Kinetic and system bias as drivers of metabotropic glutamate receptor 5 allosteric modulator pharmacology

Abstract

Allosteric modulators of the metabotropic glutamate receptor subtype 5 (mGlu₅) have been proposed as potential therapies for various CNS disorders. These ligands bind to sites distinct from the orthosteric (or endogenous) ligand, often with improved subtype selectivity and spatio-temporal control over receptor responses. We recently revealed that mGlu₅ allosteric agonists and positive allosteric modulators exhibit biased agonism and/or modulation. To establish whether negative allosteric modulators (NAMs) engender similar bias, we rigorously characterized the pharmacology of eight diverse mGlu₅ NAMs. Radioligand inhibition binding studies revealed novel modes of interaction with mGlu₅ for select NAMs, with biphasic or incomplete inhibition of the radiolabeled NAM, [³H]methoxy-PEPy. We assessed mGlu₅-mediated intracellular Ca²⁺ (iCa²⁺) mobilization and inositol phosphate (IP₁) accumulation in HEK293A cells stably expressing low levels of mGlu₅ (HEK293A-rat mGlu₅-low) and mouse embryonic cortical neurons. The apparent affinity of acetylenic NAMs, MPEP, MTEP and dipraglurant, was dependent on the signaling pathway measured, agonist used, and cell type (HEK293A-rat mGlu₅-low versus mouse cortical neurons). In contrast, the acetylenic partial NAM, M-5MPEP, and structurally distinct NAMs (VU0366248, VU0366058, fenobam), had similar affinity estimates irrespective of the assay or cellular background. Biased modulation was evident for VU0366248 in mouse cortical neurons where it was a NAM for DHPG-mediated iCa²⁺ mobilization, but neutral with DHPG in IP₁ accumulation assays. Overall, this study highlights the inherent complexity in mGlu₅ NAM pharmacology that we hypothesize may influence interpretation when translating into preclinical models and beyond in the design and development of novel therapeutics for neuropsychiatric and neurological disorders.

1. Introduction

The metabotropic glutamate receptor subtype 5 (mGlu₅) is a G protein-coupled receptor (GPCR) widely expressed throughout the brain and has been proposed as a therapeutic target in various central nervous system (CNS) disorders, ranging from anxiety and depression to Parkinson’s disease and autism (Gregory et al., 2013). mGlu₅ is a well-established G_q-coupled receptor, with activation leading to production of inositol-1,2,3-trisphosphate (IP₃) and mobilization of intracellular calcium (iCa²⁺) (Niswender and Conn, 2010). mGlu₅ is one of eight mGlu subtypes, subdivided into group I (mGlu_1/5), group II (mGlu_2/3) and group III (mGlu_4,6–8), which share a highly conserved glutamate binding site. Recent approaches in targeting mGlu₅ have therefore focused on ligands that bind to less conserved, topographically distinct allosteric binding sites (Sengmany and Gregory, 2016). Allosteric ligands may modulate the activity of orthosteric ligands by influencing the affinity and/or efficacy (a property termed cooperativity), to either enhance (positive allosteric modulator; PAM), or diminish (negative allosteric modulator; NAM) endogenous receptor responses (Changeux and Christopoulos, 2016). Some PAMs also activate the receptor in the absence of orthosteric ligand and are categorized as PAM-agonists, while neutral allosteric ligands (NALs) bind to allosteric sites without influencing orthosteric ligand activity (Changeux and Christopoulos, 2016). In the absence of endogenous agonist, “pure” PAMs and NAMs offer the additional advantage of spatial and temporal fine-tuning of receptor responses – a desirable clinical outcome within the delicate CNS network.

Fine-tuning neurotransmitter receptor activity can also be achieved through biased agonism and modulation (Kenakin and Christopoulos, 2013), whereby individual ligands may differentially activate/modulate different receptor responses to the relative exclusion of others and, as such, have a unique “signaling fingerprint”. Biased agonism and modulation is operative across a wide range of GPCRs, from opioid and endocannabinoid systems to adrenergic and adenosine receptors, to name a few (Baltos et al., 2017; da Silva Junior et al., 2017; Khajehali et al., 2015; Priestley et al., 2017; Violin et al., 2014). Thus, determining ligand signaling fingerprints for different effectors offers the invaluable opportunity to design ligands that bias receptor signaling towards certain pathways over others. Such an approach may result in ligands that have improved efficacy and lower adverse effect liability.

While the notion of bias is gaining traction, the continued use of high-throughput single-assay drug screening approaches is not capable of capturing the full scope of such ligand activity. There remains a need to decipher which pathways, or bias profiles, are linked to preclinical or clinical efficacy, to stimulate a change in approach. Importantly, it is increasingly evident that mGlu₅ is pleiotropically coupled to multiple G proteins and signaling partners (Francesconi and Duvoisin, 1998; Joly et al., 1995; Mao et al., 2005; Peavy et al., 2001; Rush et al., 2002; Thandi et al., 2002). For allosteric modulators, bias may be evident as pathway-dependent changes in three distinct parameters that dictate allosteric modulator activity; namely affinity, cooperativity and intrinsic efficacy. Indeed, we have clearly shown biased agonism and modulation to be operative amongst several chemotypes of mGlu₅ ligands broadly classified as PAMs or PAM-agonists (Sengmany et al., 2017). Importantly, we revealed how previous classification of mGlu₅ allosteric ligands based solely on their activity in iCa²⁺ mobilization assays failed to recognize the robust agonism characteristic of most mGlu₅ allosteric ligands in IP₁ accumulation and ERK1/2 phosphorylation assays (Sengmany et al., 2017). Thus, simply categorizing an mGlu₅ allosteric ligand as a PAM or NAM based on a single functional assay precludes recognition of the rich complexity each ligand may offer.

While mGlu₅ NAMs have been broadly classified as such based on their activity in iCa²⁺ mobilization assays, an insufficient understanding of mGlu₅ modulator pharmacology is likely to have contributed to recent clinical failures. For instance, fenobam showed promise some 30 years ago in the treatment of generalized anxiety disorder (Pecknold et al., 1982), but it also impairs learning at therapeutic doses in preclinical models (Jacob et al., 2009). Cognition impairment, as well as abuse and psychoactive potential, remain common adverse effects amongst several mGlu₅ NAMs including prototypical compounds such as MPEP and MTEP (Abou Farha et al., 2014; Dekundy et al., 2011; Hughes et al., 2013; Swedberg et al., 2014; Swedberg and Raboisson, 2014). It has been proposed that “partial NAMs”, i.e. NAMs with limited negative cooperativity, may offer the advantage of reduced adverse effect liability due to incomplete blockade of mGlu₅ responses (Nickols et al., 2016). However, there remains a need to better quantify and assess the influence of chemically and pharmacologically diverse NAMs on mGlu₅ activity to truly appreciate the underlying mechanisms of action that contribute to therapeutically beneficial versus adverse effects.

Here we aimed to determine the signaling profiles of select mGlu₅ NAMs through rigorous assessment of interactions between mGlu₅ orthosteric and negative allosteric ligands in iCa²⁺ mobilization and IP₁ accumulation assays in both recombinant and neuronal systems. We show that some mGlu₅ NAMs exhibit differential apparent affinities depending on cell background, pre-equilibration time, orthosteric agonist used, and, in mouse cortical neurons, co-application of CPCCOEt. Cooperativity of select NAMs was also influenced by co-application of CPCCOEt, and biased cooperativity was evident for VU0366248. In all, this study highlights the importance of robust evaluation of allosteric modulatory activity to appreciate the inherent complexity of ligand pharmacology. This study also highlights the potential pitfalls of applying standard high-throughput assays, with previously unappreciated activity at other signaling pathways potentially contributing to the efficacy and safety profiles of mGlu₅ allosteric modulators in preclinical models.

2. Methods

2.1. Materials

Dulbecco’s modified Eagle’s medium (DMEM), Neurobasal medium, Fluo-4-AM and antibiotics were purchased from Invitrogen (Carlsbad, CA). Fetal bovine serum (FBS) was sourced from Thermo Electron Corporation (Melbourne, Australia). IP-ONE HTRF® assay kit was purchased from Cisbio, Genesearch (Arundel, Australia). Select mGlu₅ ligands: 2-(1,3-benzoxazol-2-ylamino)-4-(4-fluorophenyl)pyrimidine-5-carbonitrile (VU0366058), 2-[2-(3-methoxyphenyl)ethynyl]-5-methylpyridine (M-5MPEP), 3-Fluoro-N-(4-methyl-2-thiazolyl)-5-(5-pyrimidinyloxy)benzamide (VU0409106), and N-(3-chloro-2-fluorophenyl)-3-cyano-5-fluorobenzamide (VU0366248) were synthesized as previously described (Felts et al., 2013; Mueller et al., 2012; Rodriguez et al., 2005; Sharma et al., 2008). (S)-3,5-dihydroxyphenylglycine (DHPG), 3-((2-methyl-1,3-thiazol-4-yl)ethynyl)pyridine hydrochloride (MTEP), and 7-(hydroxyimino)cyclopropa[b]chromen-1a-carboxylate ethyl ester (CPCCOEt) were purchased from Tocris Bioscience (Melbourne, Australia) and dipraglurant (ADX 48621) from ApexBio (Houston, TX). [³H]methoxy-PEPy was custom synthesized by Quotient Bioresearch (Rushden, Northamptonshire, UK) using the previously reported synthetic route (Cosford et al., 2003). Unless otherwise stated, all other reagents were purchased from Sigma-Aldrich (St. Louis, MO) and were of analytical grade.

2.2. Recombinant cell culture

HEK293A cells stably transfected with wild-type rat mGlu₅ at low levels (HEK293A-rat mGlu₅-low) comparable to those observed in primary cortical astrocytes (Noetzel et al., 2013) were maintained at 37°C and 5% CO₂ in DMEM supplemented with 5% FBS, 16 mM HEPES and 500 µg/mL G418. A recombinant cell line with low mGlu₅ expression was used as over-expression can result in pharmacological profiles that are not recapitulated in native cells, e.g. pure PAMs behaving as PAM-agonists (Noetzel et al., 2013). One day prior to experimentation, cells were plated onto poly-D-lysine coated, clear-bottom 96 well plates in assay buffer (glutamine-free DMEM supplemented with 5% dialyzed FBS, 16 mM HEPES and 500 µg/mL G418) at 40,000 cells/well.

2.3. Primary cell culture

All animal experiments were approved by the Monash Institute of Pharmaceutical Sciences Animal Ethics Committee (Protocol no. MIPS.2014.37). 8-week old female Asmu:Swiss wild-type mice were provided by the Monash Animal Research Platform (Clayton, VIC, Australia). Pregnant female mice were humanely sacrificed and E16 embryos collected for primary cell culture. Cortices were dissected from E16 Asmu:Swiss wild type mice and mechanically dissociated in Hank’s Balanced Salt Solution (HBSS: KCl 5.3 mM, KH₂PO₄ 0.44 mM, NaHCO₃ 4.17 mM, NaCl 137.93 mM, Na₂HPO₄ 0.34 mM, D-glucose 5.56 mM). Neurons were plated on poly-D-lysine, FBS-coated clear-bottom 96 well plates in Neurobasal media supplemented with 2 mM L-glutamine, 1 x B-27®, 50 U/mL penicillin, 50 U/mL streptomycin, 1.25 µg/mL Fungizone® antimycotic, at a density of 100,000 cells/well. Neurons were maintained at 37°C and 5% CO₂ for 5–7 days prior to experimentation.

2.4. iCa²⁺ mobilization

iCa²⁺ mobilization was measured as previously described (Gregory et al., 2012). Briefly, changes in fluorescence of the Ca²⁺ indicator dye, Fluo-4-AM, were measured using a Flexstation I or III, with mGlu₅ allosteric ligands added either 1 min or 30 min prior to orthosteric agonist. Experiments were conducted at room temperature (RT) for HEK293A-rat mGlu₅-low cells, and at 37°C for mouse cortical neurons. A 5-point smoothing function was applied to raw fluorescence traces, with peak fluorescence normalized to the maximal responses of either glutamate (HEK293A-rat mGlu₅-low) or the group I mGlu agonist DHPG (mouse cortical neurons).

2.5. Inositol monophosphate (IP₁) accumulation assay

Recombinant cells or primary mouse cortical neurons were washed with phosphate buffered saline (PBS; 1.1 mM KH₂PO₄, 155 mM NaCl, 3 mM Na₂HPO₄, pH 7.4) and incubated with stimulation buffer (HBSS, with 20 mM HEPES, 30 mM LiCl₂, 1.2 mM CaCl₂, pH 7.4) supplemented with 1–10U/mL glutamic pyruvic transaminase and 6 mM sodium pyruvate for 1 h at 37°C and 5% CO₂, followed by compound addition. After 1 h ligand incubation, cells were aspirated and lysed with lysis buffer (HTRF® IP-one assay kit) and IP₁ levels detected as per kit instructions. Fluorescence was measured using the Envision plate reader (PerkinElmer), and expressed as fold over basal.

2.6. Whole cell radioligand binding

[³H]methoxy-PEPy whole-cell inhibition binding assays on mouse embryonic cortical neurons were performed in 24-well plates. For binding studies in HEK293A-rat mGlu₅-low, cells were plated onto white-walled, clear bottom poly-D-lysine coated 96-well isoplates at a density of 40,000 cells/well in assay buffer (as described in section 2.2). Inhibition of ~2 nM [³H]methoxy-PEPy binding by increasing concentrations of the various allosteric ligands was assessed at RT for 1 h in HEPES-based binding buffer (145 mM NaCl, 10 mM D-glucose, 5 mM KCl, 1 mM MgSO₄, 10 mM HEPES, 1.3 mM CaCl₂, 15 mM NaHCO₃, pH 7.4). The concentration of DMSO (0.3%) was kept constant throughout. Non-specific binding was determined using 10 µM MPEP. Assays were terminated by washing three times with ice-cold 0.9% NaCl. For mouse cortical neurons, cells were lysed with 250 µl/well of 0.2 M NaOH, lysates transferred to scintillation vials, 4 mL UltimaGold scintillation cocktail added and incubated for >2 h. For HEK293A-rat mGlu₅-low cells, Microscint20 (40 µl/well) was added directly to isoplates, plates sealed and incubated for >2 h. Bound radioactivity was measured using either a MicroBeta2 plate counter or Tri-Carb 2900TR liquid scintillation counter (PerkinElmer, Waltham, MA).

2.7. Data analysis

Inhibition of [³H]methoxy-PEPy binding were fitted to either a one-site or two-site inhibition binding model as previously described (Gregory et al., 2012; Lazareno and Birdsall, 1995) and estimates of dissociation constants (K_B) were derived using the Cheng-Prusoff equation (Cheng and Prusoff, 1973). For ligands that did not fully displace radioligand, inhibition binding curves were fitted with a modified allosteric ternary complex model to estimate K_B (Lazareno and Birdsall, 1993):

$$\frac{Y}{Y_{max}}=\frac{[D]}{[D]+\frac{K_D(1+\frac{[B]}{K_B})}{(1+\frac{\propto [B]}{K_B})}}$$ (1)

where Y/Y_max is the fraction specific binding, the molar concentration of radioligand is [D] and K_D is the radioligand equilibrium dissociation constant, [B] is the molar concentration of unlabeled allosteric modulator, and α is the affinity cooperativity factor between the radioligand and unlabeled allosteric modulator.

Allosteric modulation of glutamate or DHPG-mediated responses were fitted to the operational model of allosterism (Leach et al., 2007):

$$Effect=\frac{E_m{({\tau }_A[A](K_B+\alpha \beta [B])+{\tau }_B[B]K_A)}^n}{{([A]K_B+K_A{K}_B+K_A[B]+\alpha [A][B])}^n+{({\tau }_A[A](K_B+\alpha \beta [B])+{\tau }_B[B]K_A)}^n}$$ (2)

where [A] is the molar concentrations of orthosteric agonist (glutamate or DHPG). β is a scaling factor that denotes the magnitude of effect an allosteric modulator has on orthosteric agonist efficacy. In this respect, β is not a reciprocal efficacy cooperativity factor (Giraldo TIPS 2015). [B], K_B, and α are as defined above. Here, affinity cooperativity (α) was assumed to be neutral between the orthosteric agonist and allosteric modulators as validated previously (Gregory et al., 2012) and thus constrained to a value of 1. K_A is the equilibrium dissociation constant of the orthosteric agonist and was constrained to reported affinity estimates determined from inhibition binding assay as validated previously (Gregory et al., 2012; Mutel et al., 2000; Schoepp et al., 1994). Introducing this constraint to our analyses was necessary given that DHPG and glutamate behave as full agonists in the systems tested and therefore could not derive K_A estimates directly from the functional data sets. τ_A and τ_B are operational measures of the respective ligand’s intrinsic efficacy, while E_m and n represent the maximal system response and the transducer slope respectively.

Affinity and cooperativity estimates were also derived by globally fitting an orthosteric agonist concentration response curve (equation 3) and an allosteric modulator titration curve in the presence of a single concentration of agonist (equation 4):

$$y=basal+\frac{E_m-basal}{1+{\left[\frac{K_A+[A]}{{\tau }_A[A]}\right]}^n}$$ (3)

$$y=basal+\frac{(E_m-basal){[{\tau }_A[A](K_B+\alpha \beta [B])]}^n}{{[{\tau }_A[A](K_B+\alpha \beta [B])]}^n+{([A]K_B+K_A{K}_B+K_A[B]+\alpha [A][B])}^n}$$ (4)

where K_A, τ_A, E_m and basal values were shared across analyses. For equation 4, [A] was held constant to the molar agonist concentration added, that is, ~EC₈₀ response in the particular assay.

All parameters were estimated as logarithms and expressed as mean ± SEM. Statistical analyses of binding data were performed as indicated using an extra sum-of-squares F test to determine the preferred model (one-site versus two-site binding) for each data set. Statistical analyses of functional assays were performed as indicated by using unpaired Student’s t-test or one-way analysis of variance (ANOVA) with Tukey’s or Sidak’s post hoc test, to compare affinity and cooperativity estimates between signaling assays and cell backgrounds.

3. Results

Eight mGlu₅ allosteric ligands previously reported as NAMs of glutamate-mediated iCa²⁺ mobilization were chosen for this study (Fig. 1). All eight ligands have been proposed to interact with a “common allosteric mGlu₅ site” located in the 7 transmembrane (7TM) spanning domain (Gregory et al., 2012; Porter et al., 2005). MPEP, a disubstituted alkyne, is a prototypical mGlu₅ NAM (Gasparini et al., 1999) from which MTEP, reported to have greater mGlu₅ selectivity (over mGlu₁) and potency, was derived (Cosford et al., 2003b; Iso et al., 2006). M-5MPEP, also derived from MPEP, has limited negative cooperativity (also referred to as partial NAM activity) with glutamate (Nickols and Conn, 2014; Rodriguez et al., 2005). VU0409106, VU0366058 and VU0366248 represent chemotypes distinct from MPEP – namely the aryl ether series (Felts et al., 2013), 5-cyanopyrimides (Mueller et al., 2012) and 3-cyano-5-fluoro-N-arylbenzamides (Felts et al., 2010) respectively. Previous reports indicate VU0409106 and VU0366058 are full NAMs of glutamate-mediated iCa²⁺ mobilization, while VU0366248 is a partial NAM (Felts et al., 2013; Gregory et al., 2012). Fenobam and dipraglurant were included to represent mGlu₅ NAMs that have progressed to clinical trials (Pecknold et al., 1982; Tison et al., 2016).

Figure 1. Inhibition of [³H]methoxy-PEPy binding to HEK293A-rat mGlu₅-low cells and mouse cortical neurons.

Using intact adherent cells, inhibition of [³H]methoxy-PEPy binding was determined in HEK293A-rat mGlu₅-low cells (A & B) and mouse cortical neurons (C & D). Data were fitted to either a one-site or two-site inhibition-binding model as determined by an F-test on each individual experiment. Data are mean + SEM from 3–6 independent experiments performed in duplicate.

3.1. Affinity estimation for diverse mGlu₅ NAMs using radioligand binding studies in native and recombinant cells.

Radioligand inhibition binding studies were conducted on intact adherent HEK293A-rat mGlu₅-low cells and primary mouse cortical neurons (Fig. 1). A recombinant cell line with comparable expression to native cells was selected as over-expression of mGlu₅ can influence allosteric modulator pharmacology (Noetzel et al., 2013). Interestingly, MPEP and MTEP inhibition binding curves were best fitted to a two-site model in both recombinant and native cells. Fenobam inhibition curves were also consistent with two-site binding in recombinant cells, but best fitted to a one-site allosteric ternary complex model in neurons. M-5MPEP, VU0366248, VU0409106 and dipraglurant inhibition curves were consistent with one-site binding in both cell types. VU0366058, VU0366248, VU0409106 and dipraglurant did not completely displace [³H]methoxy-PEPy binding in either cell type and curves were fitted with an allosteric ternary complex model to derive affinity (pK_B) and cooperativity (logα) estimates (Table 1). With the exception of MTEP, allosteric ligands had similar pK_B estimates between HEK293A-rat mGlu₅-low and mouse cortical neurons (Table 1).

Table 1. Binding affinities of mGlu₅ allosteric ligands derived from inhibition of [³H]methoxy-PEPy binding in intact and adherent HEK293A-rat mGlu₅-low cells and mouse cortical neurons.

Data represent mean ± SEM of 3–6 independent experiments performed in duplicate.

	HEK293A-rat mGlu5-low pKB				Mouse cortical neurons pKB
	High	Low	Fraction High	Logα	High	Low	Fraction High	Logα
MPEP	8.26 ±0.23	6.25 ±0.11	0.62 ±0.11	n.a.	8.38 ±0.30	5.92 ±0.20	0.73 ±0.11	n.a.
fenobam	7.15 ±0.28	4.66 ±0.22	0.71 ±0.09	n.a.	7.53 ±0.37a	n.a.	1a	−0.66 ±0.10
VU0409106	7.47 ±0.07	n.a.	1	−1.12 ±0.12	7.48 ±0.15	n.a.	1	−0.78 ±0.13
VU0366248	6.94 ±0.13	n.a.	1	−0.77 ±0.06	6.72 ±0.06	n.a.	1	−0.54 ±0.06
VU0366058	7.26 ±0.10	n.a.	1b	−0.82 ±0.07	6.76 ±0.23	n.a.	1b	−0.71 ±0.07
M-5MPEP	7.13 ±0.07	n.a.	1	n.a.	6.43 ±0.25	n.a.	1	n.a.
dipraglurant	7.31 ±0.11	n.a.	1	−1.24 ±0.17	6.98 ±0.14	n.a.	1	−0.89 ±0.08
MTEP	7.20 ±0.28	5.45 ±0.30	0.83 ±0.04	n.a.	8.16 ±0.14*	6.09 ±0.45	0.74 ±0.04	n.a.

^aOne individual experiment (from n=6) was best fitted to a two-site binding curve (F test p>0.05).

^bTwo individual experiments (from n=5–6) were best fitted to a two-site binding curve.

^*Denotes p<0.05, One-way ANOVA, Sidak’s multiple comparisons test, compared to HEK293A-rat mGlu₅-low

n.a. not applicable due to one-site or competitive binding curve fit

3.2. mGlu₅ allosteric ligands are NAMs of glutamate-mediated iCa²⁺ mobilization in HEK293A-rat mGlu₅-low cells.

In agreement with previous studies (Felts et al., 2010; Felts et al., 2013; Gregory et al., 2012; Mueller et al., 2012; Porter et al., 2005), all eight allosteric ligands were NAMs of glutamate-mediated iCa²⁺ mobilization in HEK293A-rat mGlu₅-low cells after 1 min pre-incubation (Fig. 2). In particular, MPEP, MTEP, fenobam, dipraglurant and VU0409106 were full NAMs, while M-5MPEP and VU0366248 displayed limited negative cooperativity. VU0366058 is fluorescent, which limited the testing of concentrations above 1 µM, however, inhibition of glutamate-mediated iCa²⁺ mobilization was consistent with high negative cooperativity. In order to quantify functional affinity and cooperativity estimates, NAM interactions with glutamate were fitted to an operational model of allosterism (Leach et al., 2007). With the exception of dipraglurant, functional affinity estimates (pK_B values) were in good agreement with binding estimates (for the high affinity site where applicable) determined from inhibition binding assays in HEK293A-rat mGlu₅-low cells (Tables 1 & 2). The affinity of dipraglurant was significantly higher in iCa²⁺ mobilization assays (7 fold) relative to the inhibition binding estimate. All NAMs had similar magnitudes of negative cooperativity with glutamate as previously reported (Table 2, Gregory et al., 2010).

Figure 2. Negative allosteric modulation of glutamate-mediated iCa²⁺ mobilization in HEK293A-rat mGlu₅-low cells.

Concentration response curves for glutamate mediated iCa²⁺ mobilization in the absence or presence of indicated allosteric ligands following 1 min pre-incubation. Data are expressed as mean + SEM of 3–9 experiments performed in duplicate. Error bars not shown lie within the dimensions of the symbols.

Table 2. Comparison of affinity and cooperativity estimates for allosteric modulation of DHPG and glutamate-mediated iCa²⁺ mobilization following 1 min vs 30 min pre-incubation of mGlu₅ allosteric ligands in HEK293A-rat mGlu₅-low cells.

Data represent mean ± SEM of indicated number (n) of independent experiments performed in duplicate.

	Glutamate						DHPG
	1 min			30 min			1 min			30 min
	pKBa	log βb	n	pKB	log β	n	pKB	log β	n	pKB	log β	n
MPEP	8.39±0.13	Full NAMc	7	7.90±0.08	Full NAM	4	9.08±0.14d,f	full NAM	7	8.29 ±0.16e	full NAM	5
fenobam	7.04±0.09	Full NAM	9	7.31±0.08	Full NAM	7	6.91±0.13	full NAM	5	6.98 ±0.14	full NAM	5
VU0409106	7.27±0.12	Full NAM	7	7.38±0.12	Full NAM	6	7.03±0.14	full NAM	5	7.67 ±0.21e	full NAM	3
VU0366248	7.22±0.09	−0.90±0.12	7	6.57±0.18e	−0.58±0.08	7	7.43±0.19	−1.23±0.48	5	6.45 ±0.24e	−0.40 ±0.08	4
VU0366058	6.98±0.11	Full NAM	5	7.11±0.28	Full NAM	5	6.71±0.13	full NAM	5	7.02 ±0.24	full NAM	4
M-5MPEP	7.00±0.07	−0.59±0.03	3	6.72±0.26	−0.52±0.06	3	6.70±0.15	−0.97±0.24	5	6.49 ±0.24	−0.47 ±0.09	3
dipraglurant	8.16±0.06d	Full NAM	6	7.47±0.07e	Full NAM	6	8.06±0.14d	full NAM	6	7.31 ±0.16e	full NAM	4
MTEP	7.83±0.09	Full NAM	3	6.97±0.15e	Full NAM	6	8.37±0.14d	full NAM	6	7.81 ±0.18f	full NAM	5

^apK_B, negative logarithm of the equilibrium dissociation constant determined using an operational model of allosterism

^blog β, logarithm of the efficacy cooperativity scaling factor determined using an operational model of allosterism where α was assumed to be equal to 1

^c“full NAM” denotes complete inhibition of DHPG response, such that β approaches zero.

^dDenotes p<0.05, One-way ANOVA, Tukey’s multiple comparisons test, compared to binding pK_i estimates

^eDenotes p<0.05, One-way ANOVA, Tukey’s multiple comparisons test, compared to pK_B estimate derived from iCa²⁺ mobilization assays (1 min versus 30 min paradigm)

^fDenotes p<0.05, One-way ANOVA, Tukey’s multiple comparisons test, compared to pK_B estimate derived from glutamate iCa²⁺ mobilization assays (equivalent incubation paradigm)

3.3. Ligand-receptor equilibrium influences mGlu₅ NAM apparent affinity in HEK293A-rat mGlu₅-low based on inhibition of glutamate stimulated iCa²⁺ mobilization

Differing ligand incubation times between different assays may result in potential bias within a kinetic context (Klein Herenbrink et al., 2016; Lane et al., 2017), especially for non-equilibrium assays. Non-equilibrium conditions may also contribute to differences in affinity estimates. Therefore, we extended the NAM pre-incubation period to 30 min prior to conducting glutamate-mediated iCa²⁺ mobilization assays in HEK293A-rat mGlu₅-low cells (Fig 3). With an increased pre-incubation time for dipraglurant prior to iCa²⁺ mobilization assays, the pK_B value was in much closer agreement with the binding estimate (Table 2). Extended pre-incubation with MTEP and VU0366248 resulted in lower pK_B estimates (7 and 4-fold respectively) when compared with the 1 min assay. VU0366248 also had 2-fold lower negative cooperativity with glutamate although this did not reach significance (Table 2). Cooperativity with glutamate for all other NAMs was similar between the two temporal conditions.

Figure 3. Negative allosteric modulation of glutamate-mediated iCa²⁺ mobilization in HEK293A-rat mGlu₅-low cells with extended pre-incubation.

Concentration response curves for glutamate mediated iCa²⁺ mobilization in the absence or presence of indicated allosteric ligands following 30 min pre-incubation. Data are expressed as mean + SEM of 3–7 experiments performed in duplicate. Error bars not shown lie within the dimensions of the symbols.

3.4. mGlu₅ NAMs are inverse agonists for IP₁ accumulation in HEK293A-rat mGlu₅-low cells

IP₁ accumulation was assessed as an alternative signaling endpoint, given previous observations of biased agonism/modulation for mGlu₅ PAMs between iCa²⁺ mobilization and IP₁ accumulation (Sengmany et al., 2017). Measurement of IP₁ accumulation provides insight into receptor activity at ligand equilibrium relative to transient, non-equilibrium, iCa²⁺ mobilization responses. In the presence of glutamic pyruvic transaminase (GPT), which eliminates ambient glutamate, all eight mGlu₅ NAMs decreased baseline IP₁ accumulation in a concentration dependent manner in the absence of agonist (Fig. 4). In the absence of GPT, most NAMs exhibited a greater maximal inhibition of baseline IP₁ accumulation, attributable to the presence of ambient glutamate (Supplementary Figure 1). However, in the presence of both 1U/mL and 10U/mL GPT, baseline inhibition is reduced to a similar extent. This indicates that the assay conditions have eliminated ambient glutamate and that these data are consistent with constitutive mGlu₅ activity and inverse agonism. The potencies of MTEP and dipraglurant as inverse agonists were similar to binding affinity estimates, whereas for the remaining six NAMs, potencies were 2–4 fold lower than binding pK_B values (Table 3). Due to appreciable inverse agonism for all eight NAMs for IP₁ accumulation in HEK293A-rat mGlu₅-low cells, it was not feasible to assess interactions between NAMs and glutamate using the operational model of allosterism applied here. Alternative operational models that incorporate constitutive activity are available, however, these models also require the ability to measure ligand responses in systems with different degrees of constitutive activity (Slack and Hall, 2012; Hall 2013).

Figure 4. Inverse agonism of constitutive IP₁ accumulation in HEK293A-rat mGlu₅-low cells but not in primary mouse cortical neuron cultures.

A & B) Concentration response curves for inhibition of constitutive mGlu₅-mediated IP₁ accumulation by indicated allosteric ligands in HEK293A-rat mGlu₅-low cells. Data are expressed as mean + SEM of 5–9 experiments performed in duplicate. C &D) No change in IP₁ accumulation basal levels were observed following 60 min exposure of primary mouse cortical neuron cultures to NAMs. Glutamic pyruvic transaminase (1U/mL) was included to eliminate ambient glutamate. A DHPG control curve from parallel experiments is shown for reference (closed black squares). Data are mean + SEM from 3–4 independent experiments performed in parallel. For all panels, error bars not shown lie within the dimensions of the symbols.

Table 3. Potency and efficacy of mGlu₅ NAMs as inverse agonists of IP₁ accumulation in HEK293A-rat mGlu₅-low cells.

Data represent mean ± SEM of indicated number (n) of independent experiments performed in duplicate.

	pIC50a	Imaxb	n
MPEP	7.86 ±0.27	0.74 ±0.03	8
fenobam	6.83 ±0.26	0.76 ±0.03	9
VU0409106	6.89 ±0.26	0.74 ±0.05	7
VU0366248	6.48 ±0.16	0.81 ±0.05	7
VU0366058	6.98 ±0.17	0.73 ±0.03	9
M-5MPEP	6.53 ±0.17	0.81 ±0.05	7
dipraglurant	7.36 ±0.15	0.73 ±0.04	6
MTEP	7.36 ±0.16	0.70 ±0.05	5

^anegative logarithm of the molar concentration required to give a half maximal inhibitory response

^bmaximal inhibitory response, expressed as fold over basal IP₁ accumulation levels.

3.5. mGlu₅ allosteric ligands are NAMs of DHPG-mediated iCa²⁺ mobilization in HEK293A-rat mGlu₅-low cells and cultured mouse cortical neurons.

We next sought to confirm the pharmacology of all eight ligands in primary mouse embryonic cortical neuron cultures. The complex cell background in neuronal cells limits the use of glutamate as an orthosteric agonist due to the presence of other glutamate receptors and transporters. Thus, we adopted a commonly used approach to measure mGlu signaling in response to the mGlu_1/5 orthosteric agonist, DHPG, co-added with the mGlu₁ NAM, CPCCOEt (to minimize mGlu₁ signaling) (Jong et al., 2009; Kettunen et al., 2002; Luccini et al., 2007; Sengmany et al., 2017; Viwatpinyo and Chongthammakun, 2009). The nature of allosteric interactions are dependent on the two chemical entities present, a phenomenon known as probe dependence, therefore we first assessed affinity and cooperativity profiles of the NAMs with DHPG in HEK293A-rat mGlu₅-low cells. To do so, we analyzed NAM inhibition of an EC₈₀ DHPG iCa²⁺ mobilization response, in parallel with a control DHPG concentration-response curve, using both 1 min and 30 min pre-incubation paradigms (Supplementary Figure 2 and Table 2). NAM affinity and cooperativity estimates derived from DHPG iCa²⁺ mobilization inhibition curves were generally similar to those derived from glutamate inhibition curves, although there were a few exceptions (Table 2). For instance, MPEP and MTEP had higher affinity estimates when DHPG was used as the agonist compared to glutamate, with significant differences observed for MPEP using the 1 min paradigm (3-fold) and for MTEP using the 30 min paradigm (10 fold). Differential apparent affinities for these two NAMs under the same assay conditions but derived from interaction studies with different agonists is suggestive of probe dependence.

In mouse cortical neurons, inclusion of 30 µM CPCCOEt had no effect on DHPG potency or E_max for iCa²⁺ mobilization (Supplementary Figure 3). None of the eight NAMs influenced basal iCa²⁺ mobilization in the absence of agonist in mouse cortical neurons (Supplementary Figure 4). All compounds inhibited the response to an EC₈₀ DHPG concentration in a concentration dependent manner following 1 min pre-incubation (Fig. 5A–B). In line with pharmacological profiles in recombinant cells, maximal concentrations of MPEP, fenobam, VU0409106, dipraglurant or MTEP resulted in complete inhibition of the response to 1µM DHPG (an approximately EC₈₀ concentration for iCa²⁺ mobilization), whereas VU0366248 or M-5MPEP had limited negative cooperativity, as evidenced by incomplete inhibition of DHPG-mediated iCa²⁺ mobilization (Fig 5A–B). VU0366058 also showed incomplete inhibition of DHPG but could not be definitively characterized as a limited or full NAM due to the restricted concentration range. Similar to observations in recombinant cells, quantification of these data with the operational model of allosterism revealed there were anomalies with respect to binding versus functional pK_B estimates for some ligands. For dipraglurant the pK_B estimate derived from iCa²⁺ mobilization assays was significantly greater than the binding value (5-fold, Table 4). Extending the pre-incubation time in mouse cortical neurons had no significant effect on NAM pK_B estimates, although there was a trend for reduced affinity for MTEP and dipraglurant (Fig 5C, D, Table 4). Increasing the pre-incubation time had no effect on negative cooperativity with DHPG for any of the NAMs (Table 5). Interestingly, extending the pre-incubation time resulted in VU0366058 becoming a full NAM at the concentration range tested.

Figure 5. Negative allosteric modulator activity at iCa²⁺ mobilization in embryonic mouse cortical neurons in the presence or absence of CPCCOEt 30 µM.

Concentration response curves for modulation of DHPG EC₈₀ (1 µM)-mediated iCa²⁺ mobilization with 1min (A & B) or 30 min (C & D) pre-incubation with mGlu₅ NAMs were performed in parallel with DHPG concentration-response curves. Modulation of DHPG-stimulated iCa²⁺ mobilization was also assessed in the absence of CPCCOEt with 1 min (E & F) or 30 min (G & H) pre-incubation with mGlu₅ NAMs. Data are expressed as mean + SEM of 4–8 experiments performed in duplicate. Error bars not shown lie within the dimensions of the symbols.

Table 4: Comparison of mGlu₅ NAM functional affinity estimates (pK_B) across different measures of receptor activity in primary mouse cortical neurons in the presence and absence of 30 µM CPCCOEt.

Data represent mean ± SEM of indicated number (n) of independent experiments performed in duplicate.

	With CPCCOEt						no CPCCOEt
	iCa2+ (1min)	n	iCa2+ (30min)	n	IP1 accumulation	n	iCa2+ (1min)	n	iCa2+ (30min)	n	IP1 accumulation	n
MPEP	7.61 ±0.20a	6	7.61 ±0.15	5	7.69 ±0.19	11	7.46 ±0.16a	5	7.59 ±0.17	7	7.42 ±0.22	10
fenobam	6.62 ±0.23	4	7.10 ±0.17	4	6.63±0.28	9	6.85 ±0.18	4	7.20 ±0.17	6	6.51 ±0.24	7
VU0409106	6.63 ±0.20	6	7.35 ±0.15	5	6.32 ±0.26b	10	7.08 ±0.17	4	7.28 ±0.18	6	6.68 ±0.24	8
VU0366248	5.99 ±0.35a	7	6.04 ±0.33	5	n.r.	5	5.91 ±0.62a	5	6.86 ±0.29	5	n.r.	5
VU0366058	6.43 ±0.45	6	6.55 ±0.14	6	6.36 ±0.24	10	7.00 ±0.50	6	6.94 ±0.17	6	6.24 ±0.24	8
M-5MPEP	6.47 ±0.30	6	5.64 ±0.25	8	6.24 ±0.39	10	6.93 ±0.48	8	5.98 ±0.29	5	5.94 ±0.31	11
dipraglurant	8.03 ±0.23b	5	7.39 ±0.15	5	6.94 ±0.20c	9	7.16±0.16a	5	7.28 ±0.17	6	6.88 ±0.12	7
MTEP	8.06 ±0.22	5	7.09 ±0.16	4	6.89 ±0.24c	8	7.11±0.28a	5	6.55 ±0.18a,b	6	6.90 ±0.28	8

^aDenotes p<0.05, One-way ANOVA, Sidak’s multiple comparisons test, compared to pK_B estimate derived from modulation of DHPG mediated iCa²⁺ mobilization in HEK293A-rat mGlu₅-low cells under comparable assay paradigm (see Table 2).

^bDenotes p<0.05, One-way ANOVA, Sidak’s multiple comparisons test, compared to binding estimate

^cDenotes p<0.05, One-way ANOVA, Sidak’s multiple comparisons test, compared with pK_B estimate derived from iCa²⁺ mobilization assays using a 1min paradigm

n.r. no modulatory response was evident.

Table 5: Comparison of mGlu₅ NAM cooperativity (logβ) values with DHPG across different measures of receptor activity in primary mouse cortical neurons in the presence and absence of 30 µM CPCCOEt.

Data represent mean ± SEM of indicated number (n) of independent experiments performed in duplicate.

	With CPCCOEt						no CPCCOEt
	iCa2+ (1min)	n	iCa2+ (30min)	n	IP1 accumulation	n	iCa2+ (1min)	n	iCa2+ (30min)	n	IP1 accumulation	n
MPEP	full NAMa	6	full NAM	5	−0.56 ±0.11	11	full NAM	5	full NAM	7	−0.55 ±0.09	10
fenobam	full NAM	4	full NAM	4	−0.59 ±0.17	9	full NAM	4	full NAM	6	−0.67 ±0.17	7
VU0409106	full NAM	6	full NAM	5	−0.39 ±0.09	10	full NAM	4	full NAM	6	−0.52 ±0.09	8
VU0366248	−0.60 ±0.21	7	−0.75±0.31	5	n.r.	5	−0.36±0.15	5	−0.69 ±0.09	5	n.r.	5
VU0366058	−0.59 ±0.32b	6	full NAM	6	−0.46 ±0.11	10	−0.41 ±0.15	6	full NAM	6	−0.69 ±0.20	8
M-5MPEP	−0.65 ±0.19	6	−0.88 ±0.35	8	−0.24 ±0.06	10	−0.31 ±0.08	8	−0.80 ±0.18	5	−0.33 ±0.08	10
dipraglurant	full NAM	5	full NAM	5	full NAM	9	full NAM	5	full NAM	6	full NAM	7
MTEP	full NAM	5	full NAM	4	−0.52 ±0.13	8	−0.68±0.16	5	full NAM	6	−0.38 ±0.07	8

^a“full NAM” denotes complete inhibition of DHPG response, such that β values approach zero.

^blimited concentration range was tested due to compound fluorescence and therefore cannot definitively conclude cooperativity is limited.

n.r. no modulatory response was evident.

3.6. Inhibition of mGlu₁ affects mGlu₅ NAM affinity and cooperativity with DHPG in mouse cortical neuron iCa²⁺ mobilization assays

Given the observed differences between NAM pK_B estimates determined in neuronal radioligand binding assays versus functional assay estimates derived in the presence of CPCCOEt, we sought to ensure that CPCCOEt did not influence mGlu₅ NAM affinity and/or cooperativity. This was of particular concern for two reasons: (i) we recently showed that CPCCOEt binds to mGlu₅ (with comparable affinity for mGlu₁) albeit it exhibits neutral cooperativity with mGlu₅ agonists (Hellyer et al., 2018), (ii) mGlu₁ and mGlu₅ can heterodimerize to form higher order oligomers (Correa et al., 2017), thus CPCCOEt could influence mGlu₅ across an mGlu₁/mGlu₅ dimer. Hence, we re-evaluated the pharmacology of the eight studied mGlu₅ allosteric ligands in the absence of CPCCOEt in mouse cortical neurons (Fig 5E–H). In general, in the absence of CPCCOEt, iCa²⁺ mobilization assays with either 1 or 30 min NAM pre-incubation yielded NAM pK_B estimates that were in good agreement with binding estimates (Table 4). The single exception was MTEP, which had a 41-fold lower pK_B estimate in the 30 min paradigm in the absence of CPCCOEt (Table 4). There were no significant differences between NAM pK_B estimates in the absence versus presence of CPCCOEt when equivalent paradigms were assessed. Cooperativity of NAMs with DHPG were similar in the absence and presence of CPCCOEt, with the exception of VU0366058 and MTEP which transformed from full to partial NAMs in the 1 min pre-incubation paradigm (Table 5). Thus, inclusion of CPCCOEt in the cortical neuron functional assays influenced mGlu₅ NAM pharmacology in a ligand-dependent manner. It is presently not clear whether this is due to unappreciated CPCCOEt activity at mGlu₅ alone, or mediated across an mGlu₁/mGlu₅ heteromer.

3.7. mGlu₅ allosteric ligands have differing degrees of cooperativity with DHPG when assessed in IP₁ accumulation in cultured mouse cortical neurons.

In contrast to recombinant cells, there was no evidence for inverse agonist activity for all eight mGlu₅ NAMs for IP₁ accumulation in mouse cortical neurons (Figure 4), indicating that the receptor has low or no constitutive activity in these native cells. Inclusion of CPCCOEt, had no significant effect on DHPG potency or E_max for IP₁ accumulation (Supplementary Figure 3) compared to untreated neurons. In the absence and presence of CPCCOEt, with the exception of VU0366248, all of the allosteric modulators were NAMs of DHPG-mediated IP₁ accumulation in mouse cortical neurons (Fig 6). VU0366248 showed negligible modulatory activity of DHPG-mediated IP₁ accumulation (Fig 6B, D), suggesting a loss of negative cooperativity with DHPG. Interestingly, only dipraglurant was able to completely abolish the response to an EC₈₀ DHPG concentration in both conditions, whereas all other NAMs were partial (Fig 6, Table 5). Analysis of interactions with DHPG in the presence of CPCCOEt revealed that dipraglurant and MTEP had significantly lower (~10-fold) affinity estimates in IP₁ accumulation relative to 1 min pre-incubation iCa²⁺ mobilization assays, and VU0409106 pK_B in IP₁ accumulation was 14-fold lower than the binding estimate (Table 4). In the absence of CPCCOEt, affinity estimates from IP₁ accumulation assays were not significantly different to inhibition binding for all ligands (Table 4). There were no significant differences in functional affinity estimates for any of the eight NAMs between the presence and absence of CPCCOEt, based on modulation of DHPG-stimulated IP₁ accumulation (Fig 6C–D; Table 4).

Figure 6. Inhibition of DHPG EC₈₀-mediated IP₁ accumulation by mGlu₅ NAMs in mouse cortical neurons in the presence or absence of CPCCOEt 30 µM.

Concentration response curves for modulation of DHPG EC₈₀ (20 µM)-mediated IP₁ accumulation in the presence (A & B) or absence (C & D) of CPCCOEt. Data are expressed as mean + SEM of 5–10 experiments performed in duplicate. Error bars not shown lie within the dimensions of the symbols.

3.8. Comparison of signaling fingerprints between recombinant and native cells based on affinity and cooperativity estimates.

To better visualize the differences and similarities in mGlu₅ NAM pharmacology in recombinant and native systems, we plotted the functional affinity estimates (based on modulation of DHPG) relative to binding estimates side by side (Fig. 7). Of note, when comparing affinity estimates between the different cell types in equivalent iCa²⁺ mobilization assays with the same orthosteric agonist, MPEP, MTEP, dipraglurant and VU0366248 had lower apparent affinities in mouse cortical neurons compared to recombinant cells although this was assay dependent (Table 4). The switch of VU0366248 to neutral cooperativity in cortical neuron IP₁ accumulation assays precluded estimation of affinity. Acetylenic full NAMs (MPEP, MTEP and dipraglurant) had similar pK_B fingerprints, with affinity estimates influenced by assay conditions and cell type (Fig 7). In contrast, functional affinity estimates of the acetylenic partial NAM M-5MPEP and the structurally distinct NAMs, VU0366058, fenobam and VU0409106, were generally consistent across all measures in both cell systems. Importantly, these data suggest that the differences in functional affinity estimates observed between IP₁ accumulation and iCa²⁺ mobilization assays are ligand dependent and not simply attributable to differences in assay conditions. For the most part, magnitudes of negative cooperativity between NAMs and DHPG were similar in neurons to HEK293A-rat mGlu₅-low cells (Table 5, Table 2).

Figure 7. Comparison of affinity estimates in HEK293A-rat mGlu₅-low cells and mouse cortical neurons.

The difference in affinity estimates from functional assays in HEK293A-mGlu₅ low cells or mouse cortical neurons versus inhibition binding. Black bars: 1 min iCa²⁺ mobilization, red bars: 30 min iCa²⁺ mobilization, blue bars: IP₁ accumulation. Data are mean + SEM from 3–10 experiments. Comparisons were performed using one-way ANOVA with Sidak’s post-test, where significance (*) was considered p<0.05. # denotes not determined due to lack of appreciable modulation.

4. Discussion

Inhibition of mGlu₅ is proposed to offer therapeutic potential for CNS disorders, ranging from anxiety and depression to autism and addiction (Sengmany and Gregory, 2016). However, adverse effect liability in the form of cognitive impairments (Simonyi et al., 2010), abuse and psychotomimetic potential (Abou Farha et al., 2014; Swedberg et al., 2014; Swedberg and Raboisson, 2014), highlights the need to better understand the underlying pharmacology of mGlu₅ inhibitors. A deeper understanding of mGlu₅ NAM pharmacology may provide insights into the differential profiles of these compounds in preclinical animal models and allow for the rational design of ligands that display improved efficacy and safety. Our quantitative pharmacological profiling of eight ligands, previously classified as mGlu₅ NAMs of glutamate in iCa²⁺ mobilization assays, revealed differences in apparent affinities between binding and functional assays (iCa²⁺ mobilization and IP₁ accumulation) for acetylenic full NAMs (MPEP, MTEP and dipraglurant) in recombinant and/or neuronal systems, whereas, non-acetylenic NAMs or the prototypical acetylenic “partial NAM” M-5MPEP were similar across all measures. Biased NAM pharmacology was evident at the level of cooperativity with DHPG; most strikingly in mouse cortical neurons where VU0366248 behaved as a neutral allosteric ligand in IP₁ accumulation assays but was a NAM of DHPG-mediated iCa²⁺ mobilization. Importantly, M-5MPEP and dipraglurant had similar levels of cooperativity with DHPG and glutamate, independent of cell type or assay paradigm, demonstrating an absence of bias at the level of cooperativity. Collectively, our study highlights the importance of rigorous assessment of allosteric ligands to quantify affinity and cooperativity, thereby enriching our appreciation of how aspects of mGlu₅ NAM pharmacology may contribute to efficacy and safety profiles in preclinical models.

Several orthosteric agonists of mGlu receptors are well-established biased agonists (Emery et al., 2012; Hathaway et al., 2015), while positive allosteric modulators of Class C GPCRs also engender bias with respect to agonism and/or cooperativity (Cook et al., 2015; Sengmany et al., 2017; Zhang et al., 2005). At the class C calcium sensing receptor, the NAM NPS2143 demonstrated biased modulation toward iCa²⁺ mobilization relative to phosphorylation of ERK1/2 (Davey et al., 2012; Leach et al., 2016; Leach et al., 2013). However, for many of the mGlu₅ NAMs studied herein, a single in vitro functional assay (iCa²⁺ mobilization) has been used to classify allosteric pharmacology. There are some exceptions, namely MPEP, M-5MPEP, VU0366248 and VU0366058, for which two different functional assays (iCa²⁺ mobilization and ERK1/2 phosphorylation) using glutamate as the orthosteric agonist in recombinant cells have been assessed (Gregory et al., 2012). No significant bias in functional affinity or cooperativity estimates between iCa²⁺ mobilization and ERK1/2 phosphorylation were observed (Gregory et al., 2012). In contrast, herein multiple NAMs showed differential magnitudes of cooperativity with DHPG in native cells depending on the measure (iCa²⁺ mobilization versus IP₁ accumulation). Importantly, these differences were not observed for all NAMs, with M-5MPEP and dipraglurant retaining similar cooperativity with DHPG across all measures and cell types tested in the present study. Thus, our results highlight the need to assess multiple receptor endpoints to probe the full scope of allosteric ligand pharmacology.

Our results are also distinct from previous work in demonstrating the presence of two-site inhibition binding. Inhibition of [³H]methoxy-PEPy binding by MPEP, MTEP and fenobam were best fitted to a two-site inhibition binding model, contrary to the one-site binding previously reported (Gregory et al., 2012; Lea and Faden, 2006; Porter et al., 2005). Furthermore, dipraglurant, VU0366058, VU0409106 and VU0366248 did not fully displace [³H]methoxy-PEPy binding under our assay conditions. There are multiple possible underlying explanations for these observations. An important difference in the current study was the use of intact and adherent cells as opposed to membrane preparations, which have typically been employed (Cosford et al., 2003a; Gregory et al., 2012; Lindemann et al., 2011; Porter et al., 2005; Raboisson et al., 2012; Rodriguez et al., 2010a; Rodriguez et al., 2005). Incomplete radioligand displacement is generally considered evidence for non-competitive binding interactions (Flanagan, 2016; Pagano et al., 2000). However, in light of the two-site binding observed for select ligands, incomplete displacement may be due to very low (pK_B <4.5) dipraglurant, VU0366058, VU0409106 and VU0366248 affinity for this apparent second site. The complex binding isotherms could also be attributable to allosteric ligands stabilizing distinct receptor conformations that are only evident within intact cells. Dynamic cellular processes such as G protein coupling, interactions with transducers or scaffolding partners, receptor dimerization and subcellular localization could contribute to an apparent “second site”. Indeed, each of the eight NAMs tested were also inverse agonists for IP₁ accumulation in recombinant cells, but not cortical neurons, and could conceivably be influenced by receptor-G protein coupling. Moreover, mGlu₅ is well-known to be expressed at the plasma membrane as well as on intracellular membranes (Jong et al., 2014). Differential membrane permeability or access to subcellular compartments may contribute to the complex binding isotherms. In keeping with this idea, all ligands that fully displaced [³H]methoxy-PEPy belong to a similar chemotype and have low molecular weights <250. Irrespective of mechanism, it is clear that mGlu₅ NAMs, both within and across chemotypes, can stabilize different receptor conformational states.

Assessment of ligand pharmacology within physiologically relevant systems aims to most closely predict a drug response within the body. However, complex native cell backgrounds and the need to use surrogate ligands raises additional issues: 1) probe dependence; and 2) system bias, where different cell backgrounds result in different ligand pharmacology. Probe dependence between glutamate and DHPG was evident in cooperativity of select mGlu₅ PAMs (Sengmany et al., 2017). Under certain conditions both MPEP and MTEP showed higher apparent affinity when DHPG was used as the agonist over glutamate, indicative of probe dependence. DHPG is a membrane impermeable agonist, and unlike glutamate is not actively transported into cells (Jong et al., 2005); therefore mGlu₅ modulator probe dependence may arise from differential access of ligands to subcellular compartments. Context-dependent pharmacology, or system bias, has also previously been described for mGlu₇ NAMs (Niswender et al., 2010). Here the influence of the system was evident for select mGlu₅ NAMs, with differential functional affinities observed between HEK293A cells and neurons. However, despite these caveats, distinct NAM fingerprints were evident which may be linked to the known adverse effect and/or preclinical efficacy of these agents. Acetylenic full NAMs exhibit differential affinities in both cell types, whereas different NAM chemotypes and the partial acetylenic NAM do not. MPEP and MTEP have undergone extensive preclinical testing, and have shown anxiolytic and antidepressant efficacy along with adverse effect liability (cognitive and psychotomimetic) (Balschun and Wetzel, 2002; Belozertseva et al., 2007; Campbell et al., 2004; Chojnacka-Wojcik et al., 2001; Kumar et al., 2013; Lea and Faden, 2006; Li et al., 2006; Palucha-Poniewiera et al., 2014; Schulz et al., 2001; Spooren et al., 2000; Swedberg et al., 2014; Tatarczynska et al., 2001). In a rat model of cocaine addiction, the unbiased partial NAM, M-5MPEP, does not induce adverse psychotomimetic effects at doses that show efficacy in attenuating addiction-associated behaviors (Gould et al., 2016). For the structurally diverse NAMs (VU0409106, VU0366058, VU0366248) fewer preclinical studies have been performed, therefore it remains to be seen if the NAM pK_B fingerprints established here can be used to inform future discovery efforts for mGlu₅, which may seek to identify biased NAMs that more selectively target therapeutic signaling over adverse effects. Assessment of the biochemical effects of mGlu₅ NAMs in vivo, such as inhibition of mGlu₅ mediated phosphoinositide hydrolysis or MAP kinase activation, would also provide important information regarding the translatability of these compounds into preclinical animal models and beyond.

Another contributing factor to system bias in the more complex neuronal system is the co-expression of mGlu₁, which is known to influence mGlu₅ activity via heteromerization and/or signaling cross-talk (Bonsi et al., 2005; Lujan et al., 1996; Neyman and Manahan-Vaughan, 2008). In order to minimize the mGlu₁ contribution we initially included CPCCOEt in all experiments (Jong et al., 2009; Sengmany et al., 2017). However, we recently showed that CPCCOEt has similar affinity for mGlu₁ and mGlu₅, and displays a negative allosteric interaction with the ‘MPEP site’ radioligand (Hellyer et al., 2018). Exclusion of CPCCOEt influenced estimation of MTEP, VU0409106 and dipraglurant affinity. Given that the majority of NAMs were unaffected, this suggests that CPCCOEt may allosterically modulate select mGlu₅ NAMs via heteromers, distinct sites, and/or changes in co-localization (Jensen and Brauner-Osborne, 2007; Noetzel et al., 2013; Pandya et al., 2016; Pin and Prézeau, 2007; Rodriguez et al., 2010b; Sevastyanova and Kammermeier, 2014). Collectively, the data highlight that while assessment in recombinant cells may be convenient and have greater reproducibility, it is important to recognize the potential for disconnect between differing cell backgrounds.

Another key consideration is the influence of the kinetic context on biased pharmacology (Klein Herenbrink et al., 2016; Lane et al., 2017), where ligand-receptor and receptor-effector interactions require time to achieve equilibrium. Previously, we showed that the majority of mGlu₅ PAMs had strong biased agonism toward IP₁ accumulation over iCa²⁺ mobilization, which may have been associated with differences in ligand-receptor equilibrium between the two measures (Sengmany, et al., 2017). Experiments performed at equilibrium (radioligand binding and IP₁ accumulation) for multiple, but not all, mGlu₅ NAMs yielded lower affinity estimates relative to the transient non-equilibrium iCa²⁺ mobilization assay. Higher affinity upon shorter preincubation was somewhat unexpected, however, if we appreciate that the impact of kinetics involves multiple arms – ligand binding, transducer coupling and downstream cell signaling processes (Lane et al., 2017), the paradoxical results observed for mGlu₅ NAM affinity may well be explained. While iCa²⁺ mobilization and IP₁ accumulation are endpoints traditionally linked through the G_q activation pathway, it is well established that mGlu₅ couples to ion channels (Kammermeier et al., 2000; Latif-Hernandez et al., 2016; Lu et al., 1999; McCool et al., 1998; Tu et al., 1999), resulting in a rapid influx of extracellular Ca²⁺. Therefore, the iCa²⁺ mobilization measured is a composite of both intracellular and extracellular calcium influx (Sengmany et al., 2017). Due to the rapid opening of ion channels relative to GPCR signaling cascades, it is possible that initial responses may be via extracellular calcium influx through ion channels. Therefore, select mGlu₅ NAMs (MTEP, dipraglurant, MPEP) may have higher affinity for mGlu₅ states that couple to plasma membrane channels, as opposed to the canonical G_q/11-PLC-IP₃-pathway. Thus, this study highlights the importance of recognizing different determinants of observed allosteric modulator pharmacology (allosteric ligand, orthosteric agonist, receptor, system) and how these can be dissected to reveal new insights into negative allosteric modulator activity.

Overall, this study provides a rigorous characterization of mGlu₅ NAMs, highlighting the stabilization of unique receptor conformations in both recombinant and native cells. In particular, ligand-dependent disparities in affinity and cooperativity between iCa²⁺ mobilization and IP₁ accumulation highlight differential modulation of DHPG responses by select mGlu₅ NAMs. Differing kinetic and cellular contexts also underscore the pharmacological fingerprints observed, underlining the necessity to carefully consider signaling pathways and cellular background when interpreting pharmacology. Acetylenic full mGlu₅ NAMs showed the most divergent pharmacological fingerprints. It is tempting to speculate that a biased signaling fingerprint may be linked to their propensity to engender psychotomimetic and cognitive impairments in preclinical animal models, especially because a partial and unbiased NAM from the same scaffold does not demonstrate such adverse effects. Discovery efforts targeting GPCRs for CNS disorders continue to suffer high attrition rates, attributable in part to the difficulties in translating agents with promising in vitro pharmacology to efficacy in preclinical models and extending this to the clinic. Our findings highlight the inherent complexity in quantifying allosteric modulator pharmacology, with previously unappreciated bias at the level of affinity and/or cooperativity potentially contributing to efficacy and safety profiles of mGlu₅ allosteric modulators in preclinical animal models. .

Highlights.

• -mGlu5 negative allosteric modulators can stabilise distinct affinity states
• -mGlu5 NAM affinity is both context and orthosteric probe dependent
• -biased mGlu5 modulation is evident at the level of cooperativity and affinity

Abbreviations

CNS

central nervous system

CPCCOEt

7-(hydroxyimino)cyclopropa[b] chromen-1a-carboxylate ethyl ester

DHPG

(S)-3,5-dihydroxyphenylglycine

dipraglurant

6-fluoro-2-[4-(2-pyridinyl)-3-butyn-1-yl]-Imidazo[1,2-a]pyridine

DMEM

Dulbecco’s modified Eagle’s medium

ERK1/2

extracellular signal-regulated kinases 1 and 2

FBS

fetal bovine serum

fenobam

1-(3-chlorophenyl)-3-[(2e)-1-methyl-4-oxoimidazolidin-2-ylidene]urea

GPCR

G protein-coupled receptor

GPT

glutamic pyruvic transaminase

HBSS

Hank’s Balanced Salt Solution

HEK293A

human embryonic kidney 293

iCa²⁺

intracellular calcium

IP₁

inositol 1-phosphate

M-5MPEP

2-[2-(3-methoxyphenyl)ethynyl]-5-methylpyridine

mGlu₁

metabotropic glutamate receptor subtype 1

mGlu₅

metabotropic glutamate receptor subtype 5

MPEP

2-Methyl-6-(phenylethynyl)pyridine

MTEP

2-methyl-4-(pyridin-3-ylethynyl)thiazole; 3-((2-Methyl-1,3-thiazol-4-yl)ethynyl)pyridine

NAL

neutral allosteric ligand

NAM

negative allosteric modulator

PAM

positive allosteric modulator

VU0366058

2-(1,3-benzoxazol-2-ylamino)-4-(4-fluorophenyl)pyrimidine-5-carbonitrile

VU0366248

N-(3-Chloro-2-fluorophenyl)-3-cyano-5-fluoro-benzamide

VU0409106

3-Fluoro-N-(4-methyl-2-thiazolyl)-5-(5-pyrimidinyloxy)benzamide