and C

and C.N. antiviral drugs. design and biochemical experiments. Our novel, rationally discovered inhibitors with validated binding modes and low molecular weight represent promising starting points for future hit optimization. Literature research revealed a lack of high-quality bioactivity data for ZIKVPro. Reported competitive ligands show either low potency, CY3 high molecular weight, or low stability in aqueous solutions.20,21,28 The substrate binding site of WNV protease (WNVPro) and ZIKVPro shows a sequence identity of 83% (Figure ?Figure11), and several nonpeptidomimetic ligands for WNVPro were reported with activity below 50 M8 (Supporting Information Table S1). Hence, WNVPro was used as a starting point for the identification of novel drug-like ZIKVPro inhibitors. Substrate binding sites of WNVPro and ZIKVPro only differ at three residue positions (Figure ?Figure11). The S1 and S2 subpockets (SchechterCBerger nomenclature32) are highly conserved in flaviviral species14 and accept lysine and arginine.14 S3 and S4 subpockets show sequence-variability and accept various residues. Both substrate binding sites are highly flexible,13 hydrophilic, and shallow,33 rendering the NS2B-NS3 protease a challenging target for drug discovery. In order to address binding pocket flexibility of WNVPro, we employed our novel application analysis. Pink letters and numbers indicate protease subpockets. Color code: yellow spheres and clouds, lipophilic contacts; purple rings and blue clouds, aromatic interactions; red arrows and clouds, hydrogen bond acceptors; purple stars and clouds, cationic interactions. Identified cationic interactions exploit contacts in the S1 subpocket to D129 and in the S2 subpocket to D75 and H51, while aromatic interactions are present facing Y161 and H51 in the S1 and S2 subpockets, respectively. Hydrogen bond acceptors are preserved in the essential oxyanion hole (S135, T134, G133) and in the backbone-binding region (G153, Y161). Lipophilic contacts are placed in the conserved regions of the S1 subpocket in proximity to Y161 and Y150. The resulting focused pharmacophore was used for combinatorial model library generation CY3 with in the S1 subpocket (Figure ?Figure22B) should be present in each pharmacophore model to enhance the likelihood of finding an active inhibitor. All other pharmacophore features were systematically combined and merged with the cationic feature to generate 3D pharmacophores with three to six independent pharmacophore features. This procedure resulted in a combinatorial library of 3022 different 3D pharmacophore models. The final pharmacophore ensemble was Rabbit Polyclonal to GPRIN3 retrospectively evaluated by screening a collection of 17 small molecular WNVPro inhibitors reported in the literature35?39 and 667 decoy molecules derived from the active ligands by the DUD-E server (Database of Useful Decoys: Enhanced).40 We compared the obtained early enrichment factors (EF1%) and absolute number of recovered active inhibitors for picking best performing pharmacophores (Supporting Information Figure S1). The three best performing models (C1_65, C1_397, and C1_427, Figure ?Figure33) were used for an extensive virtual screening (VS) campaign with more than 7.6 million commercially available compounds. In total 1079 virtual hits were detected (10 for C1_65, 712 for C1_397, and 357 for C1_427). Open in a separate window Figure 3 Best performing pharmacophore models obtained from combinatorial model library (yellow spheres, lipophilic contacts; purple rings, aromatic interactions; red arrow, hydrogen bond acceptor; purple star, cationic interaction). We docked obtained hits into the WNVPro substrate-binding pocket to explore plausible binding hypotheses. Subsequently, we minimized the energy of docking poses in the binding pocket using LigandScout41,42 and scored the ligand conformations based on their fit to the C1-pharmacophores (Figure ?Figure44). Open in a separate window Figure 4 Virtual screening protocol applied for screening of Zika and West Nile virus protease inhibitors. All compounds were visually inspected to exclude unfavorable virtual hit orientations, such as lipophilic groups pointing toward the solvent, or nondrug like moieties43 (e.g. quinones) yielding 15 compounds. To ensure that the hits CY3 can bind to the highly flexible NS2B-NS3, we performed MD simulations with the best-scoring ligand conformations in complex with.