Quinolizidinone Carboxylic Acids as CNS Penetrant, Selective M₁ Allosteric Muscarinic Receptor Modulators

Abstract

Positive allosteric modulation of the M₁ muscarinic receptor represents an approach to treat the cognitive decline in patients with Alzheimer’s disease. Replacement of a quinolone ring system in a quinolone carboxylic acid series of M₁ modulators with a quinolizidinone bearing a basic amine linkage led to a series of compounds with higher free fraction, enhanced CNS exposure, and improved efficacy in rodent in vivo models of cognition.

One of the pathologies of Alzheimer’s disease (AD) concerns the progressive degeneration of the neurons of the basal forebrain cholinergic system, contributing to the profound cognitive deficits manifest in AD.¹ Acetylcholinesterase inhibitors are currently the front line of treatment for AD symptoms and act by limiting the degradation of synaptic acetylcholine (ACh) levels to activate cholinergic receptors. Although these agents do offer cognitive benefit, the magnitude and duration of the effects are limited.

ACh activates both nicotinic (ligand-gated ion channels) and muscarinic (metabotropic) receptors, which belong to the class A family of G protein coupled receptors (GPCRs). There are five receptor subtypes in the muscarinic family (classed M₁ to M₅),^2,3 of which the M₁ receptor has been found to be highly expressed in the cortex and hippocampus areas that are critically involved in memory formation and are affected early in AD. As such, M₁ has been widely viewed as the subtype most likely modulating cognitive and memory effects in this region of the brain.⁴ As a result, a number of nonselective or partially selective M₁ agonists such as xanomeline have been evaluated clinically, exhibiting efficacy in improving cognitive performance in AD patients.^5,6 However, adverse events (gastrointestinal distress, salivation, sweating), thought to be mediated by activation of other peripheral muscarinic receptors, have prevented their utility.

Efforts to identify compounds highly selective for the M₁ receptor subtype have failed due to the highly conserved nature of the orthosteric acetylcholine binding site. One avenue to address this is to target allosteric binding sites on the M₁ receptor that may be less highly conserved and provide ample specificity for M₁ over the M₂−M₅ subtypes.⁷ As such, M₁ allosteric activators in the form of allosteric agonists⁸ or positive allosteric modulators⁹ could confer the high selectivity over their orthosteric counterparts.

In this regard, the quinolone carboxylic acid HTS lead 1(10,11) and related analogues¹²⁻¹⁴ have been reported as highly M₁ selective positive allosteric modulators (Figure 1). Compound 1 has modest M₁ receptor activity and brain penetration, high plasma protein binding, and low solubility in a neutral form. As a result, high plasma levels are required for in vivo efficacy. Attempts to address these issues via incorporation of basic amines extending off a biphenyl construct augmented some of these issues (Figure 1). However, this afforded compounds that were substrates for the human efflux pump P-glycoprotein (P-gp) leading to reduced CNS exposure.

Figure 1.

Accordingly, efforts were focused on the identification of an alternate scaffold for the quinolone carboxylic acid that would allow for improvement of these aforementioned issues.^15,16 Specifically, it was of interest to evaluate if a quinolizidinone ring system could replace the quinolone (Figure 2) and confer dissimilar properties. Ideally, this strategy would also allow for the incorporation of a basic amine at the benzylic position in the form of a piperidine or piperazine that could not be achieved in the quinolone context. Herein we report a series of novel brain penetrant quinolizidinone carboxylic acid M₁ positive allosteric modulators that address many of the concerns found in quinolone 1, with improved in vivo efficacy and pharmacokinetic properties suitable for further development.

Figure 2.

The chemistry employed to access target molecules examined in this communication is shown in Scheme 1. Lithiation of 2-picoline with n-butyllithium is followed by addition of ethoxymethylene diethylmalonate 3 and subsequent cyclization in refluxing o-xylene to provide quinolizidinone 4.¹⁷ Vilsmeier formylation at the 1-position afforded the requisite aldehyde 5. Addition of the appropriate Grignard reagents to 5, which were subsequently hydrolyzed using NaOH in THF led to acids 9c,d. Oxidation using IBX followed by hydrolysis yielded ketone 9b. Alternatively, deoxygentation of 6c,d produces 9c,d after ester hydrolysis. Intermediate 9d (R = 4-chlorophenyl) underwent Suzuki cross-coupling with the appropriate boronic acid to result in 9e−g. Lastly, reduction amination of 5 with the appropriate amines followed by saponification of the ethyl ester afforded target N-linked quinolizidinone carboxylic acids 10a−i.

Scheme 1.

Carbon linked homologues of the quinolone carboxylic were evaluated first as shown in Table 1. Modifications containing a hydroxyl group (9a) or ketone (9b) at the benzylic position were inactive. The benzyl analogue 9c gave an M₁ IP = 0.92 μM, while potency was enhanced with the p-chloro (9d; M₁ IP = 200 nM) and p-phenyl (9e; M₁ IP = 110 nM). However, all of these compounds were fully bound to plasma proteins. Efforts to enhance free fraction via incorporation of aryl pyridines¹² (9f,g) in lieu of the biphenyl yielded potent M₁ potentiators, but had negligible reduction on protein binding.

Table 1.

^aValues represent the numerical average of at least two experiments. Interassay variability was ±30% (IP, nM), unless otherwise noted.

Quinolizidinone carboxylic acids that are nitrogen linked at the benzylic position were evaluated as shown in Table 2. Having a simple piperidine (10a) in place gave an M₁ IP = 7.6 μM, but addition of a 4-phenyl group (10b) increased activity ∼6-fold. Ring sizes other than a 6-membered ring were not tolerated as highlighted by phenyl pyrrolidine 10c. Incorporation of a piperazine in lieu of the piperidine afforded 10d, which further improved M₁ potentiation and also gave very high rat and human free fractions (>25%). Ketopiperazine 10e gave similar M₁ potency compared to 10d, and led to further reductions in plasma protein binding. Attempts to incorporate a pyridine as exemplified by 10f led to reduction in M₁ functional activity. Substitution on the phenyl ring was tolerated, with the best results observed for groups at the para position. For example, bromide 10g was the most potent potentiator in this series with an M₁ IP = 132 nM and a reasonable free fraction in rat (∼4%). A trifluoromethyl group (10h) resulted in a similar free fraction, with a ∼2-fold loss in M₁ potency. A cyano group provided a reasonably potent potentiator (10i; M₁ IP = 520 nM), but had an improved plasma free fraction relative to 10g,h.

Table 2.

^aValues represent the numerical average of at least two experiments. Interassay variability was ±30% (IP, nM), unless otherwise noted.

As P-glycoprotein (P-gp) is considered the dominant efflux transporter at the blood brain barrier (BBB) responsible for the efflux of a number of xenobiotic substances from the CNS, select compounds from Table 2 were evaluated for their passive permeability properties (P_app) and susceptibility as P-gp substrates in order to further evaluate their CNS exposure potential. As can be seen from Table 3, most of the compounds possessed good passive permeability (P_app) and were not substrates for human (MDR1) or rat (MDR1a) P-gp with efflux ratios less than 2. A notable exception was ketopiperazine 10f, which had very poor P_app leading to very low brain and CSF levels despite having a very high rat free fraction (∼48%). Bromide 10g provided the highest plasma level and brain levels when dosed orally at 10 mpk. Trifluoromethyl piperazine 10h provided a similar CSF/U_plasma ratio compared to 10g, but with a moderately higher B/P ratio (∼2.7). The highest CSF levels were observed with nitrile 10i, which presented a CSF/U_plasma ratio of 0.42, albeit with a lower B/P ratio. The higher total brain levels of 10g−h could be attributed to the increased lipophilicity of the bromo and trifluoromethyl groups relative to the nitrile (log D = 1.4) moiety in 10i, which may lead to higher nonspecific brain tissue binding.¹⁸ In addition, this greater degree of lipophilicity may be responsible for off-target activity at the hERG cardiac channel for 10g (IC₅₀ = 2.1 μM) and 10h (IC₅₀ = 0.1 μM), compared to 10i (IC₅₀ = 7.4 μM). Based on the lower hERG affinity and high CSF levels, 10i was chosen for further evaluation.

Table 3. Permeability, P-gp, and Bioanalysis of Plasma, Brain, and CSF Levels in Rat for Selected Compounds.

				concn (nM)c
compds	Pappa	MDR1b	MDR1ab	plasma	brain	csf	brain/plasma	CSF/Uplasmad
10f	<5			2071	59	16	0.03	0.02
10g	16	0.1	0.6	3646	4581	70	1.3	0.32
10h	32	1.4	1.3	1138	3085	14	2.7	0.25
10i	22	0.7	1.9	1495	248	85	0.17	0.42

^aPassive permeability (10⁻⁶ cm/s).

^bMDR1 directional transport ratio (B to A)/(A to B). Values represent the average of three experiments, and interassay variability was ±20%.

^cSprague−Dawley rats. Oral dose 10 mg/kg in 0.5% methocel. Interanimal variability was less than 20% for all values.

^dDetermined using rat plasma protein binding from table 2.

The effect of 10i was evaluated on the affinity of acetylcholine for the M₁ receptor in a functional assay utilizing calcium mobilization as a readout. In CHO cells expressing the human M₁ receptor, increasing concentrations of compound 10i from 0.1 to 10 uM potentiate the effect of acetylcholine leading to a leftward shift in the acetylcholine M₁ dose−response curve (Figure 3). In the absence of acetylcholine, there was no direct activation (agonism) of M₁ by compound 10i up to the highest dose tested (10 uM). In addition, no agonist or antagonist activity is observed up to 100 μM on M₂, M₃, or M₄ receptors with or without acetylcholine present.

Figure 3.

Compound 10i potentiates the acetylcholine dose response curve for the M₁ receptor.

Potentiator 10i was evaluated in a mouse contextual fear conditioning model of memory (Figure 4). In this experiment, the nonselective muscarinic antagonist scopolamine was administered to block formation of the association of a novel environment with an aversive stimulus (a foot shock). Animals that learn to associate the novel context with the adverse stimulus will respond with a freezing response when they are brought back into the same context 24 h later. Quinolizidinone 10i was codosed at 3, 10, and 30 mpk (ip) with scopolamine on the training day, and produced no baseline (training day) effects. Mice codosed with 10i showed a reversal of scopolamine deficits in a dose dependent manner in terms of freezing behavior compared to vehicle treated animals. The maximal effect noted at 30 mpk corresponded to 7.4 μM plasma levels for 10i. By way of comparison, quinolone 1 required ∼33 μM for reversal of scopolamine deficit in this model.

Figure 4.

Compound 10i in the mouse model of fear conditioning.

The pharmacokinetic properties of 10i were evaluated in rat and dog (Table 4). Adequate bioavailability was observed in rat (F = 23%) with a good half-life (6.4 h) and relatively low clearance (10.8 mL/min/kg). The dog clearance was higher with a shorter half-life. Overall, 10i possessed suitable pharmacokinetic properties further highlighting the quinolizidinone as a suitable replacement for the quinolone carboxylic acid.

Table 4. Rat and Dog Pharmacokinetic Data for 10i^a.

rat			dog
F (%)	t1/2 (h)	Cl	F (%)	t1/2 (h)	Cl
23	6.4	10.8	33	2.5	13

^aF% oral bioavailability, half-life is represented in hours, Cl in mL/min/kg. Sprague−Dawley rats (n = 3). Oral dose =10 mg/kg in methocel, iv dose = 2 mg/kg in DMSO. Interanimal variability was less than 20%.

In summary, a quinolizidinone ring system has been identified as an advanced replacement for the quinolone core in a series of carboxylic acid M₁ positive allosteric modulators. The quinolizidinone scaffold allowed for incorporation of a basic amine in the form of a piperazine ring, which markedly improved the plasma free fraction. Quinolizidinone 10i exemplified this leading to a potentiator with markedly improved performance in the mouse contextual fear conditioning model. Moreover, 10i retains suitable pharmacokinetic and physicochemical properties for additional study to determine potential for clinical evaluation.

Supporting Information Available

Representative assay and experimental procedures and data for test compounds. This material is available free of charge via the Internet at http://pubs.acs.org.