Structure-activity relationships for analogs of the tuberculosis drug bedaquiline with the naphthalene unit replaced by bicyclic heterocycles

Graphical abstract


Introduction
Bedaquiline (TMC207, Sirturo Ò , Janssen Pharmaceuticals; Fig. 1; 1) is an exciting new drug for the treatment of tuberculosis (TB). It exhibits a novel mechanism of action compared to other TB drugs namely inhibition of the ATP synthase 1 of Mycobacterium tuberculosis (M.tb), the etiological agent of TB. Resistance to other TB drugs occurs primarily due to mutations in genes encoding their respective bacterial drug targets. Due to its novel drug target, bedaquiline is active against M.tb strains resistant to other drugs, and demonstrates efficacy against multi-drug-resistant (MDR) TB. 2 Bedaquiline was approved by the US FDA in 2012 for specific use in pulmonary MDR-TB, and does show improved outcomes when it is used in combination with standard MDR-TB drugs. 3 A recent study adding bedaquiline to a standard regimen for the treatment of MDR-TB showed a positive cost-benefit analysis by shortening hospital stays. 4 Potential drawbacks of bedaquiline include inhibition of the hERG (human Ether-à-go-go-Related Gene; KCNH2) potassium channel 5 (with the concomitant risk of cardiac toxicity), hepatic toxicity, 6 and possibly a risk of phospholipidosis 7 (related to its high lipophilicity [calculated clogP of 7.25]). 8 There are also potential pharmacokinetic (PK)-based drug-drug interactions with the common TB drugs rifampicin and rifapentine, which are potent inducers of CYP3A4, the major metabolising enzyme for bedaquiline. 9 The development of analogs of bedaquiline designed to improve on some of these properties is thus of high interest. We have previously explored the effects of more polar 6-substituents on the quinoline ring 10 and replacement of the unsubstituted phenyl ring with various heterocycles. 11 In this paper, we explore the effects of a variety of (mostly) more hydrophilic bicyclic C units ( Fig. 1) in place of the naphthalene unit of bedaquiline. In addition to being another way to lower the overall lipophilicity of analogs, a particular interest in exploring unit C structure is highlighted by a recent paper 12 on the 1.7 Å resolution crystal structure of bedaquiline bound to the c subunit of the ATP synthase F o of the mycobacterium M. phlei (84% sequence identity with M.tb). This shows the dimethylaminoethyl unit D making an H-bond to Glu65 in the ion-binding site of the enzyme, with the rest of the molecule demonstrating multiple hydrophobic contacts with the enzyme, including the naphthalene unit C with Tyr68 and Leu72, but clashing with Tyr70. Bicyclic  interest from a structural point of view, and in this paper, we prepare and evaluate a series of these, with the major focus on less lipophilic examples than naphthalene.

Chemistry
The 6-bromo compounds of Table 1 were prepared as previously described, 8,10 by LiTMP/LDA-mediated coupling of appropriate benzylquinoline A/B-units and 3-(dimethylamino)-1phenylpropan-1-one (Mannich base) C/D-units. The cyano derivatives were prepared by Pd-mediated cyanation of the corresponding bromo analogs (Scheme 1). The resulting diarylquinolines were formed as a racemic mixture of four diastereomers, and the desired RS,SR diastereomer (depicted) was isolated by super-critical fluid HPLC at BioDuro LLC (Beijing). Twelve different A/B units were used, apart from the unsubstituted parent; syntheses of many of these been reported previously. 10,11 The remainder were prepared as outlined in Scheme 2. In previous work, 10 we noted that the yields in the final condensation reaction (Scheme 1) seemed to correlate positively with the expected degree of electron density at the benzylic position of the AB-subunit, but this was not obvious in the present study.
The A/B subunits were synthesised by directed ortho lithiation of the methoxyquinoline (I), followed by quenching with a suitably functionalised aldehyde (II) to give the benzylic alcohols (III); subsequent reduction to the corresponding dihydro adducts (IV) was performed using Et 3 SiH/TFA (Scheme 1). Alternatively, the A/B subunits were synthesised by a Suzuki reaction between the boronic acid (V) and suitably functionalised benzyl bromides (VI).
The compounds in Table 1 cover a range of 4.4 clogP units; compare, for example, compounds 5 and 63. Within each C unit subset there was also limited variation of substituents in the B subunit, and only Br and CN substituents on the A subunit.
We have previously 10 discussed the use of 6-substituents other than Br on the quinoline ring of bedaquiline, and showed that the polar but electron-withdrawing CN substituent provided a substantial decrease in drug lipophilicity (about 1.4 clogP units), at the expense of only a 2-3-fold increase in MIC values against M. tb. The inclusion of 19 pairs of compounds where the 6-Br group of 1 was exchanged for a 6-CN group allowed a further evaluation of this trend. For the 15 pairs where we had definitive MIC 90 data for both members (Table 1), the average increase in MABA and LORA MIC 90 s were similar, being 2.76-and 2.81-fold respectively.
Using a dummy variable for the presence of the CN substituent shows that the CN group provides, on average, a further small (0.18 log) reduction in MIC 90 , after the effect of its lower lipophilicity is allowed for.  Note that in Eqs. (1) and (2) the absolute values for the logMIC (MABA) data in Table 1 were used; removal of the < and > qualified data points from the calculation did not change the nature of the result.
A representative subset of the compounds of Table 1 were also evaluated for a number of pharmacological properties, and compared against bedaquiline (1) ( Table 2). For the compounds with measured values, all except for 18, 34, and 36 had IC 50 's > 10 mM for inhibition of the common CYP 3A4 oxidative metabolizing enzyme, and all showed no toxicity toward Vero green monkey kidney cells 15 at a concentration of 10 mg/mL (data not shown). Bedaquiline itself (1) is a potent inhibitor of the hERG potassium channel (IC 50 1.6 mM in our assay, but also reported as 0.37 mM), 16 which is seen as a potential cardiovascular liability, and has prompted comments that its use should be confined to settings where carefully selected patients can be closely monitored, and that combination with other QTc-prolonging drugs should be avoided. 17 The IC 50 for inhibition of hERG current for bedaquiline, measured in the patch clamp assay we utilized for this study, was 1.6 mM, which is in the same range as the published value. Compared to this, a few of the analogs (5, 24, 28, 30) showed a significant (5-8-fold) improvement in this parameter, but there was no common feature to distinguish them from others which were much more potent hERG inhibitors (e.g., 14, 16, 34, 36, 55, 56). All of the compounds except 24, 26-28 (which were four of the most lipophilic compounds) had significantly faster clearance rates in human liver microsomes (HCl int ) and concomitantly shorter half-lives (Ht½) than did 1 (Table 2), which is in line with previous SAR observations. 11 The same trend was observed across the whole set of compounds for which defined human liver microsome  clearance data was obtained, with a modest correlation seen between lower lipophilicity and faster clearance.
While these compounds were also similar to bedaquiline in their low aqueous solubility and high plasma protein binding (Supplementary Data; Table S1), most had acceptable bioavailability (F > 25% in mice). Finally, as shown previously, 10,11 the oral bioavailability of the compounds in mice was broadly inversely correlated with compound lipophilicity.
Several of the compounds were further evaluated for efficacy in vivo against acute murine TB. For these assessments, female BALB/c mice, infected via aerosol with M. tuberculosis Erdman, 18 were treated by oral gavage with a dose of 20 mg/kg once daily for 12 continuous days, beginning on day 11 post-infection. Compounds were administered as a solution in 20% hydroxypropylbeta-cyclodextrin, in water, adjusted to pH 3. Bedaquiline, dosed orally once daily at either 10 mg/kg or at 20 mg/kg, was administered as a comparator, and vehicle-treated mice were evaluated as a negative control. Mice were sacrificed on day 25, and the numbers of colony forming units (CFU) in the lungs were determined by plating on agar containing charcoal (to absorb compound and prevent drug carryover effects) and compared with the CFUs for bedaquiline and vehicle alone-treated mice. Of the compounds evaluated, the most effective analogs, where 20 mg/kg/day effected a reduction in lung CFU at least as great as that demonstrated by 10 mg/kg/day bedaquiline, tested alongside, (i.e., 5, 7, 16, 24, 26, 28, 29, 36) had lipophilicities (clogPs) ranging from 7.41 to 5.61, including two with CN-substituted A units, again suggesting the viability of a CN substituent to lower overall drug lipophilicity without compromising in vivo efficacy.

Conclusions
This work, part of a programme seeking improved analogs of bedaquiline (1), sought to evaluate the effects of replacing the naphthalene C-unit with a variety of bicyclic moieties of widely differing structures and lipophilicities. The results, which show an MIC 90 /lipophilicity relationship (Eq. (1)) broadly similar to that seen for other series of TB drugs, suggest that changes in the C-subunit are well-tolerated for this series. Encouragingly, many analogs with lower clogP values than bedaquiline were at least as potent against M.tb. Further, where tested, most analogs with at least similar potency to bedaquiline against M.tb had higher clearance in human liver microsomes, suggesting they may show shorter terminal half-lives than bedaquiline in vivo, with less risk of tissue overaccumulation. Among the analogs, there was a significant range in potency for hERG channel block (IC 50 's from 0.4 to >10 mM). While there was no discernible SAR pattern to this, it is encouraging that such improvements in hERG liability can be seen for what is still a lipophilic aromatic strong base, and that anti-mycobacterial potency is still maintained when hERG inhibition is reduced. The four best compounds in his regard (24, 26, 28, 30) were all of one class (D-7; 7-benzofuran), suggest that exploration of further C-unit variations might be of interest in this regard. The value of the A-unit CN substituent is also confirmed for its ability to lower overall drug lipophilicity (thus contributing to desirable physiochemical properties) without compromising in vivo efficacy.

A. Supplementary data
Supplementary data (additional biological data on the compounds of Table 2) associated with this article can be found, in the online version, at https://doi.org/10.1016/j.bmc.2018.02.026. These data include MOL files and InChiKeys of the most important compounds described in this article.