O Cl Cl O N NH O OH O N Cl Computational Medicinal Chemistry Approaches for GPCR Structure-Based Drug Discovery © 2019 Heptares Therapeutics Limited Disclaimer: Sosei Heptares is a trading name. Sosei and the logo are Trade Marks of Sosei Group Corporation, Heptares is a Trade Mark of Heptares Therapeutics Limited. StaR® is a Trade Mark of Heptares Therapeutics Limited Juan Carlos Mobarec , Conor G. Scully, Lydia Siragusa * ,Francesca Deflorian, Robert T. Smith, Alicia Higueruelo, Jonathan S. Mason, Miles Congreve, Chris De Graaf Lemborexant Suvorexant EMPA dual Ox 1 /OX 2 dual Ox 1 /OX 2 OX 2 selective Targeting lipophilic hotspots and unhappy water sites determine ligand binding Trapped unhappy water in Ox 1 2.8 log units selectivity from trapped water, ligand would appear to dock fine if water energetics not considered OX 2 /Ox1 selectivity: Binding pocket Ox 1 (A 3x33 ) > Ox 2 (T 3x33 ) Examples Pharmacophore? Unpublished in-house structures Ox 1 /Ox 2 , bound to different ligands and with different water networks WaterFLAP: MIF based water networks. WaterMap: MD based hydration site thermodynamics. C sp 3 probe 1 kcal/mol C sp 2 probe -2.8 kcal/mol Water probe -5 kcal/mol GRID Probes: Pseudo apo water G: Energetically very unhappy water in A 2A Ligand perturbated G: Energetically stabilised water in A 2A WaterFLAP water energetics A 2A QSAR Collaboration S. Cross, G. Cruciani Bortolato, Mason et al. (2018) Methods Mol Biol J. Christopher et al., Unpublished Lipophilic hotspots & water networks in OX 1 ,OX 2 and A 2a SBDD from hit-ID to clinical trials: Case A 2a antagonist Partnered with: Monotherapy AZD4635 (A 2a antagonist) Combination with oleclumab (anti-CD73) Combination with durvalumab (anti-PD-L1) • SBDD used to re-engineer a virtual screening hit series to target a lipophilic hotspot deep in the pocket, leading high LE and LLE drug candidates • Atom by atom optimisation: ligand efficiency (LE) • Design polar contacts: control lipophilicity (LLE) • Multiple structures: receptor flexibility & selectivity • Druglike properties: In vivo efficacy & safety Langmead (2012) J Med Chem; Congreve (2012) J. Med. Chem FEP+ based binding affinity prediction for ligand optimization Ligand/protein tautomer/protonation state In-house structures of GPCR target with different ligands C C C C Linker changes c. x3 Oral Bioavailability RHS LHS Pose at 0 ns Pose at 20 ns H278 Pose at 0 ns Pose at 20 ns N256 6.55 H278 4a 4g 4e GCMC Pose at 0 ns Pose at 20 ns A2a - 4a N256 H278 7.43 4a Pose at 0 ns Pose at 20 ns A2a - 4a N256 6.55 H278 7.43 Pose at 0 ns Pose at 20 ns A2a – 4e N256 6.55 H278 Pose at 0 ns Pose at 20 ns A2a – 4e N256 6.55 H278 4e 4g lacking water network 4a 4g 4e No GCMC R 2 =0.95 In-house GPCR structures with representative ligand LHS and variable RHS Ring conformation sampling FEP+ guided GPCR LO example 1 A2a – 4g A2a – 4g Alternative residue rotamers FEP+ guided GPCR LO example 2 Binding site solvation F. Deflorian D. Branduardi J. Vendome Structural chemogenomics codification of GPCRome B H G C major D I E J L K minor A BioGPS: GRID based identification and comparison of GPCR binding sites across structural GPCRome. C5aR 4156 GPR40 5706 ECFP4 > 0.4 MACCS > 0.8 64 Similar Bioactive Ligands • Shared pharmacophore features to target GPCR-membrane interface. • Design ideas/rules for conserved GPCR PAM/NAM pockets. ligands Lipophilic HB donor HB acceptor Ligands L. Sygura, G. Cruciani Computational chemistry O N Cl Cl NH O OH O C H 3 CH 3 O C H 3 SBDD in allosteric binding sites: Cases GLP-1R and PAR2 C Hit 1 GLP-1R pK i 5.0 HTL26005 GLP1 pK i 8.3 MW 527, clogP 5.1 LE 0.31, LLE 3.2 MK-0893 GLP-1R pK i 7.3 GCGR pKi 8.8 HTL26119 GLP1 pK i 7.8 MW 574, clogP 6.4 LE 0.3 LLE 1.4 Clinically studied GCGR antagonist Simple novel starting point for SBDD efforts GCGR pK i 6.9 Less potent, non-selective in functional assay Selective vs GCGR in binding Less potent as a functional antagonist Virtual Screen In silico design SBDD N H O OH O N N O Cl Cl O NH O OH O Jazayeri (2016) Nature GCGR • Collaboration with AZ, fragment and HTS screening • Small molecule antagonists inhibit peptide and protease activation of the receptor • Difficult to optimise in the absence of structural understanding • Binding site identified in PAR2 X-ray structure • AZ8838 buried in small binding pocket (TM1-3/7, ECL2) • X-Chem DNA encoded library technology • Binding hits confirmed as functional PAR2 antagonists • AZ3451 binds in novel extra-helical site • Interaction with PAR2 predominately hydrophobic • Mechanism of action may be to restrict the conformational inter-helical rearrangement required for PAR2 activation • PeptiDream DELT focuses on peptide display • Successful hit generation approach for wide array of targets • Utilising the PAR2 StaR in collaboration with Heptares Peptidream have identified several series of potent cyclic peptide antagonists of PAR2 • Current efforts to improve potency and stability of these very encouraging peptide lead cpds using SBDD SLIGKV (model) AZ3451 (PDB: 5NDZ) AZ8838 (PDB: 5NDD) cyclic peptide antagonist (unreleased crystal structure) A. O’ Brien and co-workers 2019 PAR2 Cheng (2017) Nature Sosei Heptares, Steinmetz Building, Granta Park, Great Abington, Cambridge, CB21 6DG United Kingdom *Molecular Discovery Ltd. Middlesex , United Kingdom A) Major and minor (ancestral/classical TM) binding sites 95% GPCR ligands in PDB target this pocket B) G protein intracellular binding site (B) CCR2 (5T1A) CCR9 (5LWE) β 2 R (5X7D) C) Sodium pocket BLT1 (5X33) mGlu 5 (4OO9; 5CGC; 5CGD; 6FFH; 6FFI) D) TM3/4/5/EL2 PAF-lipid/OLC * (5ZKP; 5ZKQ) E) TM3/4/5/membrane C5AR (5O9H; 6C1Q; 6C1R) FFAR 1 (5KW2, 5TZY) GCGR (4L6R * ; 5EE7; 5XEZ; 5XF1) GLP-1R (5VEX; 5VEW) F) TM5/6membrane G) TM1/2/3/IL1 /Membrane H) TM3/4/IL/membrane FFAR 1 (5KW2, 5TZY) EP 4 (5YWY; 5YFI; 5YHL) TA 2 (6IIU) CCR2 (5T1A; 6GPS, 6GPX) CB1 (5TGZ, 5U09) I) TM1/7/membrane J) TM1/7/H8/Membrane K) TM1/2/3/Membrane P2Y 1 (4XNV) L) TM3/4/5/membrane PAR2 (5NDZ) An increasing number of cryo-EM and X-ray crystal structures of GPCR-ligand complexes continue to reveal previously unknown ligand binding sites. Furthermore, emerging sets of GPCR crystal structures of multiple diverse ligands bound to closely related receptors enable a protein-structure based view of how different ligands bind this major drug target class. From the analysis of GPCR structures we gather several important learnings and repercussions for computational medicinal chemistry design that should be transferable and relevant for many targets, including: A) The important roles of lipophilic hot spots and water networks as drivers of GPCR druggability, ligand binding, and selectivity. B) Diverse binding modes of similar ligands across the structural GPCRome. C) Caveats when using pharmacophore-based similarity principles for modeling receptor-ligand complexes with different ligand chemotypes. Multiple ligand binding sites on GPCR structures • GPCR binding sites can be identified with BioGPS, which utilizes GRID probes to locate and characterize protein pockets. • Pockets can be encoded into bitstrings which can be used to compare different pockets and measure similarity (e.g. Tanimoto). • Phylogenetic relationships can be build to compare pockets in the structural GPCRome. F Multiple in-house Ox 1 /Ox 2 crystal structures for SBDD: - Multiple ligand binding modes – pharmacophore models - Critical to consider water networks in docking and FEP+ - Design opportunities for smaller ligands with improved properties GCMC for exhaustive water sampling in protein binding pocket during FEP+ production Water mediated receptor-ligand interactions in crystal structures No direct polar (H-bond) interactions