Discovery of an Orally Bioavailable and Central Nervous System (CNS) Penetrant mGlu₇ Negative Allosteric Modulator (NAM) in Vivo Tool Compound: N-(2-(1H-1,2,4-triazol-1-yl)-5-(trifluoromethoxy)phenyl)-4-(cyclopropylmethoxy)-3-methoxybenzamide (VU6012962)

Abstract

Herein, we report the discovery of a new, orally bioavailable and CNS-penetrant metabotropic glutamate receptor 7 (mGlu₇) negative allosteric modulator (NAM) that achieves exposure in cerebral spinal fluid (CSF) 2.5× above the in vitro IC₅₀ at minimum effective doses (MEDs) of 3 mg/kg in preclinical anxiety models.

Graphical Abstract

■ INTRODUCTION

Metabotropic glutamate receptor subtype 7 (mGlu₇) is one of the family of eight mGlu receptors (mGlu_1–8). Human genetic and mouse data have associated mGlu₇ with anxiety, autism, ADHD, schizophrenia, epilepsy, depression, and Rett syndrome.^1–15 Early studies with first generation mGlu₇ negative allosteric modulators (NAMs), such as 1,^16,17 2,^16,17 3,¹⁸ and 4¹⁹ (Figure 1), have been reported to result in efficacy in models of anxiety but suffer from liabilities such as off target activity, the requirement of doses in some models that are quite high to induce efficacy, and context-dependent effects, which may result in unexpected pharmacology in native tissues.^16–20 These findings indicate that the pharmacology of mGlu₇ modulation is complex and suggest that the generation of additional tool compounds would be valuable to probe receptor function. Recently, we performed a high-throughput screening campaign to identify novel mGlu₇ NAM leads, and NAM 5, based on a new chemotype, was identified.²¹ While 5 was CNS penetrant in rodents, with total brain levels in excess of the in vitro IC₅₀, the estimated unbound brain levels based on a brain homogenate binding assay indicated that 5 would not achieve levels above the in vitro mGlu₇ IC₅₀ in terms of free brain concentrations. Despite this limitation, like 3, mGlu₇ NAM 5 demonstrated robust efficacy in native tissues in blocking high-frequency-stimulated long-term potentiation (LTP) in the hippocampus.²¹ Here, we disclose efforts aimed at the continued optimization of mGlu₇ NAM 5 and the development of a new, highly selective mGlu₇ NAM in vivo tool compound that achieves predicted unbound CNS levels in excess of the in vitro IC₅₀ and robust efficacy across multiple rodent anxiolytic models at low doses (3 mg/kg MEDs).

Figure 1.

Structures of reported mGlu₇ NAMs 1–5, yet the field still requires a robust in vivo tool compound for target validation studies.

■ RESULTS AND DISCUSSION

From our initial multidimensional optimization campaign around the mGlu₇ NAM, 5, the minimum pharmacophore was identified, as well as the key metabolic liability (Figure 2).²¹ SAR demonstrated that the only region tolerant of change was the 4-alkoxy moiety on the benzamide, and the amide carbonyl was the metabolic soft-spot, with high plasma clearance due to amide hydrolysis (not the expected O-dealkylation).²¹ Therefore, we elected to pursue two paths in parallel: (l) delete the carbonyl and survey analogs 6 for mGlu₇ NAM potency and PK and (2) explore a larger collection of diverse 4-ether analogs 7.

Figure 2.

Overview of mGlu₇ NAM 5 SAR and metabolic liabilities, suggesting a path forward toward a new, structurally distinct, in vivo rodent tool compound. The minimum/essential pharmacophore of 5 is in blue.

With large quantities of mGlu₇ NAMs 5 and 8 in hand, the concept of des-carbonyl, amine-linked congeners 6 could be most readily tested by reduction of amides with LiAlH₄ (Scheme 1) to afford 6a and 9, respectively.²² While 5 was a moderately potent mGlu₇ NAM (IC₅₀ = 760 nM, pIC₅₀ = 6.12 ± 0.04, 15 ± 3% L-AP4 min), both 6a (IC₅₀ = 5.9 μM, pIC₅₀ = 5.23 ± 0.12, 17 ± 3% L-AP4 min) and 9 (IC₅₀ = 7.9 μM, pIC₅₀ = 5.1 ± 0.10, 21 ± 3% L-AP4 min) lost considerable mGlu₇ NAM activity; however, these data warranted additional analogs to see if potency could be enhanced. To access homologated, more conformationally flexible analogs of 6a, the four-step route highlighted in Scheme 2 was employed, starting from commercial aldehydes 10.²² A Wittig reaction delivered 11 in high yield, followed by reduction and hydrolysis to give acid 12. A PyClU-mediated amide coupling with known 13 provided 14, and reduction of the amide carbonyl with LiAlH₄ provided homologated analogs 6. Related synthetic routes also enabled the incorporation of oxygen atoms in the homologated amine-linker (see Supporting Information).²²

Scheme 1. Synthesis of Amine-Linked Congeners 6a and 9^a.

^aReagents and conditions: (a) LiAlH₄ (3.5 equiv, 1.0 M in THF), THF, 0 °C, 2 h, 51–55%.

Scheme 2. Synthesis of Extended Amine-Linked Congeners 6^a.

^aReagents and conditions: (a) (carbethoxymethylene)triphenyl-phosphorane, CH₂Cl₂, rt, 16h, 90–94%; (b) (i) NaBH₄, AcOH, Pd/C, toluene, rt, 1 h, (ii) LiOH, THF:H₂O (1:1), rt, 3h, 86–90%; (c) 13, PyClU, DIEA, CH₂Cl₂, microwave 100 °C, 30 min, 38–53%; (d) LiAlH₄ (3.5 equiv, 1.0 M in THF), THF, 0 °C, 2 h, 45–60%.

Homologation of the amine linker did improve mGlu₇ NAM potency (Table 1) for representative examples to within less than 2-fold of 5, but disposition suffered. All analogs 6 displayed predicted hepatic clearances near hepatic blood flow in rats, suggesting the site of metabolism shifted in the absence of the amide bond.²² Moreover, this modification did not positively impact plasma protein binding (f_u = 0.01–0.03) or rat brain homogenate binding (f_u < 0.003). Related analogs of 6b–d, wherein the OCF₃ moiety was replaced with a chlorine atom, fared worse in terms of potency and disposition, and numerous analogs proved to be inactive (IC₅₀ > 10 μM). Thus, the pursuit of an in vivo tool compound refocused on analogs 7.

Table 1.

Structures and Rat mGlu₇ Activities of Analogs 6^a

compd	W	Y	mGlu7 IC50 (μM)a (pIC50 ± SEM)	% L-AP4 min ± SEM	rat CLhep (mL min−1 kg−1)
6a			5.9 (5.23 ± 0.12)	17.7 ± 3.3	ND
6b	CH2	H	1.6 (5.79 ± 0.08)	9.5 ± 1.4	68.1
6c	CH2	F	1.3 (5.88 ± 0.10)	10.2 ± 1.2	66.4
6d	O	F	1.2 (5.93 ± 0.09)	5.9 ± 0.6	66.4

^aCalcium mobilization assay with rat mGlu₇/G_α15/HEK cells performed in the presence of an EC₈₀ fixed concentration of L-AP4 (a more potent agonist at mGlu₇ compared to glutamate); values represent the mean from three (n = 3) independent experiments in triplicate. ND, not determined. CL_hep, predicted hepatic clearance.

To explore a broader range of 4-alkoxy analogs 7, commercial ester 15 was alkylated with various alkyl bromides under microwave conditions to affords analogs 16 in yields ranging from 79% to 96%.²² Ester hydrolysis smoothly afforded acid 17, which was subsequently coupled to aniline 13 under PyClU conditions to deliver final analogs 7 in good overall yields. Within this series, as before, SAR was steep (Table 2). As many mGlu allosteric ligands engage induced-fit pockets, SAR can be challenging, and the “right” fit may only be “found” by exploring libraries of analogs via an exercise in strategic serendipity. In the present case, lipophilic moieties such as trifluoroethyl (7a) and tert-butylmethyl (7b) were inactive, but isopropylmethyl (7c) restored mGlu₇ NAM activity, and cyclopropylmethyl (7d) proved optimal (IC₅₀ = 350 nM, pIC₅₀ = 6.46 ± 0.10, 12.6 ± 1.5% L-AP₄ min) and the most potent within this chemotype to date. Larger moieties, such as 7e–g, lost activity. Beyond an enhancement in mGlu₇ NAM potency, 7d also showed a significant improvement in predicted hepatic clearance (rat CL_hep = 15.9 mL min^–1 kg^–1), generating enthusiasm for the further profiling of 7d. While the route depicted in Scheme 3 was suitable for small-scale production of 7d, an improved route was required to support extensive DMPK and behavioral work. For large-scale production, a two-step route was developed starting from commercial aniline 18 and intermediate 16d (Scheme 4) via a PyClU-mediated coupling to provide 19. An Ullmann coupling reaction installed the requisite 1,2,4-triazole to deliver 7d in ~500 mg scale.²²

Table 2.

Structures and Rat mGlu₇ Activities of Analogs 7^a

Entry	R	mGlu7 IC50 (μM)a (pIC50±SEM)	%L-AP4 Min±SEM	Rat CLhep (mL/min/kg)
7a		>30000 (<4.5)	-----	ND
7b		>10000 (<5)	25.1±7.9	ND
7c		1.2 5.91±0.01	35.4±3.7	38.4
7d		0.35 6.46±0.10	12.6±1.S	15.9
7e		1.4 5.84±0.11	15.5±3.8	ND
7f		2.8 5.55±0.08	26.7±7.9	ND
7g		3.1 5.51±0.18	19.2±5.24	ND

^aCalcium mobilization assay with rat mGlu₇/G_α15/HEK cells performed in the presence of an EC₈₀ fixed concentration of L-AP4 (a more potent agonist at mGlu₇ compared to glutamate); values represent the mean from three (n = 3) independent experiments in triplicate. ND, not determined. CL_hep, predicted hepatic clearance.

Scheme 3. Synthesis of 4-Alkoxy Congeners 7^a.

^aReagents and conditions: (a) R-Br, K₂CO₃, MeCN, microwave, 170 °C, 1 h, 79–96%; (b) LiOH, THF:H₂O (1:1), rt, 3h, 88–97%; (c) 13, PyClU, DIEA, CH₂Cl₂, microwave 100 °C, 30 min, 38–53%.

Scheme 4. Scale-Up Route To Access 7d^a.

^aReagents and conditions: (a) PyClU, DIEA, CH₂Cl₂, microwave 100 °C, 30 min, 53%; (b) H-1,2,4-triazole, trans-N,N′-dimethylcyclohexane-1,2-diamine, K₃PO₄, CuI, DMF, 100 °C, 16 h, 56%.

NAM 7d possessed an acceptable molecular weight (448 g/mol), cLogP (3.8), and polar surface area (84.7 Ų). In addition to low predicted hepatic clearance in rat (CL_hep = 15.9 mL min^–1 kg^–1), 7d also showed moderate predicted hepatic clearance in mouse (CL_hep = 44.0 mL min^–1 kg^–1) and good free fraction in rat and mouse plasma (f_u = 0.028 (rat) and 0.026 (mouse)).²² In a rat plasma:brain level (PBl) IV cassette study,^22,23 NAM 7d displayed high brain penetration (rat K_p = 2.35 ([brain_tot] = 375 nM), K_p,_uu= 0.75 ([brain_unbound] = 3.4 nM), the latter diminished due to high rat brain homogenate binding (f_u = 0.009)). These data prompted an evaluation of discrete iv and po pharmacokinetics in rat, wherein 7d showed moderate clearance (CL_p = 31.5 mL min^–1 kg^–1), moderate volume (V_ss = 1.9 L/kg), and a short half-life (t_½= 40 min); however, at a po dose of 10 mg/kg, 7d displayed high oral bioavailability (F = 74.9%).²² Finally, NAM 7d was selective for mGlu₇ versus the other seven mGlu receptors (>10 μM versus mGlu_1–6,8) and largely devoid of ancillary pharmacology (compound activity at only one target, 5-HT_2B receptor, that was greater than 50% at 10 μM) in a Eurofins lead profiling panel of 68 GPCRs, ion channels, and transporters.²² On the basis of this profile, if in vivo efficacy was driven by total brain levels, 7d was suitable to advance, but we were aware that for a number of lipophilic allosteric GPCR ligands, the results from brain homogenate binding assays (and thus the estimate of unbound brain concentrations) are occasionally unreliable. For such compounds, drug levels in the cerebral spinal fluid (CSF) could be significantly higher and afford a better correlation with in vivo efficacy.^24,25 Thus, we performed a 30 mg/kg (ip) tissue distribution study in rat and assessed levels of 7d in plasma, brain, and CSF. Here, we noted a brain:plasma K_p of 1.24 (([plasma]_tot = 598 nM), [brain]_tot = 745 nM) and a K_p,uu of 0.38 ([plasma]_Unbound = 16.7 nM, [brain]_unbound = 6.4 nM); however, the CSF:plasma K_p was 2.15, with levels of 7d in CSF of 1.3 μM, or ~3.8-fold above the in vitro IC₅₀. As with recent mGlu₄ PAMs, assessing CSF exposure was critical, as opposed to simply estimating free brain levels (i.e., K_p,uu).^24,25

On the basis of the potency, efficacy, disposition, and high concentrations in CSF of 7d, we progressed NAM 7d as an in vivo rodent tool molecule. First generation mGlu₇ NAMs 3 and 4 have been previously evaluated in rodent models of anxiety but required relatively high doses to observe efficacy.^18,19 We first assessed the activity of 7d in an elevated zero maze (EZM) task in mice after intraperitoneal administration (Figure 3).²² 7d increased total time spent in the open arms at a dose of 3 mg/kg (Figure 3A). This minimum effective dose of 3 mg/kg is a 20- to 50-fold improvement in in vivo potency over the first generation tool compounds. It should be noted, however, that 10 mg/kg did cause a decrease in overall locomotion (Figure 3B).

Figure 3.

NAM 7d decreases anxiety in the elevated zero maze (EZM) assay in mice. (A) Intraperitoneal administration of 7d increases time spent in the open arms (MED = 3 mg/kg): one-way ANOVA (F[3,39] = 3.112, p = 0.0382), Bonferroni’s post hoc test, VEH vs 3 mg/kg, p < 0.05. (B). The 10 mg/kg dose decreased distance traveled: one-way ANOVA (F[3,39] = 4.613, p = 0.0079), Bonferroni’s post hoc test, 3 mg/kg vs 10 mg/kg, p < 0.05.

Having established a 3 mg/kg MED in the EZM assay, we then evaluated 7d in two other mouse anxiety models: the light/dark box and marble burying assay. The effects of 3 mg/kg 7d were compared to the selective serotonin reuptake inhibitor (SSRI) fluoxetine (FLX, Figure 4).²² Results from the light/dark box assay showed that administration of either 7d (3 mg/kg ip) or fluoxetine (15 mg/kg ip) increased total time spent in the light side of the chamber compare to vehicle (VEH) controls (Figure 4A). Similarly, both 7d and fluoxetine decreased the number of marbles buried in a mouse marble-burying assay, consistent with an anxiolytic effect, compared to vehicle-control conditions (Figure 4B). The observation of efficacy at the 3 mg/kg dose prompted us to perform a pharmacokinetic assessment at this dose and the 1 h time point used for treatment in mice. These studies revealed values of [plasma]_tot = 303 nM and [CSF] = 883 nM; this CSF level is 2.5× higher than the in vitro IC₅₀ of 350 nM. Taken together, mGlu₇ NAM 7d decreases anxiety responses in three distinct preclinical models and displays a low of MED of 3 mg/kg, highlighting improvements of this tool compound in the realms of potency, physiochemical properties, disposition, and unbound CSF/brain levels.

Figure 4.

NAM 7d is efficacious at 3 mg/kg in the light/dark box and marble burying assays and is comparable to fluoxetine. (A) Administration of 3 mg/kg 7d and 15 mg/kg fluoxetine (FLX) increased total time spent in the light side of the chamber compared to vehicle (VEH) controls. One-way ANOVA (F[2,20] = 6.160, p = 0.0092), Dunnett’s post hoc test, VEH vs 7d, p < 0.01; VEH vs FLX, p < 0.05. (B) Administration of 3 mg/kg of 7d reduced the number of marbles buried (one-way ANOVA (F[2,31] = 6.376, p = 0.0051)), Dunnett’s post hoc test, VEH vs 7d, p < 0.05; VEH vs FLX, p < 0.01. All data points were run with Grubb’s outlier test.

■ CONCLUSION

In summary, we have reported on the discovery of a new, structurally distinct mGlu₇ NAM in vivo tool compound, 7d (VU6012962), suitable for robust target validation studies. NAM 7d is potent, orally bioavailable, highly CNS penetrant and at modest doses achieves predicted unbound levels (CSF concentrations) ~4-fold above the in vitro IC₅₀. Moreover, NAM 7d is highly selective for mGlu₇ versus the other seven mGlu receptor subtypes and across large ancillary pharmacology panels. Like first generation mGlu₇ NAMs, 7d was efficacious in multiple preclinical models of anxiety but with a MED ~20- to 35-fold lower than earlier tool compounds. Further in vivo target validation studies with 7d are in progress and will be reported in due course.

■ EXPERIMENTAL SECTION

Chemistry.

All compounds were purified to 95% as determined by analytical LCMS (214 nm, 254 nm, and ELSD), ¹H and ¹³C NMR, and high-resolution MS.

N-(2-(1H-1,2,4-Triazol-1-yl)-5-(trifluoromethoxy)phenyl)-4-(cyclopropylmethoxy)-3-methoxybenzamide (7d) (VU6012962).

To a suspension of 19 (1.15 g, 2.50 mmol), 1H-1,2,4-triazole (173 mg, 2.50 mmol), potassium phosphate tribasic (1.34 g, 6.25 mmol), and copper(I) iodide (23.8 mg, 0.125 mmol) in DMF (10 mL) was added trans-N,N′-dimethylcyclohexane-1,2-diamine (39.4 μL, 0.250 mmol). The resulting suspension was degassed by vigorously bubbling argon through the mixture for 5 min. The reaction was then heated to 100 °C for 16 h, whereupon LCMS indicated complete consumption of starting material and formation of the desired product. The reaction was diluted with EtOAc and filtered over a pad of Celite. The combined organic material was washed with sat. NH₄Cl × 2, brine, dried over MgSO₄, filtered, concentrated, and purified via flash chromatography (Teledyne ISCO system, silica gel column, hexanes:EtOAc) to afford the desired product as a beige solid (785 mg, 70% yield). ¹H NMR (400 MHz, CDCl₃) δ 10.52 (bs, 1H), 8.75 (d, J = 2.2 Hz, 1H), 8.51 (s, 1H), 8.28 (s, 1H), 7.50 (d, J = 2.0 Hz, 1H), 7.43 (d, J = 8.8 Hz, 1H), 7.39 (dd, J = 8.4, 2.1 Hz, 1H), 7.07 (dd, J = 8.7, 1.7 Hz, 1H), 6.91 (d, J = 8.4 Hz, 1H), 3.95 (s, 3h), 3.92 (d, 2H), 1.39–1.32 (m, 1H), 0.69–0.65 (m, 2H), 0.40–0.36 (m, 2H); ¹³C NMR (100 MHz, CDCl₃) δ = 165.1, 153.1, 152.3, 149.7 (d, J_CF = 2.0 Hz), 149.6, 143.9, 133.7, 126.3, 123.8, 123.6, 120.5 (q, J_CF = 257.1 Hz), 120.0, 115.9, 115.5, 112.1, 111.0, 74.1, 56.2, 10.2, 3.6 ppm. HRMS (TOF, ES+) calcd for C₂₁H₁₉F₃N₄O₄, 448.1358; found, 448.1365.

N-(2-Bromo-5-(trifluoromethoxy)phenyl)-4-(cyclopropylmethoxy)-3-methoxybenzamide (19).

To a solution of aniline 18 (1.70 g, 6.64 mmol) in CH₂Cl₂ (15 mL) in a Biotage microwave vial were added 17d (1.48 g, 6.64 mmol), N,N-diisopropylethylamine (3.47 mL, 19.9 mmol), and chlorodipyrrolidinocarbenium hexafluoro-phosphate (PyClU) (2.21 g, 6.64 mmol) at room temperature. The vial was sealed and heated to 100 °C using a Biotage microwave reactor for 30 min, whereupon LCMS showed formation of the desired product. The reaction mixture was diluted with DCM and quenched with the addition of saturated NH4Cl. The layers were separated, and the aqueous layer was washed with DCM × 3. The combined organic layer was passed through a phase separator, concentrated, and purified via flash chromatography (Teledyne ISCO system, silica gel column, hexanes:EtOAc) to afford the desired product as a white solid (1.86 g, 61% yield). ¹H NMR (400 MHz, CDCl₃) δ 8.61 (d, J = 2.3 Hz, 1H), 8.47 (bs, 1H), 7.58 (d, J = 8.8 Hz, 1H), 7.54 (d, J = 2.1 Hz, 1h), 7.43 (dd, J = 8.4, 2.1 Hz, 1H), 6.93 (d, J = 8.4 Hz, 1H), 6.91–6.87 (m, 1H), 3.97 (s, 3H), 3.94 (d, 2H), 1.40–1.33 (m, 1H), 0.71–0.66 (m, 2H), 0.42–0.38 (m, 2H); ¹³C NMR (100 MHz, CDCI₃) δ = 165.0, 152.4, 149.9, 149.1, 137.3, 132.9, 126.6, 120.5 (q, J_CF = 256.6 Hz), 119.6, 117.1, 114.2, 112.2, 111.2, 110.7, 74.1, 56.3, 10.2, 3.6 ppm. HRMS (TOF, ES+) calcd for C₁₉H₁₇BrF₃NO₄, 459.0293; found 459.0296.

■ ABBREVIATIONS USED

mGlu₇

metabotropic glutamate receptor subtype 7

CRC

concentration–response curve

NAM

negative allosteric modulator

PBL

plasma:brain level

SAR

structure–activity–relationship

CSF

cerebral spinal fluid

EZM

elevated Z maze