Further optimization of the M₅ NAM MLPCN probe ML375: Tactics and challenges

Abstract

This letter describes the continued optimization of the MLPCN probe ML375, a highly selective M₅ negative allosteric modulator (NAM), through a combination of matrix libraries and iterative parallel synthesis. True to certain allosteric ligands, SAR was shallow, and the matrix library approach highlighted the challenges with M₅ NAM SAR within in this chemotype. Once again, enantiospecific activity was noted, and potency at rat and human M₅ were improved over ML375, along with slight enhancement in physiochemical properties, certain in vitro DMPK parameters and CNS distribution. Attempts to further enhance pharmacokinetics with deuterium incorporation afforded mixed results, but pretreatment with a pan-P450 inhibitor (1-aminobenzotriazole; ABT) provided increased plasma exposure.

Of the five muscarinic acetylcholinereceptors (mAChR subtypes M₁–M₅), far less in known about the neurobiological roles of M₅ due both to limited CNS expression (< 2% of all mAChR protein in rat brain and found exclusively in the ventral tegmental area [VTA] on dopamine transporter [DAT]-expressing neurons and in the substantia nigra pars compacta [SNc]) and the lack of highly selective, in vivo probe molecules.^1–10 Insight into the therapeutic potential of M₅ comes largely from genetic studies in M₅-KO mice, which exhibit reduced sensitivity to the rewarding effects of cocaine and opiates.^11–13 Recently, an association between an M₅ SNP and an addictive phenotype was observed in man, directly linking M₅ to drug abuse and reward.¹⁴ To advance the M₅ research field, small molecule probes are required to recapitulate the genetic data.

Previously, we have reported on the devleopment of several potent and selective M₅ positive allosteric modualtor (PAM) chemotypes,^15–18 as well as the first highly M₅ selctive orthosteric antagonist;¹⁹ however, DMPK properties were generally poor and these efforts failed to produce in vivo probes. Last year we disclosed results from an M₅ functional high-throughput screen that provided 1-(4-flurobenzoyl)-9b-phenyl-2,3-dihydro-1H-imidazo[2,1-a]isoindol-5(9bH)-one as an M₅ negative allosteric modulator (NAM) hit, 1 (Fig. 1).²⁰ A limited chemical optimization effort afforded ML375 (2), the first M₅-selective NAM with favorable CNS exposure (brain:plasma K_p = 1.8), moderate PK, high plasma protein binding (rat f_u = 0.029, human f_u = 0.013, rat brain f_u = 0.003) and enantiospecific activity (only the (S)-enantiomer of the 9b p-Cl phenyl wasactive).²⁰ Despite a major advance in the field, due to weak potency at rat M₅, coupled with high plasma protein and brain homogenate binding, ML375 lacked the requisite free drug exposure to serve as an in vivo tool compound.²⁰ In this Letter, we report on the continued optimization of our first-in-class M₅ NAM, and detail key tactics and noteworthy challenges en route to an M₅ NAM in vivo probe.

Figure 1.

Structures and mAChR activities of M₅ NAM HTS hit 1, and the optimized MLPCN probe ML375 (2). Inset, optimization plan for ML375 to improve rat potency and physiochemical/disposition properties. Potency values determined via a functional calcium mobilization assay in the presence of a fixed acetylcholine EC₈₀ in recombinant cells.²⁰

The synthesis of novel analogsof ML375 required a simple two-step synthesis involving condensation of ethylene diamine and an appropriately substituted 2-benzoylbenzoic acid 3 (or heteroaromatic/cyclo(hetero)alkyl congener) to provide 4, followed by a subsequent acylation reaction (Scheme 1) to deliver ML375 analogs 5–7.^20,21 However, we quickly exhausted the commercial analogs of 3. Fortunately, we were able to employ three synthetic routes to access key intermediates 3 with either diverse substituents or encompassing heterocycles.

Scheme 1.

Reagents and conditions: (a) ethylene diamine, p-TSA, toluene (+1,4-dioxane), reflux, Dean-Stark trap, or microwave irradiation 130–150 °C 4–77%; (b) Ar(Het)COCl, CH₂Cl₂, DIPEA, 16–91%; (c) RMgX, THF, −65 °C to 0 °C or rt, 8–68%; (d) R-H, AlCl₃, PhNO₂, rt, 41–88%; (e) i. RCOCl, cat. Ni(acac)₂, THF, rt, ii. Aq. NaOH, EtOH/THF, rt, 31–55%.

In the first round of library synthesis, we held the p-Cl 9b phenyl moiety of ML375 constant, and scanned alternate amides within a racemic core to provide analogs 5. Here, (Table 1) we found that heterocycles were generally not tolerated (5i-l) in the context of the p-Cl 9b phenyl core, but two amide congeners, the 4-isoproxyphenyl (5f) and the 3,4,5-trifluorophenyl (5g), displayed submicromolar human M₅ activity (hM₅ IC₅₀s of 790 nM and 610 nM, respectively), yet were less potent than racemic ML375 (5a, hM₅ IC₅₀ = 480 nM). Within a conserved series of ethers, hM₅ potency, e.g, M₅ IC₅₀s, was enhanced as steric bulk increased”- OMe (5d)<OEt (5e)<Oi-Pr (5f). Despite the lower hM₅ potency, 5d displayed a moderate improvement in clogP relative to 5a (4.6 versus 5.2), so we elected to evaluate how diminished lipophilicity would impact plasma protein binding. While racemic 5a displayed high plasma protein binding (rat f_u = 0.031, human f_u = 0.015), binding of 5f was slightly decreased rat f_u = 0.037, human f_u = 0.027). These findings then led us to pursue second generation libraries where we aimed to incorporate polar, basic and sp³-hybridized ring systems into the 9b position, while holding the 3,4-difluorobenzoyl moietyconstant, to assess if we could improve both physiochemical properties as well as hM₅ potency.

Table 1.

Structures and activities of analogs 5.

Entry	Ar(Het)	hM5 IC50 (µM)	hM5 pIC50a	Entry	Ar(Het)	hM5 IC50 (µM)	hM5 pIC50a
5a		0.48	6.32±0.03	5g		0.61	6.21±0.02
5b		>10	>5	5h		>10	>5
5c		6.6	5.18±0.16	5i		>10	>5
5d		1.8	5.74±0.08	5j		5.8	5.23±0.09
5e		0.85	6.07±0.05	5k		>10	>5
5f		0.79	6.10±0.03	5l		>10	>5

^ahM₅ pIC₅₀ reported as average ±SEM from our calcium mobilization assay; n = 3.

Following scheme 1, analogs 6 were rapidly prepared and screened against hM₅ (Table 2). Once again, SAR was shallow, with all sp³-based systems, as well as heterocycles, devoid of hM₅ activity. A similar pattern emerged for ethers between analog series 5 and 6, with 6c, the 4-OMe phenyl analog, superior potency (hM₅ IC₅₀ = 1.3 µM), but insufficient to advance as an in vivo probe. As before, 6c possessed a lower clogP (4.21), which translated into improved plasma free fraction (rat f_u = 0.064, human f_u = 0.037). Interestingly, the addition of more sp³-character, in the form of the cyclohexyl congener 6e, led to a higher clogP (5.2) and diminished free fraction (rat f_u = 0.016, human f_u = 0.008). In parallel, we replaced the phenyl ring at the 9b position with the three regioisomeric pyridines, and all were not tolerated (hM₅ IC₅₀ >5 µM), as were ring expansions and substitutions of the 1H-imidazo[2,1-a]isoindol-5(9bH)-one core.

Table 2.

Structures and activities of analogs 6.

Entry	R	hM5 IC50 (µM)	hM5 pIC50a	Entry	R	hM5 IC50 (µM)	hM5 pIC50a
6a		5.5	5.25±0.07	6g		>10	>5
6b		1.6	5.79±0.04	6h		>10	>5
6c		1.3	5.88±0.01	6i		>10	>5
6d		>10	>5	6j		>10	>5
6e		2.1	5.67±0.05	6k		>10	>5
6f		2.4	5.61±0.06	6l		>10	>5

^ahM₅ pIC₅₀ reported as average ±SEM from our calcium mobilization assay; n = 3.

At a loss for a rational, singleton approach to build-in hM₅ potency and improved physiochemical properties, we elected to pursue a 3 × 9 matrix library of analogs 7 to systematically evaluate all the possible combinations of monomers that showed either hM₅ potency enhancement or improved physiochemical properties (Table 3).^22,23 While we have generated, on numerous occasions, robust, tractable SAR within GPCR allosteric ligand chemotypes, we have also reported on numerous accounts of chemotypes that possess shallow or flat SAR,^8–11,24 and this matrix library is an example of the latter. Here, the clear stand-out was racemic 7B-6 (also referred to as VU0652483, hM₅ IC₅₀ = 517 nM, pIC₅₀ = 6.29±0.02), possessing a 3,4,5-trifluorobenzoyl amide and a 3-methyl-4-methoxy phenyl moiety in the 9b position. Due to the increased hM₅ potency, we evaluated VU0652483 potency at rat M₅ and found submicromolar activity (rat M₅ IC₅₀ = 963 nM, pIC₅₀ = 6.02±0.04) as well. As all of the activity of ML375 resided in the (S)-enantiomer, we resolved the enantiomers of VU0652483 via chiral SFC to afford (S)-7B-6 (VU6000181) and (R)-7B-6 (VU6000180); here again, the (R)-enantiomer was inactive (Fig. 2) and the (S)-enantiomer, VU6000181, possessed all of the M₅ M₅ NAM reported to date and maintaining selectivity versus M₁–M₄ (IC₅₀s > 30 µM).²⁰ Moreover, the clogP for VU6000181 (4.6) was improved over ML375 (5.2), and this once again translated into a slight improvement over ML375 (rat f_u = 0.031, human f_u = 0.013, rat brain f_u = 0.006). In addition, VU6000181 was highly centrally penetrant (brain:plasma K_p = 2.7 at 0.25 ht post-administration), yet a high clearance compound in vitro (rat hepatic microsome CL_INT = 332 mL/min/kg, predicted CL_HEP = 57.8 mL/min/kg and human hepatic microsome CL_INT = 359 mL/min/kg, predicted CL_HEP = 19.8 mL/min/kg) and in vivo (rat CL_p = 80 mL/min/kg, t_1/2 = 65 min, V_ss = 4.9 L/kg). The PK profile of VU6000181 rendered it unsuitable as an in vivo probe.

Table 3.

Structures and activities of matrix library analogs 7.

				Ar1
Entry	hM5IC50 (µM)a	hM5IC50 (µM)a	hM5IC50 (µM)a
1	3.2	>30	4.1
2	>30	2.4	1.9
3	1.3	1.0	1.1
4	3.3	>30	3.2
5	5.5	1.4	3.3
6	1.2	0.51	1.3
7	2.7	0.7	1.3
8	2.0	1.6	6.4
9	1.3	0.9	2.2

^ahM₅ IC₅₀ data, n = 1; hM₅ IC₅₀ for 7B-6, n = 6, pIC₅₀ = 6.29±0.02

Figure 2.

Structures and M₅ functional assay concentration-response-curves (CRCs) obtained in the presence of a fixed acetylcholine EC₈₀ in recombinant cells. A) rat M₅ CRCs of racemic VU0652483, the active (S)-enantiomer, VU6000181 and the inactive (R)-enantiomer, VU6000180; B) human M₅ CRCs of racemic VU0652483, the active (S)-enantiomer, VU6000181 and the inactive (R)-enantiomer, VU6000180; C) structures of VU0652483, VU6000181, VU6000180.

Previously, we productively utilized deuterium incorporation to overcome high clearance coupled with flat SAR in a series of mGlu₃ NAMs,²⁵ reducing rat in vitro and in vivo clearance by ~50% and affording an in vivo probe. Replacing the OCH₃ moiety of VU6000181 with a OCD₃ moiety (10, VU6001005 did in fact reduce human in vitro intrinsic clearance ~5-fold (human CL = 72.9 mL/min/kg) but provided negligible improvement to rat intrinsic clearance, and thus, was not a productive path towards an in vivo tool compound (Fig. 3).

Figure 3.

The impact of deuterium incorportion into VU6000181 on human, but not rat, intrinsic clearance determined in hepatic microsomes fortified with NADPH (values represent means from one independent determination performed in triplicate).

Finally, we performed an in vivo PK study by predosing rats with the pan-P450 inhibitor 1-aminobenzotriazole (ABT) in an attempt to potentially achieve therapeutically relevant drug levels and provide target validation for M₅ NAMs.²⁶ In this instance, we first pre-treated rats with an oral vehicle (10% tween 80 in 0.5% MC in water), followed 1.5 hours later by VU6000181 at 10 mg/kg, P.O. in 30% HPBCD in water (10 mg/mL). This control protocol revealed a VU6000181 mean residence time (MRT) of 6.5 hours, a T_max of 2.8 hours, an AUC_0-inf of 3,600 (hr*ng/mL) and low oral bioavailabiltiy (%F = 2.0). Pre-treatment of rats with a 56.6. mg/kg dose of ABT P.O. (10% tween 80 in 0.5% MC in water), followed 1.5 hours later by VU6000181 at 10 mg/kg, P.O. in 30% HPBCD in water (10 mg/mL), provided a ~5-fold increase in exposure (AUC_0-inf 17,700 (hr*ng/mL), %F = 9.5) relative to that in the vehicle pre-treated animals(Fig. 4). In addition, ABT pre-treatment afforded a ~3-fold increase in C_max (plasma = 2.0 µM, ~ 62 nM unbound) and, based on the a brain:plasma K_p (2.7), projected C_max levels in brain in the presence of ABT would be ~ 5.4 µM (~ 32 nM unbound). Thus, ABT pretreatment may serve as a viable strategy to increase levels of VU6000181 for future fMRI and addiction studies to test selective M₅ inhibition hypotheses.

Figure 4.

Oral plasma pharmacokinetics of VU6000181 in the absence (blue) or presence (red) of ABT pre-treatment in rats (male, Sprague-Dawley). ABT pre-treatment improved exposure ~ 5-fold and maximum concentrations ~ 3-fold.

In summary, we reported on the continued optimization of the MLPCN probe ML375, a highly selective M₅ NAM, through a combination of matrix libraries and iterative parallel synthesis. While SAR can be highly tractable for certain allosteric ligands, this chemotype was a clear example of ‘flat’ or ‘shallow’ SAR, wherein an matrix library was critical in identifying a productive lead compound, VU6000181, with improved activity at rat M₅ and disposition relative to ML375. Attempts to improve PK by deuterium incorporation substantially impacted only human in vitro CL_INT, but ABT pre-treatment significantly increased exposure in rats, potentially affording a strategy to achieve target validation for selective M₅ inhibition. Efforts continue, and work is in progress to pharmacologically probe M₅ neurobiology and therapeutic potential with small molecules.