Challenges in the development of an M₄ PAM in vivo tool compound: The discovery of VU0467154 and unexpected DMPK profiles of close analogs

Abstract

This letter describes the chemical optimization of a novel series of M₄ positive allosteric modulators (PAMs) based on a 5-amino-thieno[2,3-c]pyridazine core, developed via iterative parallel synthesis, and culminating in the highly utilized rodent in vivo tool compound, VU0467154 (5). This is the first report of the optimization campaign (SAR and DMPK profiling) that led to the discovery of VU0467154, and details all of the challenges faced in allosteric modulator programs (steep SAR, species differences in PAM pharmacology and subtle structural changes affecting CNS penetration).

Graphical abstract

M₄ (muscarinic acetylcholine receptor subtype 4) positive allosteric modulators (PAMs) represent an exciting therapeutic strategy to treat multiple domains of schizophrenia,^1-18 as well as other CNS disorders,^19,20 via a new molecular mechanism.²¹ However, the great potenital has been hampered by limited chemical diversity centered on a 3-amino-thieno[2,3-b]pyridine core, as in 1-4, (Figure 1), which engenders steep SAR, species differences (rat versus human M₄ PAM potency, affinity/cooperativity and subtype selectivity), poor solubility, and/or low CNS penetration.^1-18 Early in vivo tool compounds, such as 2,^1,2 energized the field but were not optimal. Recently, we disclosed VU0467154 (5) based on a novel 5-amino-thieno[2,3-c]pyridazine core that has proven to be a valuable rodent in vivo M₄ PAM tool compound with robust efficacy in a broad range of preclinical mouse and rat models of psychosis and cognition, as well as Huntington's disease.^11,17,18 Here, we describe for the first time the optimization campaign (SAR and intriguing DMPK profiles) that advanced PAMs 2-4, into 5 with exceptional rodent M₄ PAM potency, selectivity, DMPK profile and CNS penetration; however, a significant species disconnect (35x less potent on human M₄) precluded advancement as a clinical candidate.^11,17,18

Figure 1.

Structures of representative M₄ PAMs 1-4, highlighting the conserved 3-amino-thieno[2,3-b]pyridine chemotype, and the optimized rodent in vivo tool M₄ PAM, VU0467154 (5) with a novel pyridazine core.

Within the 3-amino-thieno[2,3-b]pyridine series, balancing M₄ PAM potency and solubility was a distinct challenge;^1-18 therefore, initial efforts focused on replacements for the pyridine ring, as very few substituents were tolerated. Regioisomeric pyridines were evaluated, along with isomeric pyrimidines, but all lost considerable M₄ PAM potency at both human and rat M₄. Based on the attractive physiochemical properties and high dielectric constant of the pyridazine ring, we prepared a 5-amino-thieno[2,3-c]pyridazine congener 6 of 2 (Figure 2). This modification proved favorable (M₄ PAM potency enhanced 4-fold, free fraction increased up to 9-fold, CL_hep improved as well as K_p,uu and cLogP with no change in molecular weight), but the PMB amide proved to be a metabolic ‘hot spot’ (CYP-mediated oxidative demethylation) and would need to be replaced with an alternate amide.²² At this point, multiple avenues of inquiry were pursued, and here we will focus on amide moieties containing sulfur (thioethers, sulfoxides, sulfones and SF₅) moieties.

Figure 2.

Structures of the first in vivo M4 PAM tool compound 2, and the new 5-amino-thieno[2,3-c]pyridazine congener 6, which displayed improved physiochemical, pharmacological and DMPK properties.

The synthesis of analogs 9 proved to be straightforward. Condensation of commercially available 3-chloro-5,6-dimethylpyridazine-4-carbonitrile 7 with methyl thioglycolateunder basic conditions smoothly affords the sodiumcarboxylate 8 in 78% yield.¹¹ A HATU-mediated amide coupling with various benzyl amines then delivered analogs 9in yields ranging from 45-92%. In all cases, either the desired benzylamine or the corresponding nitrile, which was easily reduced to the requisite benzyl amine, was commercially available.

SAR was driven on rat M₄, as the objective was an in vivo POC tool compound, but key compounds were assessed on human M₄ as well. As shown in Table 1, many potent rat M₄ PAMs were discovered, and several displayed excellent CNS pentration (K_p and K_p,uu). The direct thioether analog 9a of 6 displayed an ∼10-fold increase in rat M₄ PAM potency (EC₅₀= 11.2 nM), but diminished CNS exposure (K_p = 0.13). Oxidation to the corresponding sulfoxide 9b lost potency(EC₅₀ = 139 nM), which was restored by the methyl sulfone9c (EC₅₀ = 51.3 nM; however, CNS penetration was poor, K_p= 0.04). Moving the sulfone from the 4-position to the 3-position, as in 9d, was poorly tolerated (EC₅₀ = 417 nM), while steric bulk, as in 9e-g, retained good rat M₄ PAM potency (EC₅₀s 22-48 nM), but with low, variable K_ps (0.05 to 0.11). Thus, steric bulk alone was not sufficient to balance PAM potency and K_p, therefore increased lipophilicity, in the form of fluorine atoms, was then evaluated. Addition of a single fluorine atom alpha to the methyl sulfone (9h) maintained PAM potency (EC₅₀ = 58 nM), and CNS penetration improved (K_p = 0.51, K_p,uu = 2.78). Consecutive addition of fluorines to the methyl moiety of 9c afforded various fluoromethyl sulfones 9i-j with potent rat M₄ PAM activity (EC₅₀s 17 to 23 nM), but unexpected and highly variable CNS exposure. The mono-fluoromethyl analog 9i was not detected in the CNS, while brain distribution of the difluoromethyl congener 9j was modest (K_p = 0.11, K_p,uu = 0.10) and that of the trifluoromethyl derivative 5 was exceptional (K_p = 0.49, K_p,uu = 1.1). Finally, an unusual pentaflurosulfur (SF₅) analog, 9l, was a potent rat M₄ PAM (EC₅₀ = 30 nM) with excellent CNS distribution (K_p = 1.4, K_p,uu = 1.0). Of these, 5 and 9l stood out as potential candidates as rodent in vivo tool compounds, but we wanted to assess their activity at human M₄ (as species differences are common amongst M₄ PAM ligands) to determine if these could translate into clinical candidates.^8-11,13,14 Unfortunately, as shown in Table 2, there was a 6-to 35-fold rightward shift in human PAM potency, precluding these analogs from consideration as clinical candidates.

Table 1.

Structures and activities for rat M₄ PAM 5 and analogs 9.

Cpd	Ar	rM4EC50 (nM)a[% ACh Max ±SEM]	rM4 pEC50 (±SEM)	Rat Kp(Kp,uu)b
9a		11.2 [75.6±1.7]	7.95±0.09	0.13 (0.48)
9b		139 [74.2±1.9]	6.86±0.04	ND
9c		51.3 [72.6±2.2]	7.29±0.08	0.04 (0.14)
9d		417 [74.6±2.1]	6.38±0.17	ND
9e		47.9 [76.8±2.7]	7.32±0.06	0.07 (0.67)
9f		22.4 [67.1±3.1]	7.65±0.05	0.05 (0.29)
9g		44.7 [72.1±2.2]	7.35±0.06	0.11 (0.34)
9h		57.7 [69.2±3.5]	7.24±0.10	0.51 (2.8)
9i		21.9 [62.5±5.5]	7.66±0.04	BLQ
9j		17.8 [65.2±2.4]	7.75±0.14	0.11 (0.10)
5		17.8 [68.1±1.6]	7.75±0.06	0.49 (1.1)c
9k		23.4 [76.7±2.6]	7.63±0.05	BLQ
9l		30.2 [75.7±1.6]	7.52±0.03	1.4 (1.0)

^aCalcium mobilization assays with rM₄/Gqi5-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 brain:plasma partition coefficients determined 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. ND = not determined. BLQ = below limit of quantitation.

^cmean values obtained from discrete studies using a 1.5 hr sample time point.^11,13,14

Table 2.

Structures and human activities for M₄ PAM analogs 9.

Cpd	Ar	hM4EC50 (nM)a [% ACh Max ±SEM]	hM4 pEC50 (±SEM)
9a		60.2 [71.9±1.7]	7.22±0.09
9c		363 [73.8±2.2]	6.44 ±0.08
9i		209 [62.5±5.5]	6.68±0.04
9j		251 [72.8±2.4]	6.60±0.14
5		631 [56.4±2.4]	6.20±0.06
9l		550 [66.1±1.6]	6.26±0.03

^aCalcium mobilization assays with hM₄/Gqi5-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.

Next, we assessed physiochemical properties and DMPK profiles of 9j, 5 and 9l in battery of in vitro and in vivo assays (Table 2). Molecular weights were all similar and both the difluoromethyl sulfone 9j and the trifluoromethyl sulfone 5 had similar TPSAs, but the SF₅ analog, 9l, had greatly reduced TPSA. In vitro clearance (predicted CL_hep for both human and rat) was similar across the three PAMs, but plasma protein binding did differentiate the PAMs. 9j, with a cLogP of 1.98, had the most favorable fraction unbound (f_u (rat, human) of 0.066 and 0.053) as well as fraction unbound in brain (f_u rat brain, 0.062). In contrast, the highly lipophilic SF₅ (cLogP = 5.13) congener 9l, displayed poor fraction unbound (f_u (rat, human) of 0.004 and 0.010). 5 represented the middle ground with moderate fractionun bound in plasma (f_u (rat, human) of 0.031 and 0.019) and in brain (f_u rat brain, 0.067) with a n optimal cLogP (2.49) predictive of good CNS exposure. The CYP₄₅₀ profiles were very clean, except for weak inhibition of 2D6 by 9j and 9l. In vivo rat PK (1 mg/kg i.v.) again distinguished 5, with low CL_p (7.8 mL/min/kg), a long half-life (t_1/2 = 5.7 hours), excellent oral bioavailability (61%F, from 3 mg/kg suspension dose) and with excellent brain distribution (K_p = 0.49 and K_p,uu = 1.1. Interestingly, and despite varying in vivo DMPK profiles, all three afforded robust reversal of amphetamine-induced hyperlocomotion (AHL) when evaluated in single-point oral dosing at either 10 mg/kg (9j, 49.3% and 5, 55.3%) or at 30 mg/kg (9l, 66.3%) – our standard pharmacodynamic assay for M₄ PAM optimization. While we have recently reported results from dose-response AHL (as well as MK-801-induced hyperlocomotion) studies with 5 in rat and mouse (K_p for mouse of 0.64), here we show full dose response AHL for 9j and 9l (Fig. 3), and, while not ideal tools, both 9j and 9l are efficacious in vivo. Data for 5 was previously reported, with 46% reversal at 10 mg/kg p.o, 53% reversal at 30 mg/kg p.o. and a minimum effective dose (MED) at 3 mg/kg p.o. (33% reversal).¹¹

Figure 3.

Reversal of amphetamine-induced hyperlocomotion with A) 9j (VU0468182), B) 9l (VU0469785). The M₄ PAMs were administered 30 minutes after habituation in the chamber, and either vehicle or 0.75 mg/kg amphetamine administered s.c. at t = 60 min. For each dose group, n = 7-8 rats.

To vet these analogs as potential rodent in vivo tool compounds, we next evaluated selectivity across the other muscarinic receptors (M_1-3,5) and broader ancillary pharmacology in Eurofin 's Lead Profiling panel (68 GPCRs, ion channels and transporter radioligand binding assay panel).²³ All three M₄ PAMs, 9j, 5 and 9l, had no activity (EC₅₀s >30 μM) at both rat and human M_1-3,5 and 5 displayed no ancillary pharmacology at any of the 68 targets in the Eurofin panel (no % inhibition >50%@10 μM). Both 9j and 9l showed a less clean profile, showing significant off-target activities at multiple targets (>50% inhibition@10 μM which were then confirmed in full CRC). 9j proved to be a ligand for hCav_{1 2} (5.7 μM), DAT (175 nM), SERT (600 nM) and 5-HT₃ (1.6 μM). Similarly, 9l proved to be a ligand for DAT (150 nM), SERT (520 nM), Ghrelin receptor (6.5 μM), BDZ (1.5 μM), GABA (2.1 μM) and 5-HT₃ (1.5 μM) Therefore, 5 emerged as the ideal rodent in vivo M₄ PAM tool compound with clean ancillary pharmacology and excellent rodent PK. A subsequent, broader ancillary pharmacology panel identified single off-target activity for 5, the human adenosine transporter, with an IC₅₀ of 240 nM. The full in vitro and in vivo pharmacological profile of 5 (VU0467154) has been described in detail.¹¹

In summary, we have detailed for the first time the progression from the prototypical 3-amino-thieno[2,3-b]pyridine core M₄ PAM core to the 5-amino-thieno[2,3-c]pyridazine core, which imparts compounds with improved potency, physiochemical properties and DMPK profiles. However, the sulfone series detailed here still suffers from significant species differences in M₄ PAM potency, in favor of rat, and unpredictable variations in CNS penetration based on fluorine content and/or lipophilicity. This effort led to the study of an unusual SF₅ congener with potent PAM activity and in vivo efficacy, but the high cLogP proved problematic. Importantly, this gave rise to the highly valuable rodent M₄ PAM in vivo tool compound 5 (VU0467154), with a balanced profile and demonstrated utility in target validation studies. Further optimization efforts en route to M₄ PAM clinical candidates, with equivalent human and rat M₄ PAM potencies within this series, will be reported in due course.

Scheme 1.

Synthesis of M₄ PAM analogs 9.^a

^aReagents and conditions: (a) Methyl thioglycolate, MeOH, 1M aq. NaOH, 150 °C, microwave, 30 min, 78%; (b) NH₂CH₂Ar, HATU, DMF, DIEPA, 2 h, 45-92%.

Table 3.

In vitro and in vivo DMPK properties of 9j, 5 and 9l.

Property	9j	5	9l
MW	426.4	444.4	438.4
cLogP	1.98	2.49	5.13
TPSA	114	114	79.8
In vitro PK parameters
CLINT (mL/min/kg), rat	48.1	47.7	95.9
CLHEP (mL/min/kg), rat	28.1	28.4	40.5
CLINT (mL/min/kg), human	15.6	19.8	52.8
CLHEP (mL/min/kg), human	9.34	10.2	15.0
Rat fuplasma	0.066	0.031	0.004
Human fuplasma	0.053	0.019	0.010
Rat fubrain	0.062	0.067	0.004
Cytochrome P450 (IC50, μM)
1A2	>30	>30	>30
2C9	>30	>30	>30
2D6	20.8	>30	23.9
3A4	>30	>30	>30
In vivo PK (Sprague-Dawley rats)
IV (1 mg/kg)
CLp (mL/min/kg)	16	7.8	12
t ½ (h)	2.4	5.7	3.6
Vss (L/kg)	2.9	3.1	3.1
Tissue Distribution (1 mg/kg, IP) Sample time (0.25 h)
Kp, brain	0.11	0.49	1.0
Kpuu, brain	0.10	1.1	0.75
% reversal in AHL (10 mg/kg, p.o.)	49.3	55.3	66.3