3,5-Dialkoxypyridine analogues of bedaquiline are potent antituberculosis agents with minimal inhibition of the hERG channel

Graphical abstract


Introduction
Bedaquiline (TMC207, Sirturo, Janssen Pharmaceuticals; 1; Table 2), a diarylquinoline, is the first example of a new class of drug that has proven clinically effective against drug-resistant tuberculosis (TB). 1 It has a novel mechanism of action, selectively inhibiting the mycobacterial ATP synthase enzyme. 2 In attempts to develop improved second-generation analogues of 1, we have focused on two areas. One is to lower its very high lipophilicity (clogP 7.25), which may contribute to its long terminal elimination half-life and tissue accumulation at high doses. 3 The other aim is to attenuate its inhibition (IC 50 1.6 µM) of the cardiac potassium channel protein coded by the human ether-a-go-gorelated gene (hERG). 4 This forms the pore-forming subunit of the rapidly activating delayed rectifier potassium channel (IKr), which is important for cardiac repolarization. Dysfunction of hERG causes long QT syndrome and can increase the risk of sudden death in patients with cardiac ischemia -a potential safety issue that all drug candidates seeking regulatory approval must currently address. In our attempts to develop analogues of lower lipophilicity, we have reported on the utility of replacing the 6-Br substituent on the A-unit quinoline with a more polar cyano group, 5 replacing the B-unit phenyl with heterocycles, 6 and replacing the C-unit naphthalene with either bicyclic heterocycles 7 or substituted pyridines. 8 These studies provided a diverse range of potent diarylquinolines with much lower lipophilicity, and defined structure-activity relationships between antitubercular activity and lipophilicity. While they were, in the main, less successful in identifying active compounds with significantly attenuated potency against the hERG channel, we recently 8 reported encouraging results (hERG IC 50 values around or > 10 µM) in a small number of compounds containing a 4-pyridyl-3,5-dialkoxy C-unit, suggesting this substitution is a promising one for mitigation of hERG potency. In the present paper, we follow up on that observation with a more extensive structure-activity study of this motif, which has allowed selection of two candidate compounds for preclinical development.
Having previously established the utility of a range of substituted pyridyl B-and/or C-units 6,8 in lowering overall clogP values, in the present SAR study we report 3,5-dialkoxy-4-pyridyl C-unit analogues bearing a wide range of different B-unit substitutions, with the focus on mitigation of hERG potency. Thirty-four different A/B-units were used, apart from the unsubstituted parent; syntheses of many of these (for compounds 7, 9, 10, 13, 14, 16, 17, 23-26, 34, 40-42, 48, 50 of Table 2) have already been reported. [5][6][7][8] The remainder were prepared as outlined in Scheme 1 and Table 1.
The acid chlorides derived from I were reacted with N,O-dimethylhydroxylamine to give the N-methoxyacetamides (Weinreb amides) (II), which were converted to the required Mannich bases (III) in high yields by reaction with vinyl magnesium bromide to generate an intermediate (IIa), followed by reaction with dimethylamine. Mannich bases IV to VII were prepared to evaluate the role of varying D unit pKa on hERG inhibition (see Table 4). In the absence of an additional dialkylamine source, the N,O-dimethylhydroxylamide liberated in the formation of IIa itself reacts with IIa to give IV. The imidazole (VI) and  2 , Pd 2 (dba) 3 /P(o-tol) 3 , DMF, 50°C, then separation of the diastereomers by SFC HPLC. triazole (VII) derivatives required the isolation of crude IIa before subsequent reaction with imidazole or triazole, this avoided the competitive reaction between IIa and the more nucleophilic N,O-dimethylhydroxylamine.

Structure-activity relationships
In previous work, 8 we demonstrated the ability of a range of analogues with substituted pyridyl C-units to significantly lower lipophilicity (clogP values between about 5.5-4.0) while producing compounds only slightly less effective than 1 against both replicating and non-replicating cultures of M.tb in vitro, and in a mouse TB infection model. While these compounds, as a class, did not show useful reductions in potency in hERG channel blockade, the 3,5-dimethoxy-4-pyridyl unit did appear to have some utility, prompting the present study of 3,5-dimethoxy-4-pyridyl and 3,5-diethoxy-4-pyridyl analogues with a wide range of B-units. Table 2 provides data on analogues of 1 , where the naphthalene C-unit has been replaced by either a 3,5-dimethoxy-4-pyridyl or a 3,5-diethoxy-4-pyridyl unit. The compounds were evaluated for bacterial growth inhibition (MIC 90 values) in cultures of M.tb (strain H37Rv) under both replicating (MABA assay 10 ) and non-replicating (LORA assay 11 ) conditions.
The new bedaquiline analogues characterized in the present work have calculated clogP values between 7.50 and 3.36, with the majority between 6.25 and 4.25 (Table 2). Nearly all have MIC 90 values significantly superior (many by > 10-fold) to 1 against both replicating and non-replicating cultures of M.tb. As shown previously, 7,8 the more lipophilic compounds appear more potent, but as so many had indeterminate endpoint values (below the testing range of the assay), this could not be quantified. We have previously shown 8 that different Bunits have little specific effect on inhibitory activity (MIC 90 s) apart from their contribution to overall lipophilicity. This appears to also be the case for the analogues tested here.
The most interesting result of this study was the very positive effect of the 3,5-dialkoxy-4-pyridyl C-unit substitution on hERG potassium channel inhibition. Of the 33 compounds evaluated, 19 had IC 50 values of ≥ 10 µM, in strong contrast with previous studies of different sets of analogues, [6][7][8] where no compound had IC 50 values even approaching 10 µM. The data suggest that, for this series, potent hERG inhibition is broadly related to higher compound lipophilicity; the 14 analogues with hERG values < 10 µM had an average clogP of 4.57, whereas the 19 analogues with hERG values > 10 µM had an average clogP of 6.17 (Table 2). Put another way, 15/17 compounds with clogP > 5.5 had hERG IC 50 s > 10 µM, compared to only 4/16 with clogP < 5.5. Table 3 also provides data on the in vitro clearance of the analogues from human and mouse liver microsomes. The very slow clearance (CL int 3 μL/min/mg) and concomitantly long half-life (231 min) of 1 in human liver microsomes is improved upon 2-fold by many of the analogues but a 3-fold or more increase in human microsomal clearance was seen for a handful of analogues only. The mouse liver microsome data did not always reflect the corresponding human data. For example, for several compounds (16, 18, 22, 28, 41, 45, 46 and 50), the Cl int in human microsomes is similar to, or 2-fold higher than, that of 1, but in mouse microsomes, they show at least 3-fold higher clearance than 1. This may be due to differing relative contributions to metabolism of these compounds versus 1 for CYP isoforms in human versus mouse microsomes. The human microsome data may better predict the human clearance for these compounds, if microsomal metabolism is representative of metabolism overall, for these compounds. Studies of pharmacokinetics in several species, followed by modelling work, will be needed to provide a better prediction of human PK.
The pharmacokinetics in mice were evaluated after administration of a single oral dose and a single intravenous injection. Clearance in vivo Scheme 2. Synthesis of dialkoxypyridyl Mannich bases (III-VII) Reagents and conditions: (i) COCl 2 , DMF, DCM, then MeNH(OMe).HCl, pyridine; (ii) vinylmagnesium bromide, THF; (iii) dimethylamine, water; (iv) for IV, water; V, morpholine, water; VI and VII, isolation of crude IIa then imidazole or triazole. ranged from low (similar to that of 1) to moderate, around 50 mL/min/ kg, and did not always reflect the in vitro Cl int values. In particular, a significantly higher in vitro Cl int than 1 did not always lead to much higher in vivo clearance (eg compounds 31 and 41). Although all compounds shown demonstrated high volumes of distribution (12 L/kg and higher), compounds such as 31 and 41 showed particularly high volume of distribution values, and higher tissue levels in these cases may have contributed to lower in vivo clearance than predicted by in vitro Cl int . Bioavailability (F) ranged from moderate (20% for compound 50) to very high (94% for compound 11) and, after oral administration, a range of AUC values was seen. The AUC values correlated broadly with clearance values, suggesting that for the large part, plasma exposures varied depending on relative clearance. Regarding bioavailability, for the compounds with F of 30% or lower, while they showed clearance in mice of 20 mL/min/kg and higher (except compound 51, which showed lower clearance), they did not necessarily show the highest in vivo clearance values overall. Of note, compounds with F of 30% and below had clogP values of 5.53 and below, except for compound 51, suggesting perhaps lower clogP was one factor associated with lower %F for this series. All compounds tested demonstrated human plasma protein binding of > 99.9%, with the exceptions of compounds 18, 28, and 30, which were 99.85%, 99.78%, and 99.86% bound, respectively (data not shown). Assuming mouse plasma protein binding is similar, relative PPB presumably does not play a role in relative PK for these compounds.
Evaluation of the compounds for in vivo efficacy used a mouse model of TB. Female BALB/c mice (n = 7 per group) were dosed at 20 mg/kg daily, by oral gavage, for 12 days, beginning 10 days after M.tb H37Rv inoculation by the aerosol route. Compounds were formulated in hydroxy-beta-cyclodextrin 20%. The relative in vivo activity was assessed by the log reduction in colony-forming units (CFU) from lung homogenates, an indicator of bacterial burden in the lungs, following treatment with a test compound compared to that effected by vehicle administration only, and to that demonstrated by bedaquiline, dosed at 20 mg/kg in the same assay run. In this mouse model of TB, the CFU reach approximately 5-6 log in the lungs in the vehicle-treated control group. Hence a reduction in lung CFU of 5 of more log units implies complete clearance of the recoverable lung CFU following 12 days of treatment. Therefore, for compounds in Table 3 where a reduction of approximately 5 log CFU is shown, complete clearance was achieved, and it is not possible to further distinguish the relative efficacy between these compounds. Reductions of 4 log CFU and higher mean only very few bacteria were remaining in the lungs of mice and these compounds can therefore be considered to show similar efficacy to bedaquiline within the range of these experiments.
Gratifyingly, and in contrast to a previous study of bedaquiline analogues with differently-configured substituted C-unit pyridyl substituents, 8 most of the compounds showed in vivo activities similar to that of 1, at the same dose, with only compounds 12, 14 and 48 being less active, showing CFU reductions of < 4 log units (Table 3). It was interesting to note that the compounds demonstrating smaller CFU reductions in vivo were not the compounds with the lowest AUC values and, indeed, several compounds demonstrated CFU reductions that were at least similar to that effected by bedaquiline (within the limits of the assay), with substantially lower AUC values. For example, compounds 8, 30, 31, and 41 demonstrated CFU reductions within one log unit of the CFU reduction effected by bedaquiline, with AUC values of 2.69 µg*hr/mL or below, compared to 20.9 µg*hr/ml for 1. In general, for the tested compounds, where similar efficacy was achieved with lower AUCs, this is likely due to the lower MICs for these compounds than for bedaquiline, and consistent with previous work that has demonstrated AUC/MIC as the driver of efficacy against murine TB for bedaquiline. It may also be due to formation of metabolites that are active against M.tb, as is seen for bedaquiline. 13 One exception was compound 10, where efficacy in the range of that of bedaquiline was unit CFU reductions in mice, at an AUC at least approximately 4-fold lower than that of bedaquiline. This suggests these analogues may have potential to demonstrate efficacy against TB in patients at lower plasma levels than bedaquiline. Of these, compounds 8, 10, 24, 41, 45, 46, 50 and 51 exhibited hERG IC 50 s of 9.9 µM or higher, compared to 1.6 µM for bedaquiline. Taken together, a higher IC 50 against hERG along with a lower efficacious exposure may predict a lower risk of QTc prolongation in patients for these analogues compared to bedaquiline. Although not all of these compounds showed higher human microsome Cl int than bedaquiline, the possible lower risk of QTc prolongation along with potential for lower overall adverse effects based on a lower efficacious exposure, resulted in selection of these compounds, as well as others in Table 3 that demonstrated similar efficacy to bedaquiline with lower potency against hERG, for further evaluation. Electrocardiography studies in animals and ultimately in human subjects will be needed to determine whether the decreases in potency against hERG seen here, along with the decreased plasma levels needed for efficacy for some of these compounds, translates into an absence of QTc prolongation at a therapeutic dose and exposure.
Finally, the overall good activity profile of this series of 3,5-dialkoxy-4-pyridyl analogues was utilized to explore the results of changes in the pKa values of the D-unit side chain. This was explored previously by Guillemont et al. 9 in their original paper, where they concluded that a sidechain with a pKa of > 8 was needed to retain activity against the related mycobacterium, M. smegmatis. We were thus motivated by the idea that utilising a much weaker base in the side chain in analogue series with intrinsically higher potency might further attenuate hERG blockade while retaining good M.tb potency. The results in Table 4 suggest that a weaker base did lead to reduced hERG inhibition (compare hERG data for 12 and 54, 33 and 61), but there was also a large increase in the MICs for inhibition of M.tb.

Conclusions
In previous publications in this series, we have shown that less lipophilic analogues of 1 (clogP 7.25) retain substantial in vitro and in vivo activity against M.tb, down to compounds with a clogP of about 4; this shows that one of the potential drawbacks of 1 can be ameliorated. In the present paper, we show that substitution of the C-unit naphthalene of 1 with a 3,5-dialkoxy-4-pyridyl group provide analogues with not only higher potency and substantially lower clogP than 1, but also with a greatly reduced potential cardiotoxicity risk profile, due to significantly reduced blockade of the hERG potassium channel. This work has recently allowed the selection of analogues 8 (TBAJ-587) 14,15 and 46 (TBAJ-876) 16 as compounds for pre-clinical development. Based on the preliminary data presented in this paper, several compounds (including 8 and 46), with activity against murine TB similar to bedaquiline at the same dose, with lower clogP, higher IC 50 against hERG, and, in most cases, higher Cl values in human microsomes, were  selected to progress to further studies. Those advanced studies evaluated the PK in rodents and non-rodents, MICs against clinical isolates of M. tb, efficacy against a model of chronic murine TB, and toxicity in a preliminary rat toxicity and toxicokinetics study. Based on that data, compounds 8 and 46 were selected due to their lower MICs than bedaquiline against clinical isolates of M.tb, efficacy demonstrated against murine TB at lower exposures than bedaquiline, lower potency against hERG, predicted higher human clearance, and an acceptable safety margin, based on the safe exposure in rats compared to the efficacious exposure in mice. Further reports on the development of these two compounds will be published in due course.

Chemistry
Final products were analysed by reverse-phase HPLC (Alltima C18 5 µm column, 150 × 3.2 mm; Alltech Associated, Inc., Deerfield, IL) using an Agilent HP1100 equipped with a diode-array detector. Mobile phases were gradients of 80% CH 3 CN/20% H 2 O (v/v) in 45 mM NH 4 HCO 2 at pH 3.5 and 0.5 mL/min. Purity was determined by monitoring at 330 ± 50 nm and was ≥ 95% for all final products. Melting points were determined on an Electrothermal 9100 melting point apparatus. NMR spectra were obtained on a Bruker Avance 400 spectrometer at 400 MHz for 1 H. Low-resolution atmospheric pressure chemical ionization (APCI) mass spectra were measured for organic solutions on a ThermoFinnigan Surveyor MSQ mass spectrometer, connected to a Gilson autosampler.