for binding. A viability test with increased inhibitor concentration also established that the concentration used to block NCL did not induce cytotoxicity (Figure S6).</p><p>To determine whether the protein corona affected single-particle motion, we characterized the composition for AS1411-AuNS and AS1411-50NPs (SI Methods).26 The protein corona profiles for both nanoconstructs were almost identical (Figure S7, Table S2), with serum albumin being the most abundant protein (23% for AS1411-AuNS, 18% for AS1411-50NPs). Although alpha-2-HS-glycoprotein, known to target scavenger receptor-A,36 was found in the protein corona of both nanoconstructs (7% for AS1411-AuNS, 6% for AS1411-50NPs), this receptor is not expressed on MCF-7 cell membranes.37 Therefore, the protein corona does not influence the nanoconstruct targeting abilities to NCL or corresponding single-particle dynamics of AS1411-nanoconstructs during live-cell imaging.</p><p>Since translation of nanoconstructs after NP-receptor binding can provide insight on the lateral diffusion of surface receptors,38–40 we carried out single-particle tracking of AS1411-AuNS and AS1411-50NPs to probe shape-dependent translational dynamics. We processed the tracked particles by mean square displacement (MSD) analysis categorized into four diffusion modes: directed diffusion (DD), simple diffusion (SD), restricted diffusion (RD), and stationary (ST).26 For each nanoconstruct, the diffusion mode with the best MSD fit was assigned (Figure 3a and S8). Generally, AS1411-AuNS showed a drastic decrease in trajectory length on NCL− cells for all three types of motion: DD, SD, RD, while the trajectory lengths of AS1411-50NPs did not change significantly (Figure 3b).</p><p>Furthermore, we quantitively compared the translational motion of the two differently shaped nanoconstructs. The proportion of AS1411-AuNS under DD decreased from 51.8% to 15.8% after NCL was blocked, while the percentage of AS1411-AuNS under RD increased from 35.7% to 60.5% (Figures 4a–b). In addition, AS1411-AuNS that were ST were only observed on NCL− cells. AS1411-50NPs showed a lower percentage decrease in DD after NCL blocking (37.8% to 20.6%) compared to AS1411-AuNS; however, RD accounted for the highest population on both NCL+ and NCL− cells (37.8% and 47.1%, respectively). Different from AS1411-AuNS, ST particles were observed for AS1411-50NPs on both NCL+ and NCL− cells (Figures 4c–d). The change in diffusion mode distribution after NCL inhibition confirms that AS1411-AuNS binds to NCL, consistent with rotational dynamics results (Figure 2). We attribute the high percentage of AS1411-AuNS going through DD on NCL+ cells to NCL clustering being highly dependent on the presence of intact actin filaments41 in the cytoskeleton, which are responsible for directed movement on the cell membrane.</p><p>We further compared the diffusion parameters, including velocity and confinement length. For nanoconstructs under DD, AS1411-AuNS translated at ~1.7 times faster velocities compared to AS1411-50NPs (Figure 4e). For nanoconstructs with RD, AS1411-AuNS showed ~2 times higher confinement lengths compared to AS1411-50NPs on NCL+ cell membranes (Figure 4f). These differences in translational dynamics indicate that AS1411-AuNS and AS1411-50NPs bind to receptors with different diffusion dynamics. When NCL was blocked, the confinement lengths of AS1411-AuNS under RD decreased ~0.4 times, while spheres increased by ~1.2 times. The significant decrease in AS1411-AuNS confinement lengths on NCL− cells provides additional support that AS1411-AuNS bind specifically to NCL. For AS1411-50NPs, however, NCL inhibition did not significantly change the confinement lengths, which suggests that NCL was not the dominant receptor interacting with AS1411-50NPs. Therefore, the two nanoconstructs have different targeting selectivity towards NCL because of differences in their core shapes.</p><p>In summary, we demonstrated that nanoconstruct core shape governs single-particle dynamics during in vitro targeted ligand-receptor interactions on cancer cells. The differences in translation of AS1411-AuNS and AS1411-50NPs suggest that the two constructs show different targeting specificity towards NCL. We hypothesize that the branched structure of the AuNS core provides multiple areas of contact and ligand presentation and enables multivalent binding between AS1411 to the NCL receptor, while AS1411-50NPs only provides a single limited area of contact. Our study indicates that NP shape is a critical factor in determining ligand-receptor interactions that will affect downstream effects during targeted delivery.</p>

PubMed Author Manuscript

Nitroxyl accelerates the oxidation of oxyhemoglobin by nitrite

Angeli\xe2\x80\x99s salt (Na2N2O3) decomposes into nitroxyl (HNO) and nitrite (NO2\xe2\x88\x92), compounds of physiological and therapeutic interest for their impact on biological signaling both through nitric oxide and nitric oxide independent pathways. Both nitrite and HNO oxidize oxygenated hemoglobin to methemoglobin. Earlier work has shown that HNO catalyzes the reduction of nitrite by deoxygenated hemoglobin. In this work, we have shown that HNO accelerates the oxidation of oxygenated hemoglobin by NO2\xe2\x88\x92. We have demonstrated this HNO mediated acceleration of the nitrite/oxygenated hemoglobin reaction with oxygenated hemoglobin being in excess to HNO and nitrite (as would be found under physiological conditions) by monitoring the formation of methemoglobin in the presence of Angeli\xe2\x80\x99s salt with and without added NO2\xe2\x88\x92. In addition, this acceleration has been demonstrated using the HNO donor 4- nitrosotetrahydro-2H-pyran-4-yl pivalate, a water-soluble acyloxy nitroso compound that does not release NO2\xe2\x88\x92 but generates HNO in the presence of esterase. This HNO donor was used both with and without NO2\xe2\x88\x92 and acceleration of the NO2\xe2\x88\x92 induced formation of methemoglobin was observed. We found that the acceleration was not substantially affected by catalase, superoxide dismutase, c-PTIO, or IHP, suggesting that it is not due to formation of extramolecular peroxide, NO2 or H2O2, or to modulation of allosteric properties. In addition, we found that the acceleration is not likely to be related to HNO binding to free reduced hemoglobin, as we found HNO binding to reduced hemoglobin to be much weaker than has previously been proposed. We suggest that the mechanism of the acceleration involves local propagation of autocatalysis in the nitrite-oxygenated Hb reaction. This acceleration of the nitrite oxyhemoglobin reaction could affect studies aimed at understanding physiological roles of HNO and perhaps nitrite and use of these agents in therapeutics such as hemolytic anemias, heart failure, and ischemia reperfusion injury.

nitroxyl_accelerates_the_oxidation_of_oxyhemoglobin_by_nitrite

Introduction<!>Reagents<!>Spectroscopy<!>Nitrite Analysis<!>Experimental Reactions<!>Results<!>Discussion<!>

<p>Nitroxyl (HNO)1 is one-electron reduced from nitric oxide (NO) and is associated with several biochemical processes that, though sometimes resembling processes involving NO, possess distinct mechanisms and pathways [1; 2; 3; 4; 5; 6; 7; 8; 9; 10; 11; 12; 13]. HNO has been considered as the basis of a therapeutic strategy in cardiac systems associated with heart failure [14; 15; 16; 17; 18; 19; 20; 21], in mitochondrial regulation [13], and in other pharmacological and biological signaling contexts both in vitro and in vivo [22; 23; 24; 25; 26; 27; 28]. Thus, Angeli's salt (Na2N2O3, sodium α-oxyhyponitrite, AS) or other sources of nitroxyl (HNO) can potentially function as therapeutics for numerous conditions. The breakdown of Angeli's salt into HNO and nitrite has been studied for over a century [29; 30; 31; 32; 33; 34; 35].</p><p>Nitrite reacts with oxygenated hemoglobin (oxyHb) to form methemoglobin (metHb) [36; 37], and the reaction becomes very efficient at high nitrite concentrations, where the reaction becomes autocatalytic [38; 39; 40; 41; 42; 43; 44]. Interest in nitrite has been increasing lately due to its emerging role as a vasodilator and source of bioavailable nitric oxide [45; 46; 47; 48]. Nitrite has been shown to be a signaling molecule [49] with cytoprotective applications against ischemia-reperfusion injury [50; 51].</p><p>Angeli's salt decomposes under physiological conditions with a first order rate constant of k = 6×10−4 s−1 to yield HNO and NO2− [32]. Early work [52] confirmed the stoichiometry of the reaction of HNO with oxyhemoglobin to form metHb and nitrate as (Eq. 1)HNO+2[HbO2]2+→2[Hb]3++NO3−+HO2− through a proposed two-step mechanism of oxyHb oxidation by HNO to form NO, followed by NO oxidation of a second oxyHb to form nitrate (Eq. 2–3) (Eq. 2)HNO+[HbO2]2+→[Hb]3++NO+HO2− (Eq. 3)NO+[HbO2]2+→[Hb]3++NO3− </p><p>Peroxide formed in Eq. 2 would be expected to rapidly form hydrogen peroxide at neutral pH. It should be noted that there has been debate regarding the pathway described in (Eq. 2), and alternative pathways have been proposed [53]. The rate constant for (Eq. 2) is likely to be on the order of 107 M−1 s−1, the rate reported for the same reaction with oxymyoglobin [12]. The rate of (Eq. 3) is somewhat faster (k = 5–8 × 107 M−1 s−1) [54; 55; 56].</p><p>The overall reaction of nitrite with oxyHb can be described by (Eq. 4). (Eq. 4)4[HbO2]2++4NO2−+4H+→4[Hb]3++4NO3−+O2+2H2O This reaction was first studied in 1868 by Arthur Gamgee [36] and has since been studied by many others [40; 44; 55; 57; 58; 59; 60; 61; 62; 63]. The reaction progresses slowly at low nitrite concentrations (k = 0.21–0.33 M−1 s−1), but becomes autocatalytic at nitrite concentrations that are high relative to the oxyHb concentration.</p><p>Recent kinetic models by Keszler and coworkers [60] support a mechanism by which autocatalysis is initiated in a multi-step process. This scheme postulates the addition of nitrite to heme-bound oxygen to form a ferrous-peroxynitrate intermediate (Eq. 5), which oxidizes nitrite to form nitrate and a ferrous-peroxynitrite intermediate (Eq. 6). This proposed ferrous-peroxynitrite complex would then be reduced to nitric oxide and peroxide (Eq. 7), and the resulting NO would rapidly oxidize a second oxyHb (Eq. 3). (Eq. 5)[HbO2]2++NO2−→[HbO2NO2]+ (Eq. 6)[HbO2NO2]++NO2−→[HbOONO]++NO3− (Eq. 7)[HbOONO]++2H+→[Hb]3++NO+H2O2 Allowing the reaction to progress slowly according to this scheme, peroxide would be expected to decompose to water and molecular oxygen, yielding a net reaction equivalent to (Eq. 4). However, at sufficiently high concentrations of metHb and peroxide, these components react to form a ferryl hemoglobin radical (Eq. 8), and it is this radical that is believed to initiate propagation of autocatalysis [44; 62]. Under the current model, the ferryl hemoglobin radical reacts with nitrite to form NO2 radical and ferryl hemoglobin (Eq. 9), which reacts with an additional nitrite to form metHb and another NO2 radical (Eq. 10). NO2 radicals formed in this process can react with oxyHb to form a peroxynitrate adduct (Eq. 11), which decomposes to the ferryl hemoglobin radical and nitrate (Eq. 12). Thus, the autocatalytic mechanism is propagated by the NO2 radical and the cycling of ferryl and ferryl radical hemoglobins. (Eq. 8)[Hb]3++H2O2→[•Hb=O]2++H2O+H+ (Eq. 9)[•Hb=O]2++NO2−→[Hb=O]2++•NO2 (Eq. 10)[Hb=O]2++NO2−→[Hb]3++•NO2 (Eq. 11)[HbO2]2++•NO2→[HbOONO2]2+ (Eq. 12)[HbOONO2]2++NO2−→[•Hb=O]2++NO3−+H+ At low concentrations of nitrite, such as those used in our experiments, autocatalysis is not expected to occur, and the oxidation of hemoglobin by nitrite is predicted to progress according to (Eqs. 5–7) and (Eq. 3).</p><p>In addition to oxidizing oxyHb, nitrite has also been shown to react with deoxyHb to form metHb and iron-nitrosyl hemoglobin [42; 64; 65]. Doyle and coworkers reported observing catalysis of the nitrite reaction with deoxyhemoglobin by HNO, but no such catalysis was reported for the reaction of nitrite with oxyHb [52]. In this paper, we show that HNO also accelerates the reaction of nitrite with oxyHb. We demonstrate this phenomenon using AS and using the newly developed HNO donor 4-nitrosotetrahydro-2H-pyran-4-yl pivalate [66] belonging to the recently described family of acyloxy nitroso compounds that yield HNO upon hydrolysis [67; 68]. Importantly, these experiments have been performed with oxyHb in excess to both nitrite and HNO, as would be the case under physiological conditions.</p><!><p>Angeli's salt, DEA NONOate, and carboxy-PTIO (c-PTIO) were purchased from Caymen Chemical. Superoxide dismutase, diethylene triamine pentaacetic acid (DTPA), and phytic acid were purchased from Sigma Aldrich. Other chemicals and supplies were purchased through Fisher Scientific. Packed red blood cells used in the preparation of hemoglobin solution were purchased from Interstate Blood Bank (Memphis, TN, USA).</p><p>Hb was purified as described previously [69; 70]. Red blood cells were washed in pH 7.4 PBS and lysed by dilution with distilled deionized water. The membranes were spun out by centrifugation at 17,000 g. and the Hb was dialyzed against distilled deionized water and PBS. The Hb was pelleted in liquid nitrogen and stored at −80°C for future use.</p><p>Angeli's salt and DEA NONOate stock solutions were prepared in 10 mM NaOH. The concentration of each stock solution was confirmed by absorbance at 250 nm, using an extinction coefficient (ε) of 8 mM−1 cm−1 for Angeli's salt and 9 mM−1 cm−1 for DEA NONOate [71]. Stock solutions of 10 mM NaNO2 were also prepared in 0.01 M NaOH, with the concentration of NaNO2 being determined by mass. Stock solutions of 10 mM c-PTIO were prepared in phosphate buffered saline, with the concentration of c-PTIO being determined by mass. Stock solutions of 10 mM inositol hexaphosphate (IHP) were prepared from phytic acid by titration with sodium hydroxide to pH 7.3.</p><p>4-nitrosotetrahydro-2H-pyran-4-yl pivalate was synthesized as described previously [66]. Esterase from porcine liver (pig liver esterase, PLE) was purchased from Sigma-Aldrich.</p><!><p>Angeli's salt, DEA NONOate, and initial oxyHb concentrations were verified on a Cary 50 bio-spectrometer (Varian, Inc.). Reactions involving Hb were monitored using time-resolved spectroscopy on a Cary 100 bio-spectrometer (Varian, Inc.) with a temperature controller set to maintain sample temperatures at 37° C and a six-cell sample changer that facilitated scanning of up to six samples simultaneously under the same conditions.</p><p>Hemoglobin reactions were analyzed by spectral deconvolution using a least-squares fit to known basis spectra (Figure 1A). A sample spectrum and the corresponding fit are shown in Figure 1B. Of the species included in the basis spectra, oxyHb and metHb always accounted for a sum total amount > 97% of the Hb species present. Data reported is for metHb levels, and remaining Hb is almost exclusively oxyHb, with occasional trace amounts (< 2.5%) of deoxyHb found as the reactions progressed.</p><!><p>Nitrite levels were assessed using a Sievers Nitric Oxide Analyzer (NOA) (GE Instruments) according to standard procedures for a NaI assay provided by the manufacturer. Nitrite levels detected in the nitric oxide analyzer were quantified through comparison with stock NaNO2 samples.</p><!><p>Reactions were conducted using 1 mM oxyHb at 37°C. To one sample, 50 µM Angeli's salt was added. To another sample, 50 µM NaNO2 was added. To a third sample, both 50 µM AS and 50 µM NaNO2 were added. A fourth sample was used as a blank to account for any autoxidation and consisted of oxyHb with an amount of NaOH equivalent to that used for samples containing AS or NO2−. The procedure was repeated both with and without the metal chelator DTPA, and DTPA had no observed effect on the reaction (data not shown). Additional time-resolved absorption spectroscopy was used to investigate reactions with higher nitrite concentrations: 100 µM NaNO2, 50 µM NaNO2, and a blank. Time-resolved spectra were collected, with scans taken every 10 min for two hours and then every hour thereafter for up to twelve hours total. Acceleration of the nitrite reaction with oxyHb was determined by comparing the metHb yield as a function of time in the above reactions.</p><p>In addition to spectroscopic analysis, reactions were conducted with aliquots being removed at 10 min intervals and analyzed in the NOA for nitrite content. A zero time point concentration could not be established for samples containing AS due to extreme signal broadening. The signal was sufficiently resolved to allow determination of NO2− concentrations after 10 min. However, samples at 20 min and beyond showed NO2− concentrations that were so low that they did not have a statistically significant difference given the margin of error in the experiment. Thus, only NO2− concentrations at 10 min are reported. The NOA would be expected to detect both free NO2− as well as NO2− and HNO from Angeli's salt that had not yet decomposed; at the reported time, 95% of AS is expected to have decomposed, and any remaining AS detected by the NOA can be accounted for using an AS control without added nitrite.</p><p>To examine the reaction of HNO and oxyHb in the absence of nitrite, and to observe the change in such a reaction upon addition of nitrite, Angeli's salt was replaced with 4-nitrosotetrahydro-2H-pyran-4-yl pivalate. This HNO donor is a water-soluble acyloxy nitroso compound that does not release NO2− but does generate HNO in the presence of esterase. Optimum HNO formation from this donor occurs in the presence of 9 units of pig liver esterase per µmol of donor. The donor (0.1 µmol) was added to 2 mL of 1 mM oxyHb with 0.9 units of PLE to create a system with 50 µM of HNO donor. Time-resolved spectra were collected as before. This HNO donor is roughly 15 fold slower than Angeli's salt, having a reported half-life of 39 min. However, it provides nitrite-free generation of HNO at our experimental pH; thus, experiments using this compound provide qualitative rather than quantitative comparison of the effects of HNO on the reaction of nitrite with oxyHb. Controls containing either PLE or 4-nitrosotetrahydro-2H-pyran-4-yl pivalate alone were run simultaneously with the experimental samples and an oxyHb blank; no significant differences were observed between the blank and the samples containing either PLE or 4-nitrosotetrahydro-2H-pyran-4-yl pivalate.</p><p>To isolate the effects of HNO from the effects of the NO created in the reaction of HNO with oxyHb, the reactions described above were also repeated with DEA NONOate in place of Angeli's salt. Time-resolved spectra were collected as above.</p><p>To explore the possibility of a peroxide initiated catalytic reaction, the procedure was repeated in the presence of 50 µM catalase. To investigate the possibility of a NO2 radical propagated reaction, the procedure was repeated in the presence of 250 µM c-PTIO. The procedure was also repeated in the presence of 10 Ku/mL superoxide dismutase to investigate the possibility of a superoxide mediated reaction. To examine the potential role of allostery in the reaction, inositol hexaphosphate (IHP, in the form of phytic acid) was used to stabilize the T-state of hemoglobin and the procedure was repeated. Time-resolved spectra were collected as above.</p><!><p>The three primary reactions examined were (1) oxyHb + HNO donor, (2) oxyHb + NO2−, and (3) oxyHb + HNO donor + NO2−. In each case, the metHb yield was determined as a function of time by spectral deconvolution and least squares fitting to the basis spectra shown in Figure 1A, with a sample spectrum and fit shown in Figure 1B. We refer to the yield from each of these reactions as "[metHb] (×)," where×= 1, 2, or 3. Figure 2A shows time resolved spectra of the conversion of oxyHb to metHb by HNO and NO2−; spectra such as these were fit to basis spectra to determine changes in the metHb concentration over time. Figure 2B shows the average metHb yield measured using time-resolved absorption from the three primary reaction mixtures; AS was used as the HNO donor, and a blank was used to account for autoxidation.</p><p>If the reaction of oxyHb with NO2− were independent of HNO, then one would expect (Eq. 13)[metHb](1)+[metHb](2)=[metHb](3) assuming a sufficient excess of oxyHb. Equivalently, examining the activity of NO2− in the presence or absence of HNO, independent reactions would yield (Eq. 14)[metHb](3)−[metHb](1)=[metHb](2). Figure 2C shows a significant difference between [metHb] (3) – [metHb] (1) compared to [metHb] (2), indicating that the two reactions are not independent. OxyHb + NO2− forms metHb at a higher rate in the presence of HNO, even after the metHb generated by HNO (and NO2− present in AS) has been accounted for.</p><p>NO2− itself is consumed more rapidly by the reaction with oxyHb in the presence of HNO. Figure 2D shows NO2− levels measured using the NOA in the three primary reactions after 10 min, as well as the difference between oxyHb + AS + NO2− and oxyHb + AS, hereafter referred to as the concentration of NO2− in the presence of HNO. The concentration of NO2− in the presence of HNO (27.3 ± 4.7 µM NO2−) is only 54% of the concentration of NO2− in the absence of HNO (50.4 ± 6.2 µM). If the consumption of NO2− were independent of HNO, these two values would be expected to be the same. Moreover, the difference in the concentrations of NO2− in the presence and absence of HNO (23.1 ± 10.9 µM) is similar to the difference in metHb concentrations after 10 min between the corresponding samples (18.0 ± 8.0 µM). The difference in NO2− concentrations in the presence or absence of HNO can account for the difference in metHb concentrations under the same conditions if the NO2− reacted with oxyHb to make metHb on a one-to-one molar basis.</p><p>Angeli's salt decomposes into HNO + NO2−, so a sample initially containing 50 µM AS and 50 µM NO2− could contain up to 100 µM NO2− following AS decomposition. Thus, it is reasonable to compare samples of oxyHb + 50 µM AS + 50 µM NO2− and oxyHb + 50 µM AS to samples containing 100 µM NO2− and 50 µM NO2−. More precisely, it would be appropriate to compare the samples containing AS or AS + NO2− to samples titrated from 0 to 50 µM NO2− and from 50 to 100 µM NO2− at a rate comparable to AS decomposition. Figure 3A shows the oxidation of oxyHb to metHb by 100 µM NO2−, 50 µM NO2−, and autoxidation. Figure 3B demonstrates that the oxidation due to 50 µM NO2− is similar to the difference between the oxidation due to 100 µM NO2− and 50 µM NO2−. If anything, using samples containing only 50 µM NO2− may slightly underemphasize the acceleration of the reaction by HNO. It would be expected that values found by titrating NO2− at a rate comparable to AS decomposition would lie between these two curves. Thus, oxidation due to 50 µM NO2− is considered to be an appropriate comparison for this study.</p><p>The initial reaction of oxyHb with HNO produces NO as well as metHb (Eq. 2). It is therefore reasonable to verify that the acceleration of the nitrite-induced oxidation of oxyHb is not due to a reaction with NO rather than HNO. To that end, the NO donor DEA NONOate was used in place of Angeli's salt. DEA NONOate has a half-life of 2 min at 37°C and pH 7.4 [71], comparable to the half-life of Angeli's salt (2.3 min) under the same conditions. Figure 3C shows the conversion of oxyHb to metHb by 50 µM NO, 50 µM NO2−, and 50 µM NO + 50 µM NO2−. Figure 3D demonstrates that there is no significant difference in the concentration of metHb formed by NO2− in the presence or absence of NO.</p><p>The acceleration of the oxidation of oxyHb by NO2− in the presence of HNO can also be seen using HNO that does not originate from AS and does not inherently contain any NO2−. This was demonstrated using 4-nitrosotetrahydro-2H-pyran-4-yl pivalate in the presence of pig liver esterase (PLE), which releases HNO but does not release significant quantities of nitrite over the time period of interest. Time resolved spectra demonstrating the oxidation of oxyHb to metHb in the presence of this compound and NO2− are shown in Figure 4A. Figure 4B shows metHb yields as a function of time from the four different reaction mixtures using this compound instead of AS. HNO release from 4-nitrosotetrahydro-2H-pyran-4-yl pivalate and PLE (k = 3×10−4 s−1) is slower than AS (k = 5×10−3 s−1); thus, the quantitative results of experiments using the two HNO donors are different. However, both reactions demonstrate an increased rate of metHb formation due to NO2− in the presence of HNO. Figure 4C shows that metHb formation by NO2− in the presence of HNO 4-nitrosotetrahydro-2H-pyran-4-yl pivalate and PLE initially progresses at a faster rate than metHb formation by NO2− alone.</p><p>The addition of catalase to the reaction containing Angeli's salt had no substantial impact on the acceleration of the oxidation of oxyHb. Figure 5A shows metHb yields as a function of time from the four different reaction mixtures in the presence of 50 µM catalase. Figure 5B demonstrates that the acceleratory action remains in this system. Moreover, the addition of 250 µM c-PTIO had a similarly negligible effect. Although the overall metHb concentrations increased in the presence of c-PTIO (Figure 5C), these increases were uniform and did not prevent the acceleration of oxidation of oxyHb by nitrite (Figure 5D). Similarly, the addition of 10 Ku/mL superoxide dismutase had little impact on the system, either in the raw data (Figure 6A) or in the acceleration of oxidation of oxyHb (Figure 6B). Additionally, incubation of 1 mM oxyHb with 2 mM IHP for one hour showed no substantial effect on either the raw data (Figure 6C) or the acceleration of oxidation of oxyHb (Figure 6D).</p><p>In order to explore a role of metHb in the acceleration, the reactions were also conducted with an initial concentration of 250 µM metHb and 750 µM oxyHb (Figure 7A), and while the kinetics of the reaction were altered, the initial presence of metHb did not prevent acceleration of the oxidation of oxyHb by nitrite in the presence of HNO and the acceleration was similar to when no metHb was added initially (Figure 7B). Experiments were also performed at 200 µM oxyHb, the total amount of metHb expected from the reaction of 50 µM AS and 50 µM NO2− (Figure 7C), and while the kinetics were again affected by the change in initial conditions, the acceleration was still observed and the acceleration was similar to when no metHb was added initially (Figure 7D). In addition, we explored the possibility that HNO reacts with metHb bound nitrite or nitrite reacts with metHb bound HNO. We did experiments looking at any reactivity between nitrite, HNO, and metHb using high nitrite concentrations (>1mM). Nitrite slightly impeded the formation of HbNO from metHb of HNO, and formed nitrite-bound metHb (data not shown).</p><!><p>We have demonstrated that HNO accelerates the oxidation of oxyhemoglobin by nitrite. This acceleration affects the oxidation rate of oxyHb by Angeli's salt, as well as by sodium nitrite in the presence of HNO from a nitrite-free HNO donor. Nitrite from AS is often ignored in experiments involving heme proteins due to the assumption that the nitrite reaction is sufficiently slow compared to HNO so as to be negligible. However, our results demonstrate that this is not the case, and that acceleration of the nitrite reaction must be considered in experiments involving AS.</p><p>The use of AS as an HNO donor has previously been shown by Sulc et al. to yield different end products from HNO donors such as the Piloty's acid analogue MSHA under aerobic conditions [72]. These reactions were performed at differing pH (pH 7 for AS and pH 10 for MSHA), so differences in product formation could be due to pH as well as to the presence of nitrite. However, the general findings were consistent with earlier work by Bazylinski et al. demonstrating complicated reactivity of AS with heme proteins [73; 74].</p><p>Allosteric processes might be considered as a possible cause of the acceleration of the nitrite reaction with oxyHb. If reaction of oxyHb with HNO alters the allosteric state of the Hb, and this altered allosteric state reacts faster with nitrite than the unreacted form, allostery could play a role in our observed kinetics. However, treating the hemoglobin with excess IHP, which stabilizes the T-state and inhibits the allostery of hemoglobin, had no substantial effect on the acceleration of the nitrite reaction. This suggests that the reaction is independent of allostery. It should be noted that this experiment was also attempted using myoglobin in place of hemoglobin incubated with IHP to examine allosteric effects; however, the autoxidation of the system dominated the reaction and impeded collection of statistically significant data.</p><p>Another possible mechanism for the HNO mediated increase in the rate of the nitrite/oxyHb reaction involves formation of reaction products/intermediates that are known to contribute to autocatalysis of the nitrite oxyHb reaction, such as peroxide and•NO2. In the autocatalytic system described in (Eqs. 8–12), peroxide initiates autocatalysis through formation of a ferryl hemoglobin radical and subsequent•NO2 propagation. However, even high concentrations (50 µM) of catalase did not prevent or delay the acceleration of the nitrite reaction; in contrast, Keszler and coworkers saw that catalase markedly delayed the initiation of the autocatalytic nitrite reaction. This suggests that peroxide is not a vital initiator in the acceleration we see of the nitrite/oxyHb reaction. Superoxide dismutase had a similarly negligible effect, arguing against a superoxide driven reaction as well. Catalase and superoxide dismutase react with HNO at rates of 3×105 M−1 s−1 and 7×105 M−1 s−1 respectively [12], and thus would not be expected to react significantly with HNO under these conditions. In addition to the observation that catalase did not impede the acceleration of the reaction, initiating the reaction in the presence of 250 µM metHb had little impact on the overall reaction; what impact was observed can be explained by a combination of the formation of nitrite-bound metHb and subsequent decrease in the amount of free nitrite, and by changes to the kinetics of the reaction due to decreasing available oxyHb. We also investigated the possibility of HNO reducing metHb to HbNO by attempting to detect HbNO in our samples using electron paramagnetic resonance (EPR) spectroscopy. No HbNO was detected (data not shown), indicating that any HbNO present was below the detection threshold of about 1 µM for this technique. Reducing the amount of free oxyHb without the addition of metHb also affected the kinetics, decreasing the total yield of metHb in the reaction.</p><p>Keszler et al. also found that addition of c-PTIO blocked propagation of autocatalysis through scavenging of the•NO2 radical to form nitrite and PTIO+ as described by Goldstein et al. [75]. Addition of c-PTIO did not similarly block acceleration of the nitrite reaction in our experiments; however, this does not automatically exclude•NO2 radical as a possible mechanism. The reactions of PTIO and c-PTIO with NO,•NO2, and HNO, as well as reactions between the nitrogen oxides, constitutes a complicated chemical system. The previous experiments were conducted with 60 µM c-PTIO and only 30 µM oxyHb. In contrast, our experiments had oxyHb in excess to both nitrite and c-PTIO. Higher concentrations of c-PTIO were not useable in our experiments as they rapidly oxidized the oxyHb independent of nitrite. In our reactions, it is likely that, if•NO2 radical was formed, some of it did react with c-PTIO to form PTIO+, as the rate constant for this reaction was reported to be 1.5×107 M–1 s–1. However, as nitrite is produced by this reaction, and hemoglobin was not a limiting reagent, the nitrite produced by the oxidation of c-PTIO would have been free to react with another oxyHb. Thus, c-PTIO might not be expected to block an autocatalytic reaction at the concentrations we used, and any inhibition of the acceleration of oxidation by nitrite and HNO could have been masked by the oxidation of oxyHb by c-PTIO.</p><p>While we favor a mechanism that involves an autocatalytic pathway propagated by•NO2, we do not necessarily favor the initiation of this mechanism by peroxide. There are other possible sources of trace amounts of•NO2 that could be sufficient to start an autocatalytic reaction. HNO may lead to•NO2 formation through the creation and subsequent decay of peroxynitrite via direct reaction of HNO with molecular oxygen (Eqs. 15, 16) [76; 77; 78] (Eq. 15)HNO+O2→OONO−+H+ (Eq. 16)OONO−→•O2−+•NO2 However, it has been reported by others that this reaction does not occur under physiologically relevant conditions [79; 80]. It is also conceivable that such a reaction occurs between HNO and heme-bound oxygen, forming a peroxynitrite adduct similar to that proposed by Keszler et al. in (Eq. 7) for the autocatalysis of nitrite. Indeed, the products of such a reaction are in agreement with (Eq. 2) for the reaction of HNO and oxyHb. If such an adduct were formed and dissociated from the heme as OONO−, even in miniscule quantities, rather than oxidizing the heme as described in (Eq. 2), peroxynitrite could initiate an autocatalytic reaction rather than peroxide. Peroxynitrite has been shown to oxidize oxyHb by means of formation of an•NO2 radical [81; 82], and importantly, it was observed by Romero and coworkers that during the reaction of oxyHb and peroxynitrite, nitrite is consumed [81]. A catalytic mechanism involving•NO2 was also suggested to explain this phenomenon, invoking autocatalytic mechanisms [61] related to the autocatalysis of the reaction of nitrite and oxyHb. Thus, overall, we favor a mechanism whereby NO2 is formed and leads to autocatalysis of the nitrite/oxyHb reaction in the presence of HNO. The c-PTIO used may have been insufficient to scavenge•NO2, or perhaps•NO2 formed could act locally within the same Hb tetramer.</p><p>The acceleration of the oxidation of oxyHb by nitrite in the presence of HNO could have significant implications for HNO biology and the potential use of HNO as a therapeutic, particularly for HNO sources that also produce nitrite such as Angeli's salt. The nitrite/oxyHb reaction is generally slow under physiological conditions where nitrite is generally much lower in concentration than hemoglobin or myoglobin so autocatalysis does not occur. However, we have shown here that HNO accelerates the nitrite/oxyHb reaction even when Hb is in excess. If AS were to be used as a therapeutic, it could result in elevated concentrations of both HNO and nitrite and, in addition to the effects of these two agents individually, the acceleration of the oxyHb/nitrite reaction could also have effects on biological systems. This is also likely to be the case where an HNO-based therapeutic would affect the rate of the reaction of oxyhemoglobin with intrinsic nitrite. It should be noted that since HNO also catalyzes the nitrite/deoxyHb reaction [52], nitrite/Hb reactions would be expected to be faster due to the presence of HNO at all oxygen saturations. The extent of effects of HNO on reactivity of nitrite and Hb as well as other heme proteins merits further exploration.</p><!><p>This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.</p><p>Abbreviations: nitric oxide (NO), nitroxyl (HNO), Angeli's salt (AS), hemoglobin (Hb), oxygenated hemoglobin (oxyHb or [HbO2]2+), deoxygenated hemoglobin (deoxyHb), methemoglobin (metHb or [Hb]3+), pig liver esterase (PLE).</p>

PubMed Author Manuscript

Chemical accuracy in QM/MM calculations on enzyme-catalysed reactions

Combined quantum mechanics/molecular mechanics (QM/MM) modelling has the potential to answer fundamental questions about enzyme mechanisms and catalysis. Calculations using QM/MM methods can now predict barriers for enzyme-catalysed reactions with unprecedented, near chemical accuracy, i.e. to within 1 kcal/mol in the best cases. Quantitative predictions from first-principles calculations were only previously possible for very small molecules. At this level, quantitative, reliable predictions can be made about the mechanisms of enzyme-catalysed reactions. This development signals a new era of computational biochemistry.

chemical_accuracy_in_qm/mm_calculations_on_enzyme-catalysed_reactions

Background<!>Discussion<!>Conclusion<!>Acknowledgements

<p>Ever since the catalytic power of enzymes was first recognised, chemists have wondered and argued about how they work. Enzymes are outstandingly efficient natural catalysts. Better understanding of the mechanisms by which they achieve these catalytic properties promises technological spin-offs such as routes to new drugs (many drugs are enzyme inhibitors, which bind to enzymes and prevent them from functioning), analysis of the effects of genetic variation and mutation (for example in predicting individual metabolism of pharmaceuticals); and the design of new catalysts (for example biomimetic catalysts or engineered enzymes). There is great interest in developing protein catalysts for practical applications, for instance in the pharmaceutical, chemical and biotechnology industries. Computational modelling has a vital role to play in these developments: unstable species such as transition states and reaction intermediates are crucial to questions of reactivity, but cannot be studied directly by experiment in systems as complex as enzymes. The field of enzyme reaction modelling has grown enormously in recent years and has matured to the point that computational enzymology is increasingly recognised as essential for understanding these fascinating biological catalysts [1-4]. Recent calculations [5] bring a new level of accuracy to bear on the problem, essential for quantitative conclusions and comparisons with experiment.</p><p>Combined quantum mechanics/molecular mechanics (QM/MM) methods allow enzyme reactions to be modelled: a small region at the active site (where the reaction happens) is treated by a quantum mechanical electronic structure method; this region interacts with the protein and solvent environment, which are included more simply (though in atomic detail) by an empirical 'molecular mechanics' force field [1,2,4,6]. This approach combines the simplicity and speed of the MM treatment of the protein structure with the flexibility and power of a quantum chemical treatment (which allows modelling of bond breaking and making, and electronic polarization). Until recently, QM/MM investigations of enzymes have generally been limited to relatively low levels of QM theory, such as semi-empirical methods or density functional theory (DFT). Semi-empirical methods are computationally cheap, fast enough for QM/MM molecular dynamics simulations, but error-prone, and give reaction energies and barriers that can be in error by 10 kcal/mol or more. DFT (especially with the B3LYP hybrid functional) offers improved accuracy, and has opened new classes of enzymes (particularly metalloenzymes) to computational investigation [7], such as studies of cytochrome P450 enzymes that metabolize drugs in the body [8,9]. These DFT methods, however, lack key physical interactions, such as dispersion, which are important in the binding of ligands to proteins. DFT often gives barrier heights that are too low by several kcal/mol, and it does not offer a route to their systematic improvement or testing, making it difficult to assess the accuracy of results. Other modelling methods such as the empirical valence bond technique can give excellent results for enzyme activation energies [3,10], and have provided important insights into the basic causes of catalysis. Such empirical approaches, however, require extensive fitting to experimental data, and do not consider the electronic structure explicitly.</p><p>Enzymology has been marked by vigorous debates and controversial proposals about enzyme mechanisms, and about the physical origins of enzyme catalysis. Identifying the chemical mechanisms of enzymes has proved difficult: it is often hard to differentiate between alternative proposals, and many 'textbook' mechanisms are probably incorrect in important details. Recent controversies over enzyme catalysis include proposals of 'low-barrier' hydrogen bonds [11-14], 'near-attack conformations' [4,15,16], the role of enzyme dynamics in catalysis [2,3], quantum tunnelling [17] and entropic effects [3]. The applicability of transition state theory to enzyme reactions has been questioned. These arguments have often proved extremely difficult to resolve, because the complexity and large size of enzymes makes experimental analysis very difficult. Atomistic simulations have a potentially vital role to play in these debates, in the interpretation of experimental data, and in providing a molecular level picture of reactions in enzymes. Calculations have the potential to identify probable mechanisms, and to analyse key interactions and catalytic effects. For quantitative comparisons with experiments, and reliable predictions, high-level electronic structure methods are needed. Recent work by Claeyssens et al. has shown that it is now possible to achieve an unprecedented level of accuracy for enzyme-catalysed reactions in QM/MM calculations [5]. Calculated activation energies for two enzyme reactions agree very well with experiment; indeed the agreement is so good that, given the known properties of the high-level methods now available, it is clear that near chemical accuracy (1 kcal/mol) can be achieved in calculations on enzyme-catalysed reactions. Such quantitative predictions in first principles calculations were only previously possible for very small molecules. These findings herald a new era of computational biochemistry.</p><!><p>The calculations focused on two enzymes that have become paradigms for computational investigations: chorismate mutase (CM) and para-hydroxybenzoate hydroxylase (PHBH). Both have been studied previously with lower-level (semi-empirical and ab initio) QM/MM methods [4,15,18-23], and a wealth of experimental data is available for comparison. CM catalyses the Claisen rearrangement of chorismate to prephenate, a key step in the biosynthesis of essential aromatic amino acids. The reaction catalysed by PHBH, an electrophilic aromatic substitution involving a hydroperoxyflavin cofactor, is important in the microbial breakdown of aromatic pollutants and lignin from wood. Earlier modelling had been encouraging, for example in showing correlations between experimental rates and calculated activation energies for the key step in PHBH [24], in addition to identifying groups involved in lowering the energy barrier to reaction by transition state stabilization in both enzymes [4,22,23]. The barriers calculated at these lower levels of QM/MM treatment were, however, typically significantly different from experiment by 50% or more.</p><p>Computational chemistry is notorious for its love of acronyms, which can make judging the results of calculations difficult for non-specialists. The 'gold standard' of quantum chemistry is provided by first principles – 'ab initio' – methods that include correlation between electrons. They allow the calculation of rate constants for gas-phase reactions involving very few atoms with an accuracy similar to that obtained experimentally [25]. For example, the hierarchy of ab initio electron correlation methods MP2 (Møller-Plesset second-order perturbation theory), CCSD (coupled-cluster theory with single and double excitations) and CCSD(T) (CCSD with a perturbative treatment of triple excitations) provide a a route to converging reliably to high accuracy. These methods have previously been limited to small molecules because of their very large computational expense, which increases enormously with the size of the system studied. This large increase in computational cost is mostly due to the fact that the molecular orbitals are delocalized over the whole system. Physically, dynamic correlation between electrons is a short-ranged phenomenon in covalent molecules. The vital electron correlation effects can therefore be treated accurately and efficiently by localizing the molecular orbitals. Using a range of approximations it is possible to achieve effective linear scaling of the computational requirements of calculations with molecule size. With recent developments[26] it is now possible to treat systems of the size of typical QM regions in QM/MM calculations on enzymes (for example 24 atoms in CM; 49 atoms in PHBH).</p><p>The model of PHBH, constructed by Thiel et al., contained 7004 protein atoms, and surrounding water, altogether around 23,000 atoms. The model of CM, for the enzyme from Bacillus subtilis, contained 4192 protein atoms and 947 water molecules, in an approximate sphere of radius 25Å around the active site. To account for conformational variability of the enzymes, structures were taken from MM and QM/MM molecular dynamics simulations. (QM/MM molecular dynamics simulations, with semi-empirical QM methods, were also used to calculate activation free energies). These structures were used for reaction modelling, in which the structure is minimized at a series of set values of a reaction coordinate, defined in terms of breaking and forming bonds. Ten separate pathways were calculated for PHBH, and 16 for CM, with the results averaged over all paths. The paths were calculated at the B3LYP/MM level, which gives good structures.</p><p>Energies along the reaction pathways were calculated at the MP2, LMP2 and LCCSD(T0) levels (the L in the acronyms indicates that local approximations were used, and T0 is an approximate triples correction [27]), including the effects of the atomic point charges of the MM atoms in the calculations. These included by far the largest coupled cluster calculations ever performed (the calculations on PHBH involved 284 electrons and 1294 basis functions). Approximations included in the calculations were tested; for example, calculations were repeated with a much larger QM region for CM, showing a change in barrier of only 0.7 kcal/mol. The convergence with respect to basis set size was tested at the MP2 level, showing a change in barrier of less than 0.5 kcal/mol when very large basis sets were used. Similarly, tests showed the local approximations had an effect of less than 0.5 kcal/mol. Altogether, the errors in the best (LCCSD(T0)) barriers can be estimated to be less than 1 kcal/mol compared to extremely high (CCSD(T)) levels of quantum chemical theory, which are known, from calculations on small molecules, to give accurate barriers (i.e. typically within 1 kcal/mol of experiment. The accuracy obtained for energy barriers by Claeyssens et al. is unprecedented for enzyme reactions, and has been previously been difficult to attain even for small molecules.</p><p>How, then, do the calculated barriers compare with experimental findings for these enzymes? To make this comparison, the calculated (potential energy) barriers should be corrected for the quantum mechanical zero-point energy and thermal energy of molecular vibration: zero-point energy is calculated to reduce the barrier by 1.5 kcal/mol in CM, and by 1.1 kcal/mol in PHBH. In contrast, the thermal vibrational contribution is small (0.1–0.2 kcal/mol). With these corrections, the calculated barriers can be compared directly to experimental activation enthalpies. Experimental kinetic data for enzymes typically give steady-state rates, and so in making a comparison it is important to remember that a single chemical step may not be rate limiting – many enzymes have rates (at least partially) determined by conformational changes or product dissociation, for example. For both CM and PHBH, it has been suggested that other steps may be rate-limiting under some conditions. The chemical steps in both cases are thought to have barriers close to the apparent activation energy for the overall reaction.</p><p>The activation enthalpies calculated at the highest levels of quantum chemical theory (LCCSD(T0)) are in excellent agreement with experiment (within ~1 kcal/mol) for both enzymes. The calculated value for CM (average 13.3 kcal/mol, with a root mean square variation of 1.1 kcal/mol across 16 pathways) can be compared with the experimental value of 12.7 kcal/mol, while for PHBH the calculated and experimental activation enthalpies are 13.3 ± 1.5 kcal/mol and 12.0 kcal/mol, respectively. Only the LCCSD(T0) barriers are in close agreement with the experimental results. The LMP2 and B3LYP methods give barriers that are 3–5 kcal/mol too low. This shows that a high-level electron correlation treatment such as LCCSD(T0) is required for quantitative predictions of barrier heights (and probably other properties) in enzymes. In the transition state, more electrons are close together, and so the correlation energy is very different from that of the ground state. Electron correlation is therefore very important in determining the barrier.</p><p>The key quantity in determining reaction rates is the activation free energy (Δ‡G), not the enthalpy. To calculate free energy differences, the effects of protein motion should be included, i.e. averages over ensembles of structures. Activation free energies can be calculated from molecular simulations, such as molecular dynamics simulations. Such simulations are feasible for enzyme reactions with more approximate QM/MM methods, e.g. using semi-empirical QM techniques. In general, the best approach will be to calculate the free energy profile at the low level, and correct it based on the difference between low- and high-level QM/MM potential energy profiles. The difference between the average activation enthalpy and the activation free energy gives an estimate of the entropic contribution to the barrier. For both CM and PHBH, the estimated entropic contributions (at a temperature of 300 K) are small: 2.5 kcal/mol for CM (similar to the experimental value of 2.7 kcal/mol) and 0.4 kcal/mol for PHBH. Adding these entropic contributions to the calculated enthalpies at the highest QM/MM level gives free energy barriers very close to those obtained experiment: for CM, the (LCCSD(T0)) calculated Δ‡G is 15.6 ± 1.1 kcal/mol (versus 15.4 kcal/mol from experiment); for PHBH the calculated Δ‡G is 13.7 ± 1.5 kcal/mol, comparable to experimental values of 14–15 kcal/mol.</p><p>The agreement between calculated and experimental energy barriers is excellent for both enzymes. The comparison is made based on transition state theory, and the quality of the agreement indicates that transition state theory describes these enzyme-catalysed reactions well. Dynamical effects apparently play only a small role in determining the rate. Classical TST is known to be insufficient in some cases, but corrections for dynamical recrossing and quantum mechanical tunnelling can be included [2,17]. Despite some previous suggestions to the contrary, it seems that transition state theory provides a good general framework for understanding the rates of such enzyme-catalysed reactions.</p><p>Many challenges remain. Among these is the need for extensive conformational sampling to achieve convergence in free energies. In general, the best approach will be to calculate free energy profiles at a low level and correct them using high-level calculations. Determining reaction pathways can be difficult. Improvements to simple MM models (e.g. to include polarization and more sophisticated descriptions of electrostatics) are also likely to be necessary. The QM and MM methods should be consistent and balanced. The treatment of the QM region is only part of the challenge; it is important to have a good structural model of the surrounding enzyme (usually derived from X-ray crystallography), and to consider carefully, for example, the solvation of the protein and the protonation states of ionizable groups in the protein. The pKas of basic and acidic amino acid sidechains can be significantly altered in the enzyme environment. Using the wrong protonation states could lead to the prediction of an incorrect mechanism. It is worthwhile to test methods of pKa prediction, particularly for active site residues. Predictions of amino acid pKas in proteins can be made with simple but effective empirical methods such as PROPKA [28], or by finite-difference Poisson-Boltzmann calculations.</p><p>Given these many challenges, and the complexity of enzymes, validation of modelling methods by comparison with experiment will be very important. Many enzyme reactions involve several chemical and conformational changes [29,30], and the chemical step(s) may not always be rate limiting. Often experimental rates only for overall reaction under steady-state conditions are available. Experimental kcat rate constants do provide an upper limit on the barrier for any step in the mechanism, however. Even when the rate of a single step has been measured, this is likely to represent an average over many enzyme molecules in solution, whereas a calculated barrier is generally for a single molecule. Comparisons between experimental rates and calculated barriers should therefore be done with care. Transient kinetics and single molecule studies will be particularly useful for detailed comparisons. Detailed comparison with experiment will be a fascinating challenge, now that calculations based on first principles allow quantitative predictions on enzyme mechanisms to be made. These computational techniques also promise to make a significant contribution to other areas of biology and chemistry.</p><!><p>The results of Claeyssens et al. show that electronic structure calculations can now predict activation barriers for enzyme-catalysed reactions with 'chemical accuracy', i.e. to within 1 kcal/mol in the best cases. At this level, quantitative, reliable predictions can be made about the mechanisms of enzyme-catalysed reactions. This development signals a new era of computational biochemistry.</p><!><p>The author thanks his collaborators in the work described here. He also thanks EPSRC, BBSRC and the IBM High Performance Computing Life Sciences Outreach Programme for financial support.</p>

PubMed Open Access

Systematic evaluation of split-fluorescent proteins for the direct detection of native and methylated DNA

In order to directly detect nucleic acid polymers, we have designed biosensors comprising sequence-specific DNA binding proteins tethered to split-reporter proteins, which generate signal upon binding a predetermined nucleic acid target, in an approach termed SEquence-Enabled Reassembly (SEER). Herein we demonstrate that spectroscopically distinct split-fluorescent protein variants, GFPuv, EGFP, Venus, and mCherry, function effectively in the SEER system, providing sensitive DNA detection and the ability to simultaneously detect two target oligonucleotides. Additionally, a methylation-specific SEER-Venus system was generated, which was found to clearly distinguish between methylated versus non-methylated target DNA. These results will aid in refinement of the SEER system for the detection of user defined nucleic acid sequences and their chemical modifications as they relate to human disease.

systematic_evaluation_of_split-fluorescent_proteins_for_the_direct_detection_of_native_and_methylate

<p>Specific DNA sequences are typically detected by denaturation of double stranded (ds)DNA followed by hybridization with a labeled oligonucleotide probe, as is the case with Southern blotting, fluorescence in situ hybridization (FISH),1 and DNA microarrays.2 In contrast to these approaches, new methods have been developed to directly detect dsDNA, which include designed systems that rely on sequence-specific recognition in the grooves of the native DNA double helix, including hairpin polyamides3 and triplex forming oligonucleotides (TFOs).4 Currently the utility of these techniques is constrained by limitations on the length of dsDNA that can be targeted and restrictions in the constitution of detectable sequences, respectively.5 In comparison to these design efforts, endogenous sequence-specific DNA binding proteins do not suffer from these drawbacks, while additionally providing a method for nucleic acid detection in its native environment. In our work we have focused on Cys2-His2 zinc fingers (ZFs), which comprise an important class of naturally occurring transcription factors and provide a programmable means of DNA detection.6, 7 In our method derived from split-protein assays also called protein-fragment complementation, fragments of a split-reporter protein are appended to DNA detection domains, and the binding event is monitored by signal generation arising from conditional reassembly of split-protein halves. Numerous proteins and enzymes have been genetically fragmented including ubiquitin,8 dihydrofolate reductase,9 green fluorescent protein (GFP),10 β-lactamase,11 and luciferase.12 GFP is particularly appealing as it singularly lacks any requirement for cofactors or substrates.13 By fusing split-GFP to sequence-specific ZF domains, recognition of target DNA induces GFP reassembly in a method called SEquence-Enabled Reassembly (SEER) (Figure 1).14</p><p>Currently, our reported SEER systems can specifically detect the presence of a single oligonucleotide sequence at a time.14–16 New SEER systems capable of simultaneously detecting multiple targets would be useful for numerous applications, including DNA profiling and ratiometric analysis. The GFP reporter affords the opportunity to accomplish this objective since GFP variants with distinct spectroscopic properties have been extensively studied.17 At present the repertoire of proteins derived from wild type Aequorea victoria GFP includes blue, cyan, and yellow fluorescent proteins (BFP, CFP, and YFP, respectively). Although a red-only emitting FP has not been achieved using the A. victoria scaffold,18 naturally occurring red fluorescent proteins (RFPs) have been identified. The so-called mFruits, monomeric DsRED derivatives, were recently selected and have since been employed as fusion tags in a variety of organisms, offering additional opportunities for protein complementation assays.19, 20 Thus, we constructed SEER systems incorporating five distinct fluorescent protein (FP) variants: a UV-excitable GFP, GFPuv;21 a CFP, Cerulean;22 an enhanced GFP, EGFP;23 a YFP, Venus;24 and a DsRED derived RFP, mCherry19 (Figure 2, left panels). Each N-terminal FP fragment (residues 1–157 for GFP-derived FPs and 1–168 for mCherry) was fused to ZF Zif268, while ZF PBSII was attached to each C-terminal FP fragment (residues 158–238 for GFP-derived FPs and 169–231 for mCherry). In all cases the split-FP domain was connected to the DNA binding domain via a flexible [(Gly)4-Ser]3 linker. Following isolation of these constructs, each SEER-FP system was assayed for preferential refolding in the presence of an optimized DNA target, Zif268-0-PBSII, by monitoring fluorescence emission of the reassembled complex. We were unable to generate a functional SEER-Cerulean system due to difficulties encountered during expression and purification of NCerulean-Zif268. However, all other tested split-FP constructs functioned effectively in the SEER context (Figure 2). Of note, split-mCherry was capable of DNA templated reassembly, thus providing the first success of a DsRED derived FP in the SEER context. Each SEER-FP system was optimized with respect to protein concentration, and a DNA titration (Figure 2, right panels) established sensitivities of the systems, which were qualitatively equivalent, and each reassembled FP produced signal over background at 10 nM (1.0 pmol) target DNA.</p><p>Due to the high fluorescence intensity of the SEER-Venus constructs, these proteins were selected to modify for detection of cytosine methylation, since this covalent DNA modification has emerged as a promising cancer biomarker.25 By simply replacing ZF PBSII with a methyl cytosine-guanine dinucleotide (mCpG) binding domain, MBD2, we were able to generate a system called mCpG-SEER-Venus for site-specific evaluation of the methylation status at individual CpG islands. We previously characterized two mCpG-SEER systems that conferred specificity for methylation at CpG sites, one of which employed an A. victoria FP variant.26, 27 To assess the specificity of the new system, we refolded NVenus-Zif268 and CVenus-MBD2 in the presence of target and off-target oligonucleotides (Figure 2, bottom panel). The mCpG-SEER-Venus system was shown to clearly distinguish between an mCpG site and its non-methylated equivalent with a 2.3-fold preference. This value seems slightly low since MBD2 has a 70-fold preference for binding methylated CpG islands (Kd = 2.7 nM) over the corresponding non-methylated site.28 An improvement in specificity for methylated sites may be achieved through further optimization of the mCpG-SEER-Venus system. Additional controls revealed a 4.3-fold reduction in fluorescence signal upon removal of the Zif268 binding site in the target (mCpG only) and a 3.8-fold reduction for a target with only a single guanine to thymine mutation in the Zif268 binding site (mCpG-2-Zif268 (G to T)). These results demonstrate the ability of the mCpG-SEER-Venus system to discriminate between cognate target sites and non-methylated, non-specific, or mutated DNA sequences. Finally, a DNA titration curve was generated for the mCpG-SEER-Venus system, resulting in detection of at least 25 nM DNA target (Figure 2, bottom panel).</p><p>The construction of several functional SEER-FP systems provides the potential for simultaneous detection of multiple DNA sequences, which we are currently pursuing with several new ZF DNA binding domains. An interesting alternative is the possibility of complementing different halves of the split-FPs to generate hybrid FPs with distinct spectral properties, which has previously been demonstrated for the detection of protein-protein interactions.29 We first attempted to visualize two different DNA sequences by complementing NVenus with both CVenus and CGFPuv (Figure 3A). The three fusion proteins were allowed to refold in the presence of one or two target oligonucleotides. Excitation and emission wavelengths were systematically scanned to identify major peaks. A fluorescence emission signal at 506 nm was attributed to the NVenus/CGFPuv hybrid and indicated the presence of the Zif268-0-PBSII target, while a 528 nm emission signal resulted from NVenus reassembling with CVenus in the presence of the Zif268-0-PE8B target (Figure 3B). When both targets were simultaneously present in solution, two distinct fluorescence emission maxima were observed by alternating the wavelength of excitation from 395 to 515 nm, corresponding to the formation of two distinct reassembled FPs. Only a minimal degree of off-target fluorescence was observed as emission at 506 nm in the absence of the Zif268-0-PBSII target (< 20%) and 528 nm emission in the absence of the Zif268-0-PE8B target (< 15%). Therefore, by using Zif268 to direct the constructs to a common target site, detection of two distinct downstream sequences was achieved. As a second example of mixed complementation, we reassembled NVenus and CCerulean (Figure 3C). We observed characteristic Venus fluorescence in the presence of the Zif268-0-PE8B target DNA and fluorescence attributable to the NVenus/CCerulean hybrid in the presence of Zif268-0-PBSII DNA (Figure 3D). A minimum of off-target fluorescence was observed using these constructs. These two examples of mixed complementation illustrate a straightforward method for the simultaneous detection of two distinct DNA sequences using a minimum of protein constructs. Thus far, we have generated a variety of SEER-FP constructs with distinct spectroscopic properties that are capable of sequence-specifically detecting dsDNA, reporting on chemical modifications to dsDNA, and simultaneously detecting two distinct DNA sequences.</p><p>To demonstrate the specificity of our methodology, we have attempted the detection of cognate binding sites present in plasmid DNA using the SEER-Venus system. Specifically, we constructed a plasmid, pRSF Z0P10, which contains ten copies of the Zif268-0-PBSII target site, and attempted to visualize this plasmid in both a native (supercoiled) and linearized state, using the pRSF empty plasmid as background. We observed 3.1-fold signal over background for the linear pRSF Z0P10 plasmid, indicating that our system is not only capable of detecting discrete targets in solution, but also capable of detecting a target in the context of a nearly 23-fold excess of non-cognate DNA base pairs. A 2.1-fold signal over background was achieved when detecting Zif268-0-PBSII binding sites in supercoiled pRSF Z0P10. These results suggest a high level of specificity for a cognate target with little background binding to off-target DNA.</p><p>In conclusion, SEER offers a number of advantages over current approaches available for the direct detection of native dsDNA. These protein-based biosensors provide modularity, convenience of preparation, and most importantly, selectivity for targeting a user-defined nucleic acid sequence. We are currently focused on expanding the utility and broadening the applicability of our methodology by incorporating distinct signaling and detection domains. To this end, we have successfully incorporated four spectrally distinct split-FP variants, GFPuv, EGFP, Venus, and mCherry, each of which provide many opportunities for enhancing nucleic acid detection, including the ability to simultaneously detect two DNA sequences. Additionally, we determined that our SEER proteins are capable of detecting specific DNA sequences present in a plasmid DNA context, representing the high level of selectivity associated with the use of sequence-specific DNA binding proteins. These new advances in the SEER system open the door for expanding applications and further developments. For example, using appended ZFs to direct the system to a specific region of DNA, site-specific cytosine methylation has been achieved through reconstitution of a split methyltransferase.30 Additionally, real-time imaging of mitochondrial RNA localization was accomplished using an RNA targeted split-FP reassembly approach.31 Thus, we have shown that SEquence-Enabled Reassembly is an adaptable system, capable of performing a variety of functions, limited only by the availability of split-proteins and nucleic acid targeting domains and the imagination of the experimenter.</p><p>SEquence-Enabled Reassembly. The FP fusion constructs are initially non-fluorescent. The introduction of target DNA results in ZF binding, which induces reassembly of a productively fluorescent FP.</p><p>Properties of FPs utilized in SEER. The left panels display chromophore structures and wavelengths of maximum excitation (λex) and emission (λem) for each of the full length FP variants. Mutations (A. victoria numbering) are listed below the corresponding structures with chromophore mutations indicated in bold. mCherry includes additional GFP-type residues on both its N- and C-terminus.19 The right panels show a DNA titration for each SEER-FP system. The bottom panels show specificity data and a DNA titration for the mCpG-SEER system.</p><p>Mixed complementation of FPs. (A) A mixture of NVenus-Zif268, CGFPuv-PBSII, and CVenus-PE8B bind their respective targets. Wavelengths of maximum excitation (λex) and emission (λem) are indicated for each reassembled species. (B) NVenus-Zif268, CGFPuv-PBSII, and CVenus-PE8B were mixed in the presence of target DNA, and fluorescence was monitored at the indicated wavelengths. (C) SEER-Venus/Cerulean mixed complementation. The three constructs, NVenus-Zif268, CCerulean-PBSII, and CVenus-PE8B, bind their cognate sequences. Wavelengths of maximum excitation and emission are indicated next to the reassembled FPs. (D) NVenus-Zif268, CCerulean-PBSII, and CVenus-PE8B were mixed in the presence of target DNA, and fluorescence was monitored at the indicated wavelengths.</p>

PubMed Author Manuscript

An Examination of the Proteolytic Activity for Bovine Pregnancy-Associated Glycoprotein 2 and 12

The pregnancy-associated glycoproteins (PAGs) represent a complex group of putative aspartic peptidases expressed exclusively in the placentas of species in the Artiodactyla order. The ruminant PAGs segregate into two classes -the \xe2\x80\x98ancient\xe2\x80\x99 and \xe2\x80\x98modern\xe2\x80\x99 PAGs. Some of the modern PAGs possess alterations in the catalytic center that are predicted to preclude their ability to act as peptidases. The ancient ruminant PAGs in contrast are thought to be peptidases, although, no proteolytic activity has been described for these members. The goal of this present study was to investigate (1) if the ancient bovine PAGs (PAGs-2 and -12) have proteolytic activity, and (2) if there are any differences in activity between these two closely related members. Recombinant bovine PAGs-2 and -12 were expressed in a baculovirus expression system and the purified proteins were analyzed for proteolytic activity against a synthetic fluorescent cathepsin D/E substrate. Both proteins exhibited proteolytic activity with acidic pH optima. The kcat/KM for bovine PAG-2 was 2.7\xc3\x97105 M\xe2\x88\x921s\xe2\x88\x921 and for boPAG-12 it was 6.8\xc3\x97104 M\xe2\x88\x921s\xe2\x88\x921. The enzymes were inhibited by pepstatin A with a Ki of 0.56 and 7.5 nM for boPAG-2 and boPAG-12, respectively. This is the first report describing proteolytic activity in PAGs from ruminant ungulates.

an_examination_of_the_proteolytic_activity_for_bovine_pregnancy-associated_glycoprotein_2_and_12

Introduction<!>Expression and purification of recombinant boPAG-2 and -12<!>Dependence of activities on pH<!>Confirmation that observed activity was not from co-purifying insect cell peptidases<!>pH dependencies of activities of zymogen and activated forms<!>Comparison of steady-state kinetics of recombinant boPAG-2 and -12<!>Substrate preferences of boPAG-2 and -12 in comparison to porcine gastric pepsin and bovine spleen cathepsin D<!>Homology-based modeling of the structures of boPAGs -2 and -12<!>Discussion<!>Cloning and expression of recombinant bovine PAGs-2 and 12<!>Purification of recombinant boPAG-2 and 12<!>Western blot analysis<!>Determining optimal pH for activity studies<!>Estimation of contamination by insect cell-derived acid peptidases<!>Establishing pH profiles for pro-and activated forms of recombinant boPAG-2 and boPAG-12<!>Determination of steady-state enzyme kinetics of boPAG-2 and -12<!>Comparison of Aspartic Protease Specificity by using a FRET Peptide Substrate Library<!>Computational methods for generating homology models of boPAG-2 and -12<!>SDS-PAGE gels of expressed recombinant bovine PAGs-2 and 12 stained either with Coomassie blue or by Western blotting<!>The relative activity of boPAG-2 (triangles) and -12 (squares) zymogens as a function of pH<!>Activity measured after an immuno-precipitation assay performed on insect cell lysates<!>Influence of pHon PAG-2 and PAG-12 activity<!>Determination of kinetic parameters for PAG-2 and PAG-12<!>Homology models to predict substrate binding site differences between bovine PAG-2 and PAG-12<!>

<p>Aspartic peptidases (AP) are a class of proteolytic enzymes that typically require acidic conditions for optimal activity (Davies, 1990). Consequently, most mammalian AP tend to be found in low pH environments such as those present in gastric secretions (e.g. pepsin A, pepsin C, chymosin), lysosomes (e.g. cathepsin D) and other intracellular compartments (e.g. cathepsin E) within the cell (Davies, 1990; Dunn, 2002). One notable exception to this trend is renin, which is operational under physiological pH conditions (Cody, 1994).</p><p>The mammalian APs are bi-lobed proteins with both the N-terminal and C-terminal lobes being roughly symmetrical to one another (Tang et al., 1973). The lobes form a binding cleft that can accommodate a peptide substrate of seven to eight amino acids. The catalytic aspartic acid residues Asp32 and Asp215 are located at the middle of the cleft (the numbering represents that for porcine pepsin). Aspartic peptidases are produced as zymogens/pro-enzymes (Tang et al., 1973). Functional activation of the zymogens involves proteolytic processing of the amino-terminal propeptide.</p><p>One hallmark of the mammalian APs is the presence of highly conserved residues that flank the catalytic aspartic acids: hydrophobic-F/I/L-D-T-G-S in the N-terminal domain, and hydrophobic-D-T/S-GS/T in the C-terminal domain (Cooper et al., 1990). The carboxyl groups of these two aspartic acids, Asp32 and A215 interact with one another and also with a solvent water molecule by a network of hydrogen bonds. The proteolytic mechanism involves activation of the water molecule positioned between these aspartates to act as a nucleophile on the carbonyl carbon of the scissile peptide bond (Davies, 1990; Szecsi, 1992; Dunn, 2002). The current model suggests that Asp215 acts as a general base to remove a proton from the water molecule while Asp32 donates a proton to the carbonyl oxygen of the scissile bond. In the resulting tetrahedral intermediate, Asp215 is hydrogen bonded to the attacking oxygen atom (originally part of the water molecule), while the hydrogen remaining on that oxygen is hydrogen bonded to an oxygen on Asp32 (Davies, 1990; Szecsi, 1992; Dunn, 2002). The transfer of a proton from Asp215 to the nitrogen of the scissile peptide bond completes its hydrolysis.</p><p>The pregnancy-associated glycoproteins (PAGs) are a recently discovered family of genes that belong to the AP gene family (Xie et al., 1997; Green et al., 1998; Hughes et al., 2003; Telugu et al., 2009). They are expressed exclusively by trophoblasts -the outer cells of the placenta that are in direct contact with maternal tissues. The PAG gene family is found only in species within the Artiodactyla order (Green et al., 1998; Garbayo et al., 2000; Green et al., 2000; Hughes et al., 2003).</p><p>In ruminant ungulates, the PAG gene family is particularly large and complex. Dozens of distinct cDNAs, and numerous variants, have been cloned from cattle, sheep, goat and deer placentae (Szafranska et al., 1995; Xie et al., 1997; Garbayo et al., 1998; Garbayo et al., 2000; Green et al., 2000; Brandt et al., 2007; Telugu et al., 2009). The PAG gene family in ruminants is comprised of two evolutionarily distinct groups (Green et al., 2000; Hughes et al., 2000; Hughes et al., 2003). One grouping, the 'modern PAGs', is transcribed exclusively in specialized, moderately invasive trophoblasts known as 'binucleate' cells (BNC) (Green et al., 2000; Hughes et al., 2003; Wooding et al., 2005). The other grouping, known as the 'ancient PAGs', is transcribed in all trophoblast cell types (Green et al., 2000; Wooding et al., 2005).</p><p>The ancient PAGs are packaged in vesicles within both mononucleate and binucleate trophoblasts and, upon secretion, they accumulate at the microvillar junction of the maternal-fetal interface (Wooding et al., 2005). Coincident with differences in spatial expression, there are variations in temporal expression patterns as well. For example, some PAGs are expressed relatively early in gestation, while other PAGs do not appear until later in pregnancy (Green et al., 2000; Patel et al., 2004). Notably, there are also obvious differences in their levels of expression. For example, bovine (bo) PAG-2 is the most abundant transcript among all the PAGs identified to date. A closely related family member, boPAG-12, is substantially less prevalent in the placenta (Telugu et al., 2009).</p><p>Interestingly, many of the modern PAGs possess atypical residues in amino acid positions known to be involved in catalysis or in substrate-binding (Guruprasad et al., 1996; Green et al., 2000). Since PAGs are closely related to pepsin, molecular models (based on porcine pepsin and bovine chymosin crystal structures) revealed that some of the alterations within the catalytic center are likely to render some modern PAGs incapable of acting as proteolytic enzymes (Xie et al., 1995; Guruprasad et al., 1996; Green et al., 2000). On the other hand, most of the ancient PAGs have retained the characteristics of typical APs and are known or predicted to possess proteolytic activity (Green et al., 1998; Wooding et al., 2005; Telugu and Green, 2008). In those PAGs suspected to be peptidases, there are differences in residues known to contribute to catalytic activity and substrate specificity, suggesting that different members of the ancient PAG grouping probably possess distinct substrate preferences and activities (Xie et al., 1991; Guruprasad et al., 1996; Xie et al., 1997).</p><p>In the experiments described in this report, we sought to determine if some ancient PAGs in cattle are capable of proteolytic activity. Two paralogous ancient PAG members were chosen as the focus of the analysis – these were boPAG-2 and boPAG-12. Both of these proteins accumulate at the placenta-maternal interface (Wooding et al., 2005 and unpublished data). However, they also exhibit distinct temporal patterns of expression throughout pregnancy and they differ substantially in their relative level of transcript abundance in the placenta. Bovine PAG-2 is the most abundantly transcribed PAG gene in cattle, whereas boPAG-12 mRNA is much less abundant – differing by as much as two to three orders of magnitude at any given time-point during pregnancy (Telugu et al., 2009).</p><!><p>Recombinant bovine PAGs-2 and -12 were expressed in a Baculovirus insect cell expression system as fusion proteins with an N-terminal FLAG peptide engineered in-frame with the pro-peptide. The FLAG peptide allowed for affinity purification of fusion proteins by using an anti-FLAG M2 antibody matrix (Figure 1). Typical yields for the FLAG-zymogens were approximately 75 μg of purified protein per one confluent (80%) 75 cm2 culture dish. The expressed and purified proteins had the molecular weight expected of the full length fusion proteins(approximately 45,000 MW). Accompanying the regular pro-form of boPag-12, however, was some limited glycosylation, which manifested as a somewhat higher molecular weight fraction (Figure 1). This probably resulted from it being cloned into the pACMP2 transfer vector that facilitates limited cytoplasmic glycosylation (Pharmingen's Baculovirus expression manual). Both PAGs as well as the higher molecular fraction of boPAG-12 were identified in Western blots by using an anti-FLAG antibody(Figure 1). The presence of boPAG-2 was further confirmed with a polyclonal antibody raised against boPAG-2 (Wooding et al., 2005).</p><!><p>The presence of proteolytic activity and the preferred pH optima for each recombinant PAG zymogen was tested against a commercial cathepsin D/E FRET substrate. The preferred pH for maximal activity for boPAG-2 was found to be around pH 4.0, with high activity from pH 3.0 to 4.5 (Figure 2). Bovine PAG-12 had an optimal activity at pH 3.5, with high activity over a broader range of pH 2.5 to 4.5 (Figure 2). Some differences were notable. BoPAG-12 had high activity at pH 2.5 while boPAG-2 had relatively little activity at this pH. The optimal activity of boPAG-2 and 12 under acidic conditions is consistent with the pH preferences of most other aspartic peptidases (Davies, 1990; Kay and Dunn, 1992; Szecsi, 1992; Dunn, 2002).</p><!><p>An immunoprecipitation experiment was performed against several insect cell lysates (with the anti-FLAG antibody) followed by peptidase assays on the immunoprecipitated material. The lysates tested represented insect cells infected with wildtype virus and insect cells infected with baculovirus driving expression of recombinant equine TSH-beta (eTSHβ), bovine PAG-2 and boPAG-20 (a modern PAG). The activity obtained from the boPAG-2 lysate was substantially greater than was observed in the two control lysates [eTSHβ and cells infected with wild-type baculovirus (ACNPV)]. Bovine PAG-20 also exhibited some activity, albeit much less than boPAG-2 (Figure 3).</p><!><p>One hallmark of many APs is their ability to remove or displace their propeptides when placed in an acidic environment (Hazel et al., 1992; Koelsch et al., 1994; Richter et al., 1998; Wittlin et al., 1999; Barrett et al., 2004). When boPAG-2 zymogen was incubated in activation buffer (0.25 M glycine-HCl, pH 3.5) for 1 min at RT, the ability of boPAG-2 to exhibit activity at higher pH became evident. When the pre-incubated protein was compared in the assay alongside boPAG-2, which had not been subjected to a pre-incubation step, there was a moderate increase in activity in the pre-incubated form. The difference in activity was most pronounced at pH 2.5 and between pH 5.5 and 6.5 (Figure 4a). The greatest initial rate activity was recorded within the pH optima of 3.5 to 4.0.</p><p>Similar measurements were made on boPAG-12. In contrast to boPAG-2, exposing the zymogen of boPAG-12 to pH 3.5 for 1 min did not result in an increased activity. However, exposing the protein to pH 3.0 for 1 min did produce a marginal increase in activity in the range from pH 2.5 to 5.0 (Figure 4b). Also unlike the situation that was observed with PAG-2, there was no marked improvement in activity between the pre-incubated protein and the intact zymogen between pH 5.5 and 6.5. These experiments were replicated multiple times and the observations were consistent among the experiments.</p><!><p>The Michaelis-Menten kinetic parameters kcat, Km and kcat/Km were determined from progress curves for the full-length PAG-FLAG fusions (Figure 5; Table 1). The Ki for pepstatin A was also determined (Table 1). These data revealed differences in catalytic turnover rate and efficiency toward the cathepsin D/E substrate between boPAG-2 and -12. Under identical reaction conditions, the catalytic efficiency of boPAG-2 (2.72 × 105 M−1s−1) was found to be approximately 4-fold greater than boPAG-12 (6.86 × 104 M−1s−1). The kcat for boPAG-2 (0.96 ± 0.08 s−1) was approximately 3.3 fold higher than boPAG-12 (0.29 ± 0.03 s−1). The Km values of 3.53 μM and 4.2 μM for boPAG-2 and -12, respectively, being within 20% of each other indicated that the two enzymes have similar affinity for the substrate. The Ki for pepstatin A was lower by more than an order of magnitude in boPAG-2 (0.56 nM) compared to boPAG-12 (7.5 nM) (Table 1).</p><!><p>To gain insight into the primary peptide specificity for boPAG-2 and -12, both full-length forms were investigated for their activity against commercially available FRET-25X peptide libraries alongside two canonical mammalian APs, porcine pepsin (from gastric mucosa) and bovine cathepsin D (from spleen). While porcine pepsin and cathepsin D displayed good activity against the primary substrate libraries, boPAG-2 and -12 failed to produce reasonable velocities (Table 2). Of the two boPAGs, boPAG-12 exhibited a detectable, albeit minor, release of product. In contrast, boPAG-2 failed to exhibit detectable activity, even with relatively high concentrations of the enzymes (100 nM of active-site titrated enzyme) (Table 2).</p><!><p>The homology models for boPAGs-2 and -12 were created by using crystal coordinates of human pepsin bound to pepstatin A (1psn.pdb). SWISS-MODEL predicted this as one of the best templates for generating the model. Previously reported homology modeling of a modern PAG, boPAG-1, identified key residues predicted to contribute to the substrate binding pockets (Guruprasad et al., 1996).</p><p>Based on this exercise, it was determined that eight sequence positions may distinguish the active sites of pepsin, boPAG-2, boPAG-12, and homologues (Figure 6). Their differing side chains may tune their respective substrate specificities. The subsites proceed in the views of Figure 6 from S4 in the upper background towards S3′ (primed) in the lower foreground. The subsites were numbered according to the structure of the complex of pepsin with pepstatin (Fujinaga et al., 1995). Either Thr12 and Gln288 or Met12 and His288 appear at S4 in the respective homology models of boPAG-2 and -12 (toward the back at top in Figure 6). Leu115 or Ala115 contribute to both S3 and S1 in boPAG-2 and -12, respectively. Either Ser77 or Pro77 of the flap of these two proteases contribute to both S2 and S1. Either Tyr222 or His222 form the other wall of S2 on the opposite side of the cleft (Figure 6). Aromatic groups also differentiate S1′ with either His189 of boPAG -2 or Tyr189 of boPAG-12. Long aliphatic chains of either Met 128 or beta-branched Ile128 appear at S2′ in the two respective enzymes. For the flap's contribution to S3′, the bulky hydrophobic Phe74 of boPAG -2 is a notable contrast to the small polar Ser74 of boPAG-12.</p><!><p>Pregnancy-associated glycoproteins are related to APs and are abundantly expressed products of the placenta of ruminants and other species within the Artiodactyla order. Although PAGs have been studied extensively, their role(s) within the ungulate placenta remains largely obscure. Despite being structurally related to APs, some of the modern PAGs have accumulated mutations within key residues in and around the catalytic site, including the critical catalytic aspartates (Xie et al., 1991; Xie et al., 1997; Green et al., 1998). Those modern PAGs carrying such mutations are likely to be incapable of acting as peptidases. The ancient PAGs of cattle, sheep and goats, in contrast, have all the definitive characteristics of typical APs and were predicted to have proteolytic activity (Green et al., 1998; Wooding et al., 2005). Consequently, the goal of this project was to address two important questions, (1) do ancient bovine PAGs have proteolytic activity and, if so, (2) how do closely related family members compare to one another in regard to kinetic parameters? To address these fundamental questions, boPAGs-2 and -12 were chosen as candidates for these experiments. BoPAG-2 was an obvious choice since it is the most abundantly transcribed member of the bovine PAG grouping (Telugu et al., 2009). It is also among the first PAGs to have been identified and characterized in regard to its expression by trophoblasts and in regard to its abundant localization at the placenta-uterine interface (Xie et al., 1994; Wooding et al., 2005). BoPAG-12 is closely related to boPAG-2, sharing 89% identity at the nucleotide level and 83% identity in amino acid composition (Green et al., 2000; Hughes et al., 2000). However, its level of expression is orders of magnitude below that for boPAG-2, despite that it is also localized to the placenta-uterine interface (Telugu et al., 2009).</p><p>A baculovirus insect cell expression system was chosen for expression of soluble recombinant full-length PAGs. This system is amenable to expression of acid peptidases because the cytoplasm in these cells is not strongly reducing and is maintained close to a pH of 7.0 (Medina et al., 1995). Such conditions were predicted to help prevent the activation and subsequent loss of acid peptidases by autocatalysis. The inclusion of a synthetic coding sequence for the FLAG peptide permitted the isolation of relatively homogenous preparations of purified proteins by using standard anti-FLAG affinity purification protocols.</p><p>The purified full-length protein preparations were tested for optimal activity against a synthetic FRET substrate MOCAC-Gly-Lys-Pro-Ile-Leu-Phe-Phe-Arg-Leu-Lys(Dnp)-D-Arg-NH2. This peptide was derived from a consensus substrate sequence for APs [Lys-Pro-Ile-Gln-Phe*Nph-Arg-Leu(Nph: nitro-phenylalanine)] described elsewhere (Beyer et al., 2005). The fluorescent substrate was initially designed and tested for utility against other well-characterized APs, Cathepsin D and E (Yasuda et al., 1999). Both boPAG-2 and -12 exhibited activity toward this substrate (Table 1).</p><p>The optimal activity for PAG-2 and -12 was around pH 4 and 3.5 respectively (Figure 2). Such acidic pH optima are typical for most APs. Of the two candidate PAGs, boPAG-12 had greater activity at pH 2.5 compared to boPAG-2. When comparing their kinetic parameters from progress curves, it was apparent that boPAG-2 is almost four-fold more active than boPAG-12 at their pH optima, despite that they are closely related and have similar affinity for the cathepsin D/E substrate. It was determined that boPAG-2 had a catalytic turnover rate more than 3 times greater than boPAG-12 under identical experimental conditions. In addition, the affinity of boPAG-2 for pepstatin A was also more than an order of magnitude greater than boPAG-12. Finally, boPAG-2 and boPAG-12 differed from one another in regard to the increase in activity observed upon a short pre-incubation at low pH prior to performing kinetic measurements (Figure 4). In most aspartic peptidases, acidification of the protein permits an intra-or intermolecular cleavage of the propeptide (Hazel et al., 1992; Koelsch et al., 1994; Richter et al., 1998; Wittlin et al., 1999; Barrett et al., 2004). The loss of the propeptide is thought to permit more ready access of substrate to the active site and also produces an increased mobility by SDS-PAGE due to the remove of the N-terminal peptide (40–50 amino acids, depending on the protein). In the pre-incubation experiments, boPAG-2 exhibited characteristics similar to other aspartic peptidases in that an increase in activity was observed after low pH exposure. Bovine PAG-12, in contrast, exhibited little increase in activity. After extensive incubation periods at low pH it was found that boPAG-12 exhibited no shifts in mobility (i.e. no cleavage of the propeptide) by SDS-PAGE (data not shown). Bovine PAG-2 did appear to process itself after incubation periods, but the processing did not involve the loss of the N-terminus and, instead, seemed to consist of non-specific cleavage during the incubation period (data not shown). One interpretation of these observations are that boPAG-2 and boPAG-12, while peptidases, are not able to remove their own propeptides and may require pro-protein converting peptidases to perform this function.</p><p>The experiments performed with the synthetic 25-x FRET substrate libraries suggested that boPAG-2 and -12 may be relatively selective in regard to the length or composition of the substrate required for detectable activity. Both porcine pepsin and bovine cathepsin D exhibited activity toward these libraries in a manner consistent with what is already known about their substrate specificities (Abad-Zapatero et al., 1990; Scarborough and Dunn, 1994; Barrett et al., 2004). However, both boPAG-2 and -12, when used at the same molar concentration, failed to produce significant product liberation from these libraries(Table 2). This lack of activity may be reminiscent of studies performed with furin, which is fastidious when it comes to substrate sequences required for optimal activity. In those experiments, furin failed to show activity against these same substrate libraries (information provided by Peptides International). The lack of activity against these libraries could be explained if the library failed to provide peptide sequences of the appropriate length or composition for optimal activity with both PAG-2 and 12. However, it is worth mentioning that PAG-12 did produce at least some activity when compared to PAG-2, which suggests that PAG-12 might be more flexible in regard to its substrate requirements.</p><p>The fact that bovine PAG-2 and -12 have distinct pH profiles and inhibitor affinities, suggested that there are likely to be differences in residues that are known to interact with the substrate to potentially stabilize it in the substrate binding cleft. In fact, such differences were revealed from comparative modeling of these proteins based on the crystal coordinates of human pepsin (Figure 6). For example Thr12 in boPAG-2 that contributes to the S4 pocket, is substituted by Met12 in boPAG-12. Such a change may alter the preference for a hydrophobic amino acid in the substrate occupying this sub pocket. Similarly, Leu115 contributes to the S3 and S1 sites in boPAG-2, while the somewhat smaller Ala115 shares the comparable position in boPAG-12. Another major change was that of Ser77 in boPAG-2 by Pro77 in boPAG-12, this position is of crucial importance since, it forms part of the hairpin 'flap' structure. The flap is a conserved fold in the mature molecule that arches over the substrate binding cleft and interacts and stabilizes the substrate within the substrate binding cleft (Hartsuck et al., 1992; Guruprasad et al., 1996; Richter et al., 1998; Dunn, 2002). It makes contact with numerous residues in the substrate. Those residues at position 77 contribute to the S2 and S1 pockets. The replacement of Ser77 in boPAG-2 with Pro77 in boPAG-12 marks another deviation between the two PAGs that might alter the dynamics of substrate stabilization within the substrate binding cleft. Another change within the flap was the substitution of the bulky hydrophobic aromatic amino acid Phe74 in boPAG-2 with the small polar Ser74. These residues contribute to the S3′ pocket. Collectively, these differences, as well as those described in the Results section, suggest the following speculations on potential specificity differences between boPAG-2 and -12: BoPAG-12 might be able to accommodate a bulkier side chain at P3 and/or P1 of putative substrates since the Leu115 of boPAG-2 fills these pockets more than does the Ala115 of boPAG-12. BoPAG-12 would be predicted to accommodate a bulkier or more polar side chain at P3′ since it presents Ser74 (instead of Phe74) at S3′.</p><p>Although the two PAGs investigated in these experiments are active peptidases, there was little indication that they act as general peptidases, like pepsin. These data suggested that they may have somewhat refined roles in cleaving protein substrates, either within the trophoblast itself or at the fetal-maternal interface. The PAGs are secreted glycoproteins, hence they are exposed to acidic environments (potentially as low as pH 5.2) within the secretory pathway (Paroutis et al., 2004). The PAGs might have a major role within the secretory pathway, where they might be activated by an unidentified peptidase or by auto-activation (Davies, 1990; Dunn, 2002). These activated PAGs might then participate in proteolytic processing of other proteins within the secretory vesicles before they exit the cell. Alternatively, they might be diverted towards the endo-lysosomal pathway where they could function by digestion of endocytosed agonist-bound receptors or proteins taken up from uterine secretions (uterotroph). Since both the identified PAGs seem to be somewhat selective in regard to substrate preference, it may be safe to hypothesize that the ancient PAGs do not function as general degradative enzymes like cathepsin D and E, which are lysosomal and endosomal peptidases, respectively (Godbold et al., 1998; Ishidoh and Kominami, 2002). Bovine PAG-2/-12 are constitutively secreted and accumulate at the maternal-fetal interface (Wooding et al., 2005). Therefore, they might play roles as 'sheddases' to activate latent growth factors or as peptidases involved in inactivating growth factor binding proteins or other proteins within the pericellular microenvironment (Munger et al., 1998; Rifkin et al., 1999). Some data do exist to indicate that this environment may be slightly acidic (pH 6.5) (Punturieri et al., 2000). In either situation, such proteolytic conversion of trophic factors may have an important role in promoting trophoblast/placentomal growth and differentiation in the ruminant placenta (Ko et al., 1991; Mathialagan and Roberts, 1994; Grundker and Kirchner, 1996; Yelich et al., 1997; Tanaka et al., 1998; Rifkin et al., 1999; Spencer and Bazer, 2004). Future efforts will be directed toward exploring these varied possibilities.</p><!><p>The BD Baculogold™ Baculovirus insect cell expression system (BD Biosciences Pharmingen, San Diego, CA) was used to express the recombinant proteins. BoPAG-2 was cloned into the pVL92 transfer vector by using the following oligonucleotides, sense: 5′ GAC TGA GCGGCCGCATGGATTACAAGGACGATGACGATAAGATAGTCATTTTGCCTCTA 3′ and antisense: 5′ GTCAGTC AGAGTCAGAGTCATGACTAGAGTCTAGATGACTATTACACTGCCGGAGCCAG 3′. Bovine PAG-12 was cloned into the pACMP3 transfer vector with the following oligonucleotides, sense: 5′ GACTCTAGAATGGATTACAAGGACGATGACGATAAG ATAGTCATTTTGCCTCTA 3′ and antisense: 5′GATCTATGATCTCAGTACT GCGGCCGCTCACTATTACACCTGTGCCAGGCCAAT 3′. The recombinant proteins were expressed as fusion proteins with a FLAG-tag in the N-terminus of the protein. A sequence encoding FLAG peptide (DYKDDDDK), shown as regular bold in the sense oligonucleotide, was engineered into the sequence so that this peptide would be incorporated into the N-terminus of the proforms of both PAG-2 and -12. Sequences encoding restriction enzymes (bold italicized) Not-1 and Bgl-2 (New England Biolabs, MA USA) were also engineered into the sense and antisense oligonucleotides to permit directional cloning into the corresponding transfer plasmids. Once the integrity and frame of the sequences in the transfer vectors was verified by sequencing, the vectors were transfected into Sf-9 cells along with BD Baculogold linearized Baculovirus DNA by using the BD Baculogold transfection kit according to the manufacturer's recommendations. Following transfection, the viruses were extracted, amplified and used to infect Sf-9 cells to generate recombinant proteins as described elsewhere (O'Reilly et al., 1992). Infected cells were harvested, chilled on ice, centrifuged at 600 g for 5 min at 4 °C followed by two wash cycles under similar conditions with cold 1x PBS (2.68 mM KCl, 1.47 mM KH2PO4, 136.89 mM NaCl and 8.10 mM Na2HPO4, pH 7.2). The final cell pellet was stored at −80 °C until use.</p><!><p>For purification of the recombinant protein, the corresponding frozen pellets were lysed on ice with I-Per insect cell protein extraction reagent (Pierce, IL, USA). A standard cocktail of protease inhibitors, which included 0.4 mM Pefabloc SC-AEBSF (Roche Applied Science), 5 μg/mL aprotinin, 10 μM E-64, 1 mM EDTA along with 1 mM DTT, (Sigma, MO, USA) was supplemented to the lysis buffer just before use. Following mixing and incubation with lysis buffer for at least 15 min on ice, the lysate was cleared by centrifugation at 15,000 × g for 30 min and dialyzed overnight in a 30 kD MWCO dialysis tubing in a buffer containing 20 mM Tris-HCl, pH 7.4, 250 mM NaCl at 4 °C. All purification procedures were performed in a refrigerated room at 4–6 °C. The dialysed lysate was fractionated on a Sephadex-200 size exclusion column (1.5 cm × 106 cm) equilibrated in 20 mM Tris-HCl, pH 7.4, 150 mM NaCl. All fractions that were determined to have the FLAG peptide present (by dot-blot with an anti-FLAG M2 antibody) were pooled and subsequently affinity purified by using anti-FLAG M2 agarose (Sigma, MO, USA). For affinity chromatography, the matrix was equilibrated with TBS buffer (50 mM Tris-HCl, 150 mM NaCl, pH 7.4), following which, the FLAG-containing protein samples were loaded twice onto the column by gravitational flow at approximately 0.2 mL/min. The column was then subjected to subsequent washes with 20 column volumes of wash buffer (20 mM Tris-HCl, 150 mM NaCl, pH 7.4), 20 column volumes of high salt buffer (20 mM Tris-HCl, 500 mM NaCl, pH 7.4) and finally 20 column volumes of high salt buffer supplemented with 0.1% Tween. The anti-FLAG M2 agarose was re-equilibrated with 10 column volumes of wash buffer to remove residual detergent, followed by 10 column volumes of pre-elution buffer (10 mM phosphate buffer, pH 7.2). The column was then eluted with 5 column volumes of 50 mM phosphate buffer, 2 M MgCl2, pH 7.2. The protein sample was desalted by dialysis in 20 mM Tris-HCl, 250 mM NaCl, pH 8.0 and concentrated by using an Amicon Ultra-15 with Ultra cell-30 membrane (Millipore, MA, USA). The concentrated protein samples were protected with the addition of the inhibitor cocktail and glycerol to 10% (v/v) at 4°C. In most cases, the protein samples were immediately used in the assays. For long term storage, the protein sample was stored in 50% glycerol at −80 °C.</p><!><p>For the Western blot experiments the recombinant proteins were electrophoretically transferred onto Immobilon PVDF-membrane (Millipore, MA, USA). The membranes were washed once with an excess of 1x TBST (10 mM Tris, 150 mM NaCl, 0.05% Tween, pH 7.5) and blocked in blocking buffer that consisted of 3% bovine serum albumin and 3% non fat dry milk (Sigma, St. Louis, MO, USA) in 1x TBST. The blots were subsequently incubated with either a 1:1000 dilution of monoclonal anti-FLAG antibody (Sigma) or 1:2000 polyclonal anti-boPAG-2 anti-serum (to further confirm the presence of boPAG-2) in blocking buffer. The blots were then washed and incubated with 1:2000 dilution of anti-mouse (for anti-flag) or anti-rabbit IgG (for boPAG-2 antiserum) conjugated to alkaline phosphatase for 45 min (Promega, WI, USA). The blots were finally washed and stained with a mixture of NBT (Nitro-Blue Tetrazolium chloride) and BCIP (5-bromo-4-chloro-3′-indolyphosphate p-toluidine Salt) according to the manufacturer's instructions (Promega, WI, USA). The production and characterization of the boPAG-2 antiserum has been described previously (Wooding et al., 2005)</p><!><p>To estimate the optimal pH for each PAG, the recombinant PAGs were incubated in various buffers. All the pH activity experiments were conducted at 35 °C. A synthetic FRET (fluorescence resonance energy transfer) cathepsin D/E substrate, MOCAC-Gly-Lys-Pro-Ile-Leu-Phe-Phe-Arg-Leu-Lys(Dnp)-D-Arg-NH2 [MOCAc: (7-methoxycoumarin-4-yl) acetyl; Dnp: 2,4-dinitrophenyl] (Peptides International, KY, USA) was used to investigate the activity of bovine PAG-2 and -12. The substrate was dissolved in 10% DMSO to a final concentration of 200 μM, aliquoted into 10 μL samples, and stored in the dark at −80 °C until use. Equal amounts of protein sample(approximately 500 ng in 20 μL) were incubated with 20 μM of substrate. For determination of optimal pH, the following buffers were used: 0.1 M glycine-HCl buffer for pH 2.5, 3.0 and 3.5; 0.1 M sodium citrate-citric acid buffer for pH 4 and 4.5; 0.1 M sodium acetate-acetic acid for 5 and 5.5; 0.1 M Bis-tris-HCl buffer for pH 6 and 6.5; 0.1 M HEPES for pH 7; and 0.1 M Tris-HCl for pH 7.5 and 8.0 buffers. Each reaction solution also contained NaCl at a final concentration of 0.1 M. The final volume of the reaction was maintained at 100 μL. The reaction was performed for 20 min and was terminated by addition of 900 μL of 5% trichloroacetic acid (TCA), as described previously (Yasuda et al., 1999). The resultant mixture was further diluted to 2 mL with 5% TCA and the resultant fluorescence in the mixture was read in a PC1™ Photon counting spectofluorimeter at 328 nm (excitation) and 393 nm (emission) wavelengths as described previously (Yasuda et al., 1999). All the experiments were set up in duplicate. The results from duplicate reads and from two successive experiments were used to compile the data.</p><!><p>A parallel purification experiment was performed to confirm that the activity being measured was due to recombinant PAGs and not from residual insect cell-or baculovirus-derived peptidases that might have been retained after isolation of the PAGs. The lysates used for this experiment were obtained from infected insect cells expressing boPAG-2, boPAG-20 and, as a negative control, equine-thyroid stimulating protein-beta (eTSHβ). The boPAG-20 and eTSHβ were cloned and engineered to contain a FLAG peptide in a similar approach to that used to express boPAG-2 and boPAG-12. An additional control for this experiment consisted of lysed insect cells that had been infected with wild-type baculovirus (ACNPV). One mL of each lysate was pre-cleared by incubating with 100 μL of protein G (50%) bead slurry (BioRad, USA) on a rocker shaker at 4°C for half hour. The protein G beads were removed by centrifugation at 14,000G at 4°C for 10 min. Equal amounts (500μg) of insect cell lysates were mixed with 5 μg of anti-FLAG antibody and rotated at 4°C for 2 h. Protein G agarose matrix was added at 100 μL (50% slurry) and the resulting mixture was incubated overnight. The matrix was isolated by centrifugation and washed with 20 mM Tris-HCl, 150 mM NaCl, pH 7.4, 0.1% Tween buffer three times. After the final wash, the matrix was incubated with 0.1 M glycine-HCl buffer pH 3.5 containing 0.1M NaCl and 20 μM fluorescent substrate (described above) in a final volume of 100 μL at 40°C for 1 h. The resulting solution was centrifuged at 14,000 G for 10 min to precipitate the protein G matrix and 50 μL of supernatant was mixed with equal amounts of 10% TCA and the resultant fluorescence was measured by using a Fluroscan plate reader (Ascent, USA).</p><!><p>A feature of APs is the ability to auto-activate. In most cases, this phenomenon involves removal of the propeptide by either an intra-or intermolecular mechanism (Hazel et al., 1992; Koelsch et al., 1994; Richter et al., 1998; Wittlin et al., 1999; Barrett et al., 2004). After defining activation procedures for boPAG-2 (data not shown), an optimal scheme was identified that involved adding one-third volume of 0.25 M glycine-HCl, pH 3.5 (the activation buffer) to the sample of boPAG-2 (stored in 20 mM Tris-HCl, pH 8.0, 250 mM NaCl buffer), mixing and incubating for 1 min at RT. A similar activation procedure for boPAG-12 entailed adding one-third volume of 0.25 M glycine-HCl, pH 3.0 to the sample of boPAG-12, mixing and incubating for 1 min at RT. To estimate the profile of the 'activated' protein as a function of pH and to compare it to the activity of the unactivated protein, approximately 620 ng of boPAG-2 or 540 ng of boPAG-12 total protein (either the unactivated or activated form) was incubated with 20 μM of substrate in buffers of varying pH that were described above. The samples were placed in a 96-well Costar black microtitre plate (Corning, USA) and subsequently incubated at 37°C for 10 min in a Synergy-HT Fluorescent plate reader (Biotek, USA). Following the addition of substrate, the kinetic reads were obtained for the first 10 min of the reaction and the calculated initial rates were displayed as relative fluorescent units (rfu)/min. The experiment was conducted with samples in triplicate to estimate the experimental noise. The same experiment was replicated multiple times and data from a representative experiment was shown.</p><!><p>The steady-state enzyme kinetic parameters for boPAG-2 and -12 were estimated by fitting to progress curves as described elsewhere (Palmier and Van Doren, 2007). The concentration of total protein in the purified samples was estimated by BCA protein assay (Pierce, Rockford, IL). Accurate estimates of specific activity of PAG preparations relied upon accurate active enzyme concentrations measured by active site titrations with the tight-binding inhibitor of aspartic peptidases known as pepstatin A, fitted as described (Knight, 1995; Copeland, 2000). The Ki for PAGs was estimated by fitting the data to the following equation by a procedure detailed in the reference (Knight, 1995): ν=(νo/2Eo){E0−I−Ki+[(I+Ki−E0)2+4KiE0]0.5} where Eo and Io are the initial concentration of enzyme and inhibitor respectively, νo, is the initial rate of the reaction without inhibitor and 'ν' is the observed velocity.</p><p>Accurate measurement of kcat/Km was obtained from single progress curves collected under pseudo first-order conditions of [S] ≪ Km and [S] ≫ [E], as illustrated by the slower curves in Fig. 5, fitted to the following equation: y=(Fmax∗(1−exp(−(kcat/KM)∗Et∗x)))+B where y is the relative fluorescence intensity, Fmax is the maximum change in fluorescence intensity during the reaction, Et is the enzyme active site concentration in the assay, x is time in seconds, and B is a Y axis offset correction (Orsi and Tipton, 1979; Niedzwiecki et al., 1992; Neumann et al., 2004). Steady-state kinetics parameters, kcat and Km, were determined from two progress curves, one of which was under pseudo first-order conditions of [S] ≪ Km and [S] ≫ [E]. Origin Pro 7.5 (Microcal) was used to obtain kcat and Km from global, nonlinear fitting of the pairs of progress curves as described (Palmier and Van Doren, 2007). All kinetic experiments were performed in 0.1 M sodium citrate-citric acid, 0.1 M NaCl (pH 4.0) buffer at 37 °C with the fluorescent (FRET) Cathepsin D/E substrate in 3×3 mm cuvettes to reduce the path length and inner filter effect (Matayoshi et al., 1990; Lakowicz, 1999; Y. Liu et al., 1999). The data were obtained by monitoring product development in a SLM Aminco fluorometer (model 8100) with PC1™ Photon counting upgrade (ISS Inc., IL, USA).</p><!><p>To gain insight into differences in substrate preferences, if any, between full-length boPAG-2 and -12, a synthetic 25-x FRET substrate library (Peptides International, KY, USA) was employed. The library consisted of 5 different amino acids at position Y (Pro, Tyr, Lys, Ile and Asp) and Z (Phe, Ala, Val, Glu and Arg) for each residue at position X [D-A2pr(Nma)-Gly-(Zaa)5-(Yaa)5-Xaa-Ala-Phe-Pro-Lys(Dnp)-D-Arg-D-Arg. There are 19 different primary substrate libraries with each natural amino acid at position X with the exception of cysteine. Therefore, the libraries represented 475 distinct peptides. The library is useful for delineating optimal substrate sequence for a given peptidase, based on initial rates for the 19 different primary libraries by the peptidase, followed by more refined analysis (Tanskul et al., 2003; Hiraga et al., 2005; Oda et al., 2005; Namwong et al., 2006). Pre-extracted native preparations of cathepsin D (bovine spleen) and pepsin (porcine stomach) were obtained (Sigma) and used in the assay. All the preparations were active site titrated, as for the kcat and KM determinations. Equal amounts (5 nM) of active enzyme and 20 μM of substrate were used in the assay. The assay was performed in duplicate under conditions described for the kinetic analysis [0.1 M sodium citrate -citric acid, 0.1 M NaCl, pH 4.0 at 37°C]. The initial velocities were obtained against the primary substrate libraries for first 10 min of the reaction using the fluorescent plate reader.</p><!><p>Predicted 3-D models for boPAGs-2 and -12 were constructed by using SWISS-MODEL (http://swissmodel.expasy.org//SWISS-MODEL.html) in automated mode (Guex and Peitsch, 1997; Schwede et al., 2003; Arnold et al., 2006). The crystal structure of human pepsin conjugated with pepstatin A (1psn.pdb) was used as a template to generate the model (Fujinaga et al., 1995; Westbrook et al., 2003). Multiple sequence alignments of boPAGs -2, -12, boPAG-1 and human pepsin were generated by using CLUSTALW (http://align.genome.jp). The template structure of human pepsin shares sequence identity of 50.6% with boPAG-2 and 53.4% with boPAG-12. Specific residues predicted to constitute the substrate binding pockets for boPAG-1 were reported elsewhere (Guruprasad et al., 1996). The residues contributing to specific substrate binding pockets in boPAG-2 and -12 were identified from the alignments with boPAG-1. The structural views of the homology models in Figure 6 were generated by using Pymol (DeLano).</p><!><p>Panel (A) shows Coomassie blue staining of fractions from a preparation of recombinant boPAG-2. Lane-1 contains the total proteins from the lysed insect cell pellet. Lane-2 contains the flow-through of lysate from the anti-FLAG column. Lanes 3–5 contain elution fractions of recombinant boPAG-2 in order of emergence from the column. Panels (B) and (C) show Western blot images of boPAG-2-containing fractions transferred from identically loaded gels and immuno-blotted with anti-PAG-2 polyclonal and anti-FLAG monoclonal antibodies, respectively. (D) Coomassie blue staining of fractions from a preparation of recombinant boPAG-12. Lane-1 contains total proteins from the insect cell lysate. Lane-2 contains the flow-through from the anti-FLAG column. Lanes 3–5 contain fractions in order of their elution from the column. (E) Western blot image of boPAG-12-containing fractions immunoblotted and detected with anti-FLAG monoclonal antibody.</p><!><p>Each activity point is normalized by that protease's maximum activity (relative fluorescence units) at its pH optimum. The error bars represent standard deviation in results obtained from duplicate reads from two separate experiments.</p><!><p>Anti-FLAG antibody was used as a capturing antibody and protein G matrix was used as a solid support. Immunoprecipitated proteins were reacted with the cathepsin D/E FRET substrate for one hour in pH 3.5 buffer. Liberated product was measured on a fluorescent plate reader.</p><!><p>The pH-dependence of the activity of recombinant boPAG-2 (A) and boPAG-12 (B) before activation (squares) and following activation (triangles). The initial velocities (in relative fluorescent units/min) of each PAG against the fluorescent FRET-cathepsin D/E substrate at each pH were estimated from kinetic reads for 10 min.</p><!><p>Progress curves global fitted with kcat and Km for digestion of cathepsin D/E substrate by recombinant boPAG-2 (A) and boPAG-12 (B) at pH 4.0 and 37°C. In (A), the FRET substrate and boPAG-2 concentrations for the upper and lower series are 1 μM substrate and 5 nM boPAG-2 enzyme and 0.2 μM substrate and 2.5 nM boPAG-2 enzyme, respectively. In (B), the substrate and boPAG-12 concentrations are 0.5 μM substrate and 39 nM boPAG-12 enzyme and 0.5 μM substrate and 17 nM boPAG-12 enzyme. The lines representing each progess curve are indicated in the panels.</p><!><p>Residues that distinguish the active sites of boPAG-2 (A) and boPAG-12 (B) and pepsin (not shown). The homology models were constructed by using the crystal structure of mature human pepsin bound to pepstatin (PDB code 1PSN) as template by using the SWISS-MODEL server (http://swissmodel.expasy.org//SWISS-MODEL.html) (Guex and Peitsch, 1997; Schwede et al., 2003; Arnold et al., 2006). The backbone ribbon progresses through the colors of the rainbow from blue at the N-terminus to red at the C-terminus. The conserved aspartate side chains critical for peptidase activity are red. Side chains of the active site cleft differing between boPAG-2 and -12 are plotted with standard atom colors of blue for nitrogen, red for oxygen, and green for carbon. The "flap" is light blue.</p><!><p>Steady-state enzyme kinetic parameters for bovine PAG-2 and PAG-12.</p><p>Initial velocities (RFU/min) obtained from the proteolysis of 25-X FRET substrate libraries.</p>

PubMed Author Manuscript

Silica nanoparticles as a delivery system for nucleic acid-based reagents

The transport of nucleic acid-based reagents is predicated upon developing structurally stable delivery systems that can preferentially bind and protect DNA and RNA, and release their cargo upon reaching their designated sites. Recent advancements in tailoring the size, shape, and external surface functionalization of silica materials have led to increased biocompatibility and efficiency of delivery. In this Feature Article, we highlight recent research progress in the use of silica nanoparticles as a delivery vehicle for nucleic acid-based reagents.

silica_nanoparticles_as_a_delivery_system_for_nucleic_acid-based_reagents

Introduction<!>Silica nanoparticles with external surface modifications are capable of transporting DNA to affect gene expression<!>DNA delivery has been accomplished in plants using silica nanoparticles<!>Adsorption of DNA into very large pores of silica nanoparticles<!>Organic/inorganic silica hybrid nanoparticles in DNA transport<!>Silica nanoparticles can transport antisense oligonucleotides (ASOs)<!>Silica nanoparticles may be useful for transporting small interfering RNA (siRNA)<!>Use of silica nanoparticles to enhance transfection efficiency<!>Biocompatibility of silica nanoparticles<!>Conclusions and future directions<!>

<p>Silica has proven to be a valuable material for biomedical research. One application has been to use silica nanoparticles as drug delivery vehicles. In particular, mesoporous silica nanoparticles (MSNs) formed by polymerizing silica in the presence of surfactants have many advantages for intracellular delivery, such as large surface area, tunable pore sizes and volumes, and encapsulation of drugs, proteins and biogenic molecules. Moreover, they can be tailored with a variety of surface functional groups to increase biocompatibility and generate efficient and versatile agents for molecular delivery.1 For controlled delivery systems, it has been shown that silica is able to store and gradually release therapeutics such as antibiotics2–4 and other clinically-relevant compounds.5–7 In our own lab, we have incorporated a hydrophobic anticancer drug, camptothecin (CPT), into the pores of fluorescent mesoporous silica nanoparticles (FMSNs) and delivered the drug into a variety of human cancer cells to induce cell death.5 We have suggested that MSNs might be used as a vehicle to overcome the insolubility problem of many anticancer drugs.5 Hybrids of silica and other organic/inorganic materials have also been successfully used in various applications, including enzyme encapsulation,8 magnetic bioseparations,9,10 magnetic labeling and imaging,11,12 and cell labeling and tracking.7,13,14</p><p>Silica materials themselves have therapeutic potential based on their intrinsic properties. For example, in Saos-2 cells that are similar to osteoblasts, bio-silica matrices stimulate increased hydroxyapatite deposition. This indicates that synthesized bio-silica is a promising route for tooth reconstruction in vivo.15 Similarly, MSNs affect gene expression in human mesenchymal stem cells by inducing transient osteogenic signals.16 In addition, silica nanoparticles have been used to track cells. Nanoparticles conjugated to quantum dots and functionalized with amino and phosphonate groups have been used as probes to label T-lymphocytes. By conjugating them to luminophores, MSNs have been tracked and suggested to be effective delivery vehicles.17</p><p>Recent studies opened up the possibility of using MSNs as a delivery vehicle for nucleic acid-based reagents. This is important because it provides a new way to treat genetic diseases, illnesses caused by abnormalities in genes. Although they are too numerous to list, some well known genetic disorders are hemophilia, Huntington's disease, asthma, and cancer. Some diseases, such as cancer, are caused by both a genetic predisposition and environmental factors. A potential approach to the treatment of genetic disorders is gene therapy whereby a working gene replaces its dysfunctional counterpart, thereby curing inherited and acquired diseases. This enables the body to produce the appropriate gene products to eliminate the underlying cause of the disease. The treatment potential of nanoparticle-mediated gene therapy has been of particular interestin the field of nanomedicine and cancer.18 Current treatment methods, such as chemotherapy, rely on the use of cytotoxic drugs; however, such therapy has limited efficacy due to the use of suboptimal dosages of those therapeutic agents in attempt to prevent both acute and chronic, unwanted side-effects. Gene therapy is a welcome alternative treatment because it only targets the defects that gives rise to major symptoms, and therefore avoids complications associated with chemotherapy.</p><p>In order to bind negatively charged nucleic acids and improve uptake, silica materials are typically modified with positively charged organic adjuncts, such as poly-L-lysine (PLL)19 or polyethylenimine (PEI).20–22 PEI confers increased gene delivery to nanoparticles due to its 'proton sponge effect' allowing endosomal escape.21,23,24 To be an effective delivery system, silica nanocarriers must be biocompatible, possess high affinity for their particular payload, sequester their payload from the outer environment, and avoid premature release of their contents. DNA and RNA would decompose in the highly acidic environment of the stomach if the carrier could not offer the necessary protection. As a practical therapy, the nucleic acid carrier must not degrade or leak until it reaches its intended target, ensuring the release of high local concentrations of the cargo.25</p><p>In this Feature Article, we discuss various attempts to deliver nucleic acid-based reagents using silica nanoparticles. They hold the promise to encapsulate and protect a payload of therapeutic compounds, transport them to specific locations in the body, and release them in response to either external or cellular stimuli. Thus, biocompatible silica nanoparticles represent a new solution to gene delivery.</p><!><p>In the past decade, viral-mediated delivery (infection) has been the primary method of introducing DNA into mammalian cells (Fig. 1(a)). However, due to growing concerns over the toxicity and immunogenicity of viral DNA delivery systems, DNA transport via non-viral routes has become more desirable and beneficial. For example, nucleic acid-based reagents were delivered through the use of polycation–DNA complexes by Wagner and colleagues.26 Such cationic adjuncts were applied to silica nanoparticles to bind, protect, and deliver DNA (Fig. 1(c)). Synthesized silica nanoparticles with covalently linked cationic external surface modifications were produced through a variety of methods. Kneuer et al. produced them by modification of commercially available silica particles (IPAST) or in-house synthesized silica particles with either N-(2-aminoethyl)-3-aminopropyltrimethoxysilane or N-(6-aminohexyl)-3-aminopropyltrimethoxysilane.27 These nanoparticles were sized between 10 and 100 nm and displayed surface charge potentials from +7 to +31 mV at pH 7.4.27 He et al. synthesized amino-modified silica nanoparticles (45 ± 4 nm) by using the synchronous hydrolysis of tetraethoxysilane and N-(β-aminoethyl)-γ-aminopropyltriethoxysilane in water-in-oil micro-emulsion.28 Tan et al. utilized a similar method to develop uniform core/shell nanoparticles (5–400 nm), consisting of a silica layer coating and magnetite core.29 These particles are surface-modified externally so that disulfide coupling chemistry can be used for immobilization of oligonucleotides onto silica nanoparticles.29 All three of these nanoparticles possessed the ability to electrostatically bind, condense, and protect plasmid DNA from cleavage.27–29</p><p>Another study has shown that colloidal silica particles with covalently attached cationic external surface modifications with aminoalkysilanes could transfect plasmid DNA in vitro successfully.30 β-Galactosidase was chosen as the genetic payload, and its activity following delivery in Cos-1 cells was used to measure transfection efficiency. The use of silica–silane–DNA nanoplexes resulted in elevated expression of their DNA cargo.30 When chloroquine was used in conjunction with DNA nanoplexes, transfection rates increased.30 This may be due to induction of endosomolysis or delaying endosomal decomposition.30,31 Mechanistically, silica nanoparticles concentrate DNA at the surface of cells. An elevated local concentration of DNA allows efficient DNA uptake by an endosomal–lysosomal route, as confirmed previously by temperature-dependent transfection efficiency.32 This suggests that nanocomplexes are internalized by endocytosis and routed to the endosomal/lysosomal compartment (Fig. 2).</p><p>A study demonstrated the importance of temperature- and energy-dependence in the uptake of FMSNs into human cancer cells.33 Lower temperatures significantly impeded cellular uptake of FMSNs in PANC-1 cells, thus raising the possibility that FMSNs enter cells in an energy-dependent manner.33 Metabolic inhibitors, including sodium azide (which depletes intracellular ATP), sucrose (which suppresses coated pit function) and bafilomycin A (which inhibits v-ATPase function), suppressed the uptake of FMSNs into PANC-1 cells.33 This suggests that FMSN uptake is regulated by energy-dependent, clathrin-mediated endocytosis that depends upon a V-ATPase-dependent transport mechanism.33 The endocytosis inhibitor, nocadazole, also significantly disrupted FMSN uptake, suggesting that a dynamic microtubule network is also necessary.33</p><p>Polyamidoamine (PAMAM) dendrimer-capped mesoporous silica nanospheres have also successfully served as nonviral gene transfection agents.34 PAMAMs were covalently attached to the external surface of the nanospheres, and then complexed with plasmid DNA (pEGFP-C1).34 Agarose gel electrophoresis confirmed complexation between nanospheres and pEGFP-C1 and showed that the plasmid DNA is protected against enzymatic cleavage.34 Significant GFP expression was observed in fluorescence confocal micrographs of human epithelial carcinoma cells (HeLa) treated with pEGFP-C1-carrying nanoplexes.34 Thus, although there is a strong electrostatic interaction binding negatively charged DNA to positively charged PAMAM nanospheres, it is somehow overcome in mammalian cells to allow delivery and transgene expression. However, the mechanism that governs this release remains unclear.</p><!><p>Research has been done to deliver DNA to plants through the use of nanoparticles. MSNs used for animal systems are not applicable to a plant system because of the plant cell wall. However, Torney et al. circumvented this problem by utilizing the gene gun system in order to trigger gene expression in plant cells and intact leaves.35 Externally surface-modified MSNs with approximate diameter of 100–200 nm were used to bind plasmid DNA. The pores of these fluorescein-doped MSNs were capped by surface-functionalized gold nanoparticles (10–15 nm in size) which not only acted as a biocompatible capping agent,36 but more importantly added weight to each individual MSN to increase the density of the resulting complex material. This increase in MSN density improved transformation efficiency and the appearance of MSNs inside plant cells.35 The advantage of using MSNs with the gene gun is that both the DNA and small effector molecules can be delivered at the same time. Here, the chemical inducer that activates transgene expression (β-oestradiol) was contained inside the gold-capped structure. After bombardment into the plant cell, the effector molecules were released from the gold-capped structure by incubating the plant tissues on media containing dithiothreitol, a chemical that reduces the disulfide bonds that attach the gold caps to the MSNs. The encapsulated β-oestradiol was subsequently released in a controlled manner to trigger the expression of co-delivered GFP transgene in the cell.35,37 Further customization of pore size and external surface functionalization may also lead to improved targeting and delivery of multiple compounds as well.</p><!><p>It has been reported that MCM-41-type MSNs can be internalized in vitro by animal and plant cells and show minimal signs of cytotoxicity.25,38 However, the largest pore diameter available up until now has been 6 nm,39 which limits MSN's use as a transporter for large molecules such as proteins and nucleic acids. The advantage of widening the size of the pores of functionalized MSNs is a route to enhance the performance of the adsorption process.40 A study was done using acid-prepared mesoporous silica (APMS) nanoparticles, which have a distinct spherical shape. Mg2+, Ca2+, and Na+ promoted DNA adsorption onto external silica surfaces by mediating the electrostatic repulsion between the negatively charged external silica surface and DNA molecules. The synthesis of APMS is complete in less than two hours, and the particle size and pore diameter of APMS are easily controlled simply by altering a set of standard reaction conditions. The diameter of the pore was found to affect the amount of DNA that could be loaded into APMS. Materials with pores greater than 5.4 nm were found to be more favorable toward DNA adsorption, as the molecules could likely enter the pores without significant intermolecular interactions. Likewise, other authors have reported on the benefit of widening the pore size of mesoporous materials for bioimmobilization processes.39–43</p><p>Gao et al. synthesized MSNs with controlled diameter (~70–300 nm) and with very large, uniform regular pores of 20 nm (Fig. 1(d)) by a lower temperature (10 °C) synthetic method in the presence of a dual surfactant system.44 After external surface modification with aminopropyl groups, very high adsorption of DNA molecules took place. Using spectrophotometry, the authors learned that their MSN was able to adsorb more DNA per unit surface area than any other previously reported silica-based materials. Also, because of the large pore size, intermolecular interactions among DNA molecules and diffusion limitations were expected to be minimized. Finally, these large pore MSNs also conferred protection from enzymatic degradation as confirmed by agarose gel electrophoresis following exposure to restriction endonucleases.44</p><!><p>Organic/inorganic hybrid particles (Fig. 1(e)) are attractive as delivery vectors because they can be loaded with either hydrophilic or hydrophobic biomolecules, their preparation avoids the use of hydrophobic solvents such as cyclohexane, their external organic groups prevent particle precipitation in aqueous systems, and their external surfaces can modified with targeting molecules.14</p><p>ORMOSIL (organically modified silane) such as n-octyl-triethoxysilane has been found to aggregate in the form of normal micelles as well as reverse micelles in which the triethoxysilane moieties are hydrolyzed to form a hydrated silica network while the n-octyl groups are held together through hydrophobic interaction. These nanoparticles are spherical in shape with an average diameter of below 100 nm. Hybrids like ORMOSIL nanoparticles have the potential to overcome many limitations of their solely inorganic counterparts. The resulting micellar cores can be loaded with therapeutic compounds like drugs, proteins, and nucleic acids.45</p><p>Hydrated ORMOSIL nanoparticles based on the triethoxyvinysilane (VTES) precursor have been synthesized in the nonpolar core of dioctyl sodium sulfosuccinate (Aerosol-OT)/DMSO/water microemulsions. Hybrid amino-functionalized ORMOSIL nanoparticles have also been synthesized by a synchronous hydrolysis of VTES and 3-aminopropyltriethoxysilane (APTES). By varying the concentrations of Aerosol-OT and VTES, nanoparticles of various sizes (10–100 nm) were produced.14</p><p>External surface amino functionalization allows these silica nanoparticles to electrostatically bind to negatively charged DNA and protect it from enzymatic degradation as shown by agarose gel electrophoresis.14 Binding between DNA and the amino adjuncts on the external surface of these ORMOSIL nanoparticles was also confirmed by FRET analysis.14 It was postulated that as soon as the genetic material is released into the cytoplasm of the cell, it migrates to the nucleus.46 This was later confirmed by confocal microscopy with DNA labeled with EMA, a fluorescent dye, and loaded onto unlabeled hybrid nanoparticles. The fluorescent DNA was optically tracked as it was delivered into Cos-1 cells and subsequently to the nucleus.14 Finally, to confirm that the transfected DNA was still functional following transport, pEGFP was transfected into cells using hybrid nanoparticles and the resulting GFP fluorescence was confirmed by localized spectroscopy.14</p><p>Organic/silica hybrids as a non-viral vector for gene delivery have already been applied successfully to in vivo models. The same ORMOSIL nanoparticles, functionalized with amino groups and complexed with pEGFP, were used to counter neuropathy, stimulate compensatory mechanisms, and aid neurogenesis. The adult mammalian Central Nervous System (CNS) possesses limited potential to generate new neurons, making it vulnerable to long-term injury and a perfect candidate for gene therapy.47 However, the subventricular zone (SVZ) of the lateral ventricle (LZ) retains the capacity for neurogenesis.48 The SVZ contains a population of neural progenitor cells that can potentially give rise to neurons, astrocytes, and oligodendrocytes.49–51 Neural stem/progenitor cells (NSPCs) of the SVZ have been shown to proliferate once treated in vitro with exogenous FGF2. This suggests that stimulation of FGF receptors by exogenous FGF could influence the proliferation of the NSPC cells in vivo.52 In vitro, it has been found that nuclear accumulation of FGFR1 accompanies differentiation of the neural progenitor cells and is sufficient to induce cell differentiation.47,53 The following study aimed to determine if the same was true in an in vivo model.</p><p>The ORMOSIL nanoparticles complexed with pEGFP were injected into mice ventral midbrain and into the lateral ventricle. This allowed the authors to visualize and track the successfully transfected neuronal cells in the substantia nigra and areas surrounding the lateral ventricle using confocal fluorescent microscopy.54 Mice then received intraventricular injections of ORMOSIL/FGFR1 complexes followed by BrdUrd injection. BrdUrd injection was used in this study because it is likely to primarily tag the faster-proliferating progenitor cells, and thus serve as a marker for successful FGFR1 delivery by ORMOSIL nanoparticles. FGFR1 and BrdUrd immunostaining was found to be increased in the SVZ of mice transfected with FGFR1, indicating a change in the replication cycle of progenitor cells in the SVZ.54 Thus, organic/silica hybrid nanoparticles can be used to deliver FGFR1 into mammalian cells.</p><p>MSNs coupled to mannosylated PEI (MPS) have also proven to be effective non-viral transfection agents. Mannosylated PEI (MP) was synthesized by a thiourea linkage reaction between the isothiocyanate group of α-D-mannopyranosylphenyl-isothiocyanate and the primary amine group of PEI. The particle sizes of MPS/DNA complexes were analyzed by dynamic light scattering to investigate the degree of compaction with DNA. The sizes ranged from 60 to 130 nm, and the morphology was spherical with little aggregation.55</p><p>The purpose of MP functionalization was to target macrophage cells with mannose receptors and enhance transfection efficiency. These MPS were able to form complexes with DNA, protect against DNase I, and release DNA. Furthermore, the low cytotoxicity of MPS in Raw 264.7 macrophage cells and HeLa (a human cervical cancer cell line) suggests that it is a safe gene vector. MPS also led to greater transfection efficiency in macrophage cells over HeLa cells, exhibiting mannose receptor-mediated gene delivery.55 Thus, MPS is another silica hybrid nanocontainer that can safely be used in gene therapy.</p><!><p>Antisense oligonucleotides (ASOs) are short nucleotide sequences of DNA that are reverse complements of the nucleotide sequence of their target mRNA. They inhibit gene expression at both mRNA and protein levels by Watson–Crick base-pairing, where the oligo single-stranded DNA binds to its complementary mRNA.56–58 Like siRNA, ASOs show great potential as a molecular tool and therapeutic agent against diseases with underlying genetic components, such as viral infections and cancer. However, like siRNA, they show poor intracellular uptake and stability.59,60 They also interfere with normal cellular function in a non-sequence-specific manner.61 These deficiencies limit their use as potential therapeutics.</p><p>To overcome these obstacles, PLL-modified silica nanoparticles (PMS-NP) have been used to bind and protect ASOs. PMS-NP (20 ± 2 nm) were prepared in a microemulsion system, using polyoxyethylene nonylphenyl ether/cyclohexane/ammonium hydroxide.62 The delivery of PMS-NP-ASO complexes was evaluated in human nasopharyngeal carcinoma cells (HNEI) and HeLa cells and the results were compared with free ASOs by fluorescence microscopy and flow cytometry.62 The specific blocking effects of antisense constructs designed against the proto-oncogene, c-myc, were examined by determining mRNA levels with reverse transcription-PCR (RT-PCR). Furthermore, HNEI and HeLa cells were treated with various concentrations of PMS-NP in the presence of serum-free media, after which serum-containing media was added. The results indicated that PMS-NP displayed significantly low cytotoxicity.62 Only at concentrations >500 μg/mL does PMS-NP show cytotoxic effects. Overall, PMS-NP complexes were able to bind, protect, and deliver ASOs to cells where they exerted ASO-specific gene inhibition.62</p><p>In a follow-up study, positively charged amino silica nanoparticles (NH2SiNPs) were also effective as ASO carriers.63 The NH2SiNPs were synthesized by a microemulsion method utilizing synchronous hydrolysis of tetraethyl orthosilicate (TEOS) and N-(β-aminoethyl)-γ-aminopropyltriethoxysilane (AEAPS). The NH2SiNPs with an average diameter of 25 nm could combine with ASOs to form a bioconjugate favorable for cellular uptake. This was visualized by using fluorescein isothiocyanate (FITC)-labeled ASOs and NH2SiNPs doped with rhodamine 6G isothiocyanate (RITC) as fluorescent signal detectors. Compared to liposomes, NH2SiNPs were reported to be more biocompatible and had almost no cytotoxicity at the concentrations required for efficient transfection. Further, they were able to protect ASOs from degradation by DNase I.63 MTT assays and western blot analysis showed that the NH2SiNPs greatly improve the inhibition efficiently of ASOs in HeLa and A549 cells.63</p><!><p>Gene inhibition through targeted delivery of sequence-specific siRNA is a promising method of gene therapy. siRNA, or short interfering RNA, is a class of 20–25 nucleotide-long double-stranded RNA molecules that are involved in the RNA interference (RNAi) pathway, where it interferes with the expression of a specific gene. RNAi is a form of post-transcriptional gene silencing in animals and plants, initiated by double-stranded RNA (dsRNA) that is homologous to the silenced gene. siRNAs are generated from the cleavage of longer dsRNAs by the action of ribonuclease III family enzymes. The siRNA duplexes specifically suppress expression of genes in mammalian cells.64</p><p>siRNA is expected to be a powerful tool to inhibit gene function because it is easily applicable to virtually any therapeutic target including intracellular and transcription factors. However, poor intracellular uptake, instability, and non-specific immune stimulation are obstacles associated with current methods of siRNA oligonucleotide delivery.65 Because nanoparticles have been shown to be effective DNA vectors, it is only logical that they be applied to siRNA and antisense therapeutics. At the time that this Feature Article was written, silica nanoparticles have not been used to transport siRNA or antisense oligonucleotides. However, because of silica's biocompatibility and ease of functionalization, we expect numerous attempts to be made in the future.</p><p>So far, synthetic polymers have achieved success in in vitro and in vivo silencing. Cationic polymeric nanoparticles were produced by chemical synthesis of tripartite polymer conjugates. For example, the nanoparticles consisted of PEI that is PEGylated with an Arg-Gly-Asp (RGD) peptide ligand.66 The resulting nanoplex was about 100 nm in size. The purpose of this conjugate was to target tumors expressing integrins and deliver siRNA inhibiting vascular endothelial growth factor receptor-2 (VEGF R2) and thereby inhibit tumor angiogenesis. Cell delivery and activity of this nanoparticle was found to be siRNA sequence-specific, depended on the presence of peptide ligand and could be competed by free peptide. Intravenous administration into tumor-bearing mice gave selective tumor uptake, siRNA sequence-specific inhibition of protein expression within the tumor and inhibition of both tumor angiogenesis and growth rate.66 Similarly, cationic liposomal nanoparticles have been introduced into athymic mice with human xenograft tumors in which Raf-1, a protein serine/threonine kinase, is constitutively active. These liposomes delivered Raf-1 siRNA and showed anti-tumor efficacy in this in vivo model.67</p><!><p>An interesting example of the use of silica nanoparticles for gene delivery is to enhance chemical transfection protocols (Fig. 1(b)). The non-viral DNA delivery system can consist of DNA, transfection reagents, and nanoparticles as modular components. Each module can be designed to overcome roadblocks such as attenuated transfection efficiency, cellular uptake, cytotoxicity, and protein expression.68 In this modular approach, silica nanomaterials are used as a transfection enhancer rather than the primary mode of delivery.69 Dense inorganic silica nanoparticles, which by themselves do not deliver DNA, are able to enhance DNA transfection mediated by other commonly used transfection reagents.32 This three-component transfection system consists of silica nanoparticles, DNA, and transfection reagents. Each component functions so that barriers to DNA delivery, such as low uptake of DNA by cells or lack of nuclear targeting, can be tackled individually. The particular role of silica nanoparticles is to enhance uptake by physical concentration at the cell surface.32 Customizing the complex can be done at several levels, thereby conferring a wide-range of versatility to the use of silica materials in nucleic acid therapy.</p><p>SEM provided visual confirmation for the assembly of the three-component complex.68 Only in the presence of DNA did the complex form, and although some aggregation of the particles was observed, complexation appeared to prevent this event from occurring. Agarose gel electrophoresis further showed that the DNA was retained by the nanoplex.68</p><p>In Cos-7 cells, enhancement of transfection was monitored by the increase in β-galactosidase activity using plasmid DNA, pVax-LacZ1, as the genetic payload.68 Transfection enhancement with silica nanoparticles and Superfect as the transfection reagent was dependent on the concentration and size of the particles. In addition, the degree of enhancement was dependent on the transfection reagent used.68 Likewise, the three-component system developed by Gemeinhart et al. consisted of silica nanoparticles functionalized with silanes, DNA, and dendrimers as transfection reagents to enhance lysosomal escape of DNA.70,71 Dendrimers also enhance nuclear penetration of DNA in many cells because of its cationic nature similar to nuclear localization signal sequences.72 β-galactosidase activity was again used to measure transfection efficiency, and silica nanoparticles were again found to elevate transfection efficiency.73</p><p>Gemeinhart et al. elaborated upon the nanoparticle-uptake mechanism by hypothesizing that nanoparticle internalization and transport was responsible for increases in transfection efficiency.73 By using fluorescently-labeled nanoparticles (230 ± 10 nm) in flow cytometry and confocal microscopy, they were able to track the nanoparticles during transfection. It had already been suggested by earlier studies that an endosomal/lysosomal uptake route was how nanoparticles gained entry into a cell.74 This hypothesis was confirmed in this follow-up study by light microscopy and identified that fluorescently-labeled nanoparticles tend to localize near the nucleus of CHO cells.73 Gemeinhart et al. confirmed through fluorescent micrographs that nanoparticles deliver nucleic acids by entering cells via an endosomal/lysosomal route. Specifically, they localize to lysosomes and endosomes as indicated by the overlap of LysoTracker red-labeled lysosomes and green nanoparticles.73</p><!><p>Biocompatibility is of utmost importance when designing a delivery system. Since intravenous (IV) administration appears to be the most promising route for nanoparticle delivery, any cytotoxic effects that nanoparticles may have should be eliminated. The hemolysis behavior of amorphous and MSNs (100–300 nm) was investigated in rabbit red blood cells.75 It was found that amorphous nanoparticles, but not MSNs, showed hemolytic toxicity. However, replacement of external surface silanol groups on the amorphous nanoparticles with a positively charged 3-aminopropyl functionality reduced toxicity.75 A prior study had also concluded that commercial colloidal and laboratory-synthesized silica nanoparticles (20–400 nm) cause no significant genotoxicity.76</p><p>The biodistribution and urinary excretion of external surface-modified silica nanoparticles (SiNPs) has also been studied in mice in situ using in vivo optical imaging, ex vivo organ optical imaging, TEM imaging, and energy-dispersed X-ray spectrum analysis of urine samples.77 IV administration of these SiNPs followed by fluorescence tracing in vivo indicated that silica nanoparticles (SiNPs) are all cleared from systemic blood circulation, but that both the clearance time and subsequent biological organ deposition are dependent on the external surface functionalizations of the SiNPs.77 PEG-SiNPs exhibited comparatively longer blood circulation times and lower uptake by the reticuloendothelial system organs than OH-SiNPs and COOH-SiNPs. Bladder and excretion analysis revealed that all three types of IV-injected SiNPs with a size of ~45 nm were partly excreted.77</p><p>Nanoparticle aggregation is a hurdle that needs to be overcome before nanoparticles can be applied in vivo. Clusters of nanoparticles may not be filtered and excreted in living organisms, leading to chronic toxicity and other negative side effects in biodistribution. In order to reduce aggregation of nanoparticles, one solution has been to include a final incubation period in 3-hydroxysilylpropyl methylphosphonate during the synthesis process of fluorescent mesoporous silica nanoparticles (FMSNs).33 The external FMSN surface was modified with inert and hydrophilic phosphonate group to prevent aggregation caused by the interparticle hydrogen bonding interaction between the anionic silanol groups and the unreacted cationic amine groups.33</p><p>Cationic modification, such as PEI or PLL, allow for the electrostatic binding of DNA to silica nanoparticles; however, these groups are themselves cytotoxic and peaks in transfection efficiency often correlate with pronounced reductions in cell viability.21,22,70,78 Thus far, the processes that mediate in vitro toxicity of polycations are not understood. It has been proposed that the cytotoxicity and transfection efficiency of PEI is affected by its molecular weight and shape.22,79 Other studies have demonstrated that high molecular weight PEI exhibits high transfection efficiency and cytotoxicity, while low molecular weight PEI shows attenuated transfection efficiency and cytotoxicity.80–82 Recent approaches to reduce the toxicity of polymer-based transfection systems include the optimization of synthesis22,83 or the modification of the polymer side chains by glycolylation.84 This is an ongoing process and research continues on how polycations and silica structures can be modified to improve their biocompatibility.</p><!><p>In this Feature Article, we describe in detail the recent progress of utilizing silica nanomaterials as a delivery vehicle for nucleic acid-based reagents. MSNs have the potential to transport multiple therapeutic reagents simultaneously, treating diseases with a multifaceted and integrative approach. In this way, tumors can be targeted through the transport of anti-cancer drugs, such as CPT, and the delivery of siRNA to inhibit the underlying genetic component of oncogenesis. The steps by which silica materials can be transformed into intracellular nanocarriers has already been established, although future improvements in targeting and controlled release mechanisms are necessary. We have described here the passive diffusion of DNA, siRNA, and antisense oligonucleotides out of silica nanoparticles to affect gene inhibition, but an active mechanism to drive their effusion from nanopores would improve their therapeutic efficiency. For example, MSNs modified by azo-benzene derivatives, capable of storing small molecules and release them following light irradiation, have been fabricated and characterized.85 Another biocompatible controlled-release motif is the snap-top-covered silica nanocontainer (SCSN). In general, the snap-top contains guest molecules stored within nanopores, but release the guests following cleavage of the stopper cap.86 Targeting to specific tissues is also at the forefront of current research, especially in cancer therapeutics since their use can potentially avoid adverse reactions resulting from systemic drug release and absorption. Finding a particular cancer surface marker that is not expressed in normal healthy tissue has retarded progress in this area. However, progress continues to be made. For instance, the targeting and imaging of MDA-MB-231 human breast cancer cells using arginine-glycine-aspartic acid (RGD) peptide-labeled fluorescent silica nanoparticles has been achieved.87 We look forward to seeing many research breakthroughs that will lead to the further optimization of silica nanoparticles as a medium for gene therapeutics.</p><!><p>Schematic representations of the techniques used for delivery of nucleic acid-based reagents into plant and mammalian cells. Nucleic acid molecules theoretically bind to positively-charged chemical modifications on the external surface of MSNs and hybrid nanoparticles.</p><p>Schematic of the temperature- and energy-dependent process of endocytosis mediated by clathrin-coated pits. The exact mechanism and timing surrounding the release of nucleic acids from MSNs (indicated by the number 8) is currently unknown; however, transgene expression has been detected indicating that a dissociation event has occurred.</p>

PubMed Author Manuscript

Crystallizing the function of the magnetosome membrane mineralization protein Mms6

The literature on the magnetosome membrane (MM) protein, magnetosome membrane specific6 (Mms6), is reviewed. Mms6 is native to magnetotactic bacteria (MTB). These bacteria take up iron from solution and biomineralize magnetite nanoparticles within organelles called magnetosomes. Mms6 is a small protein embedded on the interior of the MM and was discovered tightly associated with the formed mineral. It has been the subject of intensive research as it is seen to control the formation of particles both in vivo and in vitro. Here, we compile, review and discuss the research detailing Mms6’s activity within the cell and in a range of chemical in vitro methods where Mms6 has a marked effect on the composition, size and distribution of synthetic particles, with approximately 21 nm in size for solution precipitations and approximately 90 nm for those formed on surfaces. Furthermore, we review and discuss recent work detailing the structure and function of Mms6. From the evidence, we propose a mechanism for its function as a specific magnetite nucleation protein and summaries the key features for this action: namely, self-assembly to display a charged surface for specific iron binding, with the curvature of the surfaces determining the particle size. We suggest these may aid design of biomimetic additives for future green nanoparticle production.

crystallizing_the_function_of_the_magnetosome_membrane_mineralization_protein_mms6

Introduction<!>The sequence of Mms6 and its activity demonstrated in vivo and in vitro<!>Mms6 in vivo<!>Producing MNPs with Mms6 in vitro<!>Summary of MNPs produced in Mms6 mediated reactions<!>Producing MNPs with Mms6 in vitro<!>Mms6 in vitro: understanding its function<!>Self-assembly<!>Summary of the self assembly and iron binding properties of Mms6 along with schematic representations of the proposed function<!>Self-assembly<!>Study through the formation process in situ<!>Fe binding<!>Discussion and conclusion: proposed mechanism of function<!>

<p>Nanoscale inorganic materials are important in an increasingly nanotechnological world. More specifically, magnetic nanoparticles (MNPs) have wide ranging uses from targeted drug delivery [1], to ultrahigh-density data storage [2]. Magnetite MNPs are particularly useful for biomedical applications such as MRI contrast enhancers for diagnostics, magnetically targeted treatments and magnetic hyperthermia therapy [1,3,4]. However, the reliable production of highly specific monodispersed MNP is a considerable challenge making new synthetic routes to precisely tailored MNPs a necessity [3].</p><p>Natural organisms carefully control the production of a vast range of inorganic minerals in a process called biomineralization [5,6]. For example organisms use calcium phosphate to form bones and teeth, calcium carbonate to make shells and diatoms produce shells and spines from silica [5,6]. Remarkably, nature offers precise genetic control over mineral formation (down to the nanoscale) using a suite of biomineralization proteins [5,6]. Harnessing these proteins presents a biological (ambient condition) synthetic approach to producing tailored MNPs.</p><p>The most studied magnetotactic bacteria (MTB) are aquatic, motile, microaerobic microbes that take up soluble iron ions and crystallize magnetite MNPs within intracellular liposomes (magnetosomes) (Figure 1b) [7,8]. However, MTB are found across the phylogenetic tree, leading to bacteria with variable phenotypes. These range from anaerobic to aerobic, micron-sized Cocci, Vibrio and Spirilla [9], to giant 10 μm rods containing 1000's of magnetosomes [10], living in environments from fresh water to saline [11], whereas some MTB even produce greigite MNP or both magnetite and greigite [12]. Early reviews, such as Bazylinski and Frankel [8] and Frankel et al. [13], offer comprehensive descriptions of MTB and their magnetosomes. Magnetosomes' size and morphology vary greatly between strains too, but is highly uniform within each strain, demonstrating the control that biomineralization proteins must have over this process. The mechanism of biomineralization in MTB enjoys extensive research and is the subject of several concise overview reviews [7,14,15], and a more specific review of the magnetosomes [16] and of their protein's predicted structure and function [17]. Briefly, the magnetosome membrane (MM) is proposed to form through invagination of the cytoplasmic membrane [18], with recruitment and insertion of unique biomineralization proteins into or on to the membrane [14,18]. These include: iron transporters [19,20], redox proteins [21] that ensure the chemistry of magnetite formation is enabled, and nucleation and shape controlling proteins [22,23] that ensure that magnetite is crystallized and grows in the correct morphology [17]. We have been interested in understanding how these proteins (particularly the latter) control magnetite MNP formation and how we can best utilize them (and their mimics) for bio-mediated MNP formation for applications. There has been considerable analysis of one such protein (magnetosome membrane specific6; Mms6), and this is the subject of this mini-review.</p><!><p>(a) Sequence alignment of the truncated Mms6 from different MTB species (the full pre-protein amino acid; numbering is shown above and the mature truncated amino acid position for M. magneticum AMB-1; Mms6 is shown below the alignment). Conserved residues are highlighted in red boxes and similar residues are in red type, showing a highly conserved truncated protein. The initial approximately 98 residues (assumed absent from the mature protein) are not shown but are less conserved (or missing in the case of M. blakemorei). The blue bar highlights the glycine–leucine repeating sequence and the yellow bar highlights the hydrophilic, acid rich, C-terminal amino acid region. (b) Demonstrates the activity of Mms6 in vivo through an mms6 knockout mutant in AMB-1 [23]. Note the MNPs formed in the cell with mms6 absent are smaller and ill formed. (c) Demonstrates the activity of Mms6 in vitro by comparing magnetite MNPs formed in a simple RTCP (protein-free control) with those formed under the same condition but with the addition of Mms6 [24]. Figures 1(b) and 1(c) reproduced from [24]: Amemiya, Y., Arakaki, A., Staniland, S.S., Tanaka, T. and Matsunaga, T. (2007) Controlled formation of magnetite crystal by partial oxidation of ferrous hydroxide in the presence of recombinant magnetotactic bacterial protein Mms6. Biomaterials 28, 5381–5389.</p><!><p>To understand the molecular elements which control magnetite biomineralization there have been a number of studies of both the genomes and proteomes of different MTB and in particular of those elements closely associated with the formation of the magnetosome itself [18,25–29]. Analysis of these sequences has shown that the majority of genes involved in magnetosome biogenesis can be grouped into four key operons (mms6, mamGFDC, mamAB and mamXY clusters) in a region of the genome termed the magnetosome island (MAI) [18,28–34]. If the MAI is lost from MTB then the magnetic properties are also lost [31] and vice versa, if these key operons are transferred to non-magnetic bacteria, then they also develop the ability to produce biogenic magnetic nanocrystals [30].</p><p>The discovery of Mms6 was reported in 2003 in a pioneering study by Arakaki et al. [35]. The magnetosomes from Magnetospirillum magneticum AMB-1 were magnetically extracted from lysed cells and the lipid membrane component of the magnetosome removed [35]. The bare magnetite nanocrystals were subjected to further treatment with detergent and heat to release proteins tightly associated. Four proteins were found: magnetosome membrane specific5 (Mms5), magnetosome membrane specific7 (Mms7), magnetosome membrane specific13 (Mms13) and Mms6, all so-called for their magnetosome membrane specific (Mms) localization and the number denotes their apparent molecular mass [35]. Mms6 has an overall net negative charge at neutral pH, in contrast with the positive charge of the other isolated Mms proteins [35]. Mms6 has been detected in the MM fraction by both 2D SDS/PAGE [25] and shotgun protein identification [18], but not in any of the other membrane fractions of the cell [18], suggesting this protein is specifically targeted to the MM.</p><p>The gene sequence of Mms6 is present in several strains of MTB with a high level of consensus, particularly in the C-terminal region (Figure 1a). The sequences code for a protein of approximately 12–15 kDa, much larger than the 6 kDa species isolated directly from magnetosomes [25,35]. The sequence harbours a glycine–leucine repeat motif (cyan in Figure 1a) and an acid rich C-terminal region (yellow in Figure 1a). It is speculated that Mms6 is processed in vivo by a specific protease to form the 6 kDa truncated protein (Figure 1a) [25,35]. It is interesting to note that in Magnetovibrio blakemorei the mms6 sequence appears to encode only the truncated form of Mms6 which lacks this N-terminal region [36], which would suggest this part of Mms6 is not required for effective magnetite biomineralization.</p><p>A number of gene knockout studies have been performed to assess the specific effect that Mms6 has on magnetite biomineralization [22,23,26]. When the mms6 gene was knocked out in M. magneticum AMB-1, the resulting particles were found to be poorly defined and smaller in size, with an average reduction in diameter of 44% [22] compared with wild-type particles. More recent and previous studies with Δmms6 strains have shown similar results with approximately 19% reduction in AMB-1 strains [23] (Figure 1b) and a 15% reduction in Magnetospirillum gryphiswaldense, indicating Mms6 is required for production of full-sized nanocrystals [26]. Discrepancies in the particle size reduction resulting from mms6 gene knockouts in different studies may be due to how these knockouts have been performed. The 44% reduction is as a result of an antibiotic resistance cassette insertion in the mms6 gene, whereas other studies use two-step recombination to delete mms6. The cassette insertion could inadvertently result in a more pronounced effect by disrupting the production of downstream gene products which have also been implicated in particle formation [23]. There also appears to be a general loss of shape control when Mms6 is absent [22,23,26]. High resolution transmission electron microscopy (HRTEM) analysis of wild-type nanoparticles shows the presence of cubo-octahedral morphology with the characteristic (100) and (111) crystal faces [22,24,37]. In contrast, a Δmms6 mutation produces particles bearing the high energy (110) face of magnetite [22], generally considered to be unstable [22]. This may demonstrate the crystallization process is unfinished in this mutated strain. A key phenotype observed in Δmms6 strains are particles with an elongated morphology (Figure 1b). Wild-type magnetosomes of M. magneticum AMB-1 produce particles with a shape factor (ratio of length to width) close to 1, but in Δmms6 cells this is 0.75 [22,23]. As well as the effects on the nanoparticle itself, the lack of Mms6 in the magnetosome also reduces the level of other magnetosome associated proteins including Mms13, 5 and 7. Mms6 is therefore clearly implicated in protein recruitment to the magnetosome [22], and the N-terminal portion of Mms6 is a likely contender for mediating these contacts.</p><!><p>Additional to its natural activity within the magnetosomes of MTB, purified Mms6 has been investigated in synthetic magnetite formation reactions to look for effects on the MNP products [24,35,38–43]. Magnetite (Fe3O4) contains both ferric (Fe3+) and ferrous (Fe2+) iron in a precise stoichiometric ratio of 2:1. There are several methods of producing magnetite synthetically but most rely on providing (or producing during the reaction) a mixture of iron of both valences and bringing about its subsequent precipitation by raising the pH. By including purified Mms6 to these syntheses, the size, shape and material purity of the resulting nanoparticles can be compared with protein-free nanoparticles prepared under identical conditions [24,35,38,40]. The nature of the magnetite precipitation process within the magnetosome has not been completely chemically resolved and remains one of the key barriers to our fuller understanding of magnetite biomineralization. However, a recent well conducted study has gone some way to answering this point [44]. Firlar et al. [44] used single particle analysis of forming magnetosomes to show that a ferric-rich amorphous precursor is formed initially, before conversion to the final magnetite species. It is therefore likely that the current approaches for studying Mms6 activity synthetically share similarities with the processes and conditions under which Mms6 would normally function (Table 1).</p><!><p>*Brackets denote the ratio of ferric to ferrous ions used.</p><!><p>The majority of studies have explored two methods of magnetite synthesis with Mms6: these are the room temperature co-precipitation (RTCP) and partial oxidation of ferrous hydroxide (POFH). In 2003, the first in vitro activity of Mms6 was reported by Arakaki et al. [35]. The authors found that addition of Mms6 to an RTCP reaction (Mms6 at 20 μg/ml) resulted in a product which contained mainly magnetite and little alternative iron oxide compared with the protein-free control sample. The alternative iron oxides formed in the control reaction suggests a relatively uncontrolled precipitation reaction occurred in this experimental system. However, this highlights the ability of Mms6 to shift the balance of products towards magnetite in such a reaction. Additionally, the Mms6 prepared particles displayed a narrower size distribution and a cuboidal appearance very similar to that observed in magnetosomes. Similar results have been replicated with lower concentrations of protein [24] (Figure 1c), different ratios of ferrous to ferric iron in the reaction mixture [41], as well as in POFH reactions [24,41] and with Mms6 immobilized on planar surfaces to mimic the magnetosome interior [40]. The general trends (Table 1) are that the addition of Mms6 in solution produces nanoparticles of approximately 21 nm while also reducing their size distribution, regardless of the protein-free particle size population [with the exception of the partial oxidation of ferrous hydroxide with ammonia and hydrazine (POFHN)]. Interestingly, the only instance when Mms6 mediated particles are of a different size is in a particular partial oxidation of ferrous hydroxide with potassium hydroxide (POFHK) reaction where the control MNPs has a huge size distribution [41]. These reactions are very sensitive to a variety of reaction parameters. In this case, the Mms6 MNPs are an average of 42 nm in size, however if the distribution is examined more closely it can be seen there is a dual population, with one peak the same size as the other Mms6 mediated MNPs (22.5 nm) and the remaining peak similar to the control particles (Table 1) [41], showing Mms6 is able to control a portion of the population, perhaps suggesting the ratio of Mms6 to iron is insufficient. Significantly, on surfaces the particles produced with Mms6 are approximately 90 nm and therefore much larger than the protein-free particles, perhaps due to the planar arrangement of Mms6 compared with the curved micelles in solution. During POFH reactions, the system is heated (to approximately 80°C), at which point most proteins would be denatured and inactive, yet Mms6 has an effect in such reactions. However, the heating is performed once all the reagents are supplied and the particles have begun to precipitate, indicating that Mms6 may either only function at the particle nucleation stage (prior to heating), or be remarkably resilient to such treatment.</p><!><p>It is clear from the previous section that Mms6 is able to control the formation of magnetite MNPs when added to a chemical precipitation. Taken together we believe we can describe the action of Mms6 in vitro and thus propose a mechanism for its function.</p><!><p>The amphiphilic nature of Mms6 (Figure 1a) implies it will form micelles in aqueous solution with the C-terminal hydrophilic regions exposed, shielding the hydrophobic N-terminal regions within the core. This structure was first quantitatively investigated by Wang et al. [47] who found through size-exclusion chromatography that the micelle was between 200–400 kDa made up of 20–40 protein subunits. Dynamic light scattering (DLS) measurements found micelles were 10.2 ± 3 nm across, equating to approximately 200 kDa, in agreement with the other analysis [47]. At pH 7.5, they have a slightly narrower elution profile than those at pH 3. The structure and size was further confirmed through SAXS experiments [48]. Using a core-corona model, the data fitted well to a hydrophobic core of radius 3.9 ± 0.4 nm and hydrophilic corona radius of 1.1 ± 0.2 nm parameters at pH 3. Again there is a difference at pH 7.5, but the modelled parameters are not given. Interestingly, they found that addition of iron caused the micelles to form higher order structures such as discs of micelles [48] (Figure 2a) presumably through iron cross-linking. Most recently, Mms6 micelles (or larger proteinaceous assemblies) have been visualized in situ in fluid cell TEM [49] where the micelles appear approximately 10-fold larger (Figure 2b). Wang et al. [47] indicated a small population of much larger protein assembly particles in SEC, so it is not unreasonable to assume these will be the easiest to visualize in the fluid cell TEM. Perhaps trace iron is causing a small amount of larger assembly.</p><!><p>Summary of the research on the (1) micellar structure (a) and (b), (2) schematic representations to describe its function and mechanism for nucleation (c) and assembly (d) and (3) iron binding (e) and (f) of Mms6. (a) shows a model for the micelle structure obtained by SAXS analysis [48]. (b) shows the nucleation and precipitation of iron oxide MNP (bright spots) on the surface of an Mms6 micelle (scale bar 20 nm) [49]. (c) An above view schematic of how Mms6 may self-assemble as a protein raft to display regular binding sites for iron ions to nucleate magnetite formation. (d) Side-on schematic to demonstrate how the curvature of the protein surfaces differs for the (i) in solution micelles, (ii) on a surface and (iii) on the MM to explain for difference seen in particle size [40]. (e) Size of chemical shifts of residues upon metal binding in Mms6 C-terminal peptide from 2D NMR analysis. Green bars represent ferrous ions [51]. (f) Ferric iron binding analysis of Mms6 (●), Mms6 with the C-terminus shuffled (▲) and Mms6 with just the acidic residues in the C-terminus shuffled (□) [47]. Figure 2(a) reproduced from [48]: Zhang, H., Liu, X., Feng, S., Wang, W., Schmidt-Rohr, K., Akinc, M., Nilsen-Hamilton, M., Vaknin, D. and Mallapragada, S. (2015) Morphological transformations in the magnetite biomineralizing protein Mms6 in iron solutions: a small-angle X-ray scattering study. Langmuir 31, 2818–2825. Figure 2(b) reproduced from [49]: Kashyap, S., Woehl, T.J., Liu, X., Mallapragada, S.K. and Prozorov, T. (2014) Nucleation of iron oxide nanoparticles mediated by Mms6 protein in situ. ACS Nano 8, 9097–9106. Figure 2(d) reproduced from [40]: Bird, S.M., Rawlings, A.E., Galloway, J.M. and Staniland, S.S. (2016) Using a biomimetic membrane surface experiment to investigate the activity of the magnetite biomineralisation protein Mms6. RSC Adv. 6, 7356–7363. Figure 2(e) reproduced from [51]: Rawlings, A.E., Bramble, J.P., Hounslow, A.M., Williamson, M.P., Monnington, A.E., Cooke, D.J. and Staniland, S.S. (2016) Ferrous iron key to Mms6 magnetite biomineralisation: a mechanistic study to understand magnetite formation using pH titration and NMR. Chem. Eur. J. 22, doi:10.1002/chem.201600322. Figure 2(f) reproduced from [47]: Wang, L., Prozorov, T., Palo, P.E., Liu, X., Vaknin, D., Prozorov, R., Mallapragada, S. and Nilsen-Hamilton, M. (2012) Self-assembly and biphasic iron-binding characteristics of Mms6, a bacterial protein that promotes the formation of superparamagnetic magnetite nanoparticles of uniform size and shape. Biomacromolecules 13, 98–105.</p><!><p>It is remarkable that Mms6 is able to convey similar activity/control over magnetite MNP formation in vitro as in vivo, leading to the conclusion that there must be some degree of self-assembly in the membrane environment, similar to the aggregation seen in vitro. We propose Mms6 is not monomeric in the MM, but self-assembles to form protein rafts on the MM interior, displaying a charged C-terminal surface (schematic Figure 2c) akin to the surface of in vitro micelles (but of the opposite curvature) (schematic Figure 2d). We tested this hypothesis by enabling Mms6 to self-assemble on a surface, mimicking the membrane environment. Remarkably, the biomimetic Mms6 surface nucleated and controlled magnetite formation, whereas the C-terminal peptide alone (C20Mms6) did not [40]. C20Mms6 is missing the N-terminal region but is still able to assemble on the surface [40]. Thus, we propose that the nature of self-assembly in Mms6 is more specific than generic hydrophobic interactions. A glycine–leucine repeating sequence is present in the conserved 6 kDa protein (but absent from C20Mms6) (Figure 1a). Such motifs are common in self-assembling structural proteins e.g. silk fibroin [40,50]. We propose this knob and hole arrangement of hydrophobic residues could interlock with adjacent Mm6 molecules (and even in vivo to other Mms proteins with the same glycine–leucine motif) to form a regularly packed structure and thus regularly space the iron binding C-terminal sites across a raft surface of Mms6 (schematic in Figure 2c). It appears that without this the magnetite nucleation ability is lost, as the C20Mms6 on surfaces demonstrates (compare in Table 1) [40]. In solution, the peptide has shown an effect on particle formation, but this could be driven by some level of aggregation in solution [39]. Interestingly, a peptide constructed of a GL-C6Mms6 repeat fusion (GLM6A) shows better control over particle formation (Table 1), further demonstrating the importance of this region [39].</p><!><p>We propose Mms6 self-assembles in a uniform manner to display a regular array of the acidic C-termini, create a negatively-charged surface for iron ion binding to nucleate magnetite formation. However, tracking and analysing this process is not trivial. Recently, Kashyap et al. [49] showed iron oxide nucleation on the surface of the Mms6 micelles in situ in a fluid TEM experiment. Remarkably, ferric ion association with Mms6 from low pH can clearly be seen, and as the pH rises small iron oxide particles visibly form across the micelle surface (Figure 2b). They note some ferric ion depletion when the pH initially rises, attributed to a first step in forming a disordered pre-nucleation phase [49]. However, the pH is not quantified for each imaging stage, and only ferric ions are used so the mineral nucleated would not ultimately crystallize to magnetite. Therefore, tracking the chemistry quantitatively throughout the process is essential to understand the effect the protein is having on magnetite formation. We performed a series of pH titrations on the in vitro precipitation of magnetite with and without Mms6 and found that Mms6 had no effect on the process below pH 4 for a range of different ferric:ferrous ratios [51]. This is the stage where the more insoluble ferric ions precipitate as a ferric oxide (such as schwertmannite in our case, but could be haematite or ferrihydrite depending on the conditions). Although studies do show that ferric ions can bind to Mms6 at pH 3, it is considerably less than at pH 7 [47]. Thus, we believe binding at low pH is not the main action of Mms6, being negligible when compared with the bulk precipitation. Only after this precipitation stage does the Mms6 pH trace diverge from the protein-free reaction when the mixed valance iron oxides start to precipitate, corroborating the idea that Mms6 is most active at higher pH when the acidic groups are most available for iron binding [51]. Interestingly, we see Mms6 has the most marked effect in ferrous-rich ferric:ferrous ion ratios, indicating increased magnetite production with Mms6 (20%) compared with negligible amounts without protein, suggesting Mms6 is able to direct mineralization towards magnetite synthesis under conditions further from the ideal for magnetite formation, effectively acting as a 'mineral/ferrous ion buffer' [51]. Furthermore, it appears that Mms6's interaction with ferrous ions is potentially crucial to this process.</p><!><p>The iron binding ability of Mms6 was first reported by Arakaki et al [35]. Using a competitive radioactive ferric ion binding assay where purified recombinant Mms6 was seen to bind Fe3+, Ca2+ and Mg2+, but not Zn2+, Ni2+ or Cu2+, showing some metal selectivity [35]. Most Mms6 iron binding studies have been performed with ferric ions [35,47,49] which have limited solubility at physiological pH. Chelators such as citrate are therefore required to solubilize Fe3+ for analysis at neutral pH. High affinity ferric ion binding (Kd=10−16 M) was established at pH 7.5 in this way, whereas mutants (with scrambled C-termini) show no significant ferric binding, demonstrating the importance of the amino acid sequence at the C-termini (Figure 2f) [47]. To facilitate an increase in ferric ion concentration, further assays were performed at pH 3 where ferric ions are soluble. A significantly lower binding affinity of Kd=0.58 ± 0.03 μM was reported [47]. At low pH, the acidic groups of Mms6 are likely to be mostly protonated (pKa of free glutamic acid is approximately 4) reducing the capacity to bind ferric ions. As the pH increases to 7 (where mixed valence iron minerals precipitate), deprotonation will produce negatively-charged acidic groups compatible with iron ion binding. Mms6 shows variations in iron ion binding and micelle morphology between low and neutral pH demonstrating the significance that pH has [48]. pH measurements during RTCP with Mms6 further demonstrates this point, showing Mms6 has minimal effect at pH<4 [51]. NMR analysis of the C20Mms6 peptide in the presence and absence of ferric ions (at pH 7) revealed only small chemical shift differences in the peptide side chains [51]. However, in the presence of ferrous ions significant (5-fold) chemical shift differences are seen (Figure 2e) indicating stronger, more specific binding of ferrous than ferric iron [51]. Molecular modelling suggests ferric ions may bind non-specifically (drawn to the areas of greatest charge) so the binding does not significantly change C20Mms6's conformation, but ferrous ions display specific multisite binding suggesting C20Mms6 is a specific multidentate ferrous ion ligand [51].</p><!><p>In vitro Mms6 shows negligible activity in RTCP experiments below pH 5 as determined by pH monitoring in situ [51]. When deprotonated the 10–12 nm sized Mms6 micelles are likely to display negatively-charged surfaces for iron binding. We propose that Mms6 binds both ferric and ferrous ions under these conditions; ferrous seemingly specifically [51], whereas the highly charged Fe3+ binds more indiscriminately and abundantly [47–49]. The acidic residues of the C-terminal region of Mms6 may concentrate mixed valence iron on the surface in the correct 1:2 ratio to nucleate magnetite. However, the C20Mms6 peptide appears unable to nucleate magnetite as effectively [40], suggesting Mms6 function requires some degree of ordered self-assembly. The structural glycine–leucine repeat sequence may provide this by achieving interlocked packing between Mms6 subunits to bring about a large charged surface for the specific positional binding of Fe2+ to 2x Fe3+ to encourage the nucleation of magnetite [39,40].</p><p>The action of Mms6 in vivo may be similar to that observed in vitro. Instead of micellar assembly, Mms6 may assemble in the MM in a raft-like form [Figures 2c and 2d (iii)]. Although the pH inside magnetosomes has not been determined, it must be high enough to enable magnetite to precipitate and thus for Mms6 to be active. It is thought that iron is transported into magnetosomes as Fe2+ [52–54] with subsequent partial oxidation to Fe3+ by oxidase enzymes. In in vitro experiments, Mms6 is most influential in ferrous-rich conditions [51], where magnetite is chemically more challenging to produce. This may reflect the conditions within the magnetosome.</p><p>Whether or not Mms6 is a nucleating or shape controlling protein is debated [17,22,39,40,49]. Iron ion binding data [47–49] and poor binding activity between magnetite surfaces and Mms6 [40] show it is only tightly bound if involved at the nucleation stages. But in vivo, mms6 knockout studies show poorly formed, smaller, magnetite crystals, supporting morphology controlling activity [22,23]. However, the more recent studies have found less clear effects on particle morphology when mms6 is deleted [23]. One possibility to account for these conflicting reports is that neighbouring genes, in particular mmsF, may be affected by the gene knockout in the earlier study. MmsF has been described as a master regulator of magnetite biomineralization in vivo [23]. However, it is likely that morphology and nucleation activities are coupled; if a particle is not nucleated properly, it may not form to the desired morphology. Equally nucleation from a specific crystal plane will guide the eventual morphology of the final nanoparticle.</p><p>It is clear that Mms6 regulates the size of particles in vitro; with consistent size across both RTCP and POFHK routes (21 nm) when nucleated on Mms6 micelles in solution [24,35,39,41], and particles approximately 90 nm in size when nucleated by Mms6 assembled on planar surfaces [40,45,46], whereas MNPs within magnetosomes are typically 40–50 nm [55]. The key difference between all these surfaces is curvature, from convex to flat to concave respectively. We suggest this difference in degree and angle of contact between the protein assembly surface and the mineral (along with nucleation physics) is responsible for the difference in particle sizes (Figure 2d and Table 1) [40].</p><p>Mms6 activity in vitro holds promise for biokleptic synthesis for nanotechnology [56]. However, Mms6 is not trivial to produce, making scale-up for commercial processes unlikely. An understanding of Mms6 informs the design of additives for MNP production to mimic the function of Mms6. We propose that the key elements for design should be: (1) negatively-charged carboxylate-rich surface, (2) a precisely packed assembly of this surface and (3) MNP size may be tuneable by controlling surface curvature.</p><!><p>Membrane Proteins From A to Z: Held at University of Leeds, U.K., 16–17 December 2015</p><p>C-terminus peptide (20 amino acids)</p><p>dynamic light scattering</p><p>the most acidic 12 amino acid region of Mms6 C terminus fused to the LG repeat region</p><p>high resolution transmission electron microscopy</p><p>magnetosome island</p><p>magnetosome membrane</p><p>magnetosome membrane specific</p><p>magnetosome membrane specific6</p><p>magnetic nanoparticle</p><p>magnetotactic bacteria</p><p>partial oxidation of ferrous hydroxide</p><p>partial oxidation of ferrous hydroxide with potassium hydroxide</p><p>partial oxidation of ferrous hydroxide with ammonia and hydrazine</p><p>room temperature co-precipitation</p><p>size exclusion chromatography</p>

PubMed Open Access