Mitigating hERG Inhibition: Design of Orally Bioavailable CCR5 Antagonists as Potent Inhibitors of R5 HIV-1 Replication

Abstract

A series of CCR5 antagonists representing the thiophene-3-yl-methyl ureas were designed that met the pharmacological criteria for HIV-1 inhibition and mitigated a human ether-a-go-go related gene (hERG) inhibition liability. Reducing lipophilicity was the main design criteria used to identify compounds that did not inhibit the hERG channel, but subtle structural modifications were also important. Interestingly, within this series, compounds with low hERG inhibition prolonged the action potential duration (APD) in dog Purkinje fibers, suggesting a mixed effect on cardiac ion channels.

The marketing approvals of the first in class chemokine receptor antagonists, maraviroc¹ in 2007 targeting CCR5 and plerixafor^2,2b in 2008 targeting CXCR4, mark an important milestone for chemokine receptor drug development. Although both of these receptors were identified as important coreceptors required by the HIV-1 virus to infect cells,³ most of the research effort has focused on developing inhibitors of CCR5, since the most commonly transmitted HIV-1 strains (M-tropic) utilize the CCR5 coreceptor.

Recently, we reported⁴ on the identification of pyridin-2-ylmethylaminopiperidin-1-ylbutyl amides (1) as a novel class of CCR5 antagonists that exhibited potent inhibition of R5 HIV-1 (BaL) replication in PBMC (Figure 1). These compounds were identified based on our initial redesign efforts of Schering C.⁵ Although these compounds met most of the criteria for further development, they were found to have a significant hERG liability (also reported by the GSK⁶ and Pfizer⁷ groups), which we sought to mitigate by reducing the lipophilicity through the design of compounds 2 (Figure 1). The antiviral activity was used to determine a structure–activity relationship (SAR), and it was measured utilizing the R5 HIV-1 PBMC assay.⁸ In addition, we also utilized a gp120/CCR5 cell fusion assay⁹ as a first pass screen to select compounds for antiviral screening since the fusion data were found to be a reliable predictor of antiviral activity.

Figure 1.

Structure of compounds 1 and 2.

On the basis of the recognition that compound 1a containing an anilino nitrogen (R¹ = phenyl) had potent antiviral activity, we reasoned that the most expedient way to address the hERG liability was to replace the substituted phenyl group on the left had side with an amide or urea that would reduce the lipophilicity¹⁰ as measured by cLog P. Initially, the 4-methyl pyridine moiety on the top left-hand side was retained, but these compounds were substantially less potent than the corresponding thiophene analogues,⁴ so the thiophene moiety was utilized for further optimization (Table 1). For example, comparison of the tetrahydropyranyl esters 1b and 2b showed that the thiophene 2b provided a 44-fold enhancement in potency as compared to the 4-methyl pyridine 1b based on CCR5 fusion inhibition. Similarly, comparison of the pyridinyl ureas 1c and 2c showed a 130-fold increase in potency for 2c as compared to the 4-methyl pyridine 1c.

Table 1. Comparative in Vitro Pharmacology and hERG Inhibition of 4-Methyl Pyridine and Thiophene Compounds^a.

^aAssays were performed in duplicate, and values represent the mean with standard deviations <30% of the mean. Bracketed values represent the number of experiments.

^bViral gp120/CCR5 binding was assessed by a cell fusion assay between P4R5 cells expressing CCR5, CD4, and a LTR-β-gal construct (LacZ gene under the control of the HIV-1 LTR promoter) and CHO-tat cells expressing the HIV-1 JRFLenv and HIV-1 Tat. The P4R5/CHO-tat, CCR5/gp120 interaction was assessed by measuring the resultant β-galactosidase activity. IC₅₀ is the concentration of compound required to inhibit cell fusion by 50%.

^cFold AGP fusion shift is the shift in the CCR5 fusion assay in the presence of 1 mg/mL of α-acid glycoprotein.

^dPeripheral blood mononuclear cells (PBMC) were isolated from healthy volunteers and infected by R5 HIV-1 (BaL strain) as reported in ref (8). IC₅₀ is the concentration of the compound required to inhibit viral replication by 50% as measured by p-24 HIV-1-specific Ag ELISA.

^eCC₅₀ is the concentration required to reduce the viability of PBMC by 50%.

^fPlasma protein binding was calculated by diluting a 25 mM stock solution in DMSO to 1 μM in human plasma, dialyzing for 5 h against PBS (pH 7.4) at 37 °C, and determining the concentration remaining in each compartment by LC.

^ghERG inhibition was measured at 10 μM in CHO cells stably expressing the recombinant human hERG channel subunit (IKr) using an automated patch clamp platform (QPatch-Sophion Biosciences).

Several amide analogues of compound 2 were prepared (Table 2). Aside from maintaining antiviral potency, the important properties that were measured to distinguish the analogues was fusion inhibition shift in the presence of α-1-acid glycoprotein (AGP), human plasma protein binding, and hERG inhibition based on % inhibition and/or IC₅₀. The simple methyl amide 2d containing the optimum 2-chloro-4,6-dimethyl pyridine amide group⁴ demonstrated that activity could still be maintained in the absence of a hydrophobic aromatic group. Extending the amide using a methoxymethyl group 2e provided an approximate 4-fold enhancement in antiviral activity and a hERG IC₅₀ of 8.4 μM. The cyanomethyl moiety (2f) further reduced lipophilicity and resulted in a 2-fold increase in potency and a hERG IC₅₀ of 16.0 μM as compared to 2e. Other amide modifications included compounds 2b and 2g, h. Of these, the cyclic carbamate 2h had the best overall profile with hERG inhibition of 21% and an anti-HIV-1 activity of 1.6 nM. Interestingly, use of the cyano pyridine amide 2g provided a 1 Log reduction in cLog P as compared to 2b but only a marginal decrease in hERG inhibition (61 vs 67%), suggesting that discrete structural modifications also played an important role.¹¹ However, because many of these compounds showed evidence of conformational isomerism around the amide bond, the amide analogues were not further pursued.

Table 2. In Vitro Pharmacology and hERG Inhibition of Compounds 2^a.

^aSee Table 1.

As a consequence, we shifted our attention to preparing a series of urea analogues (Table 2). It was encouraging to observe that the methyl urea 2i was found to have good potency, although the hERG IC₅₀ of 1.6 μM (cLog P of 2.6) suggested that even in this series other factors were impacting hERG. Interestingly, the dimethyl analogue 2j was inactive, indicating the importance of NH in binding to CCR5. Incorporation of a hydroxyl urea 2k was tolerated based on the CCR5 fusion assay, but this did not correlate to anti-HIV-1 activity (IC₅₀ = 344 nM). However, the corresponding methoxy urea analogue 2l provided encouraging in vitro properties including a hERG IC₅₀ of 5.5 μM and an anti-HIV-1 activity of 14.8 nM. Additional analogues incorporating the meythoxy methyl urea moiety (2m–p) were prepared in which the right-hand side pyridyl amide was modified in an attempt to further improve the antiviral potency and hERG inhibition. An 8-fold increase in potency (IC₅₀ = 1.9 nM) was achieved using the 2,6-dicholoro-4-methyl pyridine amide 2m (cLog P = 2.6), but this was accompanied by a 20-fold increase in hERG inhibition (IC₅₀ = 0.3 μM). However, this liability was overcome by using the corresponding N-oxide 2o that displayed a hERG inhibition of 21% while still maintaining favorable in vitro properties. In addition, using either an iso-propyl amide 2n or a cyano group 2p in place of the 2-chloro substituent of the pyridyl amide improved the hERG inhibition with IC₅₀ values of >30 and 10.8 μM, respectively, and anti-HIV-1 potencies of 3–5 nM. The hERG liability could also be mitigated by modifications to the left-hand side urea moiety. For instance, the methoxyethoxy urea 2q hERG IC₅₀ was >40 μM. On the basis of a combined modeling mutagenesis approach¹² where it was postulated that addition of an appropriately tethered carboxylic acid to the urea would pick up an ionic interaction with Lys¹⁹¹, compound 2r was prepared in which antiviral potency was maintained at 6.9 nM, but more importantly, the hERG IC₅₀ was >50 μM.

The thiophene compounds (Tables 1 and 2) were prepared as outlined in Scheme 1 other than compound 2a.⁴ The substituted cyano pyridine 4a, used for the preparation of compounds 2g and 2p, was prepared from commercially available 2-chloro-4-methyl nicotinic acid by nucleophilic addition of the cyano group to the corresponding N-oxide 3a. The advanced intermediate 8, prepared by reductive amination of 3-aminomethyl thiophene 7(13) with ketone 6, was used to prepare the analogues 2a–r. Nitrile reduction of enantiomerically pure 5(4,14) under Raney-Ni-catalyzed hydrogenation conditions followed by Boc protection of the amine afforded 6 in good yield. Installation of the amide moiety was readily accomplished using standard peptide coupling procedures, whereas the urea moiety was installed using commercially available isocyanates or amines. Boc deprotection under mildly acidic conditions then afforded the compounds 2b–m and 2o–r in good overall yields. The iso-Pr amide 2n was prepared by base hydrolysis of the precursor nitrile followed by standard coupling procedures (Scheme 1).

Scheme 1.

The pharmacokinetic parameters of some compounds that had an acceptable hERG profile were evaluated in rat and/or dog (Table 3). Many of the ureas that were evaluated in dog PK had good oral bioavailability and exposure other than the N-oxide 2o. Compounds 2e and 2l had good oral bioavailability in rat and dog and were selected for further evaluation as were compounds 2n and 2q that had good dog PK.

Table 3. Pharmacokinetics of 2e, 2l, and 2n–p in Rat and Dog^a.

compd	species	Cmax (μM)	AUC0-inf (h μM)	CL (mL/min/kg)	V (L/kg)	T1/2 (h)	F (%)
2e	rat	2.54	5.59	85.0	18.6	2.5	29
2l		3.36	13.40	87.1	53.0	7.0	73
2e	dog	4.9	1.64	89.6	9.8	1.2	72
2l		3.3	5.89	52.6	23.9	5.2	151
2n		2.62	3.71	29.5	7.7	4.1	52
2o		0.26	0.37	40.3	2.6	0.7	7
2p		1.94	3.24	39.2	6.3	1.8	60

^aClearance (CL), volume of distribution (V), and half-life (T_1/2) calculated following a 10 μmol/kg iv dose in rat and 5 μmol/kg iv dose in dog. Oral bioavailability (F) calculated following solution doses of 100 μmol/kg in rat and 12.5 μmol/kg in dog.

The selectivity of 2e, 2l, 2n, 2q, and related analogues was evaluated in Ca²⁺ flux assays against a series of other closely related G-protein-coupled receptors (GPCRs), which included CCR1, CCR2b, CCR4, CXCR1, CXCR2, and CXCR4, and were found to be noninhibitory at concentrations of 5 μM. When tested against a panel of five isoforms of CYP 450, these compounds were found to be noninhibitory at concentrations >10 μM. Compounds 2e, 2l, and 2q progressed to a 7 day safety study in rat. The NOAEL for all three compounds was determined to be at ≥400 mg/kg (the highest test dose).

In addition to 7 day safety dog studies (no telemetry studies were conducted), these compounds and others were evaluated in the canine Purkinje fiber assay¹⁵ to further understand the risk for drug-induced arrhythmias (Table 4). Surprisingly, all of the thiophene compounds tested at 10 μM using the three standard stimulation frequencies resulted in significant prolongation of the APD in a reverse rate-dependent fashion, which is consistent with blocking the hERG potassium channel.¹⁶ However, this observation was not supported by the in vitro hERG inhibitory data. For example, compounds 2f, 2h, and 2q inhibited hERG at IC₅₀ values of 16 to >40 μM, but the % change in APD₆₀ at a basic cycle length (BCL) of 2 s was significant (18–43%). In addition, 2l and 2n also showed a significant % change in APD₆₀, 55 and 37%, respectively, with a hERG IC₅₀ of 5.5 and >30 μM. There was no occurrence of early after depolarizations (EADs), and as the prolongation of APD₆₀ and APD₉₀ was similar at every stimulation cycle, there was no evidence of triangulation of the action potential. Given this observed discrepancy between hERG channel inhibition and APD, we postulated that these compounds were likely having a mixed effect on cardiac ion channels. The lack of effect on the resting membrane potential (RMP) or the rate of depolarization (Vmax) suggests that sodium channels are not inhibited at the concentrations tested. In addition to hERG (IKr), there is significant contribution to phase III repolarization provided by the slowly activating delayed rectifier potassium channel (IKs).¹⁷ Inhibition of this channel provides a possible mechanism for the APD observed in this study; however, none of these compounds were specifically tested for inhibition of this channel.

Table 4. Dog Purkinje Fiber Data^a and hERG Data for Compounds 2f, 2h, 2l, 2n, and 2q.

		Δ%		ΔmV
compdb	BCL (s)	APD60	APD90	RMP	APA	Vmax (Δ%)	hERG IC50 (μM)
2f	2	42.9 ± 4.2*	47.6 ± 7.5*	–1.5 ± 1.3	4.3 ± 3.1	29.1 ± 9.5	16
	1	31.5 ± 1.9*	31.9 ± 4.6*	–1.9 ± 1.1	3.4 ± 2.3	21.1 ± 7.0
	0.5	15.7 ± 0.9*	17.7 ± 2.0*	–0.9 ± 1.6	2.9 ± 1.5	18.4 ± 5.2*
2h	2	18.4 ± 8.3	14.2 ± 2.1*	–1.8 ± 1.8	13.7 ± 8.2	57.1 ± 42.5	na
	1	14.5 ± 8.4	10.6 ± 2.1*	–2.2 ± 2.9	15.2 ± 7.9	81.7 ± 56.9
	0.5	7.7 ± 7.1	4.5 ± 1.4	–1.5 ± 3.2	11.3 ± 6.4	48.4 ± 27.4
2l	2	55.1 ± 4.5*	57.0 ± 6.6*	–0.6 ± 0.7	–3.6 ± 1.8	–8.3 ± 15.2	5.5
	1	42.6 ± 5.6*	42.2 ± 5.5*	–0.5 ± 0.9	–3.8 ± 2.3	–8.0 ± 15.6
	0.5	24.9 ± 3.0*	27.3 ± 3.3*	–1.3 ± 0.5	–2.5 ± 1.6	–7.4 ± 17.2
2n	2	37.1 ± 11.2*	34.2 ± 5.9*	–1.4 ± 0.5	1.2 ± 2.6	4.5 ± 13.1	33
	1	34.5 ± 10.5*	31.6 ± 5.5*	–1.9 ± 0.6	2.2 ± 2.1	7.8 ± 11.1
	0.5	24.8 ± 5.7*	24.2 ± 3.4*	–2.3 ± 0.8	2.6 ± 1.8	9.0 ± 11.5
2q	2	24.2 ± 4.7*	22.6 ± 3.7*	–2.7 ± 3.1	2.4 ± 2.7	4.0 ± 6.4	>40
	1	18.1 ± 2.7*	16.5 ± 2.7*	–2.8 ± 2.8	1.3 ± 2.3	–4.0 ± 5.4
	0.5	12.2 ± 3.1	10.5 ± 2.9*	–3.6 ± 3.2	1.2 ± 2.5	–4.1 ± 3.9

^aAction potential parameters are reported as means ± SDs percent change (Δ%) increase (+) or decrease (−) in test compound-exposed Purkinje fibers over those measured in time-matched vehicle control. APD₆₀ and APD₉₀ = action potential duration at 60 and 90% repolarization, respectively; RMP = resting membrane potential; APA = action potential amplitude; V_max, maximum velocity of phase 0 depolarization; and BCL = basic cycle length.

^bCompounds tested at 10 μM.

In conclusion, we have described our efforts to mitigate a hERG liability present in a series of compounds by decreasing lipophilicity. Although we were able to identify compounds with acceptable hERG inhibition, these compounds prolonged the APD in canine Purkinje fibers, suggesting a more complex mixed cardiac ion channel effect. As a consequence, the development of these compounds was halted, and in due course, we will describe our efforts in redesigning this chemotype.

Supporting Information Available

Experimental procedures and characterization data for the synthesis of compounds 2b, 2e,f, 2h, 2j–r, 3a,b, 4a, 6, and 8. This material is available free of charge via the Internet at http://pubs.acs.org.

The authors declare no competing financial interest.