Lead optimization of the VU0486321 series of mGlu₁ PAMs. Part 1. SAR of modifications to the central aryl core

Abstract

This letter describes the lead optimization of the VU0486321 series of mGlu₁ positive allosteric modulators (PAMs). While first generation PAMs from Roche were reported in the late 1990s, little effort has focused on the development of mGlu₁ PAMs since. New genetic data linking loss-of-function mutant mGlu₁ receptors to schizophrenia, bipolar disorder and other neuropsychiatric disorders has rekindled interest in the target, but the ideal in vivo probe, e.g, with good PK, brain penetration and low plasma protein binding, for robust target validation has been lacking. Here we describe the first modifications to the central aryl core of the VU0486321 series, where robust SAR was noted. Moreover, structural variants were identified that imparted selectivity (up to >793-fold) versus mGlu₄.

Graphical abstract

Recently, deleterious non-synonymous single nucleotide polymorphisms (nsSNPS) in the GRM1 gene, which encodes metabotropic glutamate receptor subtype 1 (mGlu₁), were discovered that correlated with a higher incidence of neuropsychiatric disease, including schizophrenia and bipolar disorder.^1,2 Our lab then demonstrated that signalling downstream of these loss-of-function mutant receptors could be partially rescued with mGlu₁ PAMs;³ however, neither the first generation Roche mGlu₁ PAM (1, Ro 07-11401)⁴ nor our chemically distinct series (2, VU0483605),³ derived from an mGlu₄ PAM scaffold (3, VU0400195, ML182)⁵ via ‘molecular switches’⁶ (Fig. 1), had the requisite PK, CNS penetration and free fraction to serve as the ideal in vivo tool compounds for robust target validation and potential adverse effect liability assessment.^3,7 Subsequent SAR studies in our lab identified key substituents on the phthalimide moiety that engendered improved plasma stability, along with a critical 3-methyl furyl amide that provided 4 (VU0486321), a potent mGlu₁ PAM with moderate rat PK (CL_p = 13.3. mL/min/kg, t_1/2 = 54 min), good free fraction (human fu = 0.05, rat fu = 0.03) and excellent CNS penetration (K_p = 1.02).⁷ Despite this notable advance, we sought an in vivo proof-of-concept tool compound with a longer half-life and ideally greater selectivity versus mGlu₄ (4 is 35-fold selective). In this Letter, we describe, for the first time, SAR for modifications to the central aryl ring of 4, and the impact on mGlu₁ and mGlu₄ PAM activity with analogs 5.

Figure 1.

Structures of reported mGlu₁ PAMs 1–4, and the novel targeted analogs 5, to be accessed through iterative parallel synthesis.

In order to access analogs 5 and survey the SAR for the two regions depicted in blue in Figure 1, a general five step synthetic route was developed. As shown in Scheme 1, commercial, functionalized p-amino nitroarenes/heteroarenes 6 were bis-Boc protected to provide 7, followed by reduction of the nitro moiety to afford 8. Analogs 8 were then acylated with 3-methylfuran-2-carbonyl chloride to provide congeners 9. Finally, acid-mediated deprotection liberated the free aniline which was then condensed with various phthalic anhydrides to afford final analogs 5. The 3-methylfuranyl amide was held constant in this campaign, as it was the lone, standout amide from our earlier efforts,⁷ where SAR proved steep.

Scheme 1.

Reagents and conditions: (a) Boc₂O, DMAP, THF, r.t., 59–97%; (b) H₂, Pd/C, EtOH, rt, 94–99-%; (c) 3-methylfuran-2-carbonyl chloride, DIEA, DCM, r.t., 39–98%; (d) i. TFA, DCM, r.t. ii. phthalic anhydrides, AcOH, reflux, 53–94%.

The first core modification examined focused on an aryl moiety substituted with H, Me, Cl and F in the 2-position, relative to the phthalimide, analogs of 5 indicated as 10 (Table 1). All of the analogs 10 were potent mGlu₁ PAMs with EC₅₀s less than 500 nM, and most under 100 nM, with excellent efficacy (% Glu Max >90), a welcomed departure from the steep SAR that plagued the discovery of 4.⁷ Also, all compounds showed considerable agonism at concentrations above 10 µM, a feature previously observed with parent compound 4, with the exception of the unsubstituted phenyl ring and 2-trifluoromethyl analogs which had a pure PAM profile. Interestingly, SAR for mGlu₄ varied greatly with respect to mGlu₄ PAM potency (EC₅₀s from 260 nM to >10 µM) and efficacy (% Glu Max from 35 to 122). While lack of activity at mGlu₄ is key for an mGlu₁ PAM in vivo probe, a dual mGlu₁/mGlu₄ PAM is intriguing in an antipsychotic agent, as mGlu₄ PAM activity can alleviate catalepsy, a major adverse event with the standard treatment of care.⁸ PAM 10a, with both an un-functionalized central phenyl core and unsubstituted phthalimide moiety, proved to be not only potent (mGlu₁ EC₅₀ = 48 nM), but also >208-fold selective versus mGlu₄ (mGlu₄ EC₅₀ >10 µM). All other modifications to either the phenyl core or the phthalimide moiety engendered more significant activity at mGlu₄, but mGlu₁ was still preferred. Relative to 10a, the addition of substituents at either R¹ or R² provided potent mGlu₁ PAMs, but the SAR possessed little discernable texture with electron-donating, electron-withdrawing and/or more lipophilic moieties proving effective. One notable trend was that 2-CF₃ on the phenyl core was less effective, affording a diminution in mGlu₁ PAM potency relative to 10a, providing analogs such as 10o, with an mGlu₁ EC₅₀ of 530 nM.

Table 1.

Structures and activities for analogs 10.

Cpd	R1	R2	hmGlu1 EC50 (µM)a [% Glu Max ±SEM]	mGlu1 pEC50 (±SEM)	hmGlu4 EC50 (µM) [% Glu Maxb]	Fold selective versus mGlu4
10a	H	H	0.048 [82±6]	7.32±0.13	>10 [−]	>208
10b	Me	H	0.016 [108±6]	7.79±0.09	0.35 [62]	21.7
10c	Cl	H	0.029 [101±14]	7.54±0.10	0.483 [35]	16.7
10d	F	H	0.044 [96±4]	7.36±0.05	0.382 [34]	8.6
10e	H	F	0.022 [101±4]	7.65±0.14	0.265 [65]	11.9
10f	Me	F	0.026 [107±3]	7.58±0.10	0.287 [102]	11.0
10g	Cl	F	0.028 [100±5]	7.56±0.10	0.264 [93]	9.5
10h	F	F	0.035 [104±6]	7.46±0.12	0.674 [78]	19.3
10i	H	Me	0.089 [105±7]	7.05±0.10	0.885 [75]	9.9
10j	Me	Me	0.021 [100±7]	7.67±0.13	0.422 [122]	19.8
10k	Cl	Me	0.049 [117±10]	7.31±0.20	1.01 [127]	20.5
10m	F	Me	0.049 [95±3]	7.31±0.12	1.10 [86]	22.4
10n	H	CF3	0.123 [108±6]	6.91±0.11	6.66 [86]	54.1
10o	Me	CF3	0.530 [117±11]	6.27±0.16	5.76 [109]	10.8
10p	Cl	CF3	0.106 [105±4]	6.98±0.08	2.47 [120]	23.3
10q	F	CF3	0.155 [110±8]	6.81±0.11	2.87 [80]	18.5
10r	H	OMe	0.046 [96±3]	7.33±0.08	1.18 [50]	25.6
10s	Me	OMe	0.049 [94±6]	7.31±0.15	0.882 [107]	18
10t	Cl	OMe	0.018 [93±3]	7.74±0.14	0.647 [103]	36.1
10u	F	OMe	0.144 [103±6]	6.84±0.10	1.93 [73]	13.4

^aCalcium mobilization mGlu₁assays, values are average of three (n=3). independent experiments performed in triplicate.

^bGlu Max is expressed as % of PHCCC response, which is used as a normalization control as the % max values vary on a daily basis for the mGlu₄assay.

The lack of SAR texture with analogs 10 then led us to explore heterocyclic and bicyclic congeners of 5, indicated as analogs 11 (Table 2), wherein the phenyl core was replaced with either a pyridine, pyrimidine or naphthalene, while holding the amide constant and surveying substituents on the phthalimide. Here, the SAR possessed texture, with 2-pyridyl analogs 11a–d maintaining good mGlu₁ PAM potency (EC₅₀s 35 nM to 387 nM) and a range of selectivity versus mGlu₄ (~18- to 52-fold). Installation of a 2-pyrimidinyl core, analogs 11e–h, lost significant activity at mGlu₁ (EC₅₀s 200 nM to > 5 µM), and selectivity versus mGlu₄ was eroded. Indeed, both 11e and 11h, were equipotent mGlu₁ and mGlu₄ PAMs. Replacement of the phenyl core with a naphthyl bicyclic ring system, e.g., 11i–m, afforded potent mGlu₁ PAMs (EC₅₀s 16 nM to 23 nM), irrespective of the phthalimide substituent, with modest selectivity versus mGlu₄ (17.4- to 41-fold). Thus, from these two efforts, while mGlu₁ potency and efficacy were attractive, the overall profile and selectivity versus mGlu₄ were not appropriate for an in vivo proof of concept tool compound. However, the overall SAR developed with analog libraries 10 and 11 was insightful and overcame the steep SAR previously encountered for this series.

Table 2.

Structures and activities for analogs 11.

Cpd	R1	Het/ bicycle	hmGlu1 EC50(µM)a [% Glu Max ±SEM]	mGlu1 pEC50 (±SEM)	hmGlu4 EC50 (µM) [% Glu Maxb]	Fold Selective versus mGlu4
11a	H		0.387 [80±4]	6.41±0.11	>10 [−]	>25.8
11b	Me		0.095 [105±6]	7.02±0.12	5.00 [100]	52.7
11c	Cl		0.035 [91±8]	7.46±0.09	1.113 [44]	31.9
11d	F		0.198 [93±3]	6.70±0.08	3.55 [19]	17.9
11e	H		4.61 [89±4]	5.33±0.22	4.67 [23]	1.01
11f	Me		0.751 [111±5]	6.12±0.22	4.47 [104]	5.95
11g	Cl		0.208 [84±5]	6.68±0.09	2.53 [45]	12.2
11h	F		5.33 [99±6]	5.27±0.19	>10 [24]	>1.88
11i	H		0.020 [99±8]	7.69±0.14	0.353 [54]	17.4
11j	Me		0.016 [88±6]	7.80±0.10	0.429 [103]	27.3
11k	Cl		0.023 [92±8]	7.65±0.13	0.513 [102]	22.8
11m	F		0.020 [90±3]	7.71±0.09	0.809 [83]	41.1

^aCalcium mobilization mGlu₁assays, values are average of three (n=3) independent experiments performed in triplicate.

^bGlu Max is expressed as % of PHCCC response.

Based on these data, we elected to survey the impact of regioisomers of 10 and 11, a region of chemical space not previously explored, wherein the substituents were in the 3-position with respect to the phthalimide, e.g, analogs 12.. Analogs 12 were accessed as shown in Scheme 1, and these analogs afforded the greatest selectivity versus mGlu₄ to date (>793 fold), despite possessing an unsubstituted phthalimide moiety. Here, the fluoro analog, 12a, was a potent mGlu₁ PAM (EC₅₀ = 12.6 nM, 84% Glu Max) with >793-fold selectivity versus mGlu₄ (EC₅₀ > 10 µM), and 12b, the chloro congener, was similarly potent (mGlu₁ PAM EC₅₀ = 29.1 nM, 68% Glu Max) with >344-fold selectivity versus mGlu₄ (EC₅₀ > 10 µM). Also, these compounds act like pure PAMs and did not show any observable agonistic effect in mGlu₁ as well as no activity at mGlu₅. Interestingly, the CF₃ analog 12d was inactive, as were the regioisomeric pyridine and pyrimidine analogs 12f and 12g, further highlighting the non-obviousness and challenges in allosteric modulator SAR.

From these efforts, 12a and 12b emerged as mGlu₁ PAMs with the requisite potency and selectivity versus mGlu₄ to serve as in vivo proof-of-concept tools; thus, they were profiled in a battery of in vitro and in vivo Drug metabolism and pharmacokinetic (DMPK) assays⁹ to further assess their potential. In terms of in vitro predicted hepatic clearance measures, both 12a and 12b display moderate hepatic clearance in both human (CL_HEP = 9.65 mL/min/kg and 9.25 mL/min/kg, respectively) and rat (CL_HEP = 48.5 mL/min/kg and 48.7 mL/min/kg, respectively) hepatic microsomes. Unfortunately, both compounds suffered from high plasma protein binding (F_u <1% in human and rat) as well as high rat brain homogenate binding (F_u <1%). CNS penetration varied significantly between the two PAMs. In our standard rat plasma:brain level (PBL) cassette study (0.2 mg/kg, 10%EtOH:40% PEG 400:50% DMSO, 15 min time point), 12a displayed a K_p of 1.57, while 12b, the Cl-congener, showed very poor CNS penetration (K_p = 0.12). While CNS exposure of 12a was attractive, the low free fraction precluded it from further consideration or advancement as an in vivo mGlu₁ PAM tool compound.

In conclusion, our first exploration of central core SAR of the VU0486321 series of mGlu₁ positive allosteric modulators (PAMs) provided a combination of both steep and tractable SAR, unexpected findings and new avenues for optimization. Interestingly, analogs where the halogen on the central phenyl core was located adjacent to the 3-furly amide eliminated activity at mGlu₄, providing analogs such as 12a (>793-fold selective versus mGlu₄) and 12b (>344-fold selective versus mGlu₄). DMPK properties hindered these analogs’ utility as in vivo probes, but lessons learned will be applied to other scaffolds, and these results will be reported in due course.

Table 3.

Structures and activities for analogs 12.

Cpd	R1 or Het/	hmGlu1 EC50(µM)a [% Glu Max ±SEM]	mGlu1 pEC50 (±SEM)	hmGlu4 EC50 (µM) [% Glu Maxb]	Fold Selective versus mGlu4
12a	F	0.013 [84±2]	7.90±0.08	>10 [−]	>793
12b	Cl	0.029 [68±5]	7.54±0.09	>10 [−]	>344
12c	Me	0.015 [90±2]	7.82±0.09	0.609 [52]	9.8
12d	CF3	>10 [−]	>5.00	>10 [−]	-
12e	OMe	0.330 [80±6]	6.48±0.19	>10 [−]	>30.3
12f		>10 [−]	>5.00	>10 [−]	-
12g		>10 [−]	>5.00	>10 [−]	-

^aCalcium mobilization mGlu₁assays, values are average of three (n=3) independent experiments performed in triplicate.

^bGlu Max is expressed as % of PHCCC response.