Allosteric Molecular Switches in Metabotropic Glutamate Receptors

Abstract Metabotropic glutamate receptors (mGlu) are class C G protein‐coupled receptors of eight subtypes that are omnipresently expressed in the central nervous system. mGlus have relevance in several psychiatric and neurological disorders, therefore they raise considerable interest as drug targets. Allosteric modulators of mGlus offer advantages over orthosteric ligands owing to their increased potential to achieve subtype selectivity, and this has prompted discovery programs that have produced a large number of reported allosteric mGlu ligands. However, the optimization of allosteric ligands into drug candidates has proved to be challenging owing to induced‐fit effects, flat or steep structure‐activity relationships and unexpected changes in theirpharmacology. Subtle structural changes identified as molecular switches might modulate the functional activity of allosteric ligands. Here we review these switches discovered in the metabotropic glutamate receptor family..


Introduction
Heterotrimeric G protein-coupled receptors (GPCRs) constitute the largest class of membrane proteins in the human genome, and are responsible for conveying extracellular to intracellular signals within a broad range of physiological contexts. [1][2][3][4] The metabotropic glutamate receptors belong to the class C GPCRs, and they can be classified into eight subtypes divided into three groups based on their pharmacology, sequential homology and G protein coupling. Group I receptors (mGlu 1 , mGlu 5 ) are preferentially coupled to G αq , and are typically found postsynaptically, Group II receptors (mGlu 2 , mGlu 3 ), and Group III receptors (mGlu 4 , mGlu 6 , mGlu 7 and mGlu 8 ) are coupled to G i/o , and are found presynaptically except mGlu 6 receptor which is solely found in the retina. [4][5][6][7] Each mGlu monomer contains a so-called Venus flytrap domain (VFT), a cysteine-rich domain (CRD) and a seven trans-membrane domain (7TMD). [8] The first accommodates the extracellular orthosteric binding site. Allosteric binding sites were reported; however, mostly, [9,10] but not exclusively, [11] at the 7TMD region of the protein including the one that corresponds to the orthosteric binding site of class A GPCRs. [12] mGlus form mandatory dimers through a disulfide bond at the top of the VFT domains, which are mostly homodimers; [13,14] however, heterodimerization was also observed in several cases with other subtypes [15][16][17][18] and different GPCRs. [12,[19][20][21][22] MGluRs can modulate the release of glutamate and its postsynaptic response, as well as the activity of other synapses. [23,24] These receptors have been recognized as ther-apeutic targets in a number of central nervous system (CNS) diseases, like Parkinson's disease, schizophrenia, epilepsy, ischemia, pain and anxiety. [23,25,26] Early attempts to modulate mGlu receptor activity were directed to the orthosteric binding pocket; however, despite some promising attempts in group II [27,28] and III receptors, [29] selectively targeting this site proved to be difficult. This can be attributed to the highly conserved nature of the orthosteric binding pocket across the eight receptors. Moreover, as the orthosteric ligands are mostly glutamate derivatives, the bioavailability and blood-brain barrier permeability of these compounds are usually less than satisfactory. [5][6][7] Later on, allosteric modulation of mGlus emerged as a more viable strategy to achieve receptor subtype selectivity [13,[30][31][32] owing to the lower sequence similarity of these sites across mGlus. Moreover, pure allosteric modulators are active only in the presence of an orthosteric ligand, and the modulating effect of these modulators is saturable, which reduces the risk of over sensitization. However, allosteric modulators can promote a global change in the receptor conformation, and in this way they can modulate the affinity, potency or efficacy of the orthosteric ligand in negative or positive direction, and they might also inhibit or increase the G protein coupling. [4,[32][33][34][35] Allosteric modulators are able to exert their modulating effect through different pharmacological modes of action. Negative allosteric modulators (NAMs) weaken and positive allosteric modulators (PAMs) potentiate the effect of the endogenous ligand, while silent allosteric modulators/neutral allosteric ligands (SAMs/NALs) occupy the allosteric binding pocket without detected pharmacological function. Partial antagonists (PAs) are NAMs that fully occupy the allosteric binding site and induce partial reducing effect, while allosteric agonists (AAs) are able to activate the receptor in the absence of orthosteric ligand by binding to an allosteric site and inducing an active conformation of the receptor. Ago-potentiators (ago-PAMs) are functioning as both PAMs and AAs [4,36] by inducing allosteric agonism, to varying degrees in the absence of orthosteric

Structural Basis of Functional Mode Switching
The detailed understanding of the molecular mechanism of mGlu modulation is challenging owing to inconsistent SAR, mode and subtype switching, and biased modulation. However, available site-directed mutagenesis results (e. g., in refs. [77][78][79][80][81][82][83]), the release of X-ray structures for mGlu group I proteins' 7TM region in complex with NAMs, [51,[84][85][86] and the cryo-electronmicroscopic structure [87] of the full-length active like (potentiat- ing antibody and agonist bound) and inactive mGlu 5 significantly contributed to our understanding of the mechanism of allosteric modulation.
Based on the available structural information, the homodimers of mGlus are crosslinked only through the Venus flytrap domain in the apo form, and activation results two major changes in this domain. The first is the closure of the two VFT lobes, and the second includes an inter-subunit reorientation, which brings the cysteine rich domains (CRDs) to close proximity to each other. [88] As the CRD is fairly rigid, this conformational change involves the approach of the two 7TM regions while they are rotated by 20°. [87] This movement results in the establishment of a TM6-TM6 interface (Figure 2).
These observations are in line with the proposal that both inter-and intrasubunit rearrangements are required for full activation, [89] however, they do not fully elucidate the connection among the formation of the interface, G protein coupling and signaling.
Before the first available X-ray structure within the mGlu family the identification of key interactions of mGlu allosteric ligands with the 7TM of the protein in PAM and NAM complexes were addressed by site-directed mutagenesis stud-ies. These studies pointed out the presence of at least one common allosteric binding site between mGlu subtypes, because several residues were reported to be crucial for both positive and negative allosteric modulators in different subtypes. For example, positions 3.36a, 3.40c; 5.43a, 5.43c; 6.48a, 6.50c were essential in mGlu 1 , mGlu 2 , mGlu 4 and mGlu 5 , [78,[80][81][82][90][91][92][93][94][95] residues 5.44a, 5.44c; 5.47a, 5.47c; 6.55a, 6.57c were important in mGlu 2 and in group I mGlu receptors, [78][79][80]82,[91][92][93][94][96][97][98][99][100][101] 6.51a,6.53c was relevant in mGlu 4 and group I receptors. [78][79][80][90][91][92]95] (GPCRdb generic residue numbering is used throughout the manuscript.. [102,103] ) The impact of mutations on the effect of allosteric modulators were combined with affinity data obtained with radiolabeled allosteric modulators. Affinity and cooperativity determinants were mapped with the usage of the most well characterized representatives of allosteric modulator scaffolds in mGlu 1 and mGlu 5 structures. [97] The most significant amino acids that modulated the affinity and cooperativity upon mutations in both mGlu 1 and mGlu 5 are: 5.43a, 5.43c; 5.44a, 5.44c; 6.48a, 6.50c; 6.51a, 6.53c; 6.55a, 6.57c, while residues at positions 3.36a, 3.40c; 3.40a, 3.44c; 5.47a, 5.47c; 6.52a, 6.54c; 6.55a, 6.57c; 7.45a, 7.40c; 7.46a, 7.41c are affecting only mGlu 5 ligands, and mutation at position 7.38a, 7.33c changes the affinity and cooperativity only of mGlu 1 ligands. [78,86,96] Interestingly mutations on several amino acids resulted in switch in allosteric modulator cooperativity (functional mode switch). Mutation at 6.48a, 6.50c results in cooperativity of NAMs of glutamate to positive, [78] mutation at position 6.51a, 6.53c causes inverse cooperativity from the original both in the case of PAMs and NAMs. [91,92] Moreover, point mutations at position 3.40a 3.44c; 6.44a, 6.46c; 7.45a; 7.40c also switch acetylenic PAMs to have neutral or negative cooperativity at mGlu 5 . [78,104] The appearance of X-ray structures of group I mGlus (4OO9, [84] 5CGC, 5CGD, [85] 6FFI, 6FFH, [51] 4OR2 [86] ) opened the possibility to investigate the detailed mechanism of the 7TM intra-subunit rearrangement needed for receptor activation. Since then, numerous structure based calculations have been applied. [13,47,51,52,96,[105][106][107][108] These calculations were performed for mGlu proteins in complex with allosteric modulators. The most common and best-characterized allosteric binding pocket within the mGlu family can be found in the 7TM region surrounded by the so-called "trigger switch" (3.36a, 3.40c; 5.47a, 5.47c; 6.48a, 6.50c) and "transmission switch" (3.40a, 3.44c; 5.50a, 5.50c; 6.44a, 6.46c) amino acids, which were proposed to be crucial in the allosteric activation. [105,106,[109][110][111] Molecular dynamics simulations showed that the 3.44c amino acid has a direct or water-mediated interaction with 6.46c in the case of NAM and NAL binding. [52,105,106] This water molecule was observed in all available mGlu 5 X-ray structures and was found to have increased stability in complexes compared to the apo protein. [51,84,85] Nonetheless, these interactions cannot be formed in the PAM complexes owing to the bending of the TM6 in the active structure (Figure 3). 3.40c, a member of the trigger switch, was found to move toward TM6 [105] upon activation of mGlu 2 . In mGlu 5 , ionic interactions were observed among residue pair 3.50c-6.35c in the NAM structure, whereas it was not present in the PAM complex. [108] The destabilization of the ionic lock was also observed in the mGlu 4 PAM complex. [111] Although these observations were reported for mGlu receptor monomers, they might be applicable to the dimers as well, because on the one hand, the computational results are in line with the site-directed mutagenesis-based experimental results detailed above, on the other hand, the active like cryo-electronmicroscopic structure of mGlu 5 shows that the establishment of TM6-TM6 interface only affects the top of the 7TM region far from the reported common allosteric binding pocket ( Figure 2). The allosteric binding site of mGlus can be found in a functional water channel, and hence water is likely to play an important role in signal transduction. Therefore, calculations were also aimed at understanding the role of water molecules in ligand binding. These studies showed that most of the interactions between the ligand and the protein are water-mediated, and hence the perturbation of the water network contributes to the observed ligand affinity and functional activity. [106,108]

The Impact of Allosteric Molecular Switches on Medicinal Chemistry Programs
The detection, validation and quantification of allosteric modulation is a permanent challenge in allosteric drug discovery. Binding and functional assays in various setups are used to explore the behavior of allosteric ligands. [125,126] Binding assays are able to directly validate of allosteric mode of action and unmask the site of interaction; however, they are not able to provide information about efficacy modulation. Functional assays have the advantage of detecting a wider spectrum of allosteric behaviors including the modulation of affinity and efficacy. In addition, they are also useful to study probe dependence and saturation effects. The former describes the direction and degree of cooperativity between the allosteric modulator and the orthosteric ligand (probe), which might be, however, probe dependent. [126][127][128][129] The latter expresses the limited influence of allosteric modulator caused by the cooperativity between orthosteric and allosteric sites. This also reduces the risk of over sensitization of the receptor by allosteric modulators. [130,131] Most commonly mGlu allosteric ligand detection relies on the determination of modulator concentration-response curves with a single agonist concentration to acquire approximate modulator potency. [43,125] Nevertheless, the potency of an Figure 3. The 7TM regions of active and inactive mGlu 5 with allosteric ligands 6 (magenta) and M-MPEP (green), respectively. The active structure (the homology model of which was prepared based on the μ-opioid receptor-Gi protein complex (PDB ID: 6DDE [112] ) as is written in ref. [52]) is depicted in cyan, the inactive (PDB ID: 6FFI) is in blue. Trigger switch, transmission switch, and "ionic lock" amino acids are represented as cyan and grey sticks.

ChemMedChem
Minireviews doi.org/10.1002/cmdc.202000444 allosteric modulator depends on the allosteric ligand affinity, cooperativity and intrinsic efficacy; moreover, it is influenced by the orthosteric agonist concentration. [132] Therefore the determination of ligand affinity and cooperativity at target receptor and related subtypes during drug discovery is essential to achieve optimal selectivity. [43] As many class C GPCRs lack selective radioligands, which obstructs the determination of ligand affinity by radioligand binding-based methods, functional assays are used to evolve affinity and cooperativity estimates and to assess and optimize selectivity. [125,133] Moreover, many studies use only a single orthosteric ligand for a single signaling pathway during ligand development, therefore only a limited part of the full pharmacology will be discovered. However, as it has come to the fore in recent years, allosteric ligands may have signaling-pathway-dependent effects. This phenomenon referred to as "biased modulation" and has been described for many GPCRs along with group I and group III mGlus. [93,97,127,[134][135][136][137] Moreover, neutral allosteric ligands might be undetected owing to the neutral cooperativity with an orthosteric ligand, and in spite of their receptor affinity they may be categorized as inactive, as was demonstrated in the discovery of several neutral allosteric ligands for mGlu 5 . [10,59,125,137] These data illustrate that the use of efficacy-driven approaches and functional studies are inappropriate to describe allosteric modulator pharmacology and subtype selectivity completely, and that selectivity of allosteric modulators for class C GPCRs might be largely driven by cooperativity.
Mode switching affects the development and optimization of primary assays and the complexity of screening cascades, moreover, it influences the strategy of medicinal chemistry programs. Mode switching, together with the location and properties of the allosteric sites and the often steep or flat SAR make the optimization of GPCR allosteric modulators complex. [43] As these effects are hardly predictable and the properties of allosteric sites are often challenged the ADME properties of the ligands, multidimensional parallel optimization strategies are typically considered. [138] The implementation of this iterative, multidimensional parallel synthesis strategy has been recently exemplified by the optimization of an mGlu5 NAM to clinical candidate. [139] The procedure starts with the retrosynthetic deconvolution of the starting point to identify regions to be optimized ( Figure 4). Next, scanning libraries are used to explore the optimal set of substituents in each region. Having the optimized set of structural moieties identified, their combined effect is investigated by synthesizing and testing matrix libraries. Finally, most promising members of the matrix libraries are evaluated further and their head-to-head comparison provides candidates.

Summary
Allosteric modulation of mGlus has distinct advantages over orthosteric ligands in terms of subtype selectivity and reduced risk of receptor over sensitization, nonetheless, the optimization of these ligands proved to be challenging owing to often observed sharp and inconsistent SAR, functional selectivity, and molecular switches modulating the modes of pharmacology and subtype selectivity. Herein, we have reviewed allosteric molecular switches causing pharmacological mode and subtype switching, and summarized the available information on mGlu receptor-activation mechanisms based on experimental and computational studies. It is emphasized that the allosteric binding site of mGlus might contain water molecules playing a significant role in the activation mechanism and mode switching that makes allosteric pharmacology poorly predictable. Recent developments in the structural biology of mGlus together with the availability of effective computational protocols might facilitate the discovery of novel allosteric ligands with designed pharmacology. . Implementation of the multiple parallel synthesis approach for the optimization mGluR5 NAMs. Retrosynthetic analysis (level 1) of the scaffold identified three regions (I, II and III) for further evaluation. The preferred substituents in region II (7-fluoro and 8-fluoro) were identified by the scanning library (level 2). Next, regions I and III were explored by matrix libraries (levels 3a and 3b). The most promising compounds identified from the matrix libraries were further characterized and optimized (level 4) to yield compounds profiled for candidate selection (level 5). Adapted with permission from ref. [139]. Copyright: 2017, American Chemical Society.