Biased allosteric agonism and modulation of metabotropic glutamate receptor 5: implications for optimizing preclinical neuroscience drug discovery

Abstract

1. Introduction

In recent years, the metabotropic glutamate receptor subtype 5 (mGlu₅), a G protein-coupled receptor (GPCR) ubiquitously expressed throughout the brain, has emerged as a promising drug target for multiple central nervous system (CNS) disorders, including schizophrenia, autism, Parkinson’s disease and depression (Gregory et al., 2013c; Niswender and Conn, 2010). Drug design efforts for mGlu₅ are largely focused on targeting allosteric sites, that is, receptor binding sites topographically distinct from that of the endogenous ligand. Allosteric ligands offer the promise of increasing receptor selectivity, maximizing therapeutic effects and minimizing adverse effects. Further, modulation of orthosteric agonists by allosteric ligands without intrinsic agonism, allows for spatiotemporal fine-tuning of receptor responses – a favorable therapeutic mechanism within the complex CNS environment (Christopoulos and Kenakin, 2002; Leach et al., 2007; Melancon et al., 2012). Allosteric ligands that enhance receptor responses are called positive allosteric modulators (PAMs), while some allosteric compounds may also activate the receptor in the absence of endogenous ligand (PAM-agonists). Similar to orthosteric agonists, PAM-agonists do not have the benefit of temporal control of receptor activity. Allosteric ligands that diminish endogenous responses are termed negative allosteric modulators (NAMs), and those that neither increase nor decrease responses are referred to as neutral allosteric ligands (NALs) (Gentry et al., 2015). Discovery efforts for mGlu₅ allosteric modulators have been particularly fruitful. Diverse modulator chemotypes recognize the ‘MPEP’ site, an allosteric pocket located within the receptor’s seven transmembrane bundle (Dore et al., 2014; Gregory et al., 2014; Gregory et al., 2013b; Wu et al., 2014). Of note, mGlu₅ PAMs (classified based on glutamate stimulation of mGlu₅-intracellular Ca²⁺ (iCa²⁺) mobilization in recombinant cells) have demonstrated efficacy in preclinical models of psychosis and cognition enhancement (Kinney et al., 2005; Rodriguez et al., 2010; Gregory et al., 2013a). However, recent studies have also reported adverse effects for select mGlu₅ enhancers, such as seizure activity induced by VU0424465 (a PAM-agonist for mGlu₅ mediated iCa²⁺ mobilization) (Rook et al., 2013) and neurotoxicity in rats treated with 5PAM523 (a pure PAM for mGlu₅-iCa²⁺ mobilization stimulated by glutamate) (Parmentier-Batteur et al., 2014).

It has also become increasingly evident that, like many GPCRs, the mGlu₅ receptor is pleiotropically coupled, activating multiple signaling pathways. In particular, while the receptor is predominantly coupled to G_q and subsequent changes in intracellular calcium (iCa²⁺) mobilization, mGlu₅ couples to iCa²⁺-independent signaling pathways, such as extracellular signal-regulated kinases (ERK) 1 and 2 phosphorylation, cAMP, mTOR/PI3K, and can interact with, and modulate the activity of, other GPCRs and ion channels (Sengmany and Gregory, 2015). It is conceivable therefore that mGlu₅ activation by chemically diverse ligands may engender unique receptor conformations, such that the receptor favors distinct subsets of physiological responses. This concept is known as ‘biased agonism’ (Kenakin and Christopoulos, 2013). Biased agonism offers the opportunity to design drugs that are tailored to activate the desired complement of receptor responses linked to therapeutic outcomes while avoiding those linked to adverse effects. Biased agonism is operative for orthosteric agonists at the related Group I mGlu₁ receptor between G-protein and β-arrestin pathways in both recombinant and natively expressing cells (Emery et al., 2010; Emery et al., 2012; Hathaway et al., 2015). Several studies have suggested that mGlu₅ biased agonism may be linked to distinct physiological responses (Gregory et al., 2013a; Gregory et al., 2012; Noetzel et al., 2013; Zhang et al., 2005). However, the vast majority of mGlu₅ allosteric modulators have been classified as PAMs, PAM-agonists, NAMs or NALs, based on modulation of glutamate-mediated iCa²⁺ mobilization alone. To date, there have been no rigorous analyses of the mGlu₅ signaling fingerprints of diverse allosteric modulators across different pathways. It is possible that preferential activation or inhibition of distinct mGlu₅ signaling pathways could lead to different behavioral and/or toxicological outcomes.

One means to avoid the neurotoxicity and seizure liability associated with certain mGlu₅ PAM and/or PAM-agonists (Rook et al., 2013; Parmentier-Batteur et al., 2014) and retain anti-psychotic and pro-cognitive effects may be via the development of biased mGlu₅ modulators (Rook et al., 2015). Moreover, hitherto unappreciated bias may contribute to observations that markedly different doses of the same modulator are required for efficacy in different preclinical behavioral models. For example, DPFE, an mGlu₅ PAM-agonist of glutamate-mediated iCa²⁺ mobilization and ERK1/2 phosphorylation in HEK293A cells, required vastly different doses to elicit cognition-enhancement versus reversal of amphetamine-induced hyperlocomotion in rats (Gregory et al., 2013a). Therefore, in the current study, we undertook the first rigorous analysis of mGlu₅ bias for diverse PAMs in both HEK293A cells and primary cortical neuron cultures. Allosteric ligands were selected based on i) known interaction at the ‘MPEP’ allosteric site, ii) previous classification as PAMs or PAM-agonists in iCa²⁺ assays and iii) existing data in preclinical behavioral models. Propensity for biased allosteric agonism was determined relative to the orthosteric agonist DHPG across four signaling pathways that were amenable to assessment in both recombinant and neuronal cell cultures. We demonstrate that mGlu₅ allosteric modulators display biased agonism and cooperativity, providing proof-of-concept that the development of biased mGlu₅ allosteric modulators may be a means to tailor enhancement of mGlu₅ activity to elicit therapeutically beneficial effects, such as anti-psychotic and cognition enhancement, while avoiding adverse effects.

2. Materials and Methods

3. Results

Eight allosteric ligands previously classified as either PAMs or PAM-agonists based on glutamate stimulation of mGlu₅-mediated iCa²⁺ mobilization representing four distinct chemotypes were selected for this study to probe the potential for biased agonism and biased modulation of mGlu₅ (Supplementary Fig. 1). In preclinical studies, the structurally related VU0403602 and VU0424465 have known adverse effect liability (Bridges et al., 2013; Rook et al., 2013) and provide a reference for modulators with undesirable properties. We also included two other ligands from this scaffold, VU0360172 and VU0405398, both of which have lower cooperativity with glutamate in iCa²⁺ assays, and minimal agonism compared to VU0403602 or VU0424465 (Gregory et al., 2012). CDPPB and VU29 were included as structurally distinct allosteric modulators from the same chemotype class, where CDPPB has been reported to be a PAM-agonist and VU29 a ‘pure’ PAM (Chen et al., 2007; Lindsley et al., 2004). CDPPB has previously been tested in vivo, showing efficacy in models of Huntington’s disease, fear extinction, psychosis and addiction (Cleva et al., 2011; Doria et al., 2015; Ganella et al., 2014; Gass et al., 2014; Uslaner et al., 2009). Recently, VU0409551 was described as a biased modulator on the basis that VU0409551 positively modulates mGlu₅ activity in vitro but does not enhance mGlu₅ modulation of NMDA current or NMDA receptor dependent plasticity in the hippocampus (Rook et al., 2015). Importantly, VU0409551 has efficacy in preclinical models of antipsychotic-like activity, cognition-enhancement and a genetic model of schizophrenia (Balu et al., 2016; Conde-Ceide et al., 2015; Rook et al., 2015). DPFE, a structurally distinct ligand relative to the three previous classes, was also included to offer further insight into the relationship between chemical scaffold and pharmacological outcome. As with CDPPB & VU0409551 behavioral studies have been conducted with DPFE (Gregory et al., 2013a; Peters et al., 2015), thereby allowing the potential to link in vivo efficacies with in vitro pharmacological fingerprints.

We first confirmed the pharmacological phenotypes for each of the eight allosteric modulators as potentiators of glutamate stimulation of mGlu₅-mediated iCa²⁺ mobilization in HEK293A-mGlu₅-low cells. In agreement with previous studies, compared to the orthosteric agonists glutamate and DHPG, VU0424465, VU0403602, and DPFE were efficacious partial allosteric agonists, achieving maximal responses 50–80% of the glutamate response. VU0360172, VU0409551 and CDPPB were also partial allosteric agonists. VU29 and VU0405398 had no appreciable allosteric agonist activity (Fig. 1A & B). The rank order of agonist efficacies was: glutamate = DHPG > VU0424465 > VU0403602 > DPFE ≥ VU0409551 > VU0360172 > CDPPB (see Supplementary Table 1 for E_max values). The rank order of agonist potencies (Supplementary table 1) was: VU0403602 > CDPPB ≥ VU0424465 > glutamate > VU0360172 > DHPG = VU0409551 > DPFE.

Open in a separate window

Figure 1

mGlu₅ allosteric ligands have agonist activity for iCa²⁺ mobilization, IP₁ accumulation and ERK phosphorylation in HEK293A-mGlu₅-low

Concentration-response curves for indicated mGlu₅ orthosteric and allosteric ligands for iCa²⁺ mobilization (A & B), IP₁ accumulation (C & D), ERK1/2 phosphorylation (E & F) and iCa²⁺ mobilization in the absence of 1.2mM CaCl₂ (G & H; % glu max represents maximal glutamate response in the presence of 1.2mM CaCl_2, curve from panel A shown in grey). In panels B, D, F and H, the DHPG curve is plotted as dotted lines for ease of reference. Responses were normalized to either the glutamate maximal response (iCa²⁺ mobilization) or represented as fold over basal (IP₁ accumulation and ERK phosphorylation). IP₁ and ERK1/2 phosphorylation experiments were performed in the presence of 10 U/mL GPT to minimize contribution of ambient glutamate. Data are expressed as mean + SEM of 3–17 experiments performed in duplicate, error bars not shown lie within the dimensions of the symbols.

All eight compounds had different degrees of positive cooperativity (logβ) with glutamate for iCa²⁺ mobilization, causing concentration-dependent increases in glutamate potency that approached a limit (Supplementary Figure 1). Modulator effects on glutamate concentration-response curves for iCa²⁺ mobilization were fitted to an operational model of allosterism to quantify the functional affinity (pK_B) and efficacy modulation (logβ) for each of the ligands (Table 1). The resulting affinity and cooperativity estimates agree with those determined in previous studies (Bridges et al., 2013; Gregory et al., 2013a; Gregory et al., 2012; Manka et al., 2010).

3.1. Select mGlu₅ PAMs show ‘probe dependence’ for potentiation of mGlu₅-iCa²⁺ mobilization

Although glutamate is the cognate agonist, it is a sub-optimal ligand to study the effects of mGlu₅ alone in primary neuronal cultures due to the presence of multiple glutamate receptor subtypes and transporters in these native cells. Thus, we chose the more selective DHPG as the reference orthosteric agonist for assessing ligand pharmacology in primary neuronal cultures. We first confirmed that all eight allosteric ligands potentiated DHPG-mediated iCa²⁺ mobilization in HEK293A-mGlu₅-low cells. DHPG concentration-response curves for iCa²⁺ mobilization in the absence or presence of different concentrations of each allosteric ligand (Fig. 2) were analyzed using the operational model of allosterism. As expected, affinity estimates were similar, irrespective of which orthosteric agonist was used (Table 1). Cooperativity estimates (logβ) for the interaction with DHPG were also similar to those determined with glutamate for VU0403602, VU0360172, VU0409551 and CDPPB. In contrast, the cooperativity between VU0424465, DPFE or VU0405398 with DHPG was ~2-fold higher than with glutamate. VU29 had ~5-fold higher positive cooperativity with DHPG compared to glutamate.

Open in a separate window

Figure 2

Allosteric modulation of DHPG-stimulated iCa²⁺ mobilization in HEK293A-mGlu₅-low cells

DHPG concentration-response curves for iCa²⁺ mobilization in the absence and presence of indicated concentrations of allosteric ligands. Interaction studies for DPFE, VU0409551, VU0424465, and VU0403602 were performed using simultaneous addition of both ligands, to minimize allosteric ligand-induced acute desensitization due to intrinsic agonist activity. VU29, VU0405398, CDPPB and VU0360172 were added 1 min prior to addition of glutamate. Data sets were globally fitted to an operational model of allosterism to estimate affinity and cooperativity. Curves represent the best fit of the data. Data are mean + SEM of n=3–10 experiments performed in duplicate. Error bars not shown lie within the dimensions of the symbol.

3.4. Assessment of agonist activity of mGlu₅ allosteric ligands in cortical neurons

The same four measures of mGlu₅ activity that we investigated in recombinant cells were next examined in cortical neurons to assess the potential of biased allosteric agonist activity in a native system (Fig. 3). Cortical neurons represent a physiologically relevant system to assess mGlu₅ allosteric ligand activity, and can serve as a vital translational link for validating allosteric modulator pharmacology prior to preclinical assessments. To eliminate the influence of mGlu₁, which is also expressed in the cortex, the mGlu₁ selective negative allosteric modulator CPCCOEt 30 μM, was used in conjunction with DHPG for all experiments in cortical neurons. GPT had no effect on basal IP₁ levels or allosteric agonism (Supplementary Fig. 3) and was not included in these experiments.

Open in a separate window

Figure 3

mGlu₅ allosteric ligands have agonist activity for iCa²⁺ mobilization, IP₁ accumulation and ERK phosphorylation in cultured cortical neurons

Concentration-response curves for indicated mGlu₅ orthosteric and allosteric ligands for iCa²⁺ mobilization (A&B), IP₁ accumulation (C & D), ERK1/2 phosphorylation (E&F) and iCa²⁺ mobilization in the absence of 1.2mM CaCl₂ (G & H; % DHPG max represents maximal DHPG response in the presence of 1.2mM CaCl₂, curve from panel A is shown in grey). In panels B, D, F and H, the DHPG curve is plotted as a dotted line for ease of reference. Responses were normalized to either the glutamate maximal response (iCa²⁺ mobilization) or represented as fold over basal (IP₁ accumulation and ERK phosphorylation). All assays were performed in the presence of 30 μM CPCCOEt to eliminate the contribution of mGlu₁. Data are expressed as mean + SEM of 3–22 experiments performed in duplicate, error bars not shown lie within the dimensions of the symbols.

DHPG was ~10 fold less potent for IP₁ accumulation relative to iCa²⁺ (in absence and presence of 1.2 mM extracellular Ca²⁺) and ERK1/2 phosphorylation (Supplementary Table 3). In iCa²⁺ mobilization assays relative to DHPG, VU0424465 and VU0403602 were nearly full agonists, whereas VU0360172, DPFE, VU29 and CDPPB were all partial agonists (Fig. 3A, B, Supplementary Table 2). VU0405398 also showed agonist activity, however, the maximal response was highly variable and restricted to concentrations above 3 μM. VU0409551 had no appreciable agonist activity for iCa²⁺ mobilization. The rank order of agonist efficacy: DHPG > VU0424465 = VU0403602 > CDPPB = VU29 = VU0360172 > VU0405398 ≫ VU0409551, was different to that observed in HEK293A cells.

For IP₁ accumulation the relative efficacies of the allosteric ligands were similar to that observed for iCa²⁺ assays, with the exception of VU0409551 which had appreciable agonist activity for IP₁ accumulation (Fig. 3C–D). The time course for mGlu₅-mediated ERK1/2 phosphorylation by both orthosteric and allosteric ligands was characterized by sustained elevations or inhibition in cortical neurons, a markedly different profile to the transient response observed in HEK293A cells (Supplementary Figure 4). VU0424465, VU0403602, VU0360172, VU29, CDPPB and VU0409551 were all agonists for pERK1/2 in cortical neurons eliciting maximal responses to similar to DHPG (Fig. 3E–F). Allosteric agonism (iCa²⁺, pERK1/2 and IP₁) in cortical neurons was inhibited by pre-exposure to the mGlu₅ selective neutral allosteric ligand 5MPEP confirming that agonism was mediated via interaction with mGlu₅ (Supplementary Fig. 5). Interestingly, while DPFE and VU0405398 were partial agonists for IP₁ accumulation, they exhibited a bell-shaped concentration-response relationship for pERK1/2, elevating pERK1/2 at 1μM and dropping below basal at concentrations above 1μM (Fig. 3E–F). Similar to observations in HEK293A-mGlu₅-low cells, the DHPG maximal response was significantly reduced in the absence of extracellular calcium (Fig. 3G). There were no significant differences in potency or maximal response of VU0424465, VU0403602, VU0360172, or CDPPB between differing calcium conditions (Fig. 3G–H). However, agonist activity by VU29, DPFE and VU0405398 was lost.

3.5. mGlu₅ allosteric ligands are biased agonists

In order to rigorously and quantitatively assess the apparent bias of mGlu₅ allosteric agonists in HEK293A-mGlu₅-low cells and cortical neurons, observational bias between assays, and systems bias of the cell background should be minimized – thereby resulting in a more faithful estimate of true (conformationally-driven) biased agonism (Kenakin et al., 2012). From each assay, a transduction coefficient log(τ/K_A) (Supplementary Table 4), that is, the composite of ligand efficacy and affinity, was derived as previously described (Kenakin et al., 2012) and normalized to the reference ligand DHPG (Δlog(τ/K_A)). Subsequent normalization of Δlog(τ/K_A) between pathways, ΔΔlog(τ/K_A), allows for quantification of the degree of bias between pathways (Figure 4). Each allosteric ligand displayed a unique signaling bias fingerprint in HEK293A-mGlu₅-low cells and neurons, although there were some notable trends. Relative to DHPG, VU0424465 showed significant bias away from iCa²⁺ mobilization and toward IP₁ accumulation (110-fold) and ERK1/2 phosphorylation (9-fold) in HEK293A-mGlu₅-low cells, a profile that was largely retained in neurons. VU0403602 was biased toward IP₁ over pERK1/2 (11-fold) and iCa²⁺ mobilization (14-fold) in HEK293A-mGlu₅-low cells, however this wasn’t reflected in the neurons. VU29 and CDPPB were biased (3–25 fold) toward IP₁ over pERK1/2 and/or iCa²⁺ mobilization in both HEK293A cells and neurons. DPFE, VU0409551 and VU0405398 showed no discernible bias in HEK293A cells, but had distinct bias fingerprints in cortical neurons. VU0405398 and DPFE were biased toward IP₁ over iCa²⁺ (12 and 26-fold respectively) coupled with biphasic pERK1/2 responses that could not be analyzed using the operational model. VU0409551 was the only ligand to show significant bias (24-fold) between IP₁ and pERK1/2 in cortical neurons.

Open in a separate window

Figure 4

mGlu₅ allosteric ligands are biased agonists relative to DHPG in HEK293A-mGlu₅-low and cortical neurons

The transduction coefficient (log(τ/K_A)) was derived by applying equation 3 to agonist concentration-response curves in HEK293A-mGlu₅-low (A, curves in Figure 1) and cortical neurons (B, curves in Figure 2) and normalized to DHPG (Δlog(τ/K_A)). To calculate the degree of bias evident for different ligands between different pathways, Δlog(τ/K_A) values were subtracted from one another ΔΔ ( log(τ/K_A)) to determine Log bias factors. Data for Δlog(τ/K_A) represent the mean ± SEM, whereas Log bias estimates are mean only. * denotes significantly different comparisons, p<0.05, one-way ANOVA with Tukey’s post-test.

3.6. CDPPB, DPFE, VU0405398 are biased allosteric modulators of DHPG in cortical neurons

We were unable to perform interaction studies between the modulators with DHPG to quantify affinity and cooperativity in HEK293A-mGlu₅-low cells, because of the robust agonism for IP₁ accumulation and ERK1/2 phosphorylation displayed by each allosteric ligand in their own right. However, in cortical neurons, a number of allosteric ligands had minimal or partial agonist effects, such that a sufficient response window remained to allow for allosteric interaction experiments with DHPG in this physiologically relevant system. Thus, the resulting DHPG concentration-response curves for IP₁ and iCa²⁺ in the absence and presence of increasing concentrations of VU0360172, VU29, CDPPB, DPFE, VU0409551 or VU0405398 were constructed and analyzed using an operational model of allosterism (Fig. 5, Supplementary Fig. 6). Affinity estimates were similar for each allosteric ligand across the two pathways, with the exception of VU0405398, which had higher affinity in the IP₁ assay (Table 1). For all six allosteric modulators, the magnitude of logβ was lower for IP₁ versus iCa²⁺ (Fig. 5D); this reached significance for DPFE (3.8-fold), VU0360172 (3.6-fold), VU0409551 (6.5-fold) and VU0405398 (2.8-fold). Given that VU0360172, VU0405398 and DPFE were also partial agonists for ERK1/2 phosphorylation, we assessed modulation of the DHPG concentration-response curve by these three modulators. In cortical neurons, VU0360172 was a PAM-agonist for DHPG-stimulated ERK1/2 phosphorylation with similar affinity and cooperativity estimates (pK_B: 6.49±0.20 and logβ: 0.69±0.18) to those determined in iCa²⁺ assays (Fig. 5D). In contrast, DPFE and VU0405398 showed a unique bell-shaped modulatory response in ERK1/2 phosphorylation. While 1 μM of each ligand potentiated DHPG responses, 10 μM of each ligand resulted in decreased DHPG maximal responses for ERK1/2 phosphorylation (Fig. 5).

Open in a separate window

Figure 5

mGlu₅ allosteric ligands are biased modulators of DHPG responses in native cortical neurons

DHPG concentration response curves for iCa²⁺ mobilization (A), IP₁ accumulation (B) and ERK1/2 phosphorylation (C) in the absence and presence of VU0360172 and DPFE. Both VU0360172 and DPFE produced leftward shifts in DHPG potency for iCa²⁺ and IP₁ assays. Data sets were globally fitted to an operational model of allosterism to estimate affinity and cooperativity. Curves represent the best fit of the data. In ERK1/2 phosphorylation assays, DPFE produced a leftward shift with 1 μM however reduced the maximal response of DHPG at 10 μM. D) Cooperativity estimates (Logβ) of allosteric ligands were plotted to enable comparison of modulation across pathways. For DPFE and VU0405398, which showed mixed modulatory activity, Log β values were plotted as zero to highlight the implicit bias; the absolute numerical value in this instance has no meaning. Data are mean ± SEM of n=3–9 experiments performed in duplicate. Error bars not shown lie within the dimensions of the symbol.

4. Discussion

Allosteric modulators of mGlu₅ offer considerable therapeutic potential for the treatment of numerous psychiatric and neurological disorders. In particular, mGlu₅ PAMs have demonstrated efficacy in preclinical models of psychosis, cognition-enhancement and NMDA receptor hypofunction (Ayala et al., 2009; Gregory et al., 2013a; Horio et al., 2013; Rook et al., 2015; Uslaner et al., 2009). However, a number of recent studies have reported adverse effects with select mGlu₅ allosteric modulators (Rook et al.,, 2013; Parmentier-Batteur et al.,, 2014). Therefore in order to realize the therapeutic potential of mGlu₅ allosteric modulators, there is a need to better understand the full scope of allosteric drug action. We tested the hypothesis that structurally diverse PAMs had unappreciated biased pharmacology. We found that mGlu₅ allosteric ligands had diverse biased signaling fingerprints in both recombinant and native systems. Furthermore, this bias extended to modulation of orthosteric agonist activity, where the magnitude of cooperativity between allosteric and orthosteric ligands differed depending upon the measure of receptor activation. The prevalence of biased agonism and modulation for diverse mGlu₅ PAM chemotypes highlights the paucity in our understanding of the functional consequences of mGlu₅ allosteric modulation, and raises new avenues for therapeutic discovery. Armed with a better appreciation of the full scope of mGlu₅ allosteric modulator action, it may be possible to link activation (and/or enhancement) of specific pathways to different physiological outcomes to promote therapeutic effects and avoid adverse effects.

In particular, many of the studied mGlu₅ allosteric ligands, which previously were largely classified based on modulation of glutamate-stimulated iCa²⁺ mobilization, showed significant differences in coupling efficiency to iCa²⁺ mobilization, IP₁ accumulation and ERK1/2 phosphorylation. Most strikingly, each allosteric ligand tested had greater efficacy for IP₁ accumulation relative to iCa²⁺ mobilization. Therefore the ligands tested cannot be considered as pure mGlu₅ PAMs, and do not posses the safety advantages of pure allosteric modulators, i.e., the advantage of spatiotemporal fine-tuning of receptor responses and saturable effects. Classically, iCa²⁺ mobilization is considered to be downstream of inositol phosphate hydrolysis, however, our data show that IP₁ accumulation does not necessarily result in iCa²⁺ mobilization via mGlu₅. It is possible that these differences between IP₁ and iCa²⁺ are related to the different kinetics of response coupling measurements for these two assays. A lack of equilibrium being achieved by allosteric ligands relative to orthosteric agonists in the transient iCa²⁺ assay may also be a contributing factor. Indeed, underlying temporal differences between assays and ligands influences bias at the dopamine D2 receptor, such that kinetic context may also play a crucial role in understanding biased agonism (Klein Herenbrink et al., 2016). However, it is widely accepted that the mGlu₅ receptor is coupled to various calcium channels (Gao et al., 2013; Kammermeier et al., 2000; Lu et al., 1999; McCool et al., 1998; Ribeiro et al., 2010a; Tu et al., 1999), thereby resulting in Ca²⁺ mobilization from both intracellular and extracellular stores. In order to differentiate the sources of Ca²⁺, we compared iCa²⁺ mobilization in the presence and absence of 1.2 mM Ca²⁺ in HEK293A-mGlu₅-low cells. Interestingly, while glutamate and DHPG had reduced maximal responses in the absence of extracellular Ca²⁺, all studied mGlu₅ allosteric ligands were unaffected. By applying a rigorous analytical approach to quantify bias, we found that VU0424465 showed significant bias between IP₁ and extracellular Ca²⁺-free iCa²⁺ assays (20-fold) and between IP₁ and iCa²⁺ assays (110-fold). These data suggest that allosteric agonists couple to a different complement of effectors than orthosteric agonists to influence intracellular Ca²⁺ levels.

Our data clearly demonstrate that allosteric modulator activity in an mGlu₅-iCa²⁺ mobilization assay is not necessarily predictive of activity for other mGlu₅-mediated effects. Our data contribute to a growing body of evidence for biased agonism at metabotropic glutamate receptors including for mGlu₁ (Emery et al., 2012; Hathaway et al., 2015) and group III mGlu receptors (Jalan-Sakrikar et al., 2014). Further, we observed probe-dependence for potentiation of mGlu₅-iCa²⁺ mobilization in HEK293A cells, where the magnitude of cooperativity for select allosteric ligands was dependent upon the orthosteric agonist present. Probe dependence has not previously been observed for mGlu₅, however, is known to be prevalent for other mGlu subtypes (Jalan-Sakrikar et al, 2014). The phenomenon of probe dependence is a crucial consideration, particularly when translating findings from in vitro assays to native preparations, such as slice electrophysiology, where surrogate agonists are often used. Interestingly, while VU29 had higher positive cooperativity with DHPG compared to glutamate, CDPPB potentiated both agonists to a similar extent. Similarly, for the three picolinamide acetylenes studied, VU0424465 and VU0405398 had differential cooperativities with DHPG and glutamate whereas VU0403602 did not. These data demonstrate that the probe dependence observed is not linked to a particular chemotype, or consistent between ligands from the same scaffold.

While probe dependence is an important consideration when moving from recombinant to physiologically relevant systems, the change in cell background in itself may be a confounding factor in understanding drug pharmacology. Indeed, in this study, we observed clear differences in the temporal profile for mGlu₅-ERK1/2 phosphorylation between HEK293A-mGlu₅-low and cortical neurons. While ERK1/2 phosphorylation was transient in HEK293A-mGlu₅-low cells, in cortical neurons pERK1/2 levels remained elevated over 30min. These data suggest that distinct intracellular effectors are mediating pERK1/2 in response to mGlu₅ activation in cortical neurons versus HEK293A cells. Moreover, sustained ligand-induced responses suggest that the consequences of receptor activation will continue after a ligand has vacated the receptor. It is also possible that cellular context is driving differential compartmentalization and/or internalization of mGlu₅ in these two cell types. Of note, 50–80% of mGlu₅ receptors are thought to be located on intracellular membranes (Hubert et al., 2001; Kumar et al., 2008), and subcellular localisation of mGlu₅ can dictate the overall cellular response (Jong et al., 2009; Kumar et al., 2012). It remains to be determined whether or not the mGlu₅ allosteric ligands tested cross the plasma membrane, induce receptor internalization or are internalized with the receptor. DHPG is not actively transported across the plasma membrane, only activating receptors at the cell surface (Jong et al., 2009), this suggests that the sustained pERK1/2 response in cortical neurons originates from cell surface receptors, however, it is possible that the continued signaling may be driven by internalised cell surface receptors within endosomes.

Comparison of the biased agonism signaling fingerprints for all eight allosteric ligands revealed that VU0424465 had the most divergent profile in both HEK cells and neuronal cultures. Our earlier study linked VU0424465 intrinsic efficacy (for mGlu₅-iCa²⁺ in HEK cells) to seizure activity and behavioral convulsions (Rook et al., 2013). However, comparison of the bias signaling fingerprint of VU0424465 with that of DPFE, VU0409551 and CDPPB, three ligands that have efficacy in models of cognition and psychosis but no reported adverse effects (Gregory et al., 2013a; Uslaner et al., 2009; Balu et al., 2016; Rook et al., 2015), suggests that the pronounced bias (10–100 fold) of VU0424465 toward IP₁ and ERK1/2 relative to iCa²⁺ in HEK293A-mGlu₅-low cells may be a better prediction of adverse effect liability. It is worth noting that all experiments in cortical neurons were performed in the presence of CPCCOEt (an mGlu₁ negative allosteric modulator) to eliminate DHPG activation of mGlu₁. However, mGlu₁ and mGlu₅ are known to heteromerize (Sevastyanova and Kammermeier, 2014). Heteromerization of metabotropic glutamate receptors (mGlu₂ and mGlu₄) can influence allosteric modulator pharmacology (Yin et al., 2014). In the future it would be of interest to determine if mGlu₁/mGlu₅ heteromers influence the biased agonism observed for mGlu₅ selective PAM-agonists. Understanding the bias profile of different ligands, and linking this to known preclinical profiles, offers the opportunity to design therapeutics that avoid signaling pathways associated with adverse effects.

We found that allosteric ligand bias profiles differed between HEK293A-mGlu₅-low cells and cortical neuronal cultures. Of note, VU0409551 was the only ligand to show significant bias away from pERK1/2 (relative to IP₁) and a lack of agonist efficacy for iCa²⁺ mobilization in cortical neurons. These data are consistent with the previous report that VU0409551 acting at mGlu₅ does not potentiate NMDA channel currents (Rook et al., 2015) or stimulate pERK1/2 levels in the hippocampus or prefrontal cortex when dosed chronically (Balu et al., 2016). Interestingly, DPFE and VU0405398 had distinctly different concentration-response relationships for ERK1/2 phosphorylation in cortical neurons, eliciting a bi-phasic response. While VU0405398 has not yet been tested in vivo, DPFE has a unique preclinical efficacy profile. Considerably lower doses of DPFE (0.56mg/kg) are required for cognition enhancement, whereas >10mg/kg is needed for efficacy in reversing amphetamine-induced hyperlocomotion (Gregory et al., 2013). In contrast, cognition studies with CDPPB (Horio et al., 2013; Stefani and Moghaddam, 2010) have demonstrated efficacy at doses similar those required for efficacy in reversing amphetamine-induced hyperlocomotion (Kinney et al., 2005). It is tempting to speculate that the distinct coupling profiles for DPFE and VU0409551 with respect to pERK1/2 may be predictive of in vivo efficacy that is biased toward cognition enhancement.

In addition to biased agonism, biased modulation was also demonstrated for different mGlu₅ allosteric ligands in cortical neurons. Biased modulation may be more favorable for CNS drugs, rather than direct activation of receptors, as it allows for greater control over basal glutamatergic tone. Multiple studies have demonstrated that different mGlu₅ effectors and second messengers can be perturbed in different CNS disorders. For example, disrupted mTOR and Akt signaling is associated with major depressive disorder and suicide (Dwivedi et al., 2010; Jernigan et al., 2011) whereas increases in basal pERK1/2, Akt and altered IP₃/iCa²⁺ signaling are observed in a mouse model of Huntington’s disease (Bezprozvanny and Hayden, 2004; Ribeiro et al., 2010b; Tang et al., 2005). Therefore, biased modulation may provide the means to tailor enhancement or inhibition of mGlu₅ activity to effectors associated with different disease states.

In summary, we have determined the signaling fingerprint of several mGlu₅ allosteric ligands at iCa²⁺ mobilization, IP₁ accumulation and ERK1/2 phosphorylation in both recombinant and native systems. Probe dependence, biased agonism and biased modulation were operative at this receptor. Importantly, the most divergent bias profiles were associated with ligands previously shown to possess distinct in vivo efficacy profiles. Drug discovery for neuroscience-related disorders continues to suffer from high attrition rates, in large part due to an inability to translate promising preclinical pharmacological profiles to clinically effective therapeutics. Our findings suggest that unappreciated biased agonism and/or modulation may contribute to these failures and highlight the need to examine multiple measures of receptor activity to best classify the pharmacology of ligands. In the future, these biased signaling fingerprints may provide a framework that can be used to rationally develop mGlu₅ allosteric ligands with optimal in vitro profiles that translate to desirable activity in preclinical models and avoid on-target adverse effects.

Supplementary Material

Acknowledgments

Abbreviations

Footnotes

References