Structure-activity relationships for unit C pyridyl analogues of the tuberculosis drug bedaquiline

Graphical abstract


Introduction
Bedaquiline (TMC207, Sirturo, Janssen Pharmaceuticals; Fig. 1; 1) is an exciting new drug for tuberculosis, with a novel mechanism of inhibition of the mycobacterial ATP synthase. 1 Of particular interest, it has demonstrated clinical efficacy against multidrug-resistant TB, where treatment options are limited. 2 Potential limitations of 1 include very high lipophilicity (a calculated clogP of 7.25), 3 with the attendant risks of ultra-long half-life, 4 phospholipidosis, 5  against Mycobacterium tuberculosis, but with lower lipophilicity than bedaquiline, anticipating a desirably shorter half-life, we have recently reported on the beneficial effects of replacing the lipophilic 6-Br group on the A-unit quinoline ring with a more hydrophilic cyano group, 6 replacing the B-unit phenyl group with monoheterocycles, 7 and replacing the very lipophilic C-unit naphthalene with a series of less lipophilic bicyclic heterocycles. 8 In the present work we take the latter theme further by looking at a range of even less lipophilic pyridine Cunits, with a further comparison of paired Br/CN derivatives. A recent paper 9 on the 1.7 Å resolution crystal structure of 1 bound to the c subunit of the ATP synthase Fo of the genetically-similar mycobacterium M. phlei (83.7%), found that the quinoline and naphthalene units appear to play a role in positioning the dimethylaminoethyl side chain in the ion-binding site of the enzyme. Following on from this, it was of interest to see whether analogues with smaller, more hydrophilic, monocyclic pyridyl C-units would be as effective at binding. The only other study 1 on non-naphthalene analogues of 1 showed that 2fluoro-and 2,5-difluorophenyl compounds had MIC 90 values (in Mycobacterium smegmatis) comparable to that of 1, but pyridyl analogues were not examined.
Lithium tetramethylpiperidide (LiTMP) mediated condensation of quinoline A and substituted benzaldehydes B gave the intermediate alcohols, which were deoxygenated by triethylsilane under acid conditions to give the corresponding A/B units. Alternatively, palladium mediated coupling of the boronic acid A1 with an appropriately substituted benzyl halide or (halomethyl)pyridine B1 gave the A/B units, all of which have been reported previously. 3,6,7 The Mannich bases required for the C/D units of the final compounds were prepared from the appropriate commercial carboxypyridines as shown in Scheme 2 and Table 1. Lithiation of the A/B subunit with lithium diisopropylamide (LDA) and subsequent reaction with the C/D subunit gave the bromo bedaquiline analogues. The cyano derivatives were synthesised from their corresponding bromo derivatives by palladium-catalysed cyanation.
Bedaquiline (1) and C-unit pyridyl analogues 2-62 were evaluated for their ability to inhibit bacterial growth (measured as MIC 90 values in  µg/mL against M. tb (strain H37Rv)) under both replicating (MABA) 10 or non-replicating (LORA) 11 conditions. In these assays, 1 is a potent inhibitor, with MICs of 0.08 and 0.12 µg/mL, respectively. Replacement of the naphthalene C-unit of 1 with various pyridyls produced analogues of considerably lower lipophilicity, with clogP values ranging from 6.89 down to 2.98.

Structure-activity relationships
This work explored SAR for much less lipophilic analogues of bedaquiline, with clogP values between 6.89 and 2.98 (0.36 to 4.27 log  units less lipophilic than bedaquiline). Overall, that resulted in compounds that are slightly less potent than bedaquiline (1) against both replicating (MABA) and non-replicating (LORA) M.tb in culture ( Also, as expected, there was a reasonable correlation between inhibitory activity (illustrated here for MABA data) and compound lipophilicity, with the more lipophilic compounds being more potent; an overall trend repeatedly seen in other classes of TB drugs. 12 However, this overall correlation masks a trend seen to some extent previously, 8  Representative compounds in the series were evaluated for a range of ADME and toxicological properties in vitro (Table 3). Compounds were tested for cytotoxicity in Vero green monkey-derived epithelial kidney cells, 14 and for inhibition of CYP 3A4, the major metabolising enzyme for 1. 15 All tested compounds had IC 50 s > 10 µg/mL in the Vero assay, except for compound 38, where this was not measured (the value for 1 in repeat assays was between 4 and 16 µg/mL). All compounds also had IC 50 s > 10 µM for CYP3A4 inhibition (bedaquiline IC 50 > 40 µM), except for compound 15, which had an IC 50 of 7.6 µM, and compounds 4, 27, 33, 38, 46, 48, 49, 57-59, 60-62, which were not tested in this assay.
Since bedaquiline's relatively potent inhibition of the hERG potassium channel (IC 50 1.6 µM 14 (see Table 3) has been associated with QTc prolongation observed in the clinic (cardiovascular toxicity), the analogue compounds were also evaluated for inhibition of hERG potassium current in vitro. This was determined either by a full (5-concentration) dose-response study resulting in an IC 50 , or by determining % inhibition of channel current at 0.3 and 1 µM, using the same manual patch clamp electrophysiology assay setup. While several of the compounds (38, 46, 48, 49, 57, 61, 62) were less potent inhibitors of this channel than 1, several others (e.g., 15, 33, 39, 40, 59, 60) were similar or considerably more potent. Lipophilicity did not seem to be a factor, with both 1 (clogP 7.25) and compound 27 (clogP 3.87) being similarly potent against the hERG channel. Compounds 61 and 62, bearing 4-aza, 3,5-dialkoxy C-units, were notable for their much lower hERG potencies (IC 50 values around or > 10 µM), suggesting this motif as a promising one for mitigation of hERG risk.
One of the cited issues with 1 is its very long terminal half-life in humans. 16 This can result in undesirable levels of accumulation of drug in tissues. Table 3 also shows in vitro clearance rates of the analogues in human and mouse microsomes as well as the in vivo clearance observed in mice. All of the pyridyl compounds had (desirably) higher rates of clearance in vitro than 1, with several (4, 19, 21, 22, 27) possibly being cleared too rapidly. This was also true for those pyridyl compounds evaluated in vivo, with several demonstrating clearance > 50 mL/min/ kg (19, 22, 33, 55, 57) compared to 7 mL/min/kg for bedaquiline.
Several analogues were also evaluated for oral bioavailability in mice, and while being less bioavailable than the very lipophilic 1, still had acceptable values. As seen in Table 3, in general, higher in vivo Table 3 Additional biological data for selected representative compounds of Table 2 clearance values predicted lower in vivo AUC and lower %F, suggesting oral bioavailability for these compounds is primarily driven by clearance as opposed to absorption. The volume of distribution for these compounds was at least 6 L/kg in all cases (22 L/kg for bedaquiline) and varied among the compounds (see Table 3), but did not appear to relate to clogP. All compounds tested demonstrated human plasma protein binding of > 99.9% (data not shown). Ten representative compounds with a range of MABA MICs, volume of distribution and AUC values were evaluated for efficacy in a mouse model of TB. In this model, the infected mice are dosed daily with 20 mg/kg of the test compound for 12 days, beginning 10 days after M.tb infection. Efficacy is measured by the log reduction in colonyforming units (CFU) recovered from the lungs, compared to 1 as a positive control in the same assay. The studies sought to identify compounds that demonstrated similar efficacy to 1 at the same dose (a lung CFU reduction of similar magnitude).
Of the compounds evaluated in vivo, only compounds 61 and 62 effected a lung CFU reduction similar to that of 1, at the same dose. These compounds were also the only analogues with both AUCs and MABA MICs comparable to that of 1. These two compounds had the highest AUC/MIC ratios of the novel analogues tested, and were the only compounds evaluated for efficacy with AUC/MIC similar or higher than AUC/MIC of 1. The resultant efficacy is consistent with the results of a detailed previous study that identified plasma AUC/MIC as the driver of the in vivo efficacy of 1 in a similar mouse model of TB. 16 Of interest, compounds 33 and 57 demonstrated efficacy superior to that of the other analogues (except for 61 and 62), despite low AUC/MIC ratios. It is notable that these compounds demonstrated extremely high volumes of distribution, suggesting especially high tissue levels. It is also possible that these two compounds generate active metabolites in vivo that contribute to efficacy; bedaquiline is known to generate an active metabolite that contributes to efficacy in mice. 17 Further studies will be needed to determine whether active metabolites are also present in mice, following oral dosing of the analogues described here.

Conclusions
In this study we prepared and evaluated a set of much more polar analogues of bedaquiline (1), by replacing the naphthalene C-unit with a series of substituted pyridyls. As shown previously, the potency (MIC 90 values) of these compounds against M.tb in culture correlated positively with high lipophilicity, but this study was able to set a lower bound to lipophilicity (clogP about 4), below which in vitro potency decreased sharply. Encouragingly, many of the C-pyridyl analogues proved to be slightly less potent inhibitors of the hERG potassium channel than bedaquiline itself, and two examples (61 and 62) demonstrated IC 50 values of 7.8 and > 10 µM.
This set demonstrated varying in vitro and in vivo clearance values, with some compounds demonstrating clearance values desirably higher than those of bedaquiline. However, in part due to higher clearance values and resultant lower AUCs, this set of compounds, in general, had poor in vivo activity against murine TB compared to bedaquiline given at the same dose. The exceptions were compounds 61 and 62, which had comparable in vivo activity to 1 and were the only tested compounds with AUC/MIC ratios at least as high as that of 1. Other tested compounds had lower AUCs or higher MICs or both, compared to bedaquiline, with lower AUC/MIC ratios as a result. The observation that higher AUC/MIC results in superior efficacy for this class is consistent with previous studies of bedaquiline that indicated AUC/MIC as the driver of efficacy for that compound against murine TB.
Thus, while higher clearance than bedaquiline is a goal for this program of work, it appears that this must be accompanied by lower MICs than 1 in order to maintain the AUC/MIC ratio required to produce similar efficacy to bedaquiline, at the same dose. Although the present work did not identify a compound with lower MIC and higher clearance than bedaquiline, compounds 61 and 62, which demonstrated similar efficacy to bedaquiline with similar clearance values, also showed considerably less potent hERG inhibition, suggesting that further exploration of pyridyl-based C units might be fruitful.

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. (Table 1)  Vinylmagnesium bromide solution in THF (1 M, 45 mL, 44.6 mmol) was added to a solution of N-methoxy-N,6-dimethylpicolinamide (2.68 g, 14.87 mmol) in THF (130 mL, dist. Na) at 0°C, the brown solution was warmed to 20°C for 1 h then dimethylamine in THF (2 N, 45 mL, 89.2 mmol) and water (23 mL) were added. The solution was stirred at 20°C for 1 h, then partitioned between EtOAc and water. The solution was dried and evaporated to give 3-(dimethylamino)-1-(6methylpyridin-2-yl)propan-1-one as a yellow oil (2.78 g, 97%). 1
Isomer A, white solid. 1  Each coupled product was resolved into its four optical isomers using preparative chiral HPLC at BioDuro LLC (Beijing).
The other 6-bromo analogues of Table 1 were prepared similarly: