Allosteric modulation of the muscarinic M₄ receptor as an approach to treating schizophrenia

Abstract

Current antipsychotics provide symptomatic relief for patients suffering from schizophrenia and related psychoses; however, their effectiveness is variable and many patients discontinue treatment due to side effects. Although the etiology of schizophrenia is still unclear, a leading hypothesis implicates an imbalanced dopaminergic system. Muscarinic acetylcholine (ACh) receptors regulate dopamine levels in key areas of the brain involved in psychosis, with the M₄ subtype emerging as a key regulator of dopaminergic hyperactivity. Unfortunately, no selective small molecule tools exist to provide pharmacological validation of this hypothesis. Here, we describe the discovery of a small molecule modulator, LY2033298, that is highly selective for human M₄ receptors by virtue of targeting an allosteric site on this receptor. Pharmacological assays confirmed the selectivity of LY2033298 for the M₄ receptor and revealed the highest degree of positive allosteric enhancement of ACh potency thus far identified. Radioligand binding assays also show this compound to directly potentiate agonist binding while having minimal effects on antagonist binding. Mutational analysis identified a key amino acid (D⁴³²) in the third extracellular loop of the human M₄ receptor to be critical for selectivity and agonist potentiation by LY2033298. Importantly, LY2033298 was active in animal models predictive of clinical antipsychotic drug efficacy indicating its potential use as a first-in-class, selective, allosteric muscarinic antipsychotic agent.

Schizophrenia is a complex disease presenting a broad spectrum of endophenotypes that can be generalized to symptoms in three major domains: positive (hallucinations, hearing of voices, delusions and disorganized thinking), negative (anhedonia, flat affect), and cognitive (attention and working memory deficits). The etiological basis for the disease is believed to derive predominantly from dysregulation of dopamine and glutamate neurotransmission pathways in mesocortical and mesolimbic brain areas (1–3). Atypical antipsychotics are the current frontline treatment for schizophrenia, ameliorating the positive symptoms in approximately half of the treated patient population with little efficacy at the negative and cognitive symptoms. Most atypical antipsychotics are broad spectrum G protein-coupled receptor (GPCR) antagonists, and their therapeutic action is mediated primarily through inhibition of dopamine D₂, D₃ and D₄, and serotonin 5HT_2A, receptors (4). Their complex pharmacology, however, leads to significant undesirable side effects, including movement disorders and weight gain. Therefore, developing antipsychotics through an alternative mechanism may provide better total symptom control and reduced side effects.

The family of five muscarinic acetylcholine (ACh) receptors plays a prominent role in regulating neurotransmission in the CNS and may provide a mechanism for antipsychotic drug discovery. Studies using muscarinic receptor knock out mice have provided valuable insight into the potential role of these receptors, especially the muscarinic receptor M₁ and M₄ subtypes, in the pathophysiology of schizophrenia (5, 6). Muscarinic receptors are genetically linked to schizophrenia and cognitive deficits (7, 8). Pharmacologically, relatively non-selective muscarinic antagonists exacerbate, whereas agonists ameliorate, cognitive deficits and psychotic behaviors in animal models and in patients suffering from Alzheimer's disease and schizophrenia (9–12). In particular, the M₁/M₄ preferring partial agonist, xanomeline, was found to be efficacious in animal models predictive of antipsychotic behaviors through modulation of both dopamine and glutamate transmission (13). In exploratory clinical trials, xanomeline reduced psychotic behaviors and improved cognitive measures in Alzheimer's patients (9), and significantly improved positive, negative, and cognitive symptoms in a small schizophrenia trial (12). Unfortunately, lack of receptor subtype selectivity led to undesirable side-effects (gastrointestinal disturbances, blood pressure dysregulation) that rendered xanomeline as an unsuitable candidate for further clinical development (14).

M₄ receptors are most highly expressed in brain regions rich in dopamine and dopamine receptors (15, 16). Experiments with muscarinic knock out mice have implicated the M₄ receptor subtype in regulating dopaminergic neurons involved in movement control and cognition (17, 18); therefore, we reasoned that a highly selective M₄ muscarinic receptor activator may represent a useful candidate for the development of muscarinic-based antipsychotic agents. Unfortunately, attempts at discovering selective small molecule ligands for the five muscarinic receptor subtypes have largely been unsuccessful when targeted at the receptors' orthosteric site, i.e., the binding site for the endogenous agonist ACh, which is highly conserved across all five muscarinic subtypes (19). This has been particularly true for the ACh-mimetics, such as xanomeline, sabcomeline, and melameline (20). However, it is also known that muscarinic receptors possess allosteric binding sites that have the potential to show greater sequence divergence across subtypes and, hence, may provide an alternative means of attaining receptor subtype selectivity (21, 22). Allosteric modulators also offer the additional potential of maintaining both spatial and temporal neurotransmission through modulation of physiologically relevant receptor-mediated neural regulation, as opposed to direct (continuous) agonism or antagonism (23, 24). Here, we describe the exploitation of this concept to the discovery and molecular characterization of LY2033298 [3-amino-5-chloro-6-methoxy-4-methyl-thieno(2,3-b)pyridine-2-carboxylic acid cyclopropylamide (C₁₃H₁₄ClN₃O₂S; molecular weight = 311.8)] (Fig. 1A), a highly selective positive allosteric potentiator of ACh actions at the M₄ muscarinic receptor that also shows significant efficacy in preclinical models predictive of antipsychotic drug behavior.

Fig. 1.

LY2033298 enhances ACh responses selectively at hM₄. The effect and selectivity of LY2033298 on ACh were tested by FLIPR (A) and [³⁵S]GTPγS (B) binding, using recombinant cell lines (AV12 G_α15 hM₂, or hM₄ and CHO hM₁, hM₃, or hM₅) and CHO cell membranes of hM₁–M₅, respectively. Responses in the presence of LY2033298 were normalized to the control maximal ACh response (100%) and basal response (B; 0%) for each receptor. In both assays, no significant allosteric effect was observed in hM₁, hM₃, or hM₅ cell lines. A very modest allosteric effect on hM₂ receptors was observed and a small agonist effect was detected at hM₄. (A Inset) Illustrated is the paradigm of functional screening for the concentration-dependent effect of LY2033298 on a submaximal dose of ACh (3 nM), which was obtained by plotting the data points along the dotted line.

Results

LY2033298 Is a Selective Positive Allosteric Modulator of ACh at Human Muscarinic M₄ Receptors.

Initial screening quantified intracellular calcium mobilization as a readout of muscarinic receptor activation, using cell lines stably expressing the hM₁–M₅ receptors coupled to intrinsic or engineered calcium signaling pathways. No agonist activity at any muscarinic receptor subtype was detected when LY2033298 was applied alone. However, upon addition of a submaximal concentration of ACh, an elevated calcium response was readily detected in the hM₄ cell lines relative to the effect of the same concentration of ACh in the absence of LY2033298. Subsequent assays constructed complete concentration-response curves to ACh in the presence of increasing concentrations of LY2033298. As shown in Fig. 1A, these experiments revealed a prototypical characteristic of positive allosteric interactions: a robust potentiation of agonist potency that reached a limit at the highest concentrations of allosteric modulator above which no further effect was obtained (24). Application of an allosteric ternary complex model to the data yielded K_B = 200 ± 40 nM for LY2033298 at the allosteric site on the unoccupied hM₄ receptor, and α = 35 ± 4 for the degree of allosteric enhancement when both orthosteric and allosteric sites are occupied; to our knowledge, this is the highest degree of positive cooperativity reported for any allosteric modulator of a muscarinic receptor. In addition, these experiments also revealed the high selectivity of LY2033298 for the hM₄ receptor, because there was no effect at hM_1/3/5 receptors at any concentration of modulator and a very small allosteric potentiation at the closest homolog, hM₂ (K_B = 1.0 ± 0.3 μM; α = 3.7 ± 0.5). The positive allosteric effect of LY2033298 at the hM₄ receptor was not unique to ACh, because it also selectively potentiated other full and partial orthosteric agonists, such as carbachol, oxotremorine-M, and McN-A343 (data not shown).

To confirm that the allosteric effect detected in the calcium assay was neither unique to this signaling pathway or a consequence of engineering the coupling of the hM₄ receptor to this pathway via the use of promiscuous Gα₁₅ proteins, additional experiments were performed measuring ACh-mediated [³⁵S]guanosine-γ-S–triphosphates (GTPγS) binding to native G proteins in CHO cell membranes (Fig. 1B). As with the calcium assay, selectivity for the hM₄ receptor was observed, with no quantifiable effect at the hM₂ receptor. Interestingly, a small increase in basal [³⁵S]GTPγS binding was also observed when LY2033298 was applied alone to both hM₄ and hM₂ receptors, indicating a weak degree of allosteric agonism in addition to any allosteric modulation. However, poor solubility of LY2033298 at high concentrations limited the analysis and characterization of this allosteric agonist effect in this assay. Nonetheless, because the agonist effect of LY2033298 was small, we were still able to apply the simple allosteric ternary complex model to derive apparent estimates of modulator affinity (K_B = 870 ± 310 nM) and cooperativity (α = 28 ± 7) at the hM₄ receptor.

LY2033298 Directly Enhances the Binding of Agonists but Not the Antagonist N-methyl-scopolamine (NMS).

Radioligand binding assays were performed by using the orthosteric agonist [³H]Oxotremorine-M (Oxo-M) and antagonist [³H]NMS to confirm that the allosteric effects of LY2033298 are predominantly manifested directly at the level of orthosteric ligand affinity. Fig. 2A shows that LY2033298 robustly potentiated the specific binding of [³H]Oxo-M in CHO hM₄ cell membranes and in rat striatal membranes known to express a high proportion of M₄ muscarinic receptor (F.P.B. and C.C.F., unpublished data). Interestingly, although a significant potentiation of [³H]Oxo-M binding was retained in native rat tissue, the potency of LY2033298 at the rat M₄ receptor was 5- to 6-fold lower than that at the human M₄ receptor, suggesting a possible species difference in the allosteric effect.

Fig. 2.

LY2033298 allosterically increases agonist binding to M₄ receptors. (A) The specific binding of [³H]Oxo-M to CHO hM4 cell membranes and rat striatal membranes was potentiated by LY2033298 in a dose-dependent manner. (B) The specific binding of [³H]NMS to hM4 receptors in the presence of different concentrations of unlabeled ACh and LY2033298 was measured in the presence of 0.2 mM GTP. The incubation time was 2 h to allow equilibrium to be reached. The low slope factor of the ACh inhibition curve (0.67) was unchanged in the presence of increasing concentrations of LY2033298. The presence of LY2033298 increased ACh potency up to 40-fold. At high concentrations, LY2033298 appears to weakly and negatively modulate [³H]NMS binding at hM₄ receptors. The data are from one experiment repeated three times with quantitatively similar results.

To assess the effect of LY2033298 on receptor affinity for the endogenous agonist at hM₄ receptors, ACh inhibition of [³H]NMS binding was measured at equilibrium in the presence of 0.2 mM GTP and with increasing concentrations of LY2033298 (Fig. 2B). The inhibition curve in the absence of LY2033298 had a logistic slope factor of 0.67 and, with increasing concentrations of LY2033298, the curves were progressively shifted leftward without change in slope factor. The increase in ACh potency was 40-fold at 3 μM LY2033298, illustrating the very strong positive cooperativity between LY2033298 with ACh for binding to uncoupled hM₄ receptors. In contrast, under these incubation conditions, the binding of [³H]NMS was slightly reduced at higher concentrations of LY2033298, compatible with a small negative cooperativity with [³H]NMS (α ≅ 0.5). These assays demonstrated another characteristic of the allosteric interactions of LY2033298, probe-dependency (24). No appreciable radioligand displacement by LY2033298 was detected in orthosteric binding assays for other GPCRs (e.g., dopamine, 5HT, adrenergic receptor families), major ion channels (e.g., nicotinic), signaling molecules, and enzymes (e.g., kinase family) (data not shown). Moreover, LY2033298 enhancement of [³H]Oxo-M binding was lost in M₄ knockout and not in M₂ knockout or wild-type mouse membranes supporting the selectivity for M₄ over the M₂ receptor (data not shown).

Molecular Determinants of LY2033298 Selectivity and Allosterism at the M₄ Receptor.

The orthosteric ligand binding pocket is highly conserved between muscarinic receptor subtypes and across species (19). Among these receptors, hM₂ and hM₄ share the closest sequence homology; however, we still observed significant differences in LY2033298 selectivity. Thus, we created a series of hM₄ receptor mutants [supporting information (SI) Table S1], in which divergent residues in the extracellular N terminus and loop regions were replaced with those of hM₂ receptor at the corresponding sites, and measured ACh-stimulated functional responses. All mutants showed unaltered pharmacology to orthosteric agonists and antagonists compared with control wild-type hM₄ receptors (Fig. S1). The most dramatic loss of LY2033298 activity was noted upon progressive mutation of the third extracellular loop (o3) (Fig. 3A). LY2033298-mediated potentiation, both in terms of its potency and magnitude of potentiation, was particularly reduced in mutant hM4-Δc6.1, and a little further in hM4-Δc7.1, indicating residue D⁴³² is critical for this activity. Subsequent point mutations confirmed that the size and charge of this residue are vital to the potency of LY2033298 (Fig. S2).

Fig. 3.

Sequential mutations in the o3 loop of muscarinic receptors reveal a critical residue (D⁴³²) for LY2033298-mediated potentiation. Receptor mutants were transiently expressed in AV12 G_α15 cells. The effect of LY2033298 on ACh stimulated calcium mobilization was tested upon coapplication of LY2033298 (1 nM to 1 μM) with the EC₂₅ concentration of ACh for each receptor. The results were normalized to the potentiation mediated by LY2033298 on hM₄ receptors, where 0% and 100% corresponded to zero and maximum potentiation mediated by LY2033298 on hM₄ receptors (see Fig. S4). (A) Mutants hM4-Δ6.1 and hM4-Δ7.1 were significantly less responsive to LY2033298 (*) when compared with hM₄ receptors. The data points represents mean ± S.E.M. of more than eight experiments. (B) The enhancement of ACh function by LY2033298 for the humanized rM₄ was not significantly different from rM₄ receptors, whereas ratinized hM₄ was significantly different from hM₄ but not from rM₄ receptors. The receptor selectivity of LY2033298-mediated potentiation between rM₄ and rM₂ receptors was sustained. Data shown are mean ± SEM (n ≥ 3).

Comparison of rat and human M₄ receptor sequences at this domain also identified two nonconserved residues in the o3 loop, one of which corresponded to D⁴³² in the hM₄ receptor. In agreement with the results of the [³H]Oxo-M binding assays, the rM₄ receptor demonstrated a decreased potency of LY2033298 and a reduced potentiation of ACh-mediated calcium mobilization compared with the hM₄ receptor, although the higher selectivity of LY2033298 for the M₄ subtype relative to the M₂ subtype was retained (Fig. 3A). When the two divergent residues within the third extracellular loop of the hM₄ receptor were converted to their rM₄ counterparts (“ratinized hM₄”), there was a significant loss LY2033298 activity, albeit not to the degree observed at the wild-type rM₄. However, when the reciprocal mutations were made in the rM₄ receptor (“humanized rM₄”), there was no gain of function (Fig. 3B). This discrepancy suggests that the species difference observed likely involves a complex interplay between the o3 loop and other domains/residues on the rM₄ receptor.

LY2033298 Displays in Vivo Efficacy in Preclinical Animal Models Predictive of Antispsychotic Drug Effects.

When LY2033298 was administered alone to rats in conditioned avoidance responding (CAR), and prepulse inhibition (PPI) models and when performing microdialysis sampling of brain mono-amines, we did not observe any effect- consistent with lower activity of LY2033298 at the rodent vs. human M₄ receptor. However, when coadministered with a subeffective single dose of oxotremorine, LY2033298 was active in attenuation of CAR and reversal of apomorphine-disrupted PPI in a dose-dependent manner (Fig. 4 A and B), indicating that the compound is indeed effective in vivo through a muscarinic mechanism. Similar results were observed in microdialysis experiments where LY2033298 positively modulated the dopaminergic system in the prefrontal cortex in the presence of an inactive dose of oxotremorine (Fig. S3), suggesting it had reached and affected the desired target pathway. The effectiveness of LY2033298 in rodent models predictive of antipsychotic efficacy provides compelling proof-of-concept that allosteric potentiation of the M₄ muscarinic receptor is a viable approach toward the development of muscarinic-based antipsychotic agents. Because of its significantly higher activity at the human receptor, this and related compounds may have efficacy when given alone in human subjects.

Fig. 4.

LY2033298-mediated potentiation is effective in rat CAR and PPI psychosis models. (A) LY2033298 reduced conditioned avoidance responding (CAR) in trained male HSD rats in a dose-dependent manner in the presence of an inactive dose of oxotremorine (*, P < 0.05 versus vehicle). (B) Apomorphine-induced suppression of the acoustic startle response was reversed by LY2033298 in a dose-dependent manner in the presence of a subeffective dose of Oxo. Data are mean percentage prepulse inhibition (PPI) values (V, vehicle; Apo, apomorphine; Oxo, oxotremorine; LY, 2033298) (#, P < 0.05 versus V/V/Apo; *, P < 0.05 versus V/V/V).

Discussion

This study has identified a functionally potent and selective allosteric potentiator of muscarinic receptors. The characteristics of this allosteric modulator LY2033298 include: (i) selectivity for modulating ACh actions at the M₄ receptor, attributable to epitopes on the receptor topographically distinct from the highly conserved orthosteric site; (ii) saturability in the enhancement of ACh potency as a direct consequence of cooperativity between orthosteric and allosteric sites; (iii) probe dependence in the allosteric effect, because LY2033298 selectively potentiates the actions of orthosteric agonists at M₄ receptors but interacts very weakly with the orthosteric antagonist, [³H]NMS; and (iv) signal transduction pathway independence, because the effects were observed in [³⁵S]GTPγS binding and calcium mobilization assays.

Xanomeline is pharmacologically characterized as an M₁/M₄ > M₂/M₃ > M₅ partial agonist with modest interaction at serotonin receptors (6, 25). The clinical efficacy detailed in the Introduction raises the question as to which muscarinic receptor(s) to focus on for the indication of schizophrenia. In the absence of selective pharmacology, muscarinic receptor knockout mice have provided evidence encouraging a focus on the M₄ receptor as a logical target for modulation of dopamine neurotransmission (17) in areas thought to be disrupted in schizophrenia (18). BuTAC and PTAC, potent molecules with M₂/M₄ agonist, M₁/M₃/M₅ antagonist activity, were active in models predictive of antipsychotic efficacy (26, 27). Together with our data on LY2033298, these findings support a role for M₄ as potentially sufficient to treat symptoms of schizophrenia. Because M₁ receptor involvement in cognitive processes has been well documented (28, 29), a combination of M₁ and M₄ agonistic activity may offer some additional efficacy in the clinic.

Previous investigations identified subtype-selective muscarinic allosteric enhancers (23, 30–32). However, these molecules displayed much lower affinity and cooperativity with ACh. A particularly striking finding from our functional assays was the high degree of positive cooperativity between the modulator and ACh at the hM₄ receptor (α≈30), which is the highest for muscarinic receptors reported to date and is sufficient to impart behavioral responses in animal models predictive of antipsychotic drug effects at reasonable therapeutic doses. Radioligand binding experiments using recombinant hM₄ and native rM₄ membrane preparations confirmed that the allosteric effects of LY2033298 are mediated predominantly by enhancement of agonist binding while having close to neutral effects on antagonist binding. [³⁵S]GTPγS binding experiments indicate a low intrinsic efficacy of LY2033298 at the hM₂ and hM₄ receptors in the absence of added ACh consistent with an allosteric partial agonist of very low efficacy (33, 34).

To date, limited mutagenesis studies of muscarinic receptors have focused on a common allosteric binding site on M₁, M₂, and M₅ receptor subtypes (33, 35) and M₃ (36, 37) and M₄ receptors (38) used by prototypical negative allosteric modulators, such as gallamine, alcuronium, and C₇/3-phth (21). Functional screening of a wide range of receptor mutants systematically pointed to the acidic residue D⁴³² of the hM₄ receptor as being critical to the potentiation of ACh by LY2033298 (Fig. 3A). Additional studies on the rM₄ receptor, which shows a lower cooperativity with LY2033298, confirmed the importance of the o3 region of the receptor in the actions of the modulator (Fig. 3B). However, a humanized rM₄ mutant did not show a gain-of-function in enhancing the actions of LY2033298, suggesting that a more complex network of interactions with other residues must also be involved. In general, the finding that D⁴³² was critical to the cooperativity between LY2033298 and ACh at the hM₄ receptor contrasts with residues important for the action of less robust allosteric modulators, which, for the hM₂ receptors, have been localized to the o2 loop and the top of transmembrane domain 7 (38, 39). However, an unequivocal demonstration that LY2033298 interacts with a second allosteric site that is topographically distinct from that used by prototypical muscarinic modulators requires further pharmacological and structural characterization.

A key finding of our study was the demonstration of in vivo efficacy of a muscarinic allosteric potentiator in rodent preclinical models predictive of antipsychotic drug effects. Allosteric sites are not expected to have evolved to accommodate a common endogenous neurotransmitter or hormone; they are more likely to show higher sequence divergence across receptor subtypes (which represents a therapeutic targeting advantage) but may also show variation between species, which represents a disadvantage in terms of target validation and/or preclinical efficacy and safety studies in animal models. Transgenic approaches to knock-in humanized receptor proteins into mice may be helpful for progressing allosteric drug leads showing high human receptor selectivity. Alternatively, allosteric compounds may be coadministered with inactive doses of orthosteric agonists.

Arguably, little innovation has been made in the treatment of schizophrenia for the past several decades. Frontline treatments are generally compounds with complex broad GPCR antagonist pharmacology with highest affinity as antagonists for D₂-type dopamine receptors and 5HT₂ serotonin receptors (4). GPCR agonist mechanisms have largely failed or, at best, show temporary efficacy because of issues of receptor desensitization or down regulation (40). Aripiprazole may be an exception because of its partial agonist activity at D₂ receptors, although the exact mechanism of action of this compound has been controversial (41). More recent preliminary studies suggest antipsychotic efficacy through agonism of type 2 and 3 metabotropic glutamate receptors (42). Muscarinic selective agonism with xanomeline (although not completely muscarinic receptor subtype selective) has shown interesting promise in managing Alzheimer's-related psychosis and schizophrenia symptoms, including positive, negative, and cognitive domains (6, 9). An M₄-selective drug that provides modest or self-limiting agonism may be preferable to a full orthosteric agonist, although it is not clear how much receptor agonism (or modulator affinity and cooperativity) will be required in patients to achieve maximal therapeutic efficacy and tolerable side effects. Studies to date indicate that M₄ receptors are largely unchanged in schizophrenic patients with the exception of a potential modest decrease in hippocampus, suggesting that M₄ agonist-based therapies would remain relevant during disease progression (43, 44). Our rodent studies, although preliminary, provide support for a highly M₄ receptor-selective allosteric potentiator to control symptoms without inducing sedation or catalepsy observed with atypical antagonist antipsychotics such as clozapine. The allosteric mechanism inherent in LY2033298 may appropriately modulate imbalanced dopamine and glutamate neurotransmission, thus representing a potential first-in-class muscarinic antipsychotic therapy that does not involve direct targeting of dopamine and serotonin receptors.

Materials and Methods

Reagents.

3-Amino-5-chloro-6-methoxy-4-methyl-thieno[2,3-b]pyridine-2-carboxylic acid cyclopropylamide (LY2033298; C₁₃H₁₄ClN₃O₂S; molecular weight = 311.8) (Fig. 1A) was synthesized by Eli Lilly & Co.. LY2033298 is soluble in buffer or in 100% DMSO up to 4 mM and is stable as a dry powder at room temperature or as a frozen DMSO stock solution. The human muscarinic receptors were stably expressed in cell lines as described in SI Methods.

Methods.

Signal transduction and radioligand binding assays and data analyses were performed as described in refs. 24, 45, and 46 with modifications as detailed in SI Methods. Rat behavioral assays, conditioned avoidance responding (CAR), and prepulse inhibition of the acoustic startle reflex (PPI) were performed as described in ref. 47 with modifications as detailed in SI Methods. Mutagenesis and receptor expression methods are described in SI Methods.