Metabotropic glutamate receptor subtype 5: molecular pharmacology, allosteric modulation and stimulus bias

Abstract

The metabotropic glutamate receptor subtype 5 (mGlu₅) is a family C GPCR that has been implicated in various neuronal processes and, consequently, in several CNS disorders. Over the past few decades, GPCR‐based drug discovery, including that for mGlu₅ receptors, has turned considerable attention to targeting allosteric binding sites. Modulation of endogenous agonists by allosteric ligands offers the advantages of spatial and temporal fine‐tuning of receptor activity, increased selectivity and reduced adverse effects with the potential to elicit improved clinical outcomes. Further, with greater appreciation of the multifaceted nature of the transduction of mGlu₅ receptor signalling, it is increasingly apparent that drug discovery must take into consideration unique receptor conformations and the potential for stimulus‐bias. This novel paradigm proposes that different ligands may differentially modulate distinct signalling pathways arising from the same receptor. We review our current understanding of the complexities of mGlu₅ receptor signalling and regulation, and how these relate to allosteric ligands. Ultimately, a deeper appreciation of these relationships will provide the foundation for targeted drug design of compounds with increased selectivity, not only for the desired receptor but also for the desired signalling outcome from the receptor.

Linked Articles

This article is part of a themed section on Molecular Pharmacology of G Protein‐Coupled Receptors. To view the other articles in this section visit http://onlinelibrary.wiley.com/doi/10.1111/bph.v173.20/issuetoc

Tables of Links

Targets
GPCRsa	Ligand‐gated ion channelsc
Adenosine A2A receptors	AMPA receptors
Calcium‐sensing receptors	NMDA receptors
Dopamine D2 receptors	Enzymes d
mGlu1 receptors	Akt
mGlu5 receptors	ERK1/2
μ‐opioid receptors	GRK2, G protein‐coupled receptor kinase 2
Ion channels b	p38 MAPK
TRPV1 channels

Ligands
5PAM523
CPPHA
DFB
DHPG
IP3
MPEP
NCFP

These Tables list key protein targets and ligands in this article which are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY (Pawson et al., 2014) and are permanently archived in the Concise Guide to PHARMACOLOGY 2013/14 (^{a b c d}Alexander et al., 2013a, 2013b, 2013c, 2013d).

Abbreviations

7TMs

seven transmembrane‐spanning domains

CPPHA

N‐(4‐chloro‐2‐[(1,3‐dioxo‐1,3‐dihydro‐2H‐isoindol‐2‐yl)methyl]phenyl)‐2‐hydroxybenzamide

DFB

((3‐Fluorophenyl)methylene)hydrazone‐3‐fluorobenzaldehyde

DHPG

(S)‐3,5‐DHPG, (S)‐3,5‐dihydroxyphenylglycine

iCa²⁺

intracellular calcium

LTD

long‐term depression

MPEP

2‐methyl‐6‐(phenylethynyl)pyridine

mTOR

mammalian target of rapamycin

NAL

neutral allosteric ligand

NAM

negative allosteric modulator

NCFP

N‐(4‐chloro‐2‐[(1,3‐dioxoisoindolin‐2‐yl)methyl)phenyl)picolinamide]

PAM

positive allosteric modulator

VFT

Venus flytrap

Introduction

Glutamate is the main excitatory neurotransmitter in the brain, providing the balance to the major inhibitory neurotransmitter GABA in neuronal synaptic interplay. Until the mid‐1980s, glutamate was thought to act solely via a family of ligand‐gated ion channels, the ionotropic glutamate receptors comprised three subtypes: NMDA, AMPA and kainate. It was not until this neurotransmitter was shown to stimulate inositol phosphate that the first G protein‐coupled metabotropic glutamate (mGlu) receptor was cloned (Houamed et al., 1991; Masu et al., 1991). Eight mGlu receptor subtypes have since been characterized and are subdivided based on signal transduction mechanisms, pharmacology and sequence homology (Niswender and Conn, 2010). Group I consists of mGlu₁ and mGlu₅ receptors that preferentially activate phospholipase C via G_q/11 coupling, whereas Group II (mGlu₂ and mGlu₃ receptors) and group III (mGlu₄, mGlu₆, mGlu₇ and mGlu₈ receptors) couple to G_i/o and adenylate cyclase inhibition (Niswender and Conn, 2010).

The mGlu receptors are family C GPCRs – a group that also includes GABA_B receptors, the calcium‐sensing receptor and multiple olfactory, taste and pheromone receptors, distinguished by a large extracellular N‐terminal, Venus flytrap (VFT) domain. This N‐terminal domain comprises two distinct lobes that close around glutamate upon binding, with the closure of one lobe sufficient for receptor activation, and closure of both lobes required for maximal activity (O'Hara et al., 1993; Costantino and Pellicciari, 1996; Kunishima et al., 2000; Kniazeff et al., 2004; Muto et al., 2007). Constitutive receptor dimerization occurs via a disulfide bond in an as yet unsolved loop region linking the ‘top’ of the two VFT domains (Romano et al., 1996; Kunishima et al., 2000; Romano et al., 2001; Muto et al., 2007). Linking the N‐terminal VFT to the seven transmembrane‐spanning domains (7TMs) is a cysteine rich domain. The cysteine rich domain is crucial for the transmission of conformational changes induced by glutamate binding in the VFT to the 7TMs that mediate coupling with intracellular effectors (Rondard et al., 2006).

Widely expressed and mainly postsynaptic throughout the cortex, striatum, hippocampus, caudate nucleus and nucleus accumbens, and in astrocytes, glia and peripheral sensory neurons (Shigemoto et al., 1993; Crawford et al., 2000; Walker et al., 2001), mGlu₅ receptors are crucial for synaptic plasticity and neuronal development (Jia et al., 1998; Fendt and Schmid, 2002; Bortolotto et al., 2005; Galik et al., 2008; Wijetunge et al., 2008; She et al., 2009; Xu et al., 2009; Ballester‐Rosado et al., 2010; Black et al., 2010; Chen et al., 2012; Xiao et al., 2013; Hamilton et al., 2014). mGlu₅ receptors have been implicated in the pathology or as a therapeutic target for numerous CNS disorders, including psychosis and schizophrenia (Kinney et al., 2003; Brody et al., 2004; Chen et al., 2010), motor control (Ribeiro et al., 2014), anxiety and depression (Li et al., 2006; Inta et al., 2013), reward and addiction (Bird et al., 2010; Olsen et al., 2010; Eiler et al., 2011; Stoker et al., 2012; Chesworth et al., 2013), appetite and energy homeostasis (Bradbury et al., 2005) and pain (Galik et al., 2008; Kolber et al., 2010). The immense therapeutic potential of targeting mGlu₅ receptors has been extensively summarized (Noetzel et al., 2012a; Gregory et al., 2013a; Lindsley and Stauffer, 2013; Palucha‐Poniewiera et al., 2013; Pilc et al., 2013), with candidates entering clinical trials for anxiety, depression and Fragile X syndrome with varying successes and failures (Pecknold et al., 1982; Porter et al., 2005; Levenga et al., 2011; Wieronska and Pilc, 2013; Jaeschke et al., 2015; Lindemann et al., 2015). As discussed in detail in the following, mGlu₅ receptors promiscuously couple to various G proteins and interact with complex yet fluid protein scaffolding complexes, such that our understanding of signal transduction mechanisms remains convoluted. Here we attempt to delineate the complex nature of mGlu₅ receptor‐mediated cellular responses and discuss new therapeutic modalities to target mGlu₅ receptors, providing increased receptor specificity to avoid adverse effects and improve clinical outcomes.

Signal transduction of mGlu₅ receptors

Preferentially coupled to Gα_q/11, mGlu₅ receptor activation leads to activation of phospholipase C and production of inositol‐1,4,5‐triphosphate (IP₃) and DAG, and subsequently, mobilization of intracellular calcium (iCa²⁺) (Abe et al., 1992). Calcium, in combination with DAG, leads to the activation of PKC, PLA₂, MAPK and the modulation of ion channels (Hermans and Challiss, 2001; Conn et al., 2008; Ribeiro et al., 2010) (Figure 1). Activation of this canonical signalling pathway is linked to various cell processes including synaptic plasticity (Borgdorff and Choquet, 2002; Gerrow and Triller, 2010; Kato et al., 2012). In addition to G_q coupling, mGlu₅ receptors also couple to G_s in HEK293 cells (Francesconi and Duvoisin, 1998) and are linked to cAMP formation in LLC‐PK1 cells and oocytes (Joly et al., 1995). However, when expressed in CHO cells or astrocytes, no agonist stimulated increases in cAMP accumulation were observed (Abe et al., 1992; Balazs et al., 1997; Ribeiro et al., 2010), suggesting that mGlu₅ receptor promiscuous G protein coupling is cell type‐dependent.

Figure 1.

Activation of mGlu₅ receptors triggers diverse distinct and converging signalling pathways. The group I mGlu receptor family preferentially couples to G_q, leading to activation of the canonical PLC–IP₃–DAG–Ca²⁺ pathway. As reviewed in the text, mGlu₅ receptors promiscuously couple to other G proteins as well as alternate second messengers and kinases via mechanisms that can be independent of Ca²⁺ mobilization. Protein–protein interactions with scaffolding and regulatory proteins (blue ovals) add further diversity: integrating different signalling pathways, directly mediating interactions between mGlu₅ receptors and ion channels such as the NMDA receptor and regulating the activity of the mGlu₅ receptor itself. As an additional level of complexity, the signalling pathways and effectors stimulated by mGlu₅ receptor activation also show considerable cell type dependence.

Downstream of G protein coupling, agonist stimulation of mGlu₅ receptors leads to phosphorylation of ERK1/2 (Thandi et al., 2002; Hu et al., 2007) and p38 MAPK (Peavy and Conn, 1998; Rush et al., 2002). Interestingly, mGlu₅ receptor‐mediated phosphorylation of ERK1/2 has been reported to be independent of PKC, PI3K and calcium (Peavy and Conn, 1998; Thandi et al., 2002). Additional evidence has implicated Homer 1b/c scaffolding protein (Mao et al., 2005), epidermal growth factor receptor tyrosine kinase, proline‐rich tyrosine kinase 2 and Src activation in coupling mGlu₅ receptors to ERK1/2 phosphorylation (Peavy et al., 2001; Thandi et al., 2002; Wang et al., 2007; Nicodemo et al., 2010). mGlu₅ receptor‐mediated ERK1/2 phosphorylation leads to activation of downstream transcription factors including Elk‐1, cAMP response element binding‐protein and c‐Jun, which regulate gene expression involved in long‐term depression (LTD) (Rush et al., 2002; Gallagher et al., 2004; Wang et al., 2007). Group I mGlu receptor activation of PI3K, Akt and mammalian target of rapamycin (mTOR) has also been implicated in LTD (Chan et al., 1999; Hou and Klann, 2004). The mechanism driving group I mGlu receptor‐mediated synaptic plasticity is the interplay between ERK‐MAPK and PI3K–mTOR pathways resulting in de novo protein synthesis and late‐phase LTD (Waung and Huber, 2009). Moreover, in disease states, the balance of mGlu₅ receptor signal transduction may be perturbed. For example, in a mouse model of Huntington's disease, mGlu₅ receptor coupling to Akt, ERK1/2 and intracellular Ca²⁺ release is enhanced, while IP₃ formation is decreased (Ribeiro et al., 2010). In addition to complex acute signalling outcomes, regulatory processes triggered by mGlu₅ receptor activation show a similar level of intricacy.

Regulation of mGlu₅ receptor signalling

Signal transduction arising from mGlu₅ receptors is tightly controlled by the phosphorylation status of the receptor. The frequency of glutamate‐induced iCa²⁺ oscillations are governed by a ‘dynamic uncoupling’ mechanism where the receptor undergoes rapid cycles of phosphorylation and dephosphorylation (Kawabata et al., 1996; Nash et al., 2002; Bradley and Challiss, 2011a). The major mediator of mGlu₅ receptor phosphorylation and subsequent desensitisation of mGlu₅ receptor signalling is PKC, although other kinases also play a role such as G protein‐coupled receptor kinases (GRK), tyrosine kinases and Ca²⁺/calmodulin‐dependent protein kinase II (see Mao et al., 2008). The phosphorylation status of mGlu₅ receptors regulates interactions with other proteins. Trafficking and signalling of mGlu₅ receptors are regulated by calmodulin that interacts with the C terminus of the receptor (Choi et al., 2011). PKC phosphorylation of mGlu₅ receptors inhibits calmodulin binding (Lee et al., 2008) and facilitates seven in absentia homolog 1A binding, promoting endocytosis and receptor down‐regulation via ubiquitination‐dependent mechanisms (Kammermeier and Ikeda, 2001; Moriyoshi et al., 2004; Ko et al., 2012). Calmodulin binding sites also partially overlap with that of Norbin, a neuronal protein that binds mGlu₅ receptors to promote cell surface expression and signalling (Wang et al., 2009). Another key adaptor protein that regulates mGlu₅ receptor function and expression is the Homer family (Brakeman et al., 1997; Xiao et al., 1998; Tu et al., 1999; Ango et al., 2002; Kammermeier and Worley, 2007; Lv et al., 2014). Of note, Homer interactions with mGlu₅ receptors are enhanced by receptor phosphorylation mediated by cyclin‐dependent kinase 5 (Orlando et al., 2009). Recently, mGlu₅ receptors were also shown to be a target of PKA phosphorylation, with mGlu₅ receptor phosphorylation levels being linked to elevated intracellular cAMP levels (Uematsu et al., 2015). Collectively, these data highlight the critical role mGlu₅ receptor phosphorylation plays in regulating receptor function; however, not all mechanisms rely on phosphorylation. For example, GRK2 mediates mGlu₅ receptor desensitization and internalization (Sorensen and Conn, 2003) via a mechanism that can be phosphorylation independent (Ribeiro et al., 2009). Moreover, in disease states, abnormal interactions can occur, for example, disruption of Homer scaffolds in Fragile X syndrome results in abnormal mGlu₅ receptor function (Ronesi et al., 2012). Thus, there is much diversity in both the cellular responses and regulatory mechanisms triggered by mGlu₅ receptor activation. This diversity extends further to mGlu₅ receptor‐mediated modulation of the response to other glutamate receptors.

Activation of mGlu₅ receptors modulates function of other glutamatergic receptors

Activation of group I mGlu receptors modulates the activity of various ion channels, in particular, the NMDA and AMPA ionotropic glutamate receptors (Homayoun et al., 2004; Nakamoto et al., 2007; Kato et al., 2012). Mechanisms of modulation include activation of downstream second messengers, kinases and direct protein–protein interactions (Benquet et al., 2002; Collett and Collingridge, 2004; Homayoun et al., 2004; Ribeiro et al., 2010; Moutin et al., 2012; Gao et al., 2013). The overall response of an individual cell to glutamate is highly integrated. For example, NMDA receptor activation is involved in early phase synaptic plasticity, with rapid calcium influx leading to modulation of AMPA receptor trafficking (Borgdorff and Choquet, 2002). The modulation of ionotropic receptors by mGlu₅ receptors and vice versa is complex. For example, mGlu₅ receptor activation positively modulates the NMDA receptor through increasing open channel probability, with PKC‐dependent Src signalling, and stabilization of a Homer–Shank protein anchor implicated (Lu et al., 1999; Tu et al., 1999; Benquet et al., 2002; Kotecha et al., 2003). In contrast, mGlu₅ receptor activation inhibits NMDA receptor‐mediated activation of adenylate cyclase, resulting in reduced cAMP and neuronal nitric oxide synthase activity, signal transduction mechanisms that are involved in NMDA receptor‐mediated excitotoxicity (Llansola and Felipo, 2010). The relationship between ionotropic receptors and mGlu₅ receptors has resulted in extensive efforts to design therapies targeting mGlu₅ receptors in particular for the treatment of CNS disorders related to NMDA receptor dysfunction (Conn et al., 2008; Matosin and Newell, 2013; Veerman et al., 2014). Indeed, targeting mGlu₅ receptors represents a promising approach to overcome NMDA receptor dysfunction while avoiding excitotoxicity associated with direct stimulation of NMDA receptors.

In addition to modulating ionotropic glutamate receptors, mGlu₅ receptors are known to partner with the mGlu₁ receptor. In the hippocampus, chemically induced LTD is partially blocked by antagonists of either receptor, and a combination of both mGlu₁ and mGlu₅ receptor antagonists is required to completely block the response (Volk et al., 2006). This is in spite of the fact that genetic deletion of mGlu₅ receptors alone is sufficient to prevent (S)‐3,5‐dihydroxyphenylglycine (DHPG)‐induced LTD (Huber et al., 2001), supporting the hypothesis that heterodimerization plays a role. The mechanisms underlying this remain to be fully elucidated, with both mGlu₁/ mGlu₅ receptor heterodimerization and synergistic signalling (independent of heterodimerization) proposed (Romano et al., 1996; Goudet et al., 2005; Doumazane et al., 2011; Sevastyanova and Kammermeier, 2014). Collectively, it is clear that, where expressed, mGlu₅ receptors are integral to the interplay of receptors involved in the overall cellular response to glutamate.

Activation of mGlu₅ receptors affects other ion channels and GPCRs

In addition to modulating glutamatergic receptors, mGlu₅ receptor activation influences the function of other ion channels and GPCRs. In HEK293 cells and isolated cortical neurons, mGlu₅ receptor activation inhibits N‐type and P/Q‐type calcium channels in a voltage‐dependent manner, with both βγ subunits of G_i/o and Homer scaffolding proteins involved (McCool et al., 1998; Kammermeier et al., 2000). In hippocampal neurons, mGlu₅ receptor activation facilitates L‐type voltage‐dependent calcium channels, two mechanisms have been reported, one indirect via mGlu₅ receptor‐stimulated Ca²⁺ and PKC and the other via direct receptor–channel interactions (Topolnik et al., 2009; Kato et al., 2012). Within the periphery, mGlu₅ receptors located in the dorsal root ganglia are implicated in both inflammatory and neuropathic pain, with injury to nerves, spinal cord or peripheral tissues resulting in the release of glutamate from sensory neurons (Valerio et al., 1997; Walker et al., 2001). In these tissues, mGlu₅ receptors are co‐localized with the non‐selective cation channel TRPV1, enhancing the activity of these channels via DAG formation (Varney and Gereau, 2002; Kim et al., 2009). Recent work suggests that mGlu₅ receptor‐mediated potentiation of TRPV1 is biphasic, transiently potentiating TRPV1‐mediated inward currents but inhibiting voltage‐gated calcium channels (Masuoka et al., 2015). The mechanism mediating these two different responses remains to be elucidated.

Co‐localization of mGlu₅ receptors with several different GPCRs has functional consequences on the signalling and/or pharmacology of the individual GPCR (Schroder et al., 2009; Brown et al., 2012). For example, co‐stimulation or blockade of mGlu₅ and adenosine A_2A receptors results in synergistic effects; however, the mechanisms mediating the cross‐talk between these receptors appear to be dependent on the brain region studied (Ferre et al., 2002; Nishi et al., 2003; Domenici et al., 2004; Tebano et al., 2005; Tebano et al., 2006). Intriguingly, this synergy between mGlu₅ and adenosine A_2A receptors has been proposed to involve the formation of heteromers (Ferre et al., 2002; Rodrigues et al., 2005). In fact, in addition to heteromers with mGlu₁ receptors mentioned earlier, mGlu₅ receptors heterodimerize with numerous GPCRs including calcium‐sensing receptors (Gama et al., 2001) and μ‐opioid receptors (Schroder et al., 2009). Furthermore, mGlu₅, adenosine A_2A and D₂ dopamine receptors are thought to form higher‐order oligomers (Cabello et al., 2009), capable of changing the pharmacological properties of a heteromer ‘subunit’ (Ferre et al., 1999; Popoli et al., 2001). It is clear that the signalling pathways triggered and interacting proteins modulated by mGlu₅ receptor activation are complex and highly dependent on the cellular and tissue context. Therefore, it is perhaps not surprising that ligand pharmacology can differ depending on the system under exploration.

Localization of mGlu₅ receptor signalling

Increasing evidence suggests that in addition to different signalling profiles in distinct cell types or brain regions, within a single cell mGlu₅ receptors show sub‐cellular compartmentalization that can impact signal transduction and regulatory mechanisms. On the plasma membrane, mGlu₅ receptors are associated with caveolae and lipid rafts resulting in constitutive internalization via interactions with caveolin‐1 (Francesconi et al., 2009). Importantly, mGlu₅ receptors are mostly (>50–80%) found on intracellular membranes including the inner and outer nuclear membranes and the endoplasmic reticulum (Hubert et al., 2001; Kumar et al., 2008). Because intracellular mGlu₅ receptors are oriented with the N‐terminus in the lumen of the organelle, and a cytoplasmic C‐terminus, the same complement of effector proteins is available, unless effector proteins are similarly subject to sub‐cellular localization (Jong et al., 2014). This compartmentalization of mGlu₅ receptors results in differential pharmacology by virtue of accessibility of ligands. Intracellular receptors are only able to be activated by agonists that are able to diffuse, or be actively transported across membranes (Jong et al., 2005). While activation of mGlu₅ receptors on the plasma membrane is associated with rapid transient calcium mobilization, activation of intracellular mGlu₅ receptors is associated with sustained responses arising from ERK1/2 stimulated Elk‐1 phosphorylation (Jong et al., 2009), induction of LTD in rat hippocampal slices (Purgert et al., 2014) and activation of mediators of neuronal growth, differentiation and sustained synaptic transmission (Jong et al., 2009; Kumar et al., 2012). Thus, in targeting mGlu₅ receptors, drug discovery programmes may also need to consider not only pharmacokinetics within the body, but also pharmacokinetics within the individual cell.

Targeting mGlu₅ receptors: orthosteric ligands

As highlighted earlier, mGlu₅ receptors are implicated in various neuronal processes, and selective mGlu₅ receptor drugs are attractive putative therapies for numerous CNS disorders. Traditional drug discovery strategies have targeted the endogenous ligand binding (orthosteric) site to activate or block receptor activity. Several group I mGlu selective ligands have been discovered (Table 1), with DHPG prevailing as the most commonly utilized group I mGlu selective pharmacological tool (reviewed in Brauner‐Osborne et al., 2007). However, high affinity/potency and mGlu₅ receptor selective orthosteric ligands have remained elusive, likely due to the fact that the orthosteric site is highly conserved across the mGlu receptor family. The complete activation or inhibition of receptor responses achieved by orthosteric ligands also presents an additional liability, as such compounds lack the delicate balance on neurotransmission, which may lead to undesirable or adverse effects. Thus, many research groups have turned their focus to targeting allosteric binding sites that are topographically distinct from the orthosteric site (Christopoulos and Kenakin, 2002; Leach et al., 2007).

Table 1.

Molecular pharmacology of selected mGlu₅ receptor orthosteric ligands

Compound	mGlu5 receptor pharmacology	Subtype selectivity	References
Glutamate	Agonist, affinity: 700 nM	1 > 3 ≥ 5 ≥ 2 = 4 = 8 > 6 >>7	(Brabet et al., 1998; Niswender and Conn, 2010)
Quisqualate	Agonist, affinity: 29 nM	5 = 1 > 2 ≥ 3 > group III	(Porter et al., 1992; Mutel et al., 2000)
ACPD	Agonist, affinity: 2.1 μM	2 > 5 = 3> 1 = 6 = 8> 4 >>7	(Tückmantel et al., 1997; Cartmell et al., 1998; Wu et al., 1998)
(S)‐CHPG	Agonist, affinity: 380 μM	5 ≥ 1 > group II and III	(Doherty et al., 1997; Mutel et al., 2000)
(S)‐DHPG	Agonist, affinity: 3.9 μM	5 >1 > group II and III	(Wisniewski et al., 2002; Hathaway et al., 2015)
(S)‐4CPG	Antagonist, affinity: 26 μM	1 > 5 > group II and III	(Hayashi et al., 1994; Thomsen et al., 1994; Doherty et al., 1999; Mutel et al., 2000)
(S)‐MCPG	Antagonist, affinity: 153–316 μM	5 = 1 > group II and III	(Brabet et al., 1995; Kingston et al., 1995; Doherty et al., 1999; Mutel et al., 2000)

CHPG: 2‐Chloro‐5‐hydroxyphenylglycine; 4CPG: 4‐carboxyphenylglycine; MCPG: α‐methyl‐4‐carboxylphenylglycine.

Targeting mGlu₅ receptors: allosteric ligands

GPCRs readily lend themselves to allosteric modulation due to their dynamic receptor conformations and diversity in potential allosteric binding pockets (Lagerstrom and Schioth, 2008). Binding of a ligand to an allosteric site may modulate the receptor through changes in orthosteric agonist affinity and/or efficacy, a property referred to as cooperativity (Christopoulos and Kenakin, 2002). Allosteric ligands that enhance orthosteric agonist pharmacology are referred to as positive allosteric modulators (PAMs); those that inhibit are negative allosteric modulators (NAMs). Neutral allosteric ligands (NALs) represent a third class that bind to allosteric sites yet have no discernible effect on the pharmacology of orthosteric ligands (Christopoulos et al., 2014). Advantages of allosteric modulator focussed discovery are three‐fold: first, binding to a less conserved allosteric site allows for greater selectivity. Second, as pure allosteric modulators require the presence of orthosteric agonist for activation, they possess the ability to fine‐tune temporal and spatial receptor responses. This is a key difference between CNS drugs currently available on the market, which are largely orthosteric agonists or antagonists, because with allosteric modulators, receptor physiology is maintained, rather than silenced or completely switched on (Christopoulos and Kenakin, 2002; Leach et al., 2007; Melancon et al., 2012). Third, cooperativity between allosteric and orthosteric ligands is both reciprocal and saturable; therefore, allosteric modulators have a ceiling level to their effect and are theoretically safe in the case of overdose.

The concept of allosteric modulation has evolved with increasing understanding of GPCRs, from the simple two‐state model, which dictates that the receptor exists in equilibrium between the active and inactive state, to that of the multi‐state model that highlights the ability of ligands to stabilize unique receptor conformations to induce distinct downstream responses (Vauquelin and Van Liefde, 2005). Numerous models have been applied to quantify allosteric interactions at the mGlu receptor family (Hall, 2000; Rovira et al., 2008); however, the approach that has gained the most traction is the operational model of allosterism. The operational model of allosterism combines the simple allosteric ternary complex model with the operational model of agonism (Black and Leff, 1983; Leach et al., 2007). The advantage to this approach is the potential to delineate the contribution of affinity and efficacy modulation to the overall cooperativity observed. Moreover, an increasing number of allosteric ligands for many GPCRs including mGlu₅ receptors (Kinney et al., 2005; Noetzel et al., 2012b; Gregory et al., 2013c; Rook et al., 2013) have been identified that have intrinsic efficacy in their own right (PAM‐agonists), an aspect that is also captured and quantifiable in this model (Keov et al., 2011; Christopoulos et al., 2014). Allosteric modulator discovery for mGlu₅ receptors has been particularly fruitful, yielding diverse chemotypes covering the full spectrum of allosteric pharmacology, including PAMs, PAM‐agonists, NALs, weak (or partial) NAMs and full NAMs (Table 2). Notably, mGlu₅ receptor PAMs have efficacy in models of psychosis and cognition, while mGlu₅ receptor NAMs have demonstrated efficacy in preclinical models of anxiety, depression, autism, addiction and movement disorders. Despite this success, there are now multiple reports of adverse effect liability associated with both mGlu₅ receptor NAMs and PAMs (Table 2). Moreover, mGlu₅ receptor allosteric modulator structure–activity relationships are notoriously difficult to interpret (Wood et al., 2011). This is likely attributable to three major contributing factors. First, a lack of quantitative analyses delineating how chemical modifications alter ligand affinity versus cooperativity. Moreover, the nature, direction and magnitude of cooperativity are dependent upon the two ligands under investigation. This phenomenon is referred to as ‘probe‐dependence’, with the most physiologically relevant effect being the pharmacology observed with the endogenous agonist glutamate. However, in many experimental paradigms, in particular native systems, it is not feasible to use glutamate due to the presence of other glutamate receptors and transporters such that surrogate selective orthosteric agonists are used. Second, an under‐appreciation of the full scope of drug action, with the majority of allosteric ligand‐based discovery programmes relying on glutamate‐mediated iCa²⁺ mobilization to classify ligand pharmacology. As detailed earlier, mGlu₅ receptor‐calcium mobilization represents only the tip of the iceberg with respect to the myriad of possible cellular responses that may be triggered by mGlu₅ receptor activation. Furthermore, the production and release of glutamate by cells, receptor expression levels and constitutive activity have the potential to influence the degree of agonist efficacy of allosteric ligands (Noetzel et al., 2012b). Third, a paucity in our understanding of the specific ligand–receptor interactions that govern allosteric modulation.

Table 2.

Molecular pharmacology and therapeutic potential of selected mGlu₅ receptor allosteric ligands

Ligand	Pharmacology	Therapeutic potential	References
PAMs and PAM‐agonists
VU0360172 VU0403602 VU0424465 DPFE CDPPB	PAM: glu‐Ca2+ mobilization (iCa2+); PAM‐agonist: glu‐pERK1/2. Affinity: 269–282 nM PAM‐agonist: glu‐iCa2+. Affinity: 5 nM PAM‐agonist: glu‐iCa2+ Affinity: 12 nM PAM‐agonist: glu‐iCa2+ and pERK1/2. Affinity: 8 μM PAM: quisqualate‐PI hydrolysis; PAM‐agonist: glu‐iCa2+, pERK1/2 and p‐Akt. Affinity: 224–1700 nM	Psychosis, absence seizures, cognition enhancement, traumatic brain injury, post‐traumatic stress disorder, addiction, Huntington's disease, seizures and neurotoxicity	(Rodriguez et al., 2010b; Gregory et al., 2012; D'Amore et al., 2013; Loane et al., 2014; Bridges et al., 2013; Rook et al., 2013; Gregory et al., 2013b; Gregory et al., 2013c; Kinney et al., 2005; Uslaner et al., 2009; Cleva et al., 2011; Reichel et al., 2011; Fowler et al., 2013; Horio et al., 2013; Ganella et al., 2014; Gass et al., 2014b; Gass et al., 2014a; Parmentier‐Batteur et al., 2014; Sethna and Wang, 2014; Doria et al., 2015)
NAMs
MPEPFenobam	NAM: glu‐iCa2+, DHPG and quisqualate‐PI hydrolysis; Inverse agonist: PI hydrolysis; Affinity: 3–20 nMNAM: quisqualate‐iCa2+ and PI hydrolysis; Inverse agonist: PI hydrolysis. Affinity: 31–53 nM	Analgesia, anxiety, depression, addiction, neuroprotection, movement disorders e.g. levodopa‐induced dyskinesia, Fragile X Syndrome, memory impairments, psychostimulant	(Gasparini et al., 1999; Salt et al., 1999; Bruno et al., 2000; Spooren et al., 2000a; Spooren et al., 2000b; Schulz et al., 2001; Walker et al., 2001; Balschun and Wetzel, 2002; Campbell et al., 2004; Li et al., 2006; Belozertseva et al., 2007; Gannon and Millan, 2011; Sun et al., 2012; Yeganeh et al., 2013; Pecknold et al., 1982; Porter et al., 2005; Berry‐Kravis et al., 2009; Jacob et al., 2009; Rylander et al., 2010; Crock et al., 2012; Vinueza Veloz et al., 2012; Watterson et al., 2013; Wieronska and Pilc, 2013; Huang et al., 2014; Ko et al., 2014)
NALs
5MPEP	NAL: glu‐iCa2+ and PI hydrolysis; DHPG‐Ca2+ mobilization. Affinity: 388‐875 nM	Pharmacological tools	(Rodriguez et al., 2005; Chen et al., 2007; Chen et al., 2008; Ayala et al., 2009; Gregory et al., 2013c; Gregory and Conn, 2015)
VU0478006 (ML353)	NAL: glu‐iCa2+ Affinity: 18 nM.

Italicized text in the ‘Therapeutic potential’ column denotes adverse effect liability associated with these allosteric ligand classes.

Location of mGLu₅ allosteric binding sites

Promisingly, the recent X‐ray crystal structures of both human mGlu₁ and mGlu₅ receptor 7TMs, complexed with NAMs (Dore et al., 2014; Wu et al., 2014), have paved the way for an improved understanding of the complexities of allosteric interactions within the 7TMs. In both crystal structures, the allosteric ligand‐binding pocket is extracellularly accessible and resides between TM2, TM3, TM5, TM6 and TM7 (Dore et al., 2014; Wu et al., 2014). This pocket is consistent with extensive previous mutagenesis studies of the ‘2‐methyl‐6‐(phenylethynyl)pyridine (MPEP) binding site’ or the common allosteric site that is shared across multiple mGlu₅ receptor allosteric modulator chemotypes, including NAMs, PAMs and NALs (Gregory and Conn, 2015). Indeed, the crystal structures are also consistent with structure‐function studies of allosteric pockets across multiple family C GPCRs (Bennett et al., 2015; Gregory and Conn, 2015), suggesting a common location for binding cavities in family C GPCRs that can be selectively targeted with small molecule allosteric ligands. Commensurate with this concept is the fact that some mGlu₅ receptor allosteric ligands have activity at other mGlu subtypes (Mathiesen et al., 2003; O'Brien et al., 2004; Sheffler et al., 2010). Comparison of the allosteric binding pockets in the two structures reveals that the mGlu₁ receptor pocket is located higher within the 7TMs (Bennett et al., 2015; Gregory and Conn, 2015). This suggests that although there is a common allosteric site for family C GPCRs, the overall geography of the pocket can be dramatically different for individual receptors enabling selective ligand discovery.

In addition to this well‐characterized allosteric site, at least one other allosteric site has proposed for mGlu₅ receptors. The mGlu₅ receptor PAMs of glutamate‐stimulated iCa²⁺ mobilization, N‐(4‐chloro‐2‐[(1,3‐dioxo‐1,3‐dihydro‐2H‐isoindol‐2‐yl)methyl]phenyl)‐2‐hydroxybenzamide (CPPHA), and its derivative, N‐(4‐chloro‐2‐[(1,3‐dioxoisoindolin‐2‐yl)methyl)phenyl)picolinamide] (NCFP), show non‐competitive behaviour with the MPEP‐site radioligand, [³H]methoxyPEPy (O'Brien et al., 2004; Noetzel et al., 2013). Mutagenesis studies identified Phe⁵⁸⁵ in TM1 as a possible contributor to this secondary allosteric site, and perhaps more importantly, showed no effects of mutations of this residue on the binding of MPEP allosteric site ligands, and vice versa (Chen et al., 2008). Additional chemotypes have been suggested to bind to secondary allosteric sites on mGlu₅ receptors (Hammond et al., 2010; Rodriguez et al., 2010a); however, it remains to be established whether or not these chemotypes recognize the same binding pocket as CPPHA and NCFP or if there are additional allosteric sites. To date, no ‘second site’ mGlu₅ receptor allosteric modulators have been developed that are active in vivo. Despite this, the discovery of ligands that interact with secondary allosteric binding pockets coupled with our appreciation of the complex physiology of mGlu₅ receptor activation offers the potential for the development of allosteric ligands that have differential pharmacology and in vivo profiles.

New paradigms: biased allosteric modulation and agonism

In the past two decades, it has become apparent that the consequences of GPCR activation are not linearly linked to receptor occupancy and that different agonists binding to the same receptor can activate distinct cellular responses. Referred to as stimulus‐bias, this phenomenon arises from ligands stabilizing different subsets of receptor activation states (Kenakin et al., 2012; Kenakin and Christopoulos, 2013). Stimulus‐bias has been described for different orthosteric agonists of mGlu₁ and group III mGlu receptors (Emery et al., 2012; Jalan‐Sakrikar et al., 2014). Of note, DHPG and quisqualate, two orthosteric agonists commonly used as surrogates for glutamate in native preparations, are biased compared with glutamate in recombinant cells and neurons expressing mGlu₁ receptors when comparing phosphoinositide hydrolysis, ERK1/2 phosphorylation and cytoprotection (evidenced by differential relative potencies and/or efficacies) (Emery et al., 2012; Hathaway et al., 2015). It remains to be determined whether or not different orthosteric agonists will also be biased for different outcomes of mGlu₅ receptor activation. However, comparison of relative efficacies of mGlu₅ receptor PAM‐agonists with that of glutamate revealed that allosteric agonists were biased towards ERK1/2 phosphorylation over intracellular Ca²⁺ mobilization, the opposite profile to glutamate (Gregory et al., 2012) (Table 2). This is not necessarily surprising given that mGlu₅ receptor‐induced ERK1/2 phosphorylation can occur independently of canonical Ca²⁺ signalling (Thandi et al., 2002). Biased agonists provide the opportunity to study how individual transducer pathways relate to the overall effect of a ligand in complex systems such as tissue slices or in behavioural models; however, it also add an additional layer of complexity to understanding allosteric interactions.

The simultaneous binding of an allosteric modulator to an agonist bound receptor stabilizes a different complement of receptor activation states than can be achieved by the endogenous ligand alone, giving rise to cooperativity. In light of evidence that mGlu₅ receptor allosteric agonists exhibit stimulus‐bias, it is perhaps not surprising that there is also evidence that mGlu₅ receptor allosteric modulators can differentially modulate the functional outcomes of endogenous agonist stimulation. An early study comparing the mGlu₅ receptor PAMs for iCa²⁺ mobilization, CPPHA and ((3‐fluorophenyl)methylene)hydrazone‐3‐fluorobenzaldehyde (DFB) found that while DFB was a PAM for ERK1/2 phosphorylation, CPPHA was a NAM‐agonist (Zhang et al., 2005). These differences were attributed to the fact that these two allosteric ligands interact with distinct allosteric sites. Subsequently, comparison of the cooperativities of diverse mGlu₅ receptor PAM scaffolds including CPPHA revealed that these PAMs have lower cooperativity with glutamate for ERK1/2 phosphorylation compared with iCa²⁺ mobilization (Gregory et al., 2012; Gregory et al., 2013c). These data suggest that ligands binding within the same pocket can have differential cooperativity with endogenous agonist depending upon the measure of receptor activity. Furthermore, point mutations in the common allosteric site can engender molecular switches in allosteric modulator cooperativity; however, differential effects were observed for different NAM and PAM scaffolds (Muhlemann et al., 2006; Gregory et al., 2013b; Turlington et al., 2013; Gregory et al., 2014). Therefore, it is evident that the structural determinants and receptor conformations that mediate cooperativity are subtly different for individual allosteric ligands and that these subtleties in ligand–receptor interactions may give rise to biased modulation. It should be noted, however, that iCa²⁺ mobilization and ERK1/2 phosphorylation are measured at different time points, <60 s for iCa²⁺ mobilization versus 5–10 min for ERK1/2 phosphorylation. Differences in assay timing, as well as compound kinetics, may contribute to the reported biased agonism and modulation. To date, there have been no detailed studies examining multiple proximal and distal signalling endpoints for mGlu₅ receptors to rigorously assess the prevalence of allosteric ligand bias. Further research that incorporates the appropriate controls, analysis and quantification (Kenakin and Christopoulos, 2013) to probe mGlu₅ receptor biased agonism and modulation is required to better understand the significance, prevalence and therapeutic potential of these phenomena.

Despite the preliminary evidence, recent studies suggest that biased pharmacology of mGlu₅ receptor allosteric ligands may result in distinct effects at the synaptic level. NCFP was unable to potentiate induction of LTD and LTP at the hippocampal‐SCA1 synapse – a property distinct from other mGlu₅ receptor PAMs (Noetzel et al., 2013). Despite the fact that NCFP potentiates mGlu₅ receptor activation in astrocytes and midbrain slices, similar to other mGlu₅ receptor PAMs (O'Brien et al., 2004; Rodriguez et al., 2010b; Noetzel et al., 2012b). More recently, a novel mGlu₅ receptor PAM, VU0409551, was disclosed that has efficacy in preclinical models of anti‐psychotic activity and cognition enhancement, but does not potentiate mGlu₅ receptor modulation of NMDA receptor currents in hippocampal neurons and, similar to NCFP, does not potentiate NMDA receptor mediated LTP at the hippocampal‐SCA1 synapse (Rook et al., 2015). These data suggest that modulating the interaction between mGlu₅ receptors and NMDA receptors is not necessarily required to drive in vivo efficacy of mGlu₅ receptor PAMs and that additional, yet to be determined, mGlu₅ receptor‐linked mechanisms are involved. If we consider the myriad of possible cellular responses that can be triggered by mGlu₅ receptor activation, it is perhaps not surprising that biased agonism and modulation is operative. Two recent reports described adverse effect liability (seizures and neurotoxicity) associated with on‐target activity of the mGlu₅ receptor PAMs, VU0424465 and 5PAM523, with both intrinsic efficacy and the magnitude of cooperativity implicated (Rook et al., 2013; Parmentier‐Batteur et al., 2014). However, it remains to be fully elucidated why these compounds had adverse effects, while other related compounds do not. Therefore, biased allosteric modulators offer two advantages: as tool compounds to dissect the physiology of therapeutic versus adverse effects and as an avenue for improved therapeutic effects. Encouragingly, biased ligands for other GPCRs have reached clinical trials, with angiotensin AT₁ (Felker et al., 2015) and μ opioid receptor biased ligands (Soergel et al., 2014) in early trials for the treatment of acute heart failure and moderate to severe pain respectively.

In the field of drug discovery, stimulus‐bias raises additional challenges and potential pitfalls. The use of biased orthosteric agonists as pharmacological tools in native preparations may confound interpretation of allosteric modulator pharmacology. For example, with quisqualate, a number of mGlu₅ receptor PAMs displayed affinity modulation (Bradley et al., 2011b); however, with glutamate as the agonist, there was no evidence for affinity modulation with these same PAMs and others, and only modulation of glutamate efficacy was observed (Gregory et al., 2012). Thus, in assessing allosteric modulation, and ligand pharmacology, it is essential to consider probe dependence when interpreting results from recombinant and native systems and translating this to in vivo effects. Furthermore, as noted earlier, in addition to coupling to a wide variety of effectors, mGlu₅ receptors are also subject to subcellular compartmentalization. This leads to the interesting notion of stimulus‐bias based not only on chemical pharmacophore but also location of the receptor within the cell and accessibility to different ligands. Thus, in characterizing allosteric modulator pharmacology, it is important to understand the complete pharmacological profile, as well as the impact of different orthosteric agonists to the individual signalling fingerprint.

Concluding remarks

The past decade has seen a large focus on developing mGlu₅ receptor‐targeted therapies for the treatment of often refractory CNS disorders. In appreciating the breadth of physiological possibilities, it is clear that mGlu₅ receptors indeed play a significant role in the symphony of neuronal responses, ranging from the sensation of pain to the formation of memory. The richness of possible cellular responses associated with mGlu₅ receptor activation offers a large variety of options in reaching therapeutic endpoints. While several mGlu₅ receptor allosteric modulators have moved into early phase clinical trials, several have also failed due to adverse effects and inability to meet expected clinical endpoints (Pecknold et al., 1982; Jacquemont et al., 2014; Scharf et al., 2015). Both the adverse effects and lack of efficacy can likely be attributed to an under‐appreciation of the full scope of ligand pharmacology, whether this be due to off‐target activity, a paucity in our understanding of on‐target stimulus‐bias or a lack of selectivity due to unappreciated stimulus‐bias at other subtypes not detected in traditional counter‐screening paradigms. Thus, a greater understanding of the full scope of mGlu₅ receptor allosteric ligand activity, as well as receptor function, may aid in developing therapies with minimal adverse effects and desired clinical outcomes.

Conflict of interest

None.

Sengmany, K. , and Gregory, K. J. (2016) Metabotropic glutamate receptor subtype 5: molecular pharmacology, allosteric modulation and stimulus bias. British Journal of Pharmacology, 173: 3001–3017. doi: 10.1111/bph.13281.