INTRODUCTION

The five muscarinic acetylcholine receptors (mAChRs) have been implicated in numerous diseases, including neurodegenerative and psychiatric disorders (Wess et al, 2007). In particular, the M1 and M4 mAChRs have been associated with neurological illnesses, as they are widely expressed throughout the central nervous system (Bymaster et al, 2003; Wess et al, 2003) and are involved in cognitive processes such as attention, learning, and memory (Hasselmo, 2006; Hasselmo and Giocomo, 2006); drugs that selectively target these receptors may thus be useful in the treatment of diseases such as Alzheimer's disease and schizophrenia, in which cognitive processes are disturbed.

The M4 mAChR is of particular interest with regards to its involvement in schizophrenia because it has been implicated in the regulation of dopamine levels in the brain (Gomeza et al, 1999; Tzavara et al, 2004; Zhang et al, 2002b). Indeed, M4 mAChRs and D1 dopamine receptors are coexpressed throughout the striatum (Bernard et al, 1992; Ince et al, 1997), and decreased levels of M4 (and M1) mAChRs have been found in postmortem brain tissues of schizophrenic patients (Crook et al, 2000, 2001; Dean et al, 1996; Deng and Huang, 2005; Scarr et al, 2007), suggesting a tightly controlled balance between cholinergic and dopaminergic neurotransmission that may be disrupted in schizophrenia.

Encouragingly, Alzheimer's patients treated with the mixed M1/M4-preferring mAChR agonist, xanomeline, responded with a reduction in psychotic episodes and improvements in cognitive deficits (Bodick et al, 1997a 1997b). In addition, a recent small-scale study of xanomeline administration to schizophrenic patients revealed significant attenuation of negative and cognitive symptoms (Shekhar et al, 2008). Unfortunately, xanomeline is not sufficiently selective for the M1/M4 mAChRs over other subtypes to avoid peripheral mAChR side effects (Bodick et al, 1997a, 1997b), which are a common obstacle to the clinical development of mAChR-based drugs. Indeed, the development of highly selective mAChR ligands is impeded by the fact that all mAChR subtypes share high sequence homology within the orthosteric (ie the endogenous agonist-binding site of these receptors). Consequently, recent drug discovery efforts have focused on exploiting novel allosteric-binding sites on these, and other, G protein-coupled receptors (GPCRs) in an effort to attain greater selectivity by targeting regions that are more likely to diverge in sequence from other receptor subtypes (Conn et al, 2009; Gregory et al, 2007; May et al, 2007b).

Although allosteric modulators of GPCRs offer great promise as therapeutics, the molecular mechanisms that govern their pharmacodynamic properties and how these translate to therapeutic efficacy remain relatively unexplored. In addition, reports of allosteric GPCR modulators with the potential to also activate the receptor in their own right (allosteric agonists) are increasing (Langmead and Christopoulos, 2006), highlighting another layer of complexity that must be considered when screening, validating, and quantifying the actions of such molecules. In this regard, we recently identified LY2033298 (3-amino-5-chloro-6-methoxy-4-methyl-thieno[2,3-b]pyridine-2-carboxylic acid cyclopropylamide) as a novel, selective, positive allosteric modulator of acetylcholine (ACh) at the M4 mAChR that can express agonism under certain conditions (Chan et al, 2008; Nawaratne et al, 2008). This molecule also displayed in vivo efficacy in preclinical models predictive of antipsychotic drug effects, leading us to hypothesize that selective potentiation and/or activation of M4 mAChRs may be a novel approach to treating schizophrenia (Chan et al, 2008). Despite the exciting potential of these findings, however, a number of key questions remain unresolved, including the actual location of the allosteric site with which LY2033298 interacts on the M4 mAChR; whether agonism of LY2033298 arises solely from interaction with an allosteric site, or whether the molecule uses the latter site to modulate orthosteric ligands while acting as an agonist through the receptor's orthosteric site; whether allosteric modulation differs with the signaling pathway being activated; and to what extent manifestations of allosteric modulation and/or allosteric agonism identified in recombinant systems are retained in native systems. The aim of this study, therefore, was to address these issues and provide a detailed characterization of the molecular mechanisms that govern the novel allosteric properties of LY2033298.

MATERIALS AND METHODS

Materials

NG108-15 cells—American type culture collection (Manassas); FlpIn CHO cells, Gateway cloning vectors/enzymes—Invitrogen (Carlsbad); Dulbecco's Modified Eagle media (DMEM) and fetal bovine serum (FBS)—Gibco (Gaithersburg) and JRH Biosciences (Lenexa), respectively; Bio-Rad protein assay dye—Bio-Rad (Hercules); guanosine 5′-[γ-35S]triphosphate ([35S]GTPγS) (<1000 Ci/mmol) and [3H]choline chloride (80 Ci/mmol)—GE Healthcare (Amersham, UK); [3H]N-methylscopolamine ([3H]NMS) (<70 Ci/mmol), [3H]quinuclinyl benzilate ([3H]QNB) (52Ci/mmol), UltimaGold, Starscint, Irga-Safe Plus, Optiphase supermix, Micoscint-O, and AlphaScreen beads—PerkinElmer (Massachusetts); 96-well GF/C filter plates—Millipore (Watford, UK). SureFire extracellular regulated kinase 1/2 (ERK1/2) and glycogen synthase kinase 3β (GSK-3β) phosphorylation assay kits were a generous gift from Dr Michael Crouch, TGR Biosciences (Adelaide, Australia). LY2033298 was synthesized at Eli Lilly (Indianapolis) and C7/3-phth ((heptane-1,7-bis dimethly-3′-pthalimidopropyl)–ammonium bromide) was synthesized at Monash University using methods described earlier (Nassif-Makki et al, 1999). All other chemicals were from Sigma Chemical Company (St Louis).

Cell Lines

CHO FlpIn M4 cell lines were maintained as described earlier (Nawaratne et al, 2008). HEK293 FlpIn TRex cells stably expressing the M4 mAChR were generated according to manufacturer's instructions (Invitrogen) and maintained in high glucose DMEM containing 200 μg/ml hygromycin, 10 μg/ml blasticidin, 10% tetracycline-free FBS, and 16 mM HEPES. NG108-15 cells were maintained in high glucose DMEM without sodium pyruvate containing 10% FBS, 16 mM HEPES, 5 mM sodium hypoxanthine, 20 μM aminopterin, and 0.8 mM thymidine.

Radioligand-Binding Assays

CHO FlpIn cell membranes were prepared as described earlier (Nawaratne et al, 2008). Radioligand dissociation was determined by equilibrating cell membranes (15 μg/100 μl; 1 h; 37°C) with 2 nM [3H]NMS in binding buffer (20 mM HEPES, 100 mM NaCl, 10 mM MgCl2, pH 7.4) before distribution of 100 μl of the mix into 1 ml (final) buffer containing atropine (10 μM) in the absence or presence of modulator using a reverse-time protocol. The time course of 0.2 nM [3H]NMS association in the absence and presence of LY2033298 was performed in 0.5 ml buffer at 37°C. For equilibrium binding, cell membranes (15 μg) were incubated for 3 h at 37°C with either 0.2 nM [3H]NMS or 0.1 nM [3H]QNB as described earlier (Nawaratne et al, 2008).

[35S]GTPγS Assay

Cell membranes (10 or 35 μg for the wild type (WT) or Y113C/A203G mutant M4 mAChR, respectively) were equilibrated for 1 h at 30°C with varying concentrations of ligands in binding buffer containing 1 μM GDP. [35S]GTPγS (0.1 nM) was added to a final volume of 0.2 ml (WT) or 0.5 ml (Y113C/A203G M4 mAChR) and membranes were incubated for 30 min. Termination of [35S]GTPγS binding was by rapid filtration with a Packard plate harvester onto 96-well GF/C filter plates (WT) or with a Brandel harvester onto GF/B filter paper (Y113C/A203G mutant M4 mAChR) followed by three washes with ice cold 0.9% NaCl. Plates or filter paper were dried and 40 μl Microscint-O was added to each well, or 4 ml UltimaGold added to each filter, before radioactivity was determined by liquid scintillation counting.

ERK1/2 Phosphorylation Assays

Cells were seeded at 40 000 (FlpIn CHO M4), 30 000 (NG108-15), or 100 000 (FlpIn TRex HEK293 M4) cells per well into a transparent 96-well plate and grown overnight. For FlpIn TRex HEK293 M4 and NG108-15 cells, wells were coated with poly-D-lysine (50 μg/ml) before seeding. For FlpIn TRex HEK293 M4 cells, cells were treated overnight with tetracycline (10 or 30 ng/ml) to induce expression of the M4 mAChR. For interaction studies, antagonist or negative modulator were preincubated with cells for 30 min before the addition of agonist for a further 5 min. For interaction studies in which ACh-stimulated ERK1/2 phosphorylation was measured in the presence of LY2033298, to minimize desensitization, LY2033298 was preincubated with cells for only 1 min before the addition of ACh. All other details are as described earlier (Nawaratne et al, 2008).

GSK-3β Phosphorylation Assays

Stimulation of GSK-3β phosphorylation was performed using FlpIn CHO M4 cells, as described for ERK1/2 phosphorylation assays with the following exceptions: after agonist stimulation and lysis of cells with SureFire lysis buffer, a mixture of reaction buffer, activation buffer, dilution buffer (all as provided by the manufacturer), and AlphaScreen beads was prepared at a ratio of 90 : 10 : 40 : 1 and was added to cell lysates at a ratio of 7 : 5 in a 384-well opaque Optiplate under low light conditions, for a total volume of 12 μl per well. Plates were incubated in the dark at 37°C for 2 h before the florescence signal was measured on a Fusion-α plate reader (PerkinElmer) using standard AlphaScreen settings.

Receptor Internalization

Cells were seeded into 48-well plates at 7.5 × 104 (FlpIn CHO M4 cells) or 5 × 104 cells (NG108-15 cells) per well (coated with poly-D-lysine for NG108-15 cells) and incubated overnight, washed twice with PBS, and incubated in serum-free media. Receptor internalization assays were performed as described earlier (May et al, 2005a).

[3H]ACh Release from Rat Striatal Slices

Sprague–Dawley rats (250–350 g) were killed by exposure to CO2 followed by cervical dislocation. Striata from two rats were dissected and chopped three times at 150 μm using a McIIwain tissue chopper, each time rotating the tissue through 60°C. Slices were dispersed in HEPES buffer (25 mM HEPES, 128 mM NaCl, 2.4 mM KCl, 3.2 mM CaCl22H2O, 1.2 mM KH2PO4, 1.2 mM MgSO47H2O, 10 mM glucose, 1 μM physostigmine, pH 7.4) and incubated with [3H]choline chloride (100 nM) for 30 min at 37°C. After loading, slices were plated into 96-well GF/C filter plates and washed three times over 15 min with buffer (100 μl/well). After the final wash, buffer (with or without LY2033298 and/or NMS) was added and the plates incubated at 37°C for 5 min. Buffer was then removed into a 96-well collection plate. Slices were stimulated for 5 min with 20 mM KCl (with or without LY2033298 and/or NMS), after which the stimulating buffer was removed into a 96-well collection plate. Optiphase supermix (PerkinElmer) scintillation fluid (170 μl) was added to each well of the collection plates before plates being heat sealed and radioactivity quantified using a Wallac 1450 Microbeta 96-well plate counter. Release of [3H]ACh was determined as the total counts per minute released over the 5 min of stimulation, which was converted to a percentage of the release stimulated by 20 mM KCl. Data points are shown as mean±SEM of four independent experiments (each with eight or more replicates).

Conditioned-Avoidance Response

Male M4 mAChR knockout (KO) and WT (C57/BL6 genetic background), weighing 25–45 g, were maintained on a 12 : 12 light : dark cycle with free food and water access. Conditioned-avoidance response (CAR) training occurred in sound-attenuated mouse shuttle cages using an automated software system (Graphic State Notation; Coulbourn Instruments, Allentown, PA). Adapted to mice, CAR training was performed similarly to that described by Rorick-Kehn et al (2007). Each session consisted of a 2 min adaptation period, followed by 50 trials separated by a 30-s inter-trial interval. Illumination of a houselight and opening of shuttle door signaled trial onset. If the mouse shuttled to the other side within 10 s (avoidance response), the trial ended. If an avoidance response was not made, a 0.75-mA floor shock occurred and remained on until a shuttle was made (escape response) or 10 s elapsed (escape failure).

Drug testing occurred once mice stably achieved ⩾90% avoidance responding. Drugs were tested 1–2 times per week, providing that mice achieved 90% avoidance criteria after vehicle injection on the day before. For all qualifying and drug days, mice received oral administration of vehicle or 30 mg/kg LY2033298 (suspended in 1% hydroxyethylcellulose/0.25% polysorbate 80/0.05% antifoam) 60 min before testing. Vehicle or 0.1 mg/kg oxotremorine sesquifumarate (dissolved in normal saline) were delivered subcutaneously, 30 min before testing. Drug testing order was as follows: LY2033298 alone, oxotremorine alone, and the combination of LY2033298 and oxotremorine. All drugs were administered in a volume of 10 ml/kg. Only mice that completed all drug doses were included in the final data. A vehicle dose group was calculated by averaging each of the three qualifying days.

Data Analysis

Radioligand kinetic experiments were fitted to monoexponential association or decay equations (Motulsky and Christopoulos, 2004) and rate constants determined in the presence of LY2033298 were normalized to those determined in its absence. Competition-binding curves between [3H]NMS and ACh in the absence or presence of LY2033298 were initially fitted to a two-site-binding model (Motulsky and Christopoulos, 2004) and subsequently to the following allosteric ternary complex model:

where Y is percentage (vehicle control) binding, [A], [B], and [I] are the concentrations of [3H]NMS, LY2033298, and ACh, respectively, Ka and Kb are the equilibrium dissociation constants of [3H]NMS and LY2033298, respectively, Khi and Klo are the equilibrium dissociation constants of ACh for the high- and low-affinity receptor state, respectively, FracHi is the proportion of receptors in the high-affinity receptor state, and α′ and α are the cooperativities between LY2033298 and [3H]NMS or ACh, respectively. Values of α (or α′) >1 denote positive cooperativity; values < 1 (but >0) denote negative cooperativity, and values=1 denote neutral cooperativity.

Concentration-response curves of the interaction between ACh and LY2033298 in various cell-based signaling assays were globally fitted to the following operational model of allosterism and agonism (Leach et al, 2007; see Supplemental data for derivation and meaning of parameters):

This analysis assumes that the orthosteric ligand, ACh, is a full agonist in the signaling pathway under investigation, which was the case for this study.

All affinity, potency, cooperativity, or operational efficacy parameters were estimated as logarithms (Christopoulos, 1998). Statistical analyses were by one-way analysis of variance and Tukey's multiple comparison post-test was applied, with the exception of the CAR experiments, in which a multivariate analysis of variance (MANOVA) was conducted on percent avoidance responses and escape failures; follow-up-dependent-groups t-tests were performed on significant main effects and interactions involving the independent variables genotype and dose.

RESULTS

LY2033298 Exhibits ‘Probe Dependence’ and Redistributes the Proportion of Receptor-G Protein Complexes

In contrast to orthosteric (competitive) interactions, the magnitude of which is solely determined by the respective binding affinities and concentrations of the interacting ligands, allosteric interactions can display markedly different effects depending on the type of ligand occupying the orthosteric site. This characteristic feature of allosteric interactions is called ‘probe dependence’ (Kenakin, 2007; Leach et al, 2007) because the cooperativity between the two binding sites depends on the properties of the orthosteric ligand used as a probe of receptor binding or function. A striking example of this phenomenon is shown in Figure 1, in which it can be seen that LY2033298 minimally perturbed the binding of two structurally distinct orthosteric antagonists, [3H]NMS and [3H]QNB (Figure 1a), whereas it caused a concentration-dependent increase in the potency of the agonist, ACh, to compete for the binding of [3H]NMS (Figure 1b). The binding cooperativity between LY2033298 and the antagonists can thus be characterized as neutral (vs [3H]QNB) or almost-neutral/weakly negative (vs [3H]NMS), whereas the binding cooperativity with ACh is clearly positive. To delineate the mechanistic basis of the almost-neutral cooperativity between LY2033298 and [3H]NMS, we monitored the effects of the allosteric modulator on the kinetics of association and dissociation of the orthosteric radioligand. As shown in Figure 1c, LY2033298 retarded both the dissociation and association kinetics of [3H]NMS to the same extent and over a similar concentration range. In addition to validating an allosteric mode of action for LY2033298 against [3H]NMS, the fact that the modulator changed both the association and dissociation rates of the orthosteric ligand to the same extent explains why no significant effect was noted on the radioligand's affinity at equilibrium—the latter parameter reflects the ratio of the orthosteric ligand's dissociation and association rate constants. Thus, the conformational change engendered in the M4 mAChR by the binding of LY2033298 is large enough to change the kinetics of orthosteric antagonist binding, but not enough to favor one process (eg dissociation) over another (eg association), leading to neutral cooperativity at equilibrium.

Figure 1
figure 1

LY2033298 displays probe dependence when modulating the binding of antagonists and ACh at the M4 mAChR. (a) Effects of LY2033298 on the equilibrium binding of [3H]NMS or [3H]QNB. Data points represent the mean+SEM of three experiments performed in triplicate. (b) ACh-mediated inhibition of the equilibrium binding of [3H]NMS in the absence or presence of LY2033298. Data points represent the mean+SEM of five experiments performed in triplicate. Curves drawn through the data points represent the best fit of a two-state allosteric ternary complex model (equation (1); Table 1). (c) Concentration-dependent slowing by LY2033298 of the dissociation and apparent association rate of [3H]NMS. Kinetic rate constants were determined in the presence of increasing concentrations of LY2033298 at three time points and normalized as a percentage of the rate constants determined in the absence of modulator (see Materials and methods). Data points represent the mean+SEM of three experiments performed in duplicate.

Full size image

In contrast, the positive cooperativity exerted on the binding of ACh clearly suggests that the conformational change engendered by the modulator is favorable to the binding of the endogenous agonist. Although these assays were performed in the presence of the nonhydrolyzable GTP analog, GppNHp (100 μM), to promote the ground state of the receptor, the competition between ACh and [3H]NMS was characterized by biphasic curves with Hill slopes that were significantly less than unity. Application of empirical one- vs two-site binding models, followed by F-test, found that the latter model was more appropriate, indicating that ACh retained the capacity to distinguish between a high- and low-affinity state even in the presence of guanine nucleotide. This finding may be suggestive of a tight coupling between the receptor and its G proteins, similar to what we have earlier found with the adenosine A1 receptor (May et al, 2005b).

Another important observation from the agonist competition experiments was that LY2033298 caused a redistribution of the proportion of the high- and low-affinity-binding sites, with the fraction of high-affinity-binding sites increasing with increasing concentrations of the allosteric modulator (see Supplementary Table 1). The ability of LY2033298 to enhance the proportion of high-affinity-binding sites was also apparent when [3H]QNB was used as the radioligand in ACh competition-binding experiments (data not shown). As a consequence, the competition-binding data were fitted to an allosteric ternary complex model that accommodated the existence of the receptor in two different states (equation 1)) to estimate the affinity of LY2033298 for the allosteric site and its cooperativities with ACh and with [3H]NMS. The entire family of curves was globally fitted to the model with all parameters shared across the datasets, with the exception of the fraction of high-affinity states, which was allowed to vary between curves. The results of this analysis are shown in Table 1, in which it can be seen that the effects of LY2033298 are accounted for by a mechanism whereby the modulator binds to the unoccupied receptor with a dissociation constant of ∼5 μM (pKb=5.34), potentiates the affinity of ACh for both the high- and low-affinity receptor states by a factor of α=12, and concomitantly promotes an increase in the high-affinity (presumably G protein-coupled) state of the receptor.

Table 1 Binding parameters for the Allosteric Interaction Between LY2033298 and ACh or [3H]NMS at the Human M4 mAChR
Full size table

LY2033298 Potentiates Both the Affinity and the Efficacy of ACh at the M4 mAChR

Given that the binding data suggested an ability of LY2033298 to potentiate the affinity of ACh as well as promote an increase in receptor-G protein coupling, we performed experiments monitoring the ability of [35S]GTPγS to bind to activated Gα proteins as a proximal measure of receptor activation to more directly investigate the consequences of such a mechanism. Figure 2a shows the results of these experiments, in which it can be seen that LY2033298 caused a concentration-dependent increase in receptor signaling even in the absence of ACh, thus acting as an agonist, in addition to allosterically enhancing the potency of the orthosteric agonist. These data were globally fitted to an operational model of allosterism and agonism (equation 2), to yield the parameters shown in Table 2. For this analysis to converge, we fixed the binding affinity of the modulator to the pKb value determined from the radioligand-binding assays. In the operational model, the agonism of LY2033298 is quantified by the parameter, τB, which was found to be ∼8; τ values <10 are indicative of partial agonists (Black, 1996). Interestingly, the maximal degree of positive cooperativity between the modulator and ACh from this analysis (αβ parameter) was estimated as 36 (Table 2). Given that the allosteric enhancement of ACh-binding affinity, α, was determined to be 12 from the radioligand-binding assays, our analysis indicates that the additional positive cooperativity observed in the [35S]GTPγS assay must result from an allosteric enhancement in the signaling efficacy of ACh (ie β=3).

Figure 2
figure 2

LY2033298 displays allosteric agonism and potentiates both the affinity and the efficacy of ACh. ACh-mediated [35S]GTPγS binding at the WT (a) or a functionally impaired Y113C/A203G mutant (b) human M4 mAChR in the absence or presence of LY2033298. Ligands were preincubated with membranes for 1 h at 30°C before the addition of [35S]GTPγS for a further 30 min. Data points represent mean+SEM of 3–4 experiments performed in triplicate. Curves drawn through the data points in (a) represent the best fit of an operational allosteric ternary complex model (equation (2); Table 2). Insets indicate the basal and maximal agonist-stimulated levels of [35S]GTPγS binding, respectively, for the corresponding receptor construct in dpm/mg protein.

Full size image
Table 2 Operational Model Parameters for the Functional Allosteric Interaction between ACh and LY2033298 at Human M4 mAChR
Full size table

To more directly show that LY2033298 can increase the signaling efficacy of ACh independently of its effects on orthosteric agonist affinity, these experiments were repeated at an M4 mAChR with two mutations in key conserved orthosteric site residues (Y113C and A203G) that result in an almost complete abrogation of the binding and function of ACh (Armbruster et al, 2007). As shown in Figure 2b, the significant impairment of ACh signaling at this receptor is rescued by LY2033298, manifesting as an increase in both the potency and, in particular, the maximum response to ACh. This finding is similar to one we made recently with respect to M4 mAChR-mediated ERK1/2 signaling (Nawaratne et al, 2008), indicating that the functional rescue by LY2033298 of impaired orthosteric signaling is likely to apply to multiple signaling pathways. Collectively, these findings are consistent with a mechanism whereby LY2033298 increases the number of high-affinity, G protein-coupled, M4 mAChR states to directly stimulate G protein activation as well as enhancing both the affinity and the efficacy of the endogenous agonist ACh.

LY2033298 binds to the M4 mAChR at a Region that Overlaps with the ‘Prototypical’ Allosteric-Binding Site

Extensive mutagenesis studies have determined a function for residues in the second and third extracellular loops and at the top of transmembrane domain VII to be important for the binding of prototypical allosteric modulators such as C7/3-phth and gallamine at mAChRs (Ellis et al, 1993; Huang et al, 2005; Krejci and Tucek, 2001; Leppik et al, 1994; May et al, 2007a; Prilla et al, 2006; Voigtlander et al, 2003). Preliminary mutagenesis experiments with LY2033298 also implicated extracellular loop regions in the actions of the modulator (Chan et al, 2008). However, there is another class of ‘atypical’ modulators, such as WIN51708, WIN62577, staurosporine, and KT5720, that bind to a second, currently unresolved, allosteric site on mAChRs (Lanzafame et al, 2006; Lazareno et al, 2000, 2002). Thus, to gain more definitive insight into which allosteric site is involved in the binding of LY2033298, we exploited the ability of the modulator to also act as an agonist and performed functional interaction studies between this compound and either the classic orthosteric antagonist, atropine, the ‘prototypical-site’ allosteric modulator, C7/3-phth, or the ‘second-site’ allosteric modulator, WIN51708.

Figure 3a shows the interaction between LY2033298 and atropine in [35S]GTPγS-binding assays. In addition to behaving as an inverse agonist, atropine caused a profound reduction in the maximal agonist effect of LY2033298 with minimal change in its potency. This finding is clearly inconsistent with a competitive interaction and provides further evidence that LY2033298 mediates its agonist effects from an allosteric site. When these experiments were repeated using WIN51708 as the interacting ligand, a markedly different profile of behavior was noted (Figure 3b). Specifically, WIN51708 had virtually no effect on the shape or location of the LY2033298 concentration-response curve, indicative of neutral cooperativity between two topographically distinct-binding sites. Finally, the interaction between LY2033298 and C7/3-phth was investigated (Figure 3c). As with atropine, the allosteric modulator, C7/3-phth, was able to reduce constitutive M4 mAChR-mediated [35S]GTPγS binding. However, this effect was accompanied by a significant dextral translocation of the LY2033298 concentration-response curve, suggestive of a competitive interaction. Unfortunately, because of the relatively low potency of the allosteric agonist in this assay, together with solubility limitations, complete concentration-response curves could not be established in the presence of the higher concentrations of C7/3-phth, making our conclusion of a competitive interaction speculative. To address this, we repeated the experiments using a second functional assay of receptor activation, that of ERK1/2 phosphorylation (Figure 3d), which is characterized by a higher level of stimulus-response amplification and, accordingly, an increase in the potency of LY2033298 as an agonist. This afforded a greater window for observing translocations in the agonist concentration-response curve and, as shown in Figure 3d, the data were clearly indicative of a competitive interaction. Analysis of these ERK1/2 data according to a simple competitive model of interaction (Motulsky and Christopoulos, 2004) yielded a pKb estimate for C7/3-phth of 6.26±0.11, which was in good agreement with earlier estimates for the modulator for interaction with the prototypical allosteric site on the M4 mAChR (Christopoulos et al, 1999). It is also in excellent agreement with affinity values determined from application of a simple allosteric ternary complex model (Motulsky and Christopoulos, 2004) to the interaction between C7/3-phth and ACh in the current ERK1/2 experiments (6.26±0.05; data not shown). Thus, for the first time, we have identified that LY2033298 binds to and activates the M4 mAChR through a region on the receptor that must overlap with residues that contribute to the prototypical mAChR allosteric-binding site.

Figure 3
figure 3

LY2033298 binds to the M4 mAChR through a region that overlaps with the prototypical allosteric-binding site. (a–c) LY2033298-stimulated [35S]GTPγS binding alone or in the presence of (a) atropine, (b) WIN51708, and (c) C7/3-phth. Data points represent mean+SEM of three experiments performed in triplicate. (d) LY2033298-mediated ERK1/2 phosphorylation alone or in the presence of C7/3-phth. Data points represent the mean+SEM of three experiments performed in triplicate

Full size image

LY2033298 Allosterically Engenders Functional Selectivity in the Signaling of ACh

Given that allosteric modulators can co-bind with an orthosteric agonist and promote unique changes in receptor conformation, we have recently proposed that there is no a priori reason why the resulting conformations should display equivalent degrees of allosteric modulation across different signaling pathways linked to the same receptor in the same cellular background (Leach et al, 2007). This phenomenon is referred to as modulator-engendered ‘functional selectivity’ or ‘stimulus-bias’ (Urban et al, 2007), and has significant implications for the detection and development of allosteric molecules as putative drug candidates. To investigate whether LY2033298 can impose functional selectivity on the actions of ACh, we extended our studies of ACh/LY2033298 functional interactions to incorporate assays of the phosphorylation of ERK1/2 and GSK-3β, as well as the ability of ACh to promote M4 mAChR internalization. ERK1/2 and GSK-3β were chosen because they have been implicated as important signaling pathways downstream of receptor activation in a variety of CNS-relevant disorders, including schizophrenia (Beaulieu et al, 2009, 2004; Dwivedi et al, 2001; Emamian et al, 2004; Feng et al, 2003; Ikeda et al, 2004; Lovestone et al, 2007). Receptor internalization was chosen because it is a common mechanism whereby the effects of many GPCR agonists are regulated; the net effect of any receptor-based therapy reflects interplay between acute signaling and longer-term regulatory pathways. For further comparison across pathways, we have also included an analysis of our prior study of the effect of LY2033298 on ACh-mediated intracellular Ca++ mobilization (Chan et al, 2008).

In agreement with the [35S]GTPγS-binding experiments, LY2033298 displayed a robust degree of allosteric agonism for ERK1/2 phosphorylation, GSK-3β phosphorylation, and internalization assays (Figure 4). It should be noted that this was not because of the potentiation by LY2033298 of any trace amounts of endogenous or contaminating ACh, as identical results were obtained in the presence of acetylcholinesterase (Supplementary data Figure 1). In addition to allosteric agonism, LY2033298 potentiated the actions of ACh at all pathways examined. However, application of our operational model of allosterism and agonism to the datasets revealed that the degree of potentiation varied between pathways (Table 2). For instance, in the ERK1/2 and GSK-3β phosphorylation assays, the functional cooperativity (αβ) between LY2033298 and ACh was estimated as 372 and 380, respectively. These values were significantly (P<0.05) higher than that estimated from the [35S]GTPγS-binding assays (36), and are thus suggestive of pathway-selective allosteric modulation. However, there is a theoretical rationale for the degree of positive modulation of a given pathway to be correlated with the degree of agonism displayed for that pathway by the allosteric ligand (Hall, 2000). Given that the agonism parameter, τb, estimated for the ERK1/2 and GSK-3β phosphorylation assays (52 and 18, respectively) was higher than that determined from the [35S]GTPγS-binding assays (8), it is possible that the greater degree of positive cooperativity reflects the greater degree of allosteric agonism and that, in the absence of the latter, the positive modulation may actually be less.

Figure 4
figure 4

LY2033298 engenders functional selectivity (stimulus-bias) in the signaling of ACh. ACh-mediated ERK1/2 phosphorylation (a), GSK-3β phosphorylation (b), Ca2+ mobilization (c), and receptor internalization (d) in the absence or presence of LY2033298. Cells or membranes expressing the M4 mAChR were treated with agonist and agonist-stimulated receptor activity was determined as described in Materials and methods. Data shown in (c) are re-plotted from Chan et al (2008). Data points represent the mean+SEM of 3–8 experiments performed in triplicate. Curves drawn through the points are the best fit of an operational ternary complex model (equation (2); Table 2).

Full size image

In this regard, the analysis of our internalization data and Ca++ mobilization data was particularly striking. The former assay was associated with a level of allosteric agonism (τb=4) that was less than that determined in the [35S]GTPγS-binding assays, but resulted in the greatest degree of positive modulation observed in any of the assays (αβ=794); as with the ERK1/2 and GSK-3β assays, this was significantly different (P<0.05) from that determined for the [35S]GTPγS-binding assay. With respect to the ability of LY2033298 to potentiate ACh-mediated intracellular Ca++ mobilization, we did not see any detectable agonism of the modulator (ie τb=0) in our prior study (Chan et al, 2008), but have determined a cooperativity estimate (αβ=295) that is significantly higher (P<0.05) than that observed for [35S]GTPγS binding. Taken together, these results provide the first evidence for pathway-specific allosteric modulation by LY2033298 on the actions of ACh.

Positive Cooperativity between LY2033298 and Orthosteric Agonists Is Retained In Situ, but Greatly Reduced in M4 mAChR KO Mice

To determine how the allosteric properties of LY2033298 that we identified in the FlpIn CHO cells translate to a native cellular environment, we extended our investigations of the modulator's activity to a neuronal-like cell line, to native brain tissue and finally to behavior in WT and M4 mAChR KO mice.

We first investigated the effects of LY2033298 on NG108-15 cells, which natively express both rat and mouse mAChRs. To facilitate analysis of the signaling data generated in the NG108-15 cells, we initially determined the affinity of the allosteric modulator for the mouse M4 receptor alone expressed in CHO K1 cells (data not shown); the resulting pKb value, 5.54±0.57 (n=3), was not significantly different from that obtained at the human receptor. Subsequently, we chose two signaling assays that were indicative of LY2033298-engendered functional selectivity based on our CHO FlpIn studies (Figure 4), namely ERK1/2 phosphorylation and receptor internalization, and repeated these experiments in the NG108-15 cells. As with the CHO cell studies, LY2033298 caused a robust potentiation of the ability of ACh to promote both ERK1/2 phosphorylation and internalization in the NG108-15 cells (Figure 5a and b). It was noted, however, that the degree of allosteric agonism was markedly reduced in the native cell background, being virtually absent in the internalization assay (Figure 5b). The most likely explanation for the differing strength of LY2033298 agonism between the two cell lines is the lower expression level of M4 mAChRs in the NG108-15 cells. To confirm the dependence of LY2033298 agonism on receptor expression levels, we generated a tetracycline-inducible M4 mAChR expression system using an HEK293 TRex cell line. As shown in Supplementary data Figure 2, progressive reduction in the expression level of the M4 mAChR in these cells led to a marked reduction in the efficacy of LY2033298, relative to ACh, as expected.

Figure 5
figure 5

In vitro validation of LY2033298 as an allosteric potentiator in native cells expressing rodent M4 mAChRs. (a) ERK1/2 phosphorylation and (b) internalization in NG108-15 cells by ACh alone or in the presence of LY2033298. Data points represent the mean+SEM of four experiments performed in triplicate. Curves drawn through the points are the best fit of an operational ternary complex model (equation (2); Table 3).

Full size image

Subsequent analysis of the NG108-15 cell data according to equation 2 yielded the parameter estimates shown in Table 3. Importantly, the pathway-specific allosteric modulation that was observed in the CHO FlpIn cell line expressing the human M4 receptor seemed to be retained in the NG108-15 cells; despite displaying minimal agonism for the internalization pathway (τ=0.2), LY2033298 caused the greater degree of potentiation of this response (αβ=55) relative to ERK1/2 (αβ=22). Collectively, these findings suggest that allosteric modulator-mediated functional selectivity can indeed be operative in a native cellular environment.

Table 3 Operational Model Parameters for the Functional Allosteric Interaction between ACh and LY2033298 in NG108-15 Cells
Full size table

We next determined the effects of LY2033298 in situ by monitoring [3H]ACh release in rat striatal slices, a process known to be regulated by presynaptic M4 autoreceptors (Zhang et al, 2002a). A key focus of these experiments was to delineate the extent to which direct allosteric agonism, as opposed to allosteric potentiation of endogenous ACh, was operative in the actions of the modulator in a native tissue environment. To address this, we took advantage of the unique probe-dependent properties of LY2033298 with respect to its interaction with the antagonist, NMS (Figure 1a). Specifically, because our radioligand-binding assays had indicated that NMS and LY2033298 display minimal interaction with one another, we reasoned that any reversal of the effects of the modulator by NMS must be due to inhibition of the actions of endogenous ACh; in contrast, direct allosteric agonism by LY2033298 would be resistant to inhibition by NMS. As shown in Figure 6a, LY2033298 caused a concentration-dependent inhibition of 20 mM KCl-evoked [3H]ACh release in rat striatum that was almost completely reversed by NMS, indicating that the major effect of the modulator in this tissue is to potentiate the effects of endogenous ACh on presynaptic M4 mAChRs; the antagonist alone also caused a modest increase in [3H]ACh release, consistent with an inhibition of endogenous ACh tone. Interestingly, we also noted that a small degree (∼20%) of the inhibition of ACh release by LY2033298 could not be abolished in the presence of NMS, suggesting that some direct allosteric agonism by LY2033298 may be retained in situ.

Figure 6
figure 6

In situ and in vivo validation of LY2033298 as an allosteric potentiator at rodent M4 mAChRs. (a) 20 mM KCl-stimulated [3H]ACh release from rat striatal slices in the presence of LY2033298, NMS, or NMS plus 100 μM LY2033298. Data points represent the mean+SEM of four experiments performed with eight replicates. (b) CARs and escape failures and (c) latency in response in WT and M4 mAChR KO mice after individual dosing of LY2033298 (30 mg/kg) and oxotremorine (0.1 mg/kg), or a combination of LY2033298 and oxotremorine. *P<0.05 oxotremorine alone vs vehicle treatment; **P<0.05 combination vs oxotremorine alone; #P<0.05 WT vs M4 mAChR KO mice. Data are means+SEM.

Full size image

Finally, the in vivo activity of LY2033298 was measured in a CAR paradigm. Administration of LY2033298 (30 mg/kg) alone did not suppress avoidance responses or increase latency of response in either WT or M4 mAChR KO mice (Figure 6b and c). A submaximal dose of oxotremorine (0.1 mg/kg) reduced avoidance responses in WT and M4 mAChR KO mice by ∼55 and 35%, respectively. Co-administration of both LY2033298 and oxotremorine significantly reduced avoidance responses in the WT mice, such that they were almost completely lost, and significant escape failures (40%) were observed. In contrast, there was only a slight, but significant, reduction in avoidances, and no escape failures, over and above that produced by oxotremorine alone in the M4 mAChR KOs. MANOVA results confirmed these observations: principal effects of genotype [F(1,23)=19.63, P<0.0002] and dose [F(3,69)=0.0001], as well as a significant interaction of these variables, were found on analysis of avoidance responses [F(3,69)=13.86, P<0.0001]. A significant Genotype × Dose interaction was revealed on escape failures, with only WT mice expressing escape failures when given the combination of LY2033298 and oxotremorine [F(3,69)=10.64, P<0.0001].

DISCUSSION

This study provides novel insights into the molecular mechanisms of action of LY2033298, which, together with a related series of molecules, has recently been proposed as a novel, M4-mAChR-selective, tool that may provide a new avenue for the treatment of disorders such as schizophrenia (Chan et al, 2008; Shirey et al, 2008). We have defined a mechanistic basis for the differential effects of this compound on the binding of orthosteric antagonists compared with agonists such as ACh, have validated pharmacologically that both agonism and modulation by LY2033298 arise from an allosteric site that likely overlaps with the site used by prototypical mAChR modulators, have identified, for the first time, signal pathway-selective allosteric modulation at the M4 mAChR, and validated this receptor subtype as a key target for LY2033298 in native cellular systems as well as in vivo.

One of the great challenges in the detection and validation of novel allosteric modulators as potential drug candidates is the issue of probe dependence (Leach et al, 2007). Depending on the nature of the orthosteric ligand, therefore, the resultant pharmacology can vary dramatically. In the case of LY2033298, this was evident when testing the modulator against the antagonists, [3H]NMS, or [3H]QNB, in which virtually no effect on equilibrium binding was noted, compared with the agonist, ACh, in which marked positive cooperativity was observed. These findings have a number of important implications. First, the lack of affinity modulation of the radioligands clearly indicates that the binding site of LY2033298 does not overlap with the classic orthosteric pocket; indeed, the positive cooperativity with ACh means that the modulator prefers to bind to a receptor in which this pocket is occupied by an agonist. Second, if the cooperativity between modulator and orthosteric probe is essentially neutral, as was the case with the antagonists in our study, then the assay will fail to detect the allosteric ligand as having any activity unless a different probe is used; this has important implications for modulator screening. Importantly, our [3H]NMS kinetic-binding studies clearly revealed that the modulator was binding allosterically and changing the conformation of the receptor; the mechanistic basis for the neutral cooperativity at equilibrium was that both radioligand association and dissociation were modulated to the same extent over the same concentration range of the modulator. Third, given that [3H]NMS and [3H]QNB are structurally different (tropate vs benzilate, respectively), yet both exhibited neutral cooperativity with LY2033298 in contrast to ACh, this suggests that the molecular basis for the probe dependence we observed is a sensitivity of the modulator to the conformational state of the receptor (ie active vs inactive).

Further evidence in support of a preferential interaction with an active state of the M4 mAChR by LY2033298 was evident in the ACh/[3H]NMS competition-binding studies, in which the modulator caused an increase in the proportion of the high-affinity state as well as enhancing the binding affinity of ACh. Although the mechanisms that govern the dispersion of agonist affinities across multiple GPCR states in membrane-based radioligand-binding assays remain unclear, a common interpretation is that this somehow reflects the coupling of the GPCR to its cognate G protein(s) to engender a high-affinity state (Christopoulos and El-Fakahany, 1999). Interestingly, the attendant increase in the proportion of the high-affinity ACh state with increasing concentrations of LY2033298 was also observed in our prior study (Nawaratne et al, 2008), but was not factored into our analysis of the interaction at that time because of a limited number of datasets. This likely explains the difference in α values between that study (α=60) and this study (α=12).

The ability of LY2033298 to act as an agonist in the absence of ACh in various signaling assays was another unambiguous demonstration that the modulator prefers an active receptor state. Importantly, this agonistic property of LY2033298 afforded a unique opportunity to perform functional interaction studies with various ligands to pharmacologically probe the location of the allosteric site with which LY2033298 interacts. The interaction between LY2033298 and C7/3-phth was competitive, revealing for the first time that the binding site for LY2033298 must overlap with regions in the extracellular, ‘prototypical’, allosteric mAChR site used by C7/3-phth. This finding is in general agreement with our prior mutagenesis study (Chan et al, 2008), which suggested an engagement of the modulator with extracellular loop regions of the M4 mAChR and an important function of D432 in the third extracellular loop; this residue has been implicated in the binding of the prototypical modulator, gallamine, to the M4 mAChR (Gnagey et al, 1999). In contrast, the interaction between LY2033298 and atropine was negatively cooperative, whereas that between LY2033298 and WIN51708 was neutrally cooperative; both findings clearly indicate that LY2033298 does not interact with either the orthosteric site or a second allosteric site on the M4 mAChR recognized by staurosporine, KT5720, and various WIN compounds (Lanzafame et al, 2006; Lazareno et al, 2000, 2002).

It should be noted that LY2033298 is not the first agonist of mAChRs reported to interact allosterically with this receptor family. Compounds such as AC-42, 77-LH-28-1, N-desmethylclozapine, TBPB, and McN-A-343 have been earlier classified as ‘allosteric agonists’ at M1 and/or M2 mAChRs (Birdsall et al, 1983; Jones et al, 2008; Langmead et al, 2006; May et al, 2007a; Spalding et al, 2006; Sur et al, 2003). However, the interaction between each of these ligands and orthosteric mAChR ligands is antagonistic, suggesting that they either interact with both the orthosteric and an allosteric site on the mAChR, or otherwise are highly negatively cooperative with orthosteric ligands. Given our recent demonstration that McN-A-343 is actually a ‘bitopic’ ligand, concomitantly engaging both the orthosteric and allosteric sites on the M2 mAChR to engender functional selectivity in its actions (Valant et al, 2008), it is possible that this bitopic mode may also extend to the aforementioned agonists (Valant et al, 2009). In contrast, and based on the pharmacological data presented herein, we propose that LY2033298 exerts both its allosteric agonism and its allosteric modulation solely through interaction with the same allosteric site. In this regard, the molecule is unique and the first example of a ‘pure’ allosteric agonist of mAChRs.

Another important premise of this study is that the use of quantitative models of allosterism can yield parameters that describe the nature of the interaction and furnish numbers that can guide subsequent structure-activity studies. Application of our operational model of allosterism and agonism to data derived for LY2033298 in various signaling assays revealed the novel finding that the magnitude of the modulation varied depending on the pathway under investigation, that is the allosteric modulator was ‘biasing’ the stimulus imparted to the receptor by ACh. This property is quantified by the β parameter (efficacy modulation) of our model (Leach et al, 2007). On the basis of the data summarized in Table 2, and assuming that α=12 (affinity modulation) for the interaction between LY2033298 and ACh at the human M4 mAChR, then β values can be calculated as 3, 25, 31, 32, and 66 for [35S]GTPγS binding, Ca++ mobilization, ERK1/2 phosphorylation, GSK-3β phosphorylation, and receptor internalization, respectively. To our knowledge, this is the first example of allosteric modulator-engendered functional selectivity in the signaling of ACh at any mAChR. On a practical level, these findings have significant implications because in most instances, it is not possible to link modulator effects on a specific signal pathway to the prediction of therapeutic efficacy, which in turn suggests that novel compound profiling should be as broad as possible.

One means of increasing the translational relevance of findings such as those described above is to assess the activity of novel modulators in native systems. In this regard, two advantages of the NG108-15 cells used in this study were that they endogenously express M4 mAChRs (albeit rodent receptors), and that the expression level approximates those found in brain (Crook et al, 2000, 2001; Dean et al, 1996; Oki et al, 2005; Scarr et al, 2007; Van Den Beukel et al, 1997). It was interesting, therefore, to note that although LY2033298 was an allosteric agonist in ERK1/2 phosphorylation experiments, it had lower efficacy in the NG108-15 cell line and, when used alone, it had minimal effects on receptor internalization. This is an important finding because it indicates that although LY2033298 retains its ability to stimulate certain signaling responses (ie ERK1/2 phosphorylation), when LY2033298 occupies the receptor in the absence of ACh, it does not itself cause the receptor to couple strongly to internalization, which is a key event in the process of receptor desensitization. However, even though the agonist activity of LY2033298 was lower in the NG108-15 cell line, it retained its ability to allosterically potentiate the actions of ACh on the ERK1/2 and internalization pathways in these cells, clearly indicating that its allosteric modulator activity at M4 mAChRs is fundamental to its actions irrespective of cellular background or receptor expression level. This key finding is underscored by the fact that we also observed a significant effect of LY203298 on [3H]ACh release in rat striatal slices in a largely NMS-sensitive manner, which could only occur if the predominant mode of action of LY2033298 was to potentiate endogenous ACh.

Perhaps most importantly, we have found that the in vivo efficacy of LY2033298 to potentiate mAChR agonist-mediated CAR, a model commonly used as a predictor of antipsychotic drug efficacy, was significantly attenuated in M4 mAChR KO mice. However, we also made the intriguing observation that the actions of LY2033298 in this model were not completely abolished in M4 mAChR KO mice. As described in our prior study (Chan et al, 2008), it is difficult to observe in vivo behavioral effects of LY2033298 alone in rodents because of a lower potency of the modulator in potentiating ACh at the rodent M4 mAChR compared with the human receptor and/or decreased CNS penetrance because of pharmacokinetic limitations. Consequently, this necessitates the administration of a submaximal dose of oxotremorine to provide additional cholinergic tone in vivo, thus revealing the allosteric effect. However, the probe-dependent nature of allosteric interactions means that there is a possibility of LY2033298 and oxotremorine interacting at a non-M4 mAChR subtype, even though this is not possible for the interaction between ACh and LY2033298 (Chan et al, 2008). Further studies are currently underway in our laboratory to address this issue. Nonetheless, the significant blunting of in vivo activity of LY2033298 in the M4 mAChR KO mice validates the M4 mAChR as an important, if not the sole, target for the in vivo actions of LY2033298.

In conclusion, we have shown that the novel small molecule allosteric modulator, LY2033298, exerts both agonistic and modulator effects through a common site on the M4 mAChR that overlaps with the region used by prototypical modulators such as C7/3-phth. The modulator is also able to differentially traffic stimuli imparted to the receptor by the endogenous agonist, ACh, highlighting a novel use of allosteric ligands to engender pathway selectivity in addition to orthosteric ligand and receptor subtype selectivity. Importantly, behavioral studies in M4 mAChR KO mice further validate this receptor as a novel target for the potential treatment of disorders such as psychosis.