Further optimization of the M₁ PAM VU0453595: Discovery of novel heterobicyclic core motifs with improved CNS penetration

Abstract

This letter describes the continued chemical optimization of the VU0453595 series of M₁ positive allosteric modulators (PAMs). By surveying alternative 5,6- and 6,6-heterobicylic cores for the 6,7-dihydro-5H-pyrrolo[3,4-b]pyridine-5-one core of VU453595, we found new cores that engendered not only comparable or improved M₁ PAM potency, but significantly improved CNS distribution (K_ps 0.3 to 3.1). Moreover, this campaign provided fundamentally distinct M₁ PAM chemotypes, greatly expanding the available structural diversity for this valuable CNS target, devoid of hydrogen-bond donors.

Graphical Abstract

Positive allosteric modulators (PAMs) of the muscarinic acetylcholine receptor subtype 1 (M₁) have garnered a great deal of attention as a novel therapeutic approach for the treatment of the cognitive and negative symptom domains of schizophrenia, especially targeting NMDA receptor hypofunction.^1-8 Moreover, M₁ PAMs are also of interest in general cognition enhancement and Alzheimer’s disease.^1-4,9-12 Since we reported on the discovery of the first M₁ PAM, BQCA,^13,14 numerous M₁ PAMs have been reported in the primary and patent literature (most by scaffold hopping VU and Merck series) with many conserved moieties that consistently engender low CNS penetration (K_ps < 0.3).^15-25 Our latest entry into M₁ PAMs was the result of three distinct high-thoughput screening campaigns,²⁶ which resulted in novel indole-, azaindole- and isatin-based M₁ PAM scaffolds.^25,27-29 Of these, the isatin VU0119498 (1) was a unique PAM in that it potentiated all of the G_q-coupled mAChRs (M₁, M₃ and M₅) with equivalent potency and efficacy.²⁷ Subsequent optimization efforts identfied ‘molecular switches’ that gave rise to a series of highly selective M₅ PAMs, ^30-32 as well as ML137 (2), a highly selective M₁ PAM by virtue of the fluorophenyl pyrazole moiety.²⁸ Carbonyl deletion provided lactam 3, and surveying regioisomeric lactams afforded VU0451725 (4), with improved DMPK properties over 2 and 3.³³ Finally, an aza-scan identified the the 6,7-dihydro-5H-pyrrolo[3,4-b]pyridine-5-one core of VU0453595 (5), which proved a useful in vitro and in vivo tool, demonstrating efficacy in rodent models of pharmacologically-induced NMDA hypofunction.³³ In this Letter, we detail an optimization campaign surveying alternative 5,6- and 6,6-heterobicyclic cores, alternate moieties for the pyrazole, and walking additional fluorines around the central phenyl ring to ultimately provide multiple novel M₁ PAM scaffolds with comparable or improved rat M₁ PAM potencies and improved CNS distribution (K_ps 0.4 to 3.1).

The chemistry to access new analogs, if not commercially available, was straightforward (Scheme 1).³⁴ The fluorinated heterobiaryl tail moities were readily prepared in two steps as either a benzyl chloride 7 or a benzyl mesylate 8 from commercial benzyl alcohols 6. Various 5,6- and 6,6-heterobiaryl systems were then alkylated with either 7 or 8 to provide analogs 10. A subsequent Suzuki coupling installed the heterobiaryl motif, delivering analogs 11. Quinolinone and naphthyridinone analogs 11 of 9, were made in a single step from 12, and based on our previous work, cores such as 15 were also accessed in a simple three step procedure.³⁴

Scheme 1.

Reagents and conditions: (a) Ghosez’s reagent or SOCl₂, DCM, rt, 65-78%; (b) MsCl, Et N, DCM, 0 °C, 75-88%; (c) 7 or 8, Cs₂CO₃, MeCN, 70 °C, 52-80%; (d) Het-B(OH) , Pd(dppf)Cl , Cs₂CO₃, THF:H₂0 (10:1), μw 140 °C, 22-69%; (e) ethyl glyoxalate, 51-68%; (f) Br(CH₂)₂NHBoc, Cs₂CO₂, DMF, rt, 16h, 98%; (g) HCl, 1,4-dioxane, rt, 1.5 h, 90%; (h) Na₂CO₃, 1,4-dioxane/H₂0, rt 3 hr, 90%.

SAR was steep for the diverse analogs 11, with many compounds devoid of M₁ PAM activity on both human or rat M₁, or displaying species bias towards rat M₁ PAM activity. In general, the 2,6-difluoro analogs were active whereas mono-fluoro and des-fluoro phenyl congeners were inactive as M₁ PAMs. Representative SAR is shown in Table 1 for a subset of analogs 16, possessing an N-Me-indazole attached at the 4-position to the 2,6-difluorophenyl ring. While only rat M₁ data is shown, analogs 16 were uniformly 2- to 3-fold less potent on human M₁ (with many > 10 μM). Here, various 6,6-hetrobicyclic ring systems were comparably active (rM₁ EC₅₀s 4.3-4.9 μM) across quinazolin-4(3H)-ones (16a), pyrido[3,4-d]pyrimidin-4(3H)-ones (16b), quinoxalin-2(1H)-ones (16c) and naphthyridin-5(6H)-ones (16d). These analogs possessed favorable in vitro DMPK profiles (rat and human f_us of 0.01 to 0.04) and moderate predicted hepatic clearance (CL_heps of 40-44 mL/min/kg). However, they were superior to the lead 5 in terms of brain distribution (K_p), wherein 16a-d displayed K_ps (rat brain:plasma ratios) of 0.35 to 2.16, and when corrected for fraction unbound in plasma and brain homogenate binding, the K_puus ranged from 0.3 to 0.77 – a major advance in the context of M₁ PAMs. Notably, 16c (VU0478436) afforded a >6-fold increase in CNS penetration over 5. The 5,6-congener 16e (VU0486691),based on a dihydroimidazol[1,2-c]pyrimidin-5(3H)-one core, showed enhanced M₁ PAM potency (rM₁ EC₅₀ = 1.7 μM, 50% ACh Max), improved in vitro DMPK profile (rat and human f_us of 0.08 and 0.04, respectively and moderate rat predicted hepatic clearance (CL_hep = 40 mL/min/kg)). Moreover, 16e demonstrated a rat K_p of 0.35 and a K_puu of 0.3. This finding led us to explore additional 5,6-heterobicyclic cores.

Table 1.

Structures and activities of analogs 16.

^aCalcium mobilization assays with rM₁-CHO cells performed in the presence of an EC₂₀ fixed concentration of acetylcholine; values represent means from three (n=3) independent experiments performed in triplicate.

^bTotal and calculated unbound braimplasma partition coefficients determinec at 0.25 hr post-administration of an IV cassette dose (0.20-0.25 mg/kg) to male, SD rat (n=1), in conjunction with in vitro rat plasma protein and brain homogenate binding assay data.

SAR proved steep as additional 5,6-hetrobicyclic cores were prepared and evaluated, with the vast majority devoid of M₁ PAM activity. During this effort, it was also shown that regioisomeric N-Me indazoles had a profound effect on M₁ PAM activity (Fig. 2). Interestingly, the 4-positional isomer 17 was devoid of M₁ PAM activity, while in contrast, the 5-positional isomer 18 was a potent M₁ PAM (EC₅₀ = 1.7 μM, 50% ACh Max, pEC₅₀ = 5.76+0.02) with very attractive in vitro DMPK properties (rat and human f_us of 0.06 and 0.05, respectively, low rat hepatic clearance (CL_hep = 29 mL/min/kg) and a large free fraction in rat brain homogenate binding, f_u = 0.098). Moreover, 18 (VU0484061) possessed a rat brain:plasma ratio (K_p) of 0.40 and a K_puu of 0.70. Once again, mL/min/kg). However, they were superior to the lead 5 in terms and in comparison to the known M₁ PAMs with low K_ps and K_puus, both 16c and 18 truly stand out. It is important to point out that the majority of M₁ PAMs possess two or more hydrogen bond donors (typically a trans-2-hydroxy cylohexyl amide moiety) that likely engenders the poor CNS penentration due to P-gp efflux or low permeability.^13-25

Figure 2.

The impact of positional isomers of the N-Me indazole in the context of dihydropyrazolo[1,5-a]pyrazin-4-(5H)-ones 17 and 18.

Prior to leaving this unique sub-series of M₁ PAMs, one last library of analogs was prepared with more diverse tail pieces within the 16b and 16c 6,6-heterobicyclic cores. Again, SAR was steep, and few active M₁ PAM resulted. However, this last campaign afforded three M₁ PAMs 19-21 with diverse profiles (Fig. 3). Here, the pyrido[2,3-b]pyrazin-2(1H)-one 19 (VU0486384) was a potent and efficacious rat M₁ PAM (EC₅₀ = 2.8 μM, 80±1% ACh Max, pEC₅₀ = 5.56±0.12), with a favorable fraction unbound in plasma (rat and human f_us of 0.04), high predicted hepatic clearance (CL_hep = 60 mL/min/kg), yet excellent CNS pentration (K_p = 1.64 and K_puu = 1.1). The nature of the heterobiaryl moiety played a key role in analogous pyrido[3,4-d]pyrimidin-4-(3H)-one congeners 20 and 21. The isoquinoline analog 20 was a weak rat M₁ PAM (EC₅₀ = 6.1 μM, 42±3% ACh Max, pEC₅₀ = 5.21±0.11) with equivalent plasma fraction unbound (f_us of 0.03) for rat and human, and the best K_p to date for an M₁ PAM of 3.1 (and a K_puu of 2.7). In sharp contrast, the more basic N-Me benzimidazole congener 21 was of comparable potency (EC₅₀ = 5.3 μM, 65±4% ACh Max, pEC₅₀ = 5.28±0.14), good plasma fraction unbound (rat and human f_us of 0.04 and 0.06, respectively), but no detectable CNS penetration (brain levels below the level of quantitation, BLQ). These data show that subtle pK_a modulation can dramtically impact K_p.

Figure 3.

Additional M₁ PAMs 19-21 based on 6,6-heterobicyclic cores with a diverse range of pharmacological and DMPK properties.

Finally, the concept of divergent signal bias, mediated by stabilization of unique conformers of the GPCR by the allosteric ligand, has emerged, and in many instances is critical for avoiding adverse effect liabilites.^1-4,35-38 Thus, our lab surveys the propensity of new M₁ PAM ligands to display signal bias.^38-41 For VU0453595 (5), DiscoverRX assessed activities of the M₁ PAM against human M₁ in a calcium flux assay, as well on β-arrestin recruitment and internalization.³⁹ PAM 5 was shown to be a modest human M₁ PAM (EC₅₀ = 1.9 μM, 79% ACh Max) with no effect on receptor internalization (EC₅₀ >10 μM) and modest effect on β-arrestin recruitment (EC₅₀ = 2.6 μM, 57% Max). New M₁ PAM 16b was evaluated similarly, and was found to be a modest human M₁ PAM (EC₅₀ = 5.3 μM, 65% ACh Max) with no effect on receptor internalization (EC₅₀ >10 μM), yet a submicromolar effect on β-arrestin recruitment (EC₅₀ = 980 nM, 33% Max). At this point, the in vivo ramification of these profiles across signal transduction pathways are unclear, but we are tracking and noting differences between M₁ PAMs and plan to investigate more thoroughly once a collection of M₁ PAM ligands with diverse profiles (and comparable PK) are accumulated.

In summary, we report on the further optimization of the in vivo tool M₁ PAM, VU0453595 (5). A diverse array of 5,6- and 6,6-heterobicyclic cores were developed as novel M₁ PAMs with unprecedented levels of CNS penetration (K_ps 0.3 to 3.1 and K_puus of 0.3 to 2.7) and lacking the prototypical hydrogen-bond donor motifs. While these M₁ PAMs are too weak to advance as clinical candidates, the improved disposition of these new chemoptypes represent fundamentally new starting points for further chemical optimization. Additional refinements are in progress and will be reported in due course.

Figure 1.

Evolution of the development of the VU0119498 series of M₁ PAMs, culminating in VU0453595 (5), a moderately potent PAM (rM₁ EC₅₀ = 3.2 μM, 75% ACh Max and 3x less potent on hM₁) with modest CNS penetration (K_p = 0.3). In this work, we survey alternative 5,6- and 6,6-heterobicyclic cores and pyrrole replacements in an attempt to increase CNS penetration.