Discovery and optimization of a novel, selective and brain penetrant M₁ positive allosteric modulator (PAM): the development of ML169, an MLPCN Probe

Abstract

This Letter describes a chemical lead optimization campaign directed at VU0108370, a weak M₁ PAM hit with a novel chemical scaffold from a functional HTS screen within the MLPCN. An iterative parallel synthesis approach rapidly established SAR for this series and afforded VU0405652 (ML169), a potent, selective and brain penetrant M₁ PAM with an in vitro profile comparable to the prototypical M₁ PAM, BQCA, but with an improved brain to plasma ratio.

The muscarinic acetylcholine receptors (mAChRs) are members of the family A G-protein-coupled receptors (GPCRs) and include five subtypes denoted M₁-M₅. All five of the mAChRs are known to play critical roles in multiple basic physiological processes and represent attractive therapeutic targets for a number of peripheral and CNS pathologies.^1-3 Within the mAChRs, a major challenge has been a lack of subtype selective ligands to study the specific contribution of discrete mAChRs in various disease states.^3,4 To address this limitation, we have focused on targeting allosteric sites on mAChRs as a means to develop subtype selective small molecules, both allosteric agonists and positive allosteric modulators (PAMs).^5-9 Moreover, the emerging phenomenon of ligand-biased signaling requires the development of diverse chemical scaffolds of M₁ ligands to successfully dissect of the roles of M₁ activation through multiple, discrete ligand-biased signaling pathways.^10,11

As members of the Molecular Libraries Production Center Network (MLPCN),¹² we performed a real-time cell-based calcium-mobilization assay employing a rat M₁/CHO cell line (Z′ averaged 0.7) and screened a 63,656 member MLPCN compound library following a triple-add protocol to simultaneously identify M₁ antagonists, agonists (both orthosteric and allosteric) and positive allosteric modualtors (PAMs). This screen proved to be a major success providing viable leads that were optimized into potent and highly selective M₁ ligands (Fig. 1): an M₁ antagonist (1, VU0255035, ML012),¹³ an M₁ allosteric agonist (2, VU0357017, ML071),¹⁴ and both an M₁ PAM (3, VU0366369, ML137)¹⁵ and the first M₅ PAM (4, VU0238429, ML129)¹⁶ derived from a pan-M₁,M₃,M₅-PAM (5, VU0119498).^17,18 However, the brain penetration (brain_AUC/plasma_AUC = 0.1) and efficacy (60% ACh Max) of 3 were poor, as was the brain penetration of the prototypical M₁ PAM, BQCA (6, brain_AUC/plasma_AUC = 0.1)^19-21; therefore, M₁ PAM ligands with improved physicohemical properties for in vivo studies and novel scaffolds to address ligand-biased signaling are required. In this Letter, we describe the development of VU0405652 (ML169), a highly selective M₁ PAM MLPCN probe, with a novel chemical scaffold and improved brain penetration.

Figure 1.

Structures of selective M₁ and M₅ MLPCN probes developed from hits from a triple-add functional M₁ HTS MLPCN screen (1-5) and BQCA (6).

Perusal of the HTS data, which also yielded the non-selective hit 5, identified a second weak M₁ PAM hit 7, VU0108370, with an EC₅₀ of ∼13 μM. Confirmation of 7 from fresh powder and counter-screening against M₂-M₅ increased our enthusiasm for this highly M₁ mAChR selective hit (Fig. 2); however, the CRC did not plateau, suggesting the M₁ EC₅₀ was actually >13 μM. Despite the weak potency, the confirmation of a novel M₁ PAM scaffold with high M₁ selectivity initiated a lead optimization campaign to improve M₁ potency while maintaining the high M₂-M₅ selectivity.

Figure 2.

Concentration response curves (CRCs) for M₁-M₅ for HTS hit VU0108370. M₁ EC₅₀ ∼13 μM (does not plateau) and M₂-M₅ EC₅₀ >30 μM.

Our initial optimization strategy is outlined in Figure 3, and as SAR with allosteric ligands is often shallow, we employed an iterative parallel synthesis approach, along with targeted syntheses for structures encompassing more speculative modifications. Attempted modifications of the Eastern oxazole-amide, although not extensive, met with no success, returning only compounds with undetectable activity. In a straightforward attempt to reduce molecular weight the benzyl group attached to the indole nitrogen was omitted, but met with a similar lack of success (EC₅₀ > 10 μM) as did the sulfide and sulfoxide congeners.

Figure 3.

Initial optimization strategy for VU0108370, 7.

Thus, we planned to hold the northern portion of 7 constant, and survey diversity on the southern benzyl moiety employing a library synthesis approach. As shown in Scheme 1, the key library scaffold 12 was readily prepared in 3 steps from methyl thioglycolate 8. A PyClu-mediated microwave-assisted coupling between 9 and 10 provided 11 in 71% yield, which was then oxidized to the corresponding sulfone 12 with Oxone in 88% yield. An 18-membered library of analogs 13 was then prepared by treatment with NaH and 18 diverse benzyl halides.

Scheme 1.

Reagents and conditions: (a) i. indole, I₂, KI, MeOH, H₂O; ii. 2M LiOH, THF (38%); (b) PyClu, DCE, 110 °C, 20 min, mw (71%); (c) Oxone, MeOH, H₂O (88%); NaH, DMF, BnX (50-90%).

As shown in Table 1, SAR, as with many allosteric ligands, was shallow affording only five active compounds from the eighteen synthesized. Upon resynthesis, HTS hit 7 (a 2-Cl congener) showed an EC₅₀ of 9.7 μM with 83% ACh Max; however, the CRC once again did not plateau. The other four actives, all with substituents in the 3-position, did afford sigmoidal CRCs with up to 96% ACh Max and EC₅₀s in the 3.8 to 6.5 μM range. The most potent analog was 13g, a 3-Br derivative (EC₅₀ = 3.8 μM, 91% ACh Max) which provided a significant increase in potency, but at the cost of physiochemical properties (cLogP >4 and poor solubility). Replacement of the benzyl moiety with a pyridyl analog also led to an inactive compound (data not shown). However, this first generation library indicated that substitution at the 3-position of the benzyl moiety was preferred. This result then prompted us to employ 13g as a starting material for a small Suzuki coupling library (Scheme 2) to replace the lipohilic bromide with various aryl and heteroaryl moieties to introduce basicity and/or polarity.

Table 1.

Structures and activities of analogs 13.

Cmpd	R	M1EC50 (μM)a	%ACh Maxa
7	2-Cl	9.71	83
13a	3-Cl	5.82	96
13b	4-Cl	>10	-
13c	2-OMe	>10	-
13d	3-OMe	6.54	79
13e	4-OMe	>10	-
13f	3-F	5.22	96
13g	3-Br	3.79	91
13h	3-CF3	>10	-
13i	3-CN	>10	-
13j	3,5-diBr	>10	-
13k	3,4-diCl	>10	-
13l	4-CF3	>10	-
13m	4-OCF3	>10	-
13n	2-F	>10	-
13o	2,4-diF	>10	-
13p	2-Br-4-F	>10	-

^aAverage of at least three independent determinations.

Scheme 2.

Reagents and conditions: (a) Ar-B(OH)₂ or Het-B(OH)₂, 10 mol % Pd(t-Bu)₂, 1.0 M aq Cs₂CO₃, THF, mw, 120 °C (65-90%).

As shown in Table 2, SAR was again shallow, but we identified pyrazole as a preferred heterocyclic replacement for the bromide. N-Me pyrazole 14a possessed an M₁ EC₅₀ of 3.2 μM (102% ACh Max), while the unsubstituted pyrazole congener 14b was essentially equipotent (M₁ EC₅₀ = 2.2 μM, 90% ACh Max). Moreover, both pyrazole congeners lowered cLogP a full order of magnitude relative (cLogP = 3.1) to 13g. Additional steric bulk on the pyrazole, as in the sec-butyl derivative 14c, led to an inactive analog. Phenyl derivative 14d and basic amino pyridine analogs 14e-14h, were inactive (M₁ EC₅₀s >10 μM).

Table 2.

Structures and activities of analogs 14.

Cmpd	Ar/Het	M1 EC50 (μM)a	%ACh Maxa
14a		3.21	102
14b		2.19	90
14c		>10	-
14d		>10	-
14e		>10	-
14f		>10	-
14g		>10	-
14h		>10	-

^aAverage of at least three independent determinations

To further the development of this novel series of M₁ PAMs we focused on three of the more potent analogs (13a, 13g and 14b) and applied a fine-tuning process of introducing fluorine atoms at various locations to provide analogs 15 (Table 3), an approach we found successful for multiple allosteric ligands. Across the series, substitution at the 4-position was uniformly not tolerated (15a-c), consistent with the SAR appearing in Table 1. Bis-fluorination of the indole ring eroded activity in the context of the bromine analog 15d but conversely augmented the activity of the pyrazole congener 15e. This type of subtle/confounding SAR was similarly observed with respect to fluorination at the R² position in analogs 15f-h. While the presence of a fluorine at R² was neutral or slightly beneficial in the context of the chlorine analog (15f), its presence resulted in clearly decreased activity for both the bromine and pyrazole analogs, 15g and 15h. Lastly, the introduction of a single fluorine at the R⁶ position could either be moderately detrimental in the context of pyrazole 15i or decidedly beneficial with respect to bromine analog 15j, where an almost 3-fold improvement in potency was observed. In this manner, both VU0405652 (15j) and the related difluorindole analog VU0405645 (15e) were chosen for further evaluation.

Table 3.

Structures and activities of analogs 15.

Cmpd	G	R1	R2	R4	R6	M1 EC50 (μM)a	%ACh Maxa
15a	A	H	H	F	H	>10	-
15b	Br	H	H	F	H	>10	-
15c	Cl	H	H	F	H	>10	-
15d	Br	F	H	H	H	5.37	103
15e	A	F	H	H	H	1.85	102
15f	Cl	H	F	H	H	5.10	56
15g	Br	H	F	H	H	>10	-
15h	A	H	F	H	H	4.40	103
15i	A	H	H	H	F	3.07	96
15j	Br	H	H	H	F	1.38	84

^aAverage of at least three independent determinations.

Both 15e and 15j were selective (Fig. 4A) for M₁ (>30 μM vs. M₂-M₅) and both compounds demonstrated impressive left-ward shifts of the ACh CRC (98-fold and 49-fold, respectively) in fold-shift assays at 30 μM (Fig. 4B), values comparable to BQCA. However, a combination of calculated properties (TPSA, Hacc, etc…), ancillary pharmacology and in vivo PK suggested 15j was the more optimal MLPCN probe molecule compared to 15e.

Figure 4.

A) CRC for 15j, VU0405652 at M₁-M₅ (15e, VU0405645 looked similar); B) ACh fold-shift experiment at 30 μM for VU040652 (49-fold) and VU0405645 (98-fold).

In terms of ancillary pharmacology, the Lead Profile screen at Ricerca, evaluating 68 GPCRs, ion channels and transporters in radioligand binding assays, resulted in only two significant activities (DAT, 83% at 10 μM and sodium channel site 2, 83% at 10 μM); importantly, 15j was selective versus the biogenic amines (D₂, H-HT₂B, etc…) and displayed no orthosteric binding at M₁-M₅. Based on the exciting in vitro profile, we then evaluated 15j for brain penetration in rat. A 10 mg/kg IP dose of 15j afforded a brain _AUC/plasma_AUC of 0.32 at 1 hour, providng an improvement over both 3 (ML137) and BQCA with brain_AUC/plasma_AUC of ∼0.1. Based on this profile, 15j (VU0405652) was declared an MLPCN probe, ML169.

As we have previously demonstrated with both an M₁ PAM (BQCA)²⁰ and M₁ allosteric agonists (2, ML071 and TBPB),²³ ML169 also shifted APP processing towards a non-amyloidogenic pathway.²⁴ As shown in Figure 5, 10 μM carbachol (CCh) affords a significant increase in soluble APP (APPsα), while 100 nM CCh provides a modest increase. VU0405652 (ML169) at a dose of 2 μM has no effect, but in combination with low dose CCh (100 nM), ML169 potentiates the CCh-mediated non-amyloidogenic APPsα release to the same degree as 10 μM CCh. These data once again suggest that selective activation of M₁ may have a disease modifying role in Alzheimer's disease.^14,20

Figure 5.

M₁ PAM VU04505652 (ML169) potentiates the CCh-mediated non-amyloidogenic APPsα release in TREx293-hM₁ cells.

In summary, we have developed a potent, selective and brain penetrant M₁ PAM, ML169, based on a novel indole scaffold from an MLPCN functional HTS. Further in vivo evaluation of this probe is underway and results from preclinical models of Alzheimer's disease and schizophrenia will be reported in due course. ML169 is an MLPCN probe and freely available upon request.