Allosteric Modulators - ACS Publications - American Chemical Society

Feb 18, 2018 - (7) Christopher, J. A.; Aves, S. J.; Bennett, K. A.; Doré, A. S.; Errey, J. C.; Jazayeri, A.; Marshall, F. H.; Okrasa, K.; Serrano-Vega...
2 downloads 0 Views 13MB Size
Article Cite This: J. Med. Chem. 2019, 62, 207−222

pubs.acs.org/jmc

Structure-Based Optimization Strategies for G Protein-Coupled Receptor (GPCR) Allosteric Modulators: A Case Study from Analyses of New Metabotropic Glutamate Receptor 5 (mGlu5) X‑ray Structures John A. Christopher,*,† Zoltán Orgován,‡ Miles Congreve,† Andrew S. Doré,† James C. Errey,† Fiona H. Marshall,† Jonathan S. Mason,† Krzysztof Okrasa,† Prakash Rucktooa,† Maria J. Serrano-Vega,† György G. Ferenczy,‡ and György M. Keserű*,‡ J. Med. Chem. 2019.62:207-222. Downloaded from pubs.acs.org by EASTERN KENTUCKY UNIV on 01/17/19. For personal use only.



Heptares Therapeutics Ltd., BioPark, Welwyn Garden City, Hertfordshire AL7 3AX, U.K. Medicinal Chemistry Research Group, Research Centre for Natural Sciences, Hungarian Academy of Sciences, 2 Magyar tudósok körútja, Budapest H-1117, Hungary



S Supporting Information *

ABSTRACT: Two interesting new X-ray structures of negative allosteric modulator (NAM) ligands for the mGlu5 receptor, M-MPEP (3) and fenobam (4), are reported. The new structures show how the binding of the ligands induces different receptor water channel conformations to previously published structures. The structure of fenobam, where a urea replaces the acetylenic linker in M-MPEP and mavoglurant, reveals a binding mode where the ligand is rotated by 180° compared to a previously proposed docking model. The need for multiple ligand structures for accurate GPCR structure-based drug design is demonstrated by the different growing vectors identified for the head groups of M-MPEP and mavoglurant and by the unexpected water-mediated receptor interactions of a new chemotype represented by fenobam. The implications of the new structures for ligand design are discussed, with extensive analysis of the energetics of the water networks of both pseudoapo and bound structures providing a new design strategy for allosteric modulators.



INTRODUCTION The concept of allosteric modulation can be traced back to 19651 and was first realized in the discovery of benzodiazepines that were shown to be allosteric modulators of the GABAA receptor.2 The success of these compounds together with the demonstrated advantages of allosteric modulation over orthosteric ligands make this approach increasingly popular in the field of G protein-coupled receptors (GPCRs).3 Binding sites used by allosteric ligands are outside the more conserved orthosteric pockets that have evolved for endogenous agonist signaling. Consequently, targeting allosteric sites might improve target specificity and decrease the risk of potential side effects. Allosteric modulators that lack intrinsic activity are only effective in the presence of the corresponding orthosteric agonist; thus, their limited positive or negative cooperativity controls the magnitude of their allosteric effect. On one hand, this feature might contribute to the increased safety margin of allosteric modulators, and on the other, it provides a unique opportunity for disease specific tailoring of the endogenous agonist activity. Despite these potential benefits, only a few GPCR targeted allosteric modulators have reached the clinic to date.4 One of the potential reasons is associated with the inherent challenges of their medicinal chemistry optimizations, where the already complex challenge of balancing multiple © 2018 American Chemical Society

parameters is further complicated by variables such as allosteric coupling and cooperativity, effects on signaling, and functional activity. In addition, it has been extensively discussed that allosteric modulator programs are frequently impacted by challenges including steep or flat structure−activity relationships (SAR), the limited transferability of SAR knowledge between chemotypes, and molecular switches that cause variations in affinity versus efficacy modulation.4 A further consideration is that not all allosteric modulators are devoid of intrinsic activity, especially positive allosteric modulators (PAMs), which adds complexity to their pharmacological characterization. With a lack of structural information, early optimization in most GPCR allosteric modulator programs to date have used iterative (parallel) synthesis strategies and small matrix libraries.5 This stochastic and resource intensive strategy is typically based on retrosynthetic analysis of a lead compound and independently varies the synthons to develop SAR knowledge. The optimized structural moieties are then Special Issue: Allosteric Modulators Received: November 23, 2017 Published: February 18, 2018 207

DOI: 10.1021/acs.jmedchem.7b01722 J. Med. Chem. 2019, 62, 207−222

Journal of Medicinal Chemistry

Article

Chart 1. Literature mGlu5 NAMs Mavoglurant 1, MPEP 2, M-MPEP 3, Fenobam 4, HTL14242 5, and Pyrazole 6

in numerous disease settings, including fragile X syndrome (FXS),13 Parkinson’s disease levodopa-induced dyskinesias (PD-LID), 14 anxiety, 15 gastroesophageal reflux disease (GERD),16 neuropathic pain,17 obsessive-compulsive disorder (OCD),18 migraine,19 chorea in Huntington’s disease,20 and depression.21 The crystal structure of the TMD of mGlu5 with the NAM mavoglurant (1, Chart 1) bound in the allosteric region commonly known as the “MPEP site” (named with regard to the early NAM MPEP22 (2) that binds in this site) was published in 2014, using a form of the mGlu5 receptor thermostabilized by introduction of a small number of residue mutations outside the ligand binding site (termed mGlu5-StaR), further modified by excision of flexible domains from the receptor N-terminus and insertion of T4-lysozyme into intracellular loop 2.23 Pharmacological characterization of a range of mGlu5 NAM ligands, including MPEP, M-MPEP, and fenobam (Chart 1), confirmed no differences in affinity for wild-type and thermostabilized receptors.23 Subsequent work in our laboratories led to the identification of a pyridinylpyrimidine series of mGlu5 NAMs through fragment and structure-based drug discovery (SBDD), exemplified by HTL14242 (5) and the related pyrazole 6, both of which were crystallized using the mGlu5-StaR protein.7 During the course of our mGlu5 structural biology campaign, several intriguing differences in the allosteric binding sites were observed, most notably in the region of the upper hydrophobic chamber (occupied by the octahydro-1H-indole portion of 1 and the substituted phenyl of 5 and 6), where a rotameric change in the orientation of Trp7856.50 (superscripts indicate Ballesteros−Weinstein nomenclature24) results in a significantly smaller upper portion of the allosteric site in the latter structures. Inspired by the differences between these acetylenic and nonacetylenic structures, we sought to generate further crystallographic information from complementary chemotypes. M-MPEP (3) is a high affinity mGlu5 NAM, commonly used as a tritiated radioligand,25 and has a chemical structure that shares an acetylene central linker with mavoglurant. Fenobam (4), originally progressed as an anxiolytic compound several decades ago, was progressed into clinical trials more recently for the treatment of FXS and features a urea as the linkage between two-ring systems. Both compounds yielded high-resolution

combined to explore the biologically relevant chemistry space and identify rational and serendipitous elements of the SAR. The advent of GPCR structural biology, however, provides a more rational alternative to this approach. Insights from X-ray analysis of several GPCR-allosteric modulator complexes have revealed the diverse nature of interactions that modulate receptor function,6 and an understanding of the structural details of allosteric modulation of the metabotropic glutamate receptor 5 (mGlu5) has allowed structure-based drug design to be applied to this target.7 In this article, we disclose the highresolution X-ray structures of mGlu5 cocrystallized with two negative allosteric modulators (NAMs), which enabled us to use the largest amount of available allosteric structural information (five NAMs from three chemotypes) to provide a new strategy to rationalize and overcome medicinal chemistry challenges associated with the optimization of allosteric modulators. The metabotropic glutamate (mGlu) receptors are a family of eight G protein-coupled receptors (GPCRs), subdivided into groups I, II , and III based on pharmacology, sequence homology, and preferred signaling pathways.8 L-Glutamic acid (glutamate), the major excitatory neurotransmitter in the central nervous system (CNS) of mammals, activates the receptors, and modulation of glutamate transmission has potential in the treatment of a wide range of neurological disorders. Glutamate binds to mGlu receptors in a large extracellular binding domain, commonly known as the Venus flytrap (VFT) domain, which is linked via a cysteine-rich domain to the seven transmembrane domain (TMD). Early drug discovery efforts focused on the orthosteric site in the VFT and resulted in the identification of acidic glutamate-based compounds, which failed to demonstrate appropriate profiles for clinical progression, with pharmacokinetic, CNS penetration, and selectivity considerations as the most notable challenges.9 Subsequent targeting of alternative binding sites (allosteric sites) within the TMD has yielded greater success, with a number of allosteric modulators progressing to clinical studies, as reviewed comprehensively elsewhere.4,10−12 The mGlu5 receptor is located postsynaptically and belongs to group I, coupling via the Gq/11 family of G proteins and playing a key role in regulating the activity of NMDA receptors. Molecules binding to an allosteric site that act as noncompetitive antagonists are termed negative allosteric modulators (NAMs), and mGlu5 NAMs have potential clinical utility 208

DOI: 10.1021/acs.jmedchem.7b01722 J. Med. Chem. 2019, 62, 207−222

Journal of Medicinal Chemistry

Article

Figure 1. mGlu5-StaR crystal structures with M-MPEP (3, PDB: ID 6FFI) and fenobam (4, PDB ID: 6FFH) bound in the allosteric site. Cartoon representations of the mGlu5-StaR-3 (left) and -4 (right) structures are shown, with interesting binding site residues highlighted as gray sticks. Compounds 3 (yellow, left) and 4 (purple, right) are represented as sticks together with the 1.0 σ contoured m|Fo| − d|Fc| electron density maps (green mesh) carved around the respective ligands. Interesting binding site residues are shown as sticks. The conserved binding site water molecule is shown as a blue sphere, and hydrogen bonding interactions are represented as dashed red lines.

anchored through a disulfide bridge between the ECL2 Cys733 and Cys6443.29 at the N-terminal end of TM3. The tight packing of ECL2 on top of the mGlu5-StaR TMD occludes access to the allosteric binding site located ∼8 Å from the receptor surface. The allosteric pocket is made up of a lower and an upper chamber connected by a narrow linker region. In the mGlu5-StaR-M-MPEP or fenobam complexes, this allosteric pocket houses strong density, allowing unambiguous fit of these ligands in the sigma-A weighted 2m|Fo| − d|Fc| and m|Fo| − d|Fc| electron density maps (Figure 1). For both complexes, the upper chamber Trp7856.50 adopts a “swung-in” χ1 trans-rotameric state, with its indole ring nitrogen hydrogenbonding to the Tyr6593.44 and Ser8097.39 side chains. This conformation has also been observed in the mGlu5-StaR-5 and -6 complex structures,7 as opposed to the “swung-out” rotameric state observed in the mGlu5-StaR-mavoglurant complex,23 and results in a markedly smaller binding pocket size. The Ser8057.35 side chain is oriented toward the exterior of the TM helix bundle as in the mGlu5-StaR-5 and -6 complex structures as opposed to the mGlu5-StaR-mavoglurant structure where it is involved in hydrogen bonding to the mavoglurant hydroxyl group. In brief, the side chain orientations adopted by Trp7856.50 and Ser8097.39 in the mGlu5-StaR-M-MPEP and -fenobam structures reported here are similar to those in the mGlu5-StaR-5 and -6 complex structures. This allows the MMPEP methoxybenzene group and the fenobam 3-methyl-4Himidazol-5-one, housed in the allosteric site upper chamber, to make π−π interactions with Trp7856.50 and edge to face contacts with Phe7886.53. The above moieties are further enclosed in a hydrophobic cavity lined with Ile6513.36, Leu7445.44, Ile7846.49, Met8027.32, and Ser8057.35. The lower chamber of the allosteric pocket encases the M-MPEP 2methylpyridine or the fenobam chlorobenzene. The M-MPEP 2-methylpyridine makes van der Waals contacts with Ile6252.46, Ser6543.39, Ser6583.43, Tyr6593.44, Ser8097.39, Ala8107.40, and Ala8137.43. The fenobam chlorobenzene moiety makes similar interactions to the M-MPEP 2-methylpyridine but is also in close proximity to a water molecule coordinated by Tyr6593.44, Ser8097.39, and Thr7816.46 (Figure 1). This water molecule is conserved across the different mGlu5-StaR structures reported so far7,23 and is involved in hydrogen-bonding with compounds 5 and 6.

protein−ligand structures, described below, that reveal additional unexpected features. In this article, we describe the fascinating insights that the new crystal structures bring to our understanding of the mGlu5 allosteric site and subsequently use the broadened structural understanding in a prototype study examining the SAR of published chemotypes in the context of water-mediated interactions. Our analysis indicates that increases in affinity can result from a beneficial perturbation of the water network for compounds in the low affinity range, while in addition to an optimal water network, favorable interactions between the ligand and protein are also significant in high affinity molecules.



RESULTS Structure Determination of mGlu5 in Complex with M-MPEP and Fenobam. The structures of the mGlu5 TMD complexes with the allosteric modulators M-MPEP and fenobam were determined using a thermostabilized receptor (StaR).23 mGlu5 was thermostabilized in the presence of the allosteric radioligand [3H]-M-MPEP, resulting in a StaR containing six mutations, located away from the allosteric site. The mGlu5-StaR was further modified to facilitate crystallization in lipidic cubic phase (LCP), by removing the Nterminal extracellular domain (residues 2−568) as well as the unstructured C-terminus (residues 833−1153), and finally inserting T4-lysozyme (T4L) into intracellular loop (ICL) 2 between Lys678 and Lys679. The structures of the mGlu5 TMD bound to M-MPEP (PDB ID: 6FFI) and fenobam (PDB ID: 6FFH) were determined to 2.2 and 2.6 Å, respectively, through merging diffraction data from multiple crystals grown in lipidic cubic phase (LCP). The structures were solved by molecular replacement using the mGlu5 mavoglurant X-ray structure (PDB ID: 4OO9) as the model.23 Details of data collection and refinement statistics for both structures are given in Supporting Information Table 1. For both structures, a similar overall organization for the seven transmembrane helices is observed (Figure 1) as in previously reported mGlu5 structures.7,23 ICL1 and ICL3, as well as extracellular loop (ECL) 1 and ECL3, are well-defined in electron density maps. ECL2 extensively interacts with the apical surface of the receptor, engaging in contacts with ECL1, the TM1 N-terminus, and the TM2 C-terminus, and is 209

DOI: 10.1021/acs.jmedchem.7b01722 J. Med. Chem. 2019, 62, 207−222

Journal of Medicinal Chemistry

Article

Figure 2. (A) X-ray structure of M-MPEP (PDB ID: 6FFI, carbon atoms purple) with the GRID C3 probe in gray to show the shape of the site (contoured at 1 kcal/mol) and the GRID C1 probe to show the lipophilic hotspots (contoured in light yellow at −2.8 kcal/mol, yellow at −3.8 kcal/mol, and orange at −4.1 kcal/mol). (B) Overlay of the X-ray structures of M-MPEP (carbon atoms in purple of ligand and protein) and mavoglurant (PDB ID: 4OO9) (carbon atoms of ligand and protein in green) showing the different orientation of growing vectors from the head groups and the larger binding site of mavoglurant (GRID C3 surface in green).

the site, quantifying the highly lipophilic/hydrophobic nature of this allosteric binding site. A much stronger level than normal of −3.8 kcal/mol is needed to see the lipophilic hotspots, which is shown in Figure 2A, together with the −4.1 kcal contour to highlight the most lipophilic regions. M-MPEP shares an acetylenic linker with mavoglurant, but in their crystal structures this feature does not overlay. This key observation, and thus the different vectors for substitution, is shown in Figure 2B, together with the GRID C3 surfaces highlighting the much larger binding site in the mavoglurant structure due to the flip movement of Trp7856.50 (see Figure 2B, Trp in MMPEP near the bottom of the saturated ring system of mavoglurant and in mavoglurant behind the methoxy substituent of M-MPEP). The repercussions of this for design are important, as a model of M-MPEP binding based on modifying the mavoglurant structure would give multiple reasonable looking, but ambiguous, binding modes. Considering the top 10 poses in a mavoglurant-based docking model, we identified two binding mode clusters. In the first one, we found binding modes that resemble that of the X-ray structure, in the other one, however, the methoxyphenyl group of M-MPEP is rotated by 180° (Figure 3A). This is due to the position of Trp7856.50, which is flipped out of the binding site in the X-ray structure of the larger ligand mavoglurant,7 rotating back into the site accompanied by other subtle backbone and side chain movements. It highlights that each ligand induces a different binding site conformation that is nontransferable for correct cross-docking since most of the docking tools treat the protein environment as being rigid. In our case, docking showed acceptable sampling of the potential ligand binding modes but scoring failed to identify the experimental pose. In fact, Glide scores calculated for the docking solutions were similar and less favorable than that of the experimental pose. Hence, as the docking results suggest ambiguous vectors for substitution of the phenyl ring, subsequent optimization and scaffold hopping designs based on the docked poses would be severely compromised. Figure 3B illustrates the much deeper binding site (toward the intracellular side) of the Class C M-MPEP ligand compared to the normal Class A orthosteric binding site

The M-MPEP methoxybenzene and 2-methylpyridine moieties are connected by an acetylene linker, whereas the fenobam 3-methyl-4H-imidazol-5-one and chlorobenzene groups are connected by a urea group. These linkers traverse the tunnel connecting the upper and lower chambers forming the mGlu5 allosteric binding site and interact with Val8067.36, Ser8097.39, and Pro6553.40 (Figure 1). Unexpected Binding Modes and Insights from Multiple Cocrystal Structures. The X-ray crystal structures of MMPEP and fenobam provide a number of unanticipated insights into how ligands bind to the mGlu5 receptor. The first two structures of allosteric binders (in the TMD) of mGlu receptors, mavoglurant in mGlu5 (PDB ID: 4OO9) and FITM (also known as FM9) in mGlu1 (PDB ID: 4OR2), showed that the ligands bind in different adjacent regions of the TMD part of the receptor. The mGlu5 ligand binds deeper than the mGlu1 ligand, inducing a pocket in a region that is just a narrow (water) channel in the mGlu1 structure (see later discussion and Figure 3B−D). This illustrates the importance of structures for each GPCR subtype, and the dangers of homology modeling where the mGlu5 binding pocket could be reasonably modeled in the higher region where the mGlu1 ligand binds. The two new structures described here also illustrate the importance of having multiple ligand structures for each subtype, as both ligands have been previously proposed (fenobam) or can be fitted (M-MPEP, shown here) in a reverse binding mode using the receptor structure from another ligand. The X-ray structures of M-MPEP and fenobam show that the ligands bind in a similar region to mavoglurant, but with an important flip movement of Trp7856.50 into the binding site, making the binding site much smaller. This further illustrates the importance of having multiple ligand costructures for a target, and the dangers of using a single ligand X-ray complex structure for structure-based drug design (where either the binding site is too small or is induced and too large or different for the new ligand, leading to wrong docking poses or an inability to dock). The pseudo-apo structure of M-MPEP (ligand removed from the induced fit structure) reveals that the binding site is very lipophilic, the −2.8 kcal/mol contour (as used previously26,27) of the GRID C1 lipophilic probe fills 210

DOI: 10.1021/acs.jmedchem.7b01722 J. Med. Chem. 2019, 62, 207−222

Journal of Medicinal Chemistry

Article

Figure 3. (A) X-ray structure of M-MPEP (PDB ID: 6FFI, helices and carbon atoms in purple) and the docking model of M-MPEP based on mavoglurant (carbon atoms in green). The protein X-ray structures of M-MPEP and mavoglurant are shown with the carbons colored purple and green, respectively, with the shape of the mGlu5 receptor with M-MPEP bound in mesh purple (GRID C3 probe contoured at 1 kcal/mol); three carbons of the cyclohexyl ring of mavoglurant (wireframe, green) can be seen below the phenyl ring of the docking model of M-MPEP (stick, green). (B) Overlay of the X-ray structures of M-MPEP (stick, helices, and carbon atoms of ligand and protein in purple), FITM ligand from mGlu1 X-ray structure (PDB ID: 4OR2) (stick, carbon atoms in gray), and a Class A adenosine A2A antagonist ligand (wireframe, carbon atoms in green) (triazine 4g, PDB ID: 3UZA). The shape of the mGlu1 receptor with FITM bound is shown in transparent solid gray (GRID C3 probe contoured at 1 kcal/ mol) and of the mGlu5 receptor with M-MPEP bound in mesh purple (GRID C3 probe contoured at 1 kcal/mol); it can be clearly seen that the binding site of the mGlu5 receptor with M-MPEP bound is not present in the mGlu5 structure, just a hint of a binding site around the methylpyridine part of the ligand. (C) Water molecules generated by WaterFLAP in the binding pocket of mGlu1 receptor cocrystallized with FITM (calculation done with ligand bound). High energy water molecules (red and yellow) are located in the lower chamber of the pocket; these two waters have poor enthalpic energy, with GRID water probe (OH2) energies of only −9.1 and −5.4 kcal/mol (by comparison, the GRID water probe energies for the two blue “happy” waters in the upper region are −15.1 and −13.1 kcal/mol). (D) Overlay of M-MPEP (carbon atoms in green) from the mGlu5 Xray structure and FITM from the mGlu1 X-ray structure (PDB ID: 4OR2) (carbon atoms in orange), indicating that the M-MPEP binding site in mGlu5 overlaps with the lower chamber of mGlu1 that accommodates high energy waters.

mediated, highlighting again the importance of multiple protein−ligand X-ray structures for a target and the key need to consider full water networks when designing new ligands and understanding binding modes and SAR. The urea places the aromatic rings similarly to the acetylenic linker (Figure 4A). Interestingly, the top ring carbonyl group is not exposed to solvent in the main channel but sits in a mainly lipophilic pocket and, as shown in Figure 4B, is predicted from water network calculations to make a productive H-bond interaction (2.8 Å) with a water molecule through a small channel facing the membrane. The carbonyl group of the urea does make a strong H-bond to Ser8097.39 (2.9 Å), enabling it to pierce the narrow channel defined by the lipophilic carbon atom probe. The urea NH, however, makes no direct H-bond interactions

(illustrated by a triazine adenosine A2A antagonist (PDB ID: 3UZA)).28 Also shown are the different positions of the binding site compared to the FITM ligand in the mGlu1 X-ray structure (PDB ID: 4OR2),29 which binds higher in the allosteric channel, bridging the mGlu5 M-MPEP ligand and the Class A triazine adenosine A2A antagonist. These significantly different binding locations in the TMD highlight the difficulty and dangers of homology modeling without significant appropriate experimental data (e.g., mutation data) to increase confidence in a predicted binding position and orientation. The X-ray crystal structure of fenobam is now discussed. The structure shows a binding mode that is the inverse of a previously published one, based on induced fit docking.30 The structure shows that most of the polar interactions are water 211

DOI: 10.1021/acs.jmedchem.7b01722 J. Med. Chem. 2019, 62, 207−222

Journal of Medicinal Chemistry

Article

Figure 4. (A) X-ray structure of fenobam (PDB ID: 6FFH, carbon atoms in cyan) with the GRID C3 probe in gray to show the shape of the site (contoured at 1 kcal/mol) and the GRID C1 probe in yellow to show the lipophilic hotspots (contoured at −3.8 kcal/mol). (B) X-ray structure of fenobam, as in (A), but with the addition of the water network of the complex calculated by WaterFLAP (see later section) and the GRID water probe (OH2) in green to show H-bonding hotspots (contoured at −6 kcal/mol). Some of the H-bonding network involving the ligand, waters, and protein is shown as green dashed lines.

Figure 5. (A) Overlap of the X-ray structure of mavoglurant (carbon atoms in green), M-MPEP (carbon atoms in purple), fenobam (carbon atoms in cyan), and the pyrimidine ligands 5 (PDB ID: 5GCD) and 6 (PDB ID: 5CGC) (carbon atoms in gray), with the corresponding protein structure carbon atoms colored the same as the ligand, with the GRID C3 probe in gray to show the shape of the site (contoured at 1 kcal/mol) and the GRID C1 probe in yellow to show the lipophilic hotspots (contoured at −3.8 kcal/mol), from the M-MPEP protein structure. (B) Same as (A), with the addition of the C3 shape surfaces for mavoglurant in green and 6 in blue, showing how the pyrimidine ring of 6 penetrates the surface of both the MMPEP and mavoglurant structures.

ligands 5 (PDB ID: 5GCD) and 6 (PDB ID: 5CGC) is shown in Figure 5. It can be seen that the pyrimidine linker ring of ligands 5 and 6 significantly induces a larger binding site than in fenobam or even mavoglurant (Figure 5B). Comparison of the binding modes revealed that due to the common hydrophobic/ lipophilic and aromatic pharmacophores, all ligands overlay in the upper and lower hydrophobic/lipophilic cavities. The quite limited overlap of H-bonding pharmacophoric groups, however, can also be seen, which would make pharmacophore-based modeling very challenging to interpret. Optimization against Water: SAR Studies on mGlu5 NAMs. Potency optimization typically involves the formation of new interactions between the ligand and its protein target, which are often considered to form exclusively between the interacting partners. However, an increasing number of studies reveal that binding site water molecules play an important role in these interactions.27,31−34 In many cases, ligand−protein interactions are water mediated, making SAR analyses

with the protein, and analysis of the calculated water network shows a strong H-bond (2.7 Å) to a water (see Figure 4B), which also interacts with a backbone isoleucine carbonyl group and a serine; it could also act as a donor with the ligand exocyclic NC group, particularly if it moves a little toward it along the water channel (visible in Figure 4B in the green GRID water probe hotspots). The ring NH also does not make direct H-bonds with the protein but makes an internal H-bond (2.7 Å) to the carbonyl CO of the urea, stabilizing the ligand in the binding conformation. In the absence of a ligand H-bond acceptor group, Ser8097.39 can be seen from the other X-ray structures to make H-bonds to nearby tyrosine and tryptophan residues. The fenobam structure is thus interesting due to its mainly water-mediated polar interactions that emphasize the need to consider the final water network to understand the binding and undertake rational structure-based drug design. A superposition of all the mGlu5 ligand X-ray structures, mavoglurant, M-MPEP, fenobam and the previously published 212

DOI: 10.1021/acs.jmedchem.7b01722 J. Med. Chem. 2019, 62, 207−222

Journal of Medicinal Chemistry

Article

Figure 6. (A) Water molecules generated by WaterFLAP in the pseudoapo binding pocket extracted from the mGlu5-M-MPEP complex. The orientation is consistent with Figure 5. The blue (“happy”) water molecule in the lower region is the one that is consistently present in mGlu5 X-ray structures. (B) Water molecules replaced by M-MPEP inserted into the binding pocket in the position found in the X-ray structure. Waters are colored by their ΔG values: the most “unhappy” waters are red, then yellow, gray (bulk water), and blue (“happy” waters). The pocket surface (CH3 methyl probe) colored as gray at −1.5 kcal/mol and lipophilic probe (C1 probe) as light yellow at 4.1 kcal/mol.

S8097.39) and two nonconserved (P6553.40 and A8107.40) residues play a key role in the PAM activity.35 Interestingly, all of these amino acids form interactions with a network of multiple waters, and all but P6553.40 are located in the lower chamber of the binding site (Figure 6A). These observations help to explain why small changes in the ligand structure of certain chemotypes in this region can result in switching between NAM and PAM pharmacologies and suggest that perturbation of the lower water network is possibly one component of the mode switch. Next, we used the same procedure for the mGlu5-StaR-MMPEP complex (Figure 7) to analyze the effect of the ligand on the water network. WaterFLAP generates several water molecules around M-MPEP. Moreover, it assigns ΔΔG values to waters that are the relative free energies in the complex compared to those in the empty receptor. Several waters in the binding pocket are stabilized, having lower free energy in the

extremely complex and pharmacophore modeling and docking calculations challenging. The situation can be even more difficult when ligand interactions are almost exclusively water mediated, and the potency gain can be traced back to the displacement of energetically unfavored waters and/or the perturbation of water networks in the binding pocket. Targeting functional water channels in allosteric sites in GPCRs is a good example of such a challenging optimization path. As a prototype study in this article, we now investigate mGlu5 NAM programs analyzing the published SAR in the context of water-mediated interactions formed during the optimization of these chemotypes. SAR Analysis of M-MPEP Analogues. Water molecules were generated using WaterFLAP in the binding site of the mGlu5-StaR-M-MPEP receptor complex using the highresolution mavoglurant X-ray structure23 without the ligand in order to estimate the position of water molecules together with their free energies (Figure 6A). The cocrystallized MMPEP was then inserted into the water filled binding pocket in the position observed in the X-ray structure described earlier. This procedure suggests that 13 water molecules are replaced by M-MPEP. The free energies assigned to these water molecules by WaterFLAP are all positive (“unhappy” water molecules), which is in line with the lipophilic character of the site discussed above (Figure 6A). There is a set of water molecules in the bottom of the pocket of the pseudoapo structure, some of which appear to remain in the pocket after ligand binding. It is noteworthy that one of these water molecules is consistently present in the available X-ray structures of mGlu5 complexes.7,23 WaterFLAP predicts this water molecule to have advantageous free energy, which is in line with its presence in the X-ray complexes (see Figure 6A). By contrast, the other water molecules generated by WaterFLAP in the bottom of the pocket are associated with high free energy, and their replacement or stabilization by the ligand is therefore beneficial. Indeed, M-MPEP displaces two of these water molecules (see red water molecules in the bottom of Figure 6B). It has also been proposed that the perturbation of the water network in the bottom of the pocket may affect the activation mechanism of mGlu5.23 Site-directed mutagenesis data has revealed that three conserved (Y6593.44, T7816.46, and

Figure 7. WaterFLAP generated water molecules in the mGlu5-StaRM-MPEP complex. Waters are colored by ΔΔG: the most destabilized waters are red, then yellow, gray (unchanged), and blue (stabilized). The pocket surface (CH3 methyl probe) colored as gray at −1.5 kcal/ mol and lipophilic probe (C1 probe) as light yellow at 4.1 kcal/mol. 213

DOI: 10.1021/acs.jmedchem.7b01722 J. Med. Chem. 2019, 62, 207−222

Journal of Medicinal Chemistry

Article

Table 1. Structure−Activity Data of the M-MPEP Chemotype

IC50 values are expressed as mean ± SEM, n = 2.36 IC50 values represent the ability of compounds to inhibit agonist-induced phosphoinositide hydrolysis. bGroup efficiency (see text). a

increases can be achieved by substitution at the 6-position of the pyridine ring (R2 substituent with X = N in Table 1 and bottom ring in Figure 8; compare compounds 7 and 8). It is also potentially beneficial to put a substituent in the meta position of the phenyl ring toward the extracellular region (R1 in Table 1 and top ring in Figure 8; compare the activity of compound 8 with 3 and 30 SI Table S2). Analyzing the water network in the pseudoapo protein (Figure 6A) together with the binding hot spots (Figure 8) reveals a connection between the position of high energy (red and yellow) water molecules and the beneficial (red) regions of the binding pocket. Comparing the water networks in pairs of docked complexes allows, in certain cases, the interpretation of activity differences of ligands. Comparison of the water network in the complexes of ligands 8 (IC50 = 24 nM) and 9 (IC50 = 1961 nM) reveals that extension of the 6-position substituent in the pyridine ring leads to the destabilization of a water molecule (Figure 9), which we propose contributes to the significantly lower activity of 9 with respect to 8. Another comparison can be made at the other side of the MMPEP chemotype in the context of the R1 substituent. The different size and orientation of the R1 aromatic rings in compounds 10 and 11 (Figure 10) impact the affinity of these compounds significantly (Table 1). The difference between the water networks include the presence of a stabilized water molecule in the complex of 10 (top middle), which is displaced by 11 and also the destabilization of two other water molecules by 11 (top right). SAR Analysis of MTEP Analogues. A related acetylene series of mGlu5 NAMs where the upper pyridine group is bioisosterically replaced by a thiazole has been reported, with the headline compound (18, Table 2, hereafter referred to as MTEP) displaying superior selectivity and in vivo activity to MPEP (2).40 Since the X-ray structure of the mGlu5-MTEP complex is not available, this chemical series serves to explore the limitation of our approach. Given the structural similarity

M-MPEP complex than in the absence of the ligand, which is in line with the high activity of M-MPEP. We next investigated the structure−activity relationship within the M-MPEP series of compounds (Table 1 and Supporting Information (SI) Table S2). Ligands with available mGlu5 inhibitory activity in a phosphoinositide hydrolysis assay36 were first docked into the pseudoapo mGlu5 structure obtained from the mGlu5-StaR-M-MPEP complex X-ray structure after the removal of the ligand and water molecules. Table 1 shows the ligands together with their measured activity and group efficiency values (GE).37 Ligand efficiencies (LE)38 are properties of entire molecules, whereas group efficiencies characterize the binding efficiency of parts of a molecule; hence, we have chosen to evaluate the data set in the context of GE to identify aspects of the SAR that contribute significantly to affinity. Group efficiencies were calculated with compound 7 as reference, and they were used in hot spot calculations39 that indicate the maximal contribution of binding site regions to the binding affinity (Figure 8). The analysis revealed that activity

Figure 8. Binding hot spots of mGlu5 obtained with M-MPEP analogues (compounds 3 and 7−11). Color code: blue are the least favorable and red are the most favorable hot spots. The surface was colored according to the maximum GE values for each protein atom. 214

DOI: 10.1021/acs.jmedchem.7b01722 J. Med. Chem. 2019, 62, 207−222

Journal of Medicinal Chemistry

Article

Figure 9. Water network of mGlu5 docked complexes of (A) compound 8 and (B) compound 9. Waters are colored by ΔΔG values: the most destabilized waters are red, then yellow, gray (unchanged), and blue (stabilized).

Figure 10. Water network of mGlu5 complexes of (A) compound 10 and (B) compound 11. Waters are colored by ΔΔG values: the most destabilized waters are red, then yellow, gray (unchanged), and blue (stabilized).

SI Table S3) were used to generate hot spot maps at the MTEP binding site (Figure 13). The hot spot map reflects that substitution on the methyl thiazole ring is not favorable. As in the case of many allosteric optimization programs, a “fluorine-walk” strategy was used to identify the variable region of the ligand by screening the impact of lipophilic, hydrogen bonding, and van der Waals interactions. Systematic fluorine substitution around the phenyl group revealed that the meta position is beneficial (12−14). The presence of the pyridine ring connecting to the acetylene tether is favorable as indicated by a red area in the hot spot map. Our hot spot analysis shows that substitution at the 3position of the A ring is potentially beneficial, although there is a significant variation in the activity of compounds 15, 16, 17, and 19 (further examples are 47−56 in SI Table S3). Some extended substituents are suboptimal (e.g., 16 and 17), while others result in high activity (e.g., 19 and 50 (SI Table S3)). The low activity of 16 and 17 appears to stem from the steric clash of the NHAc and CONH2 substituents with Phe7886.53, while the smaller cyano (50) and vinyl (19) groups are well accommodated in the binding pocket. Moreover, different ligands perturb the water network in different ways, and this contributes to the variations in activity. Considering the freeenergy changes of the water molecules upon complex formation with compounds 16 and 19, it can be seen in Figure 14 that 16 displaces two water molecules that are stabilized by 19 (top

between M-MPEP and MTEP, we hypothesize that MTEP binds with a similar binding mode to that of MTEP. In fact, induced fit docking of MTEP and its analogues into the pseudoapo form of the mGlu5-StaR-M-MPEP X-ray structure resulted in docked complexes with ligand positions in good accordance with the M-MPEP X-ray structure. (The RMSD of the heavy atoms common in MTEP and M-MPEP molecules is 0.148 Å.) The WaterFLAP generated water network agrees with that of Figure 6A, while the water molecules repelled by the insertion of MTEP are shown in Figure 11. MTEP displaces two water molecules with high excess energy (red) from the subpocket at the intracellular side (bottom of Figure 11) similarly to M-MPEP (Figure 6A). Binding site analysis and water network generation was also performed for the docked mGlu5-MTEP complex (Figure 12). MTEP fits well into the lipophilic part of the pocket. Although there are a number of water molecules whose energy is decreased with the binding of MTEP the presence of a few destabilized waters suggests that the activity of the ligand can be improved. Comparing the water molecules present in the complexes of M-MPEP and MTEP, several differences can be identified. It is notable that although the same set of waters appears in the intracellular part of the binding pocket, one of them is stabilized (blue) by M-MPEP (Figure 7) and energetically unchanged (gray) by MTEP (Figure 12). A number of MTEP analogues with available inhibitory activity in a phosphoinositide hydrolysis assay41 (Table 2 and 215

DOI: 10.1021/acs.jmedchem.7b01722 J. Med. Chem. 2019, 62, 207−222

Journal of Medicinal Chemistry

Article

Table 2. Structure−Activity Data of the MTEP Chemotype

Figure 11. Water molecules replaced by MTEP inserted into the binding pocket in the position obtained by docking MTEP into the mGlu5-StaR-M-MPEP complex structure after the removal of the ligand. Waters are colored by their ΔΔG values: the most “unhappy” waters are red, then yellow, gray (bulk water), and blue “happy” waters. The pocket surface (CH3 methyl probe) colored as gray at −1.5 kcal/mol and lipophilic probe (C1 probe) as light yellow at 4.1 kcal/mol.

IC50 values are expressed as mean ± SEM, n = 2.41 IC50 values represent the ability of compounds to inhibit agonist-induced phosphoinositide hydrolysis. bGroup efficiency (see text). a

right) and that there are less unfavorable waters around 19 (middle left). Comparison of the water networks around compounds 15 (methyl) and 19 (vinyl) shows that the binding of the latter compound has a beneficial effect on selected water molecules (Figure 15); in the case of 19, three water molecules have more favorable free energy (top left), and in addition, an extra “happy” water molecule (top middle) appears in the network. This reorganization supports the significantly higher activity of 19 compared to 15. In conclusion, it can be clearly seen that different ligands perturb the water network in different ways, and this can contribute significantly to the variations in activity. Thus, the 1000-fold increase in activity of compounds 19 vs 16 can be understood by considering the final water network, where 16 has many more unfavorable waters; it also displaces waters that are stabilized by 19. Water Network and Ligand−Protein Interactions. The available X-ray structures and comprehensive SAR data for MMPEP 3 (Table 1 and SI Table S2), MTEP 18 (Table 2 and SI Table S3), mavoglurant 1 (SI discussion and Table S4), HTL14242 5 and pyrazole 6 (SI discussion and Tables S5 and S6), and fenobam 4 (SI discussion and Table S7) allowed us to investigate the contribution of binding site water molecules to the affinity of analogues in six chemical series of three

Figure 12. WaterFLAP generated water network in the docked mGlu5MTEP complex. Waters are colored by ΔΔG values: the most destabilized waters are red, then yellow, gray (unchanged), and blue (stabilized). The pocket surface (CH3 methyl probe) colored as gray at −1.5 kcal/mol and lipophilic probe (C1 probe) as light yellow at 4.1 kcal/mol.

chemotypes. We have tried to demonstrate that comparative analysis of water networks helps the interpretation of SAR and that this understanding can contribute to optimization within a chemical series. This may be especially important for allosteric binders since their SAR is often flat or irreconcilable.3 We have also attempted to quantitatively characterize the role of the water network on ligand affinity differences. For each ligand, we calculated the free energy of network waters (ΔGwater) as the sum of water free energy differences between the complex and the corresponding pseudoapo structure (∑iperturbed ΔΔGi) minus the total free energy of the displaced water molecules (∑displaced ΔGj) (eq 1). j 216

DOI: 10.1021/acs.jmedchem.7b01722 J. Med. Chem. 2019, 62, 207−222

Journal of Medicinal Chemistry

Article

energy gain between the ligand and the protein importantly contributes to the affinity increase for high affinity ligands (pIC50 > 8). In this high affinity range, the water network freeenergy gain (Figure 16) appears not to significantly contribute to the affinity increase. Considering the intermediate affinity range (pIC50 between 6 and 8), both the free energy gain of water networks and the ligand−protein interaction energy show slight increases with affinity, indicating that these factors are at most partially responsible for the affinity increase and implying the importance of other factors such as favorable entropy change or conformational rearrangement of the interacting partners. Taken together, Figures 16 and 17 suggest that affinity increases in the low affinity range are the result of the beneficial perturbation of the water network, while in the high affinity range, in addition to the optimal water network, favorable interactions between the ligand and protein also play a significant role. It seems that starting points and compounds identified during the early stage of optimization can benefit from the reorganization of water networks. This includes displacing or replacing unfavored waters from or adopting favored waters into the network. Identification of the appropriate growing vectors helps to optimize the free energy gain from the water network. The obvious limitation of this gain in energy is the best available binding mode that maximizes ΔGwater within the pocket. Further improvement of the binding affinity needs more ligand protein interactions to be induced.

Figure 13. Binding hot spots of mGlu5 obtained with MTEP analogues (compounds 12−19). Color code: blue are the least favorable, and red are the most favorable hot spots. The surface was colored according to the maximum GE values for each protein atom. perturbed

ΔGwater =

∑ i

displaced

ΔΔGi −



ΔGj

j

(1)

ΔGwater is the free energy change of the water network upon ligand binding that was correlated with measured activities using six different chemical series including M-MPEP, MTEP, fenobam, mavoglurant, 5, and 6 (Figure 16 and Supporting Information). Since ligand protein interactions should also contribute significantly to the binding affinity, we calculated ligand− protein interaction energies for the same series of mGlu5 NAMs. Complex structures were therefore minimized42 without solvent, energies were calculated for each complex, and also for the separate ligand and the isolated protein taken from the complex. The interaction energy was obtained using eq 2, where E(P,L) is the energy of the complex, E(P) is the energy of the protein, and E(L) is the energy of the ligand. ΔE int = E(P , L) − (E(P) + E(L))



DISCUSSION AND CONCLUSIONS Similar to other allosteric modulator medicinal chemistry programs, the optimization of allosteric mGlu5 ligands is notoriously difficult.22,25,36,43−49 Allosteric ligands bind in the transmembrane domain and may intrude into a narrow subpocket accessible in mGlu5 but not in other mGlu subtypes. This is in line with the observed subtype selectivity of the ligands but also responsible for the steep SAR with limited transferability between different chemotypes. We have shown that induced fit effects are significant in the small volume allosteric site and that this can explain the SAR challenges. The allosteric site of the mGlu5 receptor is located in a functional water channel, and the binding pocket is therefore occupied by water molecules that are at least partially ordered and thought to play a role in signal transduction. On one hand, they might be responsible for pharmacological mode-switching, whereby

(2)

The means of the interaction energies within pIC50 bins are shown in Figure 17. Results in Figure 16 show that in the low affinity range, affinity of ligands increases with the free energy gain of water networks (going from pIC50 < 6 toward higher pIC50s). In this affinity range, the interaction energy (Figure 17) exhibits no large variation. However, Figure 17 suggests that the interaction

Figure 14. Water network of mGlu5 complexes of (A) compound 16 and (B) compound 19. Waters are colored by ΔΔG values: the most destabilized waters are red, then yellow, gray (unchanged), and blue (stabilized). 217

DOI: 10.1021/acs.jmedchem.7b01722 J. Med. Chem. 2019, 62, 207−222

Journal of Medicinal Chemistry

Article

Figure 15. Water network of mGlu5 complexes of (A) compound 15 and (B) compound 19. Waters are colored by ΔΔG values: the most destabilized waters are red, then yellow, gray (unchanged), and blue (stabilized).

Figure 16. Free-energy change (kcal/mol) of the water network upon ligand binding. Means and standard errors for pIC50 bins are shown.

Figure 17. Calculated interaction energies of ligand−protein complexes. Means and standard errors of pIC50 bins are shown.

receptor interactions. An improved interpretation of SAR can be achieved by considering the perturbation of the water network in complex formation. Ligand binding leads to the expulsion of water molecules, and the free-energy consequence of the binding is affected by this process. The removal of high free-energy (“unhappy”) water molecules gives a favorable contribution, while the removal of the low free-energy (“happy”) water molecules gives an unfavorable contribution to the binding. Moreover, ligand binding perturbs the freeenergy of the water network that remains in the binding pocket

minor structural modifications in certain chemotypes can result in a change from a NAM mode of action to PAM or vice versa, complicating optimization of the ligand. On the other hand, they are involved in water-mediated ligand interactions that are affected or potentially eliminated by subtle structural changes, and this complicates SAR interpretation. The perturbation of the water network by allosteric ligands contributes significantly to the observed ligand affinity and functional activity, a phenomenon that potentially compromises the success of SAR studies and molecular modeling focusing on ligand− 218

DOI: 10.1021/acs.jmedchem.7b01722 J. Med. Chem. 2019, 62, 207−222

Journal of Medicinal Chemistry

Article

ethanesulfonic acid (MES) across a pH range of 5.5−6.8, 100−250 mM (NH4)2HPO4, 24−34% (v/v) polyethylene glycol 400, and supplemented with either 100 μM M-MPEP (or fenobam). Crystals were harvested and flash frozen in liquid nitrogen prior to diffraction experiments. Diffraction Data Collection, Processing, Structure Solution, Refinement, and Analysis. X-ray diffraction data were collected on beamline I24 at Diamond Light Source using a 10 μm diameter beam with a Pilatus 6 M hybrid-pixel detector. Data from individual crystals were integrated using XDS,51 and then merged and scaled using AIMLESS from the CCP4 suite.52,53 Crystals for these two complexes belonged to the monoclinic space group C121. A complete data set to 2.2 Å for mGlu5-StaR-M-MPEP was obtained by merging diffraction data from four crystals, whereas data from five crystals were required to generate a complete data set to 2.65 Å for mGlu5-StaR-fenobam. Data collection statistics are reported in SI Table S1. The structure was solved by molecular replacement using Phaser54 with the mGlu5-StaR-mavoglurant costructure (PDB ID: 4OO9)23 as the search model. Iterative rounds of manual model building in COOT55 were interspersed with refinement using phenix.ref ine56 implementing a combination of positional, individual isotropic B-factor refinement, and TLS refinement. Structure quality was assessed with MolProbity.57 Refinement statistics are presented in SI Table S1. Figures were prepared using PyMOL (Schrödinger). Computational Methods. For ligand preparation, protein preparation, docking, and ligand−protein interaction calculations, ligands were prepared with Instant JChem (version: 15.12.14.0) and were converted to SDF format. Protonation states were generated using Ligprep (Schrödinger Suite, 2016−1; Schrödinger, LLC: New York, NY, 2016. Maestro version: 11.0.015) at pH = 7. The proteins were prepared using default settings of Schrö dinger Protein Preparation Wizard (Maestro version: 11.0.015). The molecular docking was carried out with Glide standard precision with default settings. Ligand−protein interaction energy calculation started with the minimization of the complex structures without solvent, and without any constraint using MacroModel (version 11.4) and OPLS 2005 force field. The energies were calculated for each complex and also for the separate ligand and protein taken from the complex. Binding Hot Spot Calculation. Group efficiencies (GE) were calculated with the formula −ΔΔG/ΔN = [ΔG(mol B) − ΔG(mol A)]/[HA(mol B) − HA(mol A)]. GE values were assigned to the protein heavy atoms within 4 Å of the proposed binding mode of the heavy atoms of the ligands. The surface of the binding pocket was colored according to the maximum GE values of atoms, omitting those atoms that are in contact with all ligands. WaterFLAP Analyses. Protein cavities were solvated using iterative water hot spot identification with molecular interaction field analysis (GRID)58,59 and short molecular dynamics simulation (WaterFLAP version 2.2.1). The energies of the water molecules were calculated using OH2 and CRY (combining C1 and DRY) probes27 and an entropy analysis of the degrees of freedom and the shape of the energy wells of the water molecules. Scoring functions have been calibrated from the analysis of numerous reference structures and Heptares in-house experience of GPCR structures.60 Complexes for water network analysis were generated by aligning the ligands with the X-ray ligand in each series using Flexible Ligand Alignment (Maestro version: 11.0.015).

and the effect of this perturbation can also significantly influence the observed ligand affinity. We have presented examples where an analysis of the free-energy change of the water network makes it possible to interpret activity differences. The importance of the water molecules is also reflected by the difference in the free-energy change of the water network corresponding to the binding of ligands as their activity improves from micromolar to high nanomolar (Figure 16). This suggests that ligand−protein interactions are not the only factors, and even they are not always decisive for the binding of ligands. However, the presence of a favorable water network and optimized ligand−protein interactions are both required to achieve high activity at the mGlu5 receptor. In summary, the strategy we have used to retrospectively evaluate the published SAR of several mGlu5 NAM chemotypes has a number of features. These include the use of crystallographic information for the exact chemotype under optimization, the implementation of computational approaches such as GE-based hot spot analyses, water network predictions and analysis, and the tandem evaluation of SAR data in light of both structural and computational data sets. We conclude that the chance of success of future GPCR allosteric modulator drug discovery is likely to be enhanced if multiple structure-based approaches are prospectively pursued in parallel.



EXPERIMENTAL SECTION

Expression, Membrane Preparation, Protein Purification, and Crystallization of Thermostabilized mGlu5. The mGlu5StaR7 carrying an N-terminal GP64 signal sequence and a C-terminal decahistidine-tag was expressed in Sf 21 cells grown in ESF921 medium supplemented with 10% (v/v) FBS and 1% (v/v) penicillin/ streptomycin using the FastBac expression system (Invitrogen). Cells were infected at a density of 2 × 106 cells/mL with baculovirus at an approximate multiplicity of infection of 1. Cultures were grown at 27 °C and harvested 48 h post-infection. Subsequent purification steps were carried out at 4 °C. Two liters of cells were resuspended in PBS buffer supplemented with protease inhibitor tablets and 5 mM EDTA before they were disrupted using a microfluidizer at 60 PSI. Membranes were isolated by ultracentrifugation at 204 700g for 1 h and then washed with PBS buffer supplemented with protease inhibitor tablets and 500 mM NaCl. Washed membranes were collected by ultracentrifugation and resuspended in 40 mM HEPES, pH 7.5, and 250 mM NaCl and stored at −80 °C until required. Prior to solubilization, membranes were thawed, homogenized, supplemented with 30 μM ligand (M-MPEP or fenobam) and 8 mM iodoacetamide, and incubated for 40 min. Membrane solubilization was done in the presence of 1.5% (w/v) DDM for 1 h. Insoluble material was removed by ultracentrifugation, and the solubilized lysate was bound to 10 mL of Ni-NTA Superflow resin (Qiagen) for 2.5 h in the presence of 10 mM imidazole. The resin was washed with a gradient of 10−50 mM imidazole in 40 mM HEPES, pH 7.5, 250 mM NaCl, 0.05% (w/v) DDM, and 30 μM ligand over 35 column volumes before bound material was eluted with 245 mM imidazole. The complexes were further purified by gel filtration in 40 mM HEPES, pH 7.5, 150 mM NaCl, 0.03% (w/v) DDM, and 30 μM ligand. Receptor purity was analyzed using SDS-PAGE and LC−MS, and receptor monodispersity was assayed by analytical SEC. mGlu5-StaR-M-MPEP (or -fenobam) complexes were concentrated to ∼40 mg/mL before crystallization trials were performed in lipidic cubic phase at 20 °C. Briefly, the mGlu5-StaR-M-MPEP (or -fenobam) complexes were mixed with monoolein (Nu-Check) supplemented with 10% (w/w) cholesterol (Sigma-Aldrich) and 100 μM ligand using the twin-syringe method.50 The final protein/lipid ratio was 40:60 (w/w). Forty nanoliters of boli was dispensed on 96-well glass bases and overlaid with 750 nL of precipitant solution using a Mosquito LCP from TTPLabtech. Forty micrometer plate-shaped crystals of mGlu5-StaR in complex with M-MPEP (or fenobam) were grown in 100 mM 2-



ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jmedchem.7b01722. Crystallographic information table (data processing and refinement), tabulated data for M-MPEP, MTEP, mavoglurant, HTL14242, compound 6, and fenobam analogues, figures of WaterFLAP analyses, details of SAR analysis of mavoglurant, HTL14242, and fenobam 219

DOI: 10.1021/acs.jmedchem.7b01722 J. Med. Chem. 2019, 62, 207−222

Journal of Medicinal Chemistry

Article

negative allosteric modulator HTL14242 (3-Chloro-5-[6-(5-fluoropyridin-2-yl)pyrimidin-4-yl]benzonitrile. J. Med. Chem. 2015, 58, 6653− 6664. (8) Niswender, C. M.; Conn, P. J. Metabotropic glutamate receptors: physiology, pharmacology, and disease. Annu. Rev. Pharmacol. Toxicol. 2010, 50, 295−322. (9) Bennett, K. A.; Doré, A. S.; Christopher, J. A.; Weiss, D. R.; Marshall, F. H. Structures of mGluRs shed light on the challenges of drug development of allosteric modulators. Curr. Opin. Pharmacol. 2015, 20, 1−7. (10) Celanire, S.; Sebhat, I.; Wichmann, J.; Mayer, S.; Schann, S.; Gatti, S. Novel metabotropic glutamate receptor 2/3 antagonists and their therapeutic applications: a patent review (2005 - present). Expert Opin. Ther. Patents 2015, 25, 69−90. (11) Lindsley, C. W.; Stauffer, S. R. Metabotropic glutamate receptor 5-positive allosteric modulators for the treatment of schizophrenia (2004−2012). Pharm. Pat. Anal. 2013, 2, 93−108. (12) Emmitte, K. A. mGlu5 negative allosteric modulators: a patent review (2013 - 2016). Expert Opin. Ther. Pat. 2017, 27, 691−706. (13) Scharf, S. H.; Jaeschke, G.; Wettstein, J. G.; Lindemann, L. Metabotropic glutamate receptor 5 as drug target for Fragile X syndrome. Curr. Opin. Pharmacol. 2015, 20, 124−134. (14) Tison, F.; Keywood, C.; Wakefield, M.; Durif, F.; Corvol, J.-C.; Eggert, K.; Lew, M.; Isaacson, S.; Bezard, E.; Poli, S.-M.; Goetz, C. G.; Trenkwalder, C.; Rascol, O. A phase 2A trial of the novel mGluR5negative allosteric modulator dipraglurant for levodopa-induced dyskinesia in Parkinson’s disease. Mov. Disord. 2016, 31, 1373−1380. (15) Bates, B. S.; Rodriguez, A. L.; Felts, A. S.; Morrison, R. D.; Venable, D. F.; Blobaum, A. L.; Byers, F. W.; Lawson, K. P.; Daniels, J. S.; Niswender, C. M.; Jones, C. K.; Conn, P. J.; Lindsley, C. W.; Emmitte, K. A. Discovery of VU0431316: A negative allosteric modulator of mGlu5 with activity in a mouse model of anxiety. Bioorg. Med. Chem. Lett. 2014, 24, 3307−3314. (16) Zerbib, F.; Bruley des Varannes, S.; Roman, S.; Tutuian, R.; Galmiche, J. P.; Mion, F.; Tack, J.; Malfertheiner, P.; Keywood, C. Randomised clinical trial: effects of monotherapy with ADX10059, a mGluR5 inhibitor, on symptoms and reflux events in patients with gastro-oesophageal reflux disease. Aliment. Pharmacol. Ther. 2011, 33, 911−921. (17) Osikowicz, M.; Mika, J.; Przewlocka, B. The glutamatergic system as a target for neuropathic pain relief. Exp. Physiol. 2013, 98, 372−384. (18) D’Antoni, S.; Spatuzza, M.; Bonaccorso, C. M.; Musumeci, S. A.; Ciranna, L.; Nicoletti, F.; Huber, K. M.; Catania, M. V. Dysregulation of group-I metabotropic glutamate (mGlu) receptor mediated signalling in disorders associated with intellectual disability and autism. Neurosci. Biobehav. Rev. 2014, 46, 228−241. (19) Waung, M. W.; Akerman, S.; Wakefield, M.; Keywood, C.; Goadsby, P. J. Metabotropic glutamate receptor 5: a target for migraine therapy. Ann. Clin. Transl. Neurol. 2016, 3, 560−571. (20) Reilmann, R.; Rouzade-Dominguez, M.-L.; Saft, C.; Süssmuth, S. D.; Priller, J.; Rosser, A.; Rickards, H.; Schöls, L.; Pezous, N.; Gasparini, F.; Johns, D.; Landwehrmeyer, G. B.; Gomez-Mancilla, B. A randomized, placebo-controlled trial of AFQ056 for the treatment of chorea in Huntington’s disease. Mov. Disord. 2015, 30, 427−431. (21) Fuxe, K.; Borroto-Escuela, D. O. Basimglurant for treatment of major depressive disorder: a novel negative allosteric modulator of metabotropic glutamate receptor 5. Expert Opin. Invest. Drugs 2015, 24, 1247−1260. (22) Gasparini, F.; Lingenhöhl, K.; Stoehr, N.; Flor, P. J.; Heinrich, M.; Vranesic, I.; Biollaz, M.; Allgeier, H.; Heckendorn, R.; Urwyler, S.; Varney, M. A.; Johnson, E. C.; Hess, S. D.; Rao, S. P.; Sacaan, A. I.; Santori, E. M.; Veliçelebi, G.; Kuhn, R. 2-Methyl-6-(phenylethynyl)pyridine (MPEP), a potent, selective and systemically active mGlu5 receptor antagonist. Neuropharmacology 1999, 38, 1493−1503. (23) Doré, A. S.; Okrasa, K.; Patel, J. C.; Serrano-Vega, M.; Bennett, K.; Cooke, R. M.; Errey, J. C.; Jazayeri, A.; Khan, S.; Tehan, B.; Weir, M.; Wiggin, G. R.; Marshall, F. H. Structure of class C GPCR

analogues, binding hot spot analyses of mGlu5 with mavoglurant and HTL14242 analogues, and tabulated data used in the generation of Tables 16 and 17 (PDF) Molecular formula strings (CSV) Accession Codes

Coordinates and structure factors have been deposited in the Protein Data Bank under the accession codes 6FFI (M-MPEP) and 6FFH (fenobam).



AUTHOR INFORMATION

Corresponding Authors

*(J.A.C.) Phone: +44 (0)1707 448160. Fax: +44 (0)1707 358640. E-mail: [email protected]. *(G.M.K.) Phone: +36-1-3826821. E-mail: [email protected]. hu. ORCID

John A. Christopher: 0000-0001-5737-4650 Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS The support of the National Brain Research Program (KTIA_NAP_13-1-2013-0001 and 2017-1.2.1-NKP-201700002) for O. Z., G.G.F., and G.M.K. is acknowledged.



ABBREVIATIONS USED M-MPEP, 2-[(3-methoxyphenyl)ethynyl]-6-methylpyridine; VFT, Venus flytrap; TMD, transmembrane domain; FXS, fragile X syndrome; PD-LID, Parkinson’s disease levodopainduced dyskinesia; OCD, obsessive compulsive disorder; MPEP, 2-methyl-6-(phenylethynyl)pyridine; LCP, lipidic cubic phase; T4L, T4-lysozyme; ICL, intracellular loop; ECL, extracellular loop; GE, group efficiency



REFERENCES

(1) Monod, J.; Wyman, J.; Changeux, J.-P. On the nature of allosteric transitions: a plausible model. J. Mol. Biol. 1965, 12, 88−118. (2) Möhler, H.; Fritschy, J. M.; Rudolph, U. A new benzodiazepine pharmacology. J. Pharmacol. Exp. Ther. 2002, 300, 2−8. (3) Conn, P. J.; Lindsley, C. W.; Meiler, J.; Niswender, C. M. Opportunities and challenges in the discovery of allosteric modulators of GPCRs for treating CNS disorders. Nat. Rev. Drug Discovery 2014, 13, 692−708. (4) Lindsley, C. W.; Emmitte, K. A.; Hopkins, C. R.; Bridges, T. M.; Gregory, K. J.; Niswender, C. M.; Conn, P. J. Practical strategies and concepts in GPCR allosteric modulator discovery: recent advances with metabotropic glutamate receptors. Chem. Rev. 2016, 116, 6707− 6741. (5) Galambos, J.; Bielik, A.; Krasavin, M.; Orgován, Z.; Domány, G.; Nógrádi, K.; Wágner, G.; Balogh, G. T.; Béni, Z.; Kóti, J.; Szakács, Z.; Bobok, A.; Kolok, S.; Mikó-Bakk, M. L.; Vastag, M.; Sághy, K.; Laszy, J.; Halász, A. S.; Balázs, O.; Gál, K.; Greiner, I.; Szombathelyi, Z.; Keserű , G. M. Discovery and preclinical characterization of 3-((4-(4chlorophenyl)-7-fluoroquinoline-3-yl)sulfonyl)benzonitrile, a novel non-acetylenic metabotropic glutamate receptor 5 (mGluR5) negative allosteric modulator for psychiatric indications. J. Med. Chem. 2017, 60, 2470−2484. (6) Congreve, M.; Oswald, C.; Marshall, F. H. Applying structurebased drug design approaches to allosteric modulators of GPCRs. Trends Pharmacol. Sci. 2017, 38, 837−847. (7) Christopher, J. A.; Aves, S. J.; Bennett, K. A.; Doré, A. S.; Errey, J. C.; Jazayeri, A.; Marshall, F. H.; Okrasa, K.; Serrano-Vega, M. J.; Tehan, B. G.; Wiggin, G. R.; Congreve, M. Fragment and structurebased drug discovery for a class C GPCR: discovery of the mGlu5 220

DOI: 10.1021/acs.jmedchem.7b01722 J. Med. Chem. 2019, 62, 207−222

Journal of Medicinal Chemistry

Article

metabotropic glutamate receptor 5 transmembrane domain. Nature 2014, 511, 557−562. (24) Ballesteros, J. A.; Weinstein, H. Integrated methods for the construction of three-dimensional models and computational probing of structure−function relations in G protein-coupled receptors. Methods Neurosci. 1995, 25, 366−428. (25) Gasparini, F.; Andres, H.; Flor, P. J.; Heinrich, M.; Inderbitzin, W.; Lingenhöhl, K.; Müller, H.; Munk, V. C.; Omilusik, K.; Stierlin, C.; Stoehr, N.; Vranesic, I.; Kuhn, R. [3H]-M-MPEP, a potent, subtypeselective radioligand for the metabotropic glutamate receptor subtype 5. Bioorg. Med. Chem. Lett. 2002, 12, 407−409. (26) Mason, J. S.; Bortolato, A.; Congreve, M.; Marshall, F. H. New insights from structural biology into the druggability of G proteincoupled receptors. Trends Pharmacol. Sci. 2012, 33, 249−260. (27) Mason, J. S.; Bortolato, A.; Weiss, D. R.; Deflorian, F.; Tehan, B.; Marshall, F. H. High end GPCR design: crafted ligand design and druggability analysis using protein structure, lipophilic hotspots and explicit water networks. In Silico Pharmacol, [Online] 2013, 1, 23. (28) Congreve, M.; Andrews, S. P.; Doré, A. S.; Hollenstein, K.; Hurrell, E.; Langmead, C. J.; Mason, J. S.; Ng, I. W.; Tehan, B.; Zhukov, A.; Weir, M.; Marshall, F. H. Discovery of 1,2,4-triazine derivatives as adenosine A2A antagonists using structure based drug design. J. Med. Chem. 2012, 55, 1898−1903. (29) Wu, H.; Wang, C.; Gregory, K. J.; Han, G. W.; Cho, H. P.; Xia, Y.; Niswender, C. M.; Katritch, V.; Meiler, J.; Cherezov, V.; Conn, P. J.; Stevens, R. C. Structure of a class C GPCR metabotropic glutamate receptor 1 bound to an allosteric modulator. Science 2014, 344, 58−64. (30) Anighoro, A.; Graziani, D.; Bettinelli, I.; Cilia, A.; De Toma, C.; Longhi, M.; Mangiarotti, F.; Menegon, S.; Pirona, L.; Poggesi, E.; Riva, C.; Rastelli, G. Insights into the interaction of negative allosteric modulators with the metabotropic glutamate receptor 5: discovery and computational modeling of a new series of ligands with nanomolar affinity. Bioorg. Med. Chem. 2015, 23, 3040−3058. (31) Michel, J.; Tirado-Rives, J.; Jorgensen, W. L. Energetics of displacing water molecules from protein binding sites: consequences for ligand optimization. J. Am. Chem. Soc. 2009, 131, 15403−15411. (32) Andrews, S. P.; Mason, J. S.; Hurrell, E.; Congreve, M. Structure-based drug design identifies potent & selective GPCR antagonists. MedChemComm 2014, 5, 571−575. (33) Bortolato, A.; Deflorian, F.; Weiss, D. R.; Mason, J. S. Decoding the role of water dynamics in ligand−protein unbinding: CRF1R as a test case. J. Chem. Inf. Model. 2015, 55, 1857−1866. (34) Cooke, R. M.; Brown, A. J.; Marshall, F. H.; Mason, J. S. Structures of G protein-coupled receptors reveal new opportunities for drug discovery. Drug Discovery Today 2015, 20, 1355−1364. (35) Gregory, K. J.; Nguyen, E. D.; Reiff, S. D.; Squire, E. F.; Stauffer, S. R.; Lindsley, C. W.; Meiler, J.; Conn, P. J. Probing the metabotropic glutamate receptor 5 (mGlu5) positive allosteric modulator (PAM) binding pocket: discovery of point mutations that engender a “molecular switch” in PAM pharmacology. Mol. Pharmacol. 2013, 83, 991−1006. (36) Alagille, D.; Baldwin, R. M.; Roth, B. L.; Wroblewski, J. T.; Grajkowska, E.; Tamagnan, G. D. Synthesis and receptor assay of aromatic ethynyl-aromatic derivatives with potent mGluR5 antagonist activity. Bioorg. Med. Chem. 2005, 13, 197−209. (37) Verdonk, M. L.; Rees, D. C. Group efficiency: a guideline for hits-to-leads chemistry. ChemMedChem 2008, 3, 1179−1180. (38) Hopkins, A. L.; Groom, C. R.; Alex, A. Ligand efficiency: a useful metric for lead selection. Drug Discovery Today 2004, 9, 430− 443. (39) Kuhne, S.; Kooistra, A. J.; Bosma, R.; Bortolato, A.; Wijtmans, M.; Vischer, H. F.; Mason, J. S.; de Graaf, C.; de Esch, I. J. P.; Leurs, R. Identification of ligand binding hot spots of the histamine H1 receptor following structure-based fragment optimization. J. Med. Chem. 2016, 59, 9047−9061. (40) Cosford, N. D. P.; Tehrani, L.; Roppe, J.; Schweiger, E.; Smith, N. D.; Anderson, J.; Bristow, L.; Brodkin, J.; Jiang, X.; McDonald, I.; Rao, S.; Washburn, M.; Varney, M. A. 3-[(2-Methyl-1,3-thiazol-4yl)ethynyl]-pyridine: A potent and highly selective metabotropic

glutamate subtype 5 receptor antagonist with anxiolytic activity. J. Med. Chem. 2003, 46, 204−206. (41) Iso, Y.; Grajkowska, E.; Wroblewski, J. T.; Davis, J.; Goeders, N. E.; Johnson, K. M.; Sanker, S.; Roth, B. L.; Tueckmantel, W.; Kozikowski, A. P. Synthesis and structure-activity relationships of 3[(2-methyl-1,3-thiazol-4-yl)ethynyl]pyridine analogues as potent, noncompetitive metabotropic glutamate receptor subtype 5 antagonists; search for cocaine medications. J. Med. Chem. 2006, 49, 1080−100. (42) Schrödinger MacroModel, v11.4; OPLS, 2005. (43) Harpsøe, K.; Isberg, V.; Tehan, B. G.; Weiss, D.; Arsova, A.; Marshall, F. H.; Bräuner-Osborne, H.; Gloriam, D. E. Selective negative allosteric modulation of metabotropic glutamate receptors − a structural perspective of ligands and mutants. Sci. Rep. 2015, 5, 13869. (44) Poon, S. F.; Eastman, B. W.; Chapman, D. F.; Chung, J.; Cramer, M.; Holtz, G.; Cosford, N. D.; Smith, N. D. 3-[3-Fluoro-5-(5pyridin-2-yl-2H-tetrazol-2-yl)phenyl]-4-methylpyridine: a highly potent and orally bioavailable metabotropic glutamate subtype 5 (mGlu5) receptor antagonist. Bioorg. Med. Chem. Lett. 2004, 14, 5477−5480. (45) Eastman, B.; Chen, C.; Smith, N. D.; Poon, S.; Chung, J.; ReyesManalo, G.; Cosford, N. D.; Munoz, B. Expedited SAR study of an mGluR5 antagonists: generation of a focused library using a solutionphase Suzuki coupling methodology. Bioorg. Med. Chem. Lett. 2004, 14, 5485−5488. (46) Roppe, J.; Smith, N. D.; Huang, D.; Tehrani, L.; Wang, B.; Anderson, J.; Brodkin, J.; Chung, J.; Jiang, X.; King, C.; Munoz, B.; Varney, M. A.; Prasit, P.; Cosford, N. D. Discovery of novel heteroarylazoles that are metabotropic glutamate subtype 5 receptor antagonists with anxiolytic activity. J. Med. Chem. 2004, 47, 4645− 4648. (47) Nógrádi, K.; Wágner, G.; Domány, G.; Bobok, A.; Magdó, I.; Kolok, S.; Mikó-Bakk, M. L.; Vastag, M.; Sághy, K.; Gyertyán, I.; Kóti, J.; Gál, K.; Farkas, S.; Keserű , G. M.; Greiner, I.; Szombathelyi, Z. Thieno[2,3-b]pyridines as negative allosteric modulators of metabotropic GluR5 receptors: lead optimization. Bioorg. Med. Chem. Lett. 2015, 25, 1724−1729. (48) Galambos, J.; Bielik, A.; Wágner, G.; Domány, G.; Kóti, J.; Béni, Z.; Szigetvári, Á .; Sánta, Z.; Orgován, Z.; Bobok, A.; Kiss, B.; MikóBakk, M. L.; Vastag, M.; Sághy, K.; Krasavin, M.; Gál, K.; Greiner, I.; Szombathelyi, Zs.; Keserű , G. M. Discovery of 4-amino-3-arylsulfoquinolines, a novel non-acetylenic chemotype of metabotropic glutamate 5 (mGlu5) receptor negative allosteric modulators. Eur. J. Med. Chem. 2017, 133, 240−254. (49) Galambos, J.; Bielik, A.; Krasavin, M.; Orgován, Z.; Domány, G.; Nógrádi, K.; Wágner, G.; Balogh, G. T.; Béni, Z.; Kóti, J.; Szakács, Z.; Bobok, A.; Kolok, S.; Mikó-Bakk, M. L.; Vastag, M.; Sághy, K.; Laszy, J.; Halász, A. S.; Balázs, O.; Gál, K.; Greiner, I.; Szombathelyi, Z.; Keserű , G. M. Discovery and preclinical characterization of 3-((4-(4chlorophenyl)-7-fluoroquinoline-3-yl)sulfonyl)benzonitrile, a novel non-acetylenic metabotropic glutamate receptor 5 (mGluR5) negative allosteric modulator for psychiatric indications. J. Med. Chem. 2017, 60, 2470−2484. (50) Caffrey, M.; Cherezov, V. Crystallizing membrane proteins using lipidic mesophases. Nat. Protoc. 2009, 4, 706−731. (51) Kabsch, W. XDS. Acta Crystallogr., Sect. D: Biol. Crystallogr. 2010, 66, 125−132. (52) Evans, P. R.; Murshudov, G. N. How good are my data and what is the resolution? Acta Crystallogr., Sect. D: Biol. Crystallogr. 2013, 69, 1204−1214. (53) Winn, M. D.; Ballard, C. C.; Cowtan, K. D.; Dodson, E. J.; Emsley, P.; Evans, P. R.; Keegan, R. M.; Krissinel, E. B.; Leslie, A. G. W.; McCoy, A.; McNicholas, S. J.; Murshudov, G. N.; Pannu, N. S.; Potterton, E. A.; Powell, H. R.; Read, R. J.; Vagin, A.; Wilson, K. S. Overview of the CCP4 suite and current developments. Acta Crystallogr., Sect. D: Biol. Crystallogr. 2011, 67, 235−242. (54) McCoy, A. J.; Grosse-Kunstleve, R. W.; Adams, P. D.; Winn, M. D.; Storoni, L. C.; Read, R. J. Phaser crystallographic software. J. Appl. Crystallogr. 2007, 40, 658−674. 221

DOI: 10.1021/acs.jmedchem.7b01722 J. Med. Chem. 2019, 62, 207−222

Journal of Medicinal Chemistry

Article

(55) Emsley, P.; Lohkamp, B.; Scott, W. G.; Cowtan, K. Features and development of Coot. Acta Crystallogr., Sect. D: Biol. Crystallogr. 2010, 66, 486−501. (56) Afonine, P. V.; Grosse-Kunstleve, R. W.; Echols, N.; Headd, J. J.; Moriarty, N. W.; Mustyakimov, M.; Terwilliger, T. C.; Urzhumtsev, A.; Zwart, P. H.; Adams, P. D. Towards automated crystallographic structure refinement with phenix.refine. Acta Crystallogr., Sect. D: Biol. Crystallogr. 2012, D68, 352−367. (57) Chen, V. B.; Arendall, W. B., 3rd; Headd, J. J.; Keedy, D. A.; Immormino, R. M.; Kapral, G. J.; Murray, L. W.; Richardson, J. S.; Richardson, D. C. MolProbity: all-atom structure validation for macromolecular crystallography. Acta Crystallogr., Sect. D: Biol. Crystallogr. 2010, 66, 12−21. (58) Goodford, P. J. A computational procedure for determining energetically favorable binding sites on biologically important macromolecules. J. Med. Chem. 1985, 28, 849−857. (59) Sciabola, S.; Stanton, R. V.; Mills, J. E.; Flocco, M. M.; Baroni, M.; Cruciani, G.; Perruccio, F.; Mason, J. S. High-throughput virtual screening of proteins using GRID molecular interaction fields. J. Chem. Inf. Model. 2010, 50, 155−169. (60) Bortolato, A.; Tehan, B. G.; Smith, R. T.; Mason, J. S. Methodologies for the Examination of Water in GPCRs. In Computational Methods for GPCR Drug Discovery; Heifetz, A., Ed.; Methods in Molecular Biology; Humana Press: New York, 2017; Vol. 1705, pp 207−232.

222

DOI: 10.1021/acs.jmedchem.7b01722 J. Med. Chem. 2019, 62, 207−222