Synthesis and evaluation of analogues of the tuberculosis drug bedaquiline containing heterocyclic B-ring units

Graphical abstract

The often late detection of tuberculosis (TB), 1 coupled with the need to use long and complex multi-drug treatment regimens, has led to an alarming increase in cases that are resistant to the standard front-line drugs (multi-drug-resistant; MDR). In 2016, about 580,000 new cases (3.9% of new cases and 21% of recurrent cases) were classified globally as MDR-TB, and this proportion has been rising rapidly. 2 Thus the discovery of bedaquiline (TMC207, Sirturo, Janssen Pharmaceuticals; Fig. 1; 1), a new TB drug, that, due to a novel mechanism of action (inhibition of the mycobacterial ATP synthase 3 ) is useful against drug-resistant tuberculosis (MDR-TB), has been of great significance. When added to standard background therapy used for MDR-TB, it demonstrated more rapid bactericidal activity than background therapy alone, 4 and was approved by the US Food and Drug Administration in 2012 for the treatment of MDR-TB. However it is very lipophilic (measured log P 7.25), 5 which likely contributes to its long terminal half-life of 5-6 months. 6 The resultant potential for over-proportional accumulation in tissue has limited the full exploration of its potential dose range. 7 More generally, highly lipophilic drugs also have a propensity for liver toxicity. 8 Bedaquiline also shows inhibition of the hERG (human Ether-à-go-go-Related Gene; KCNH2) cardiac potassium channel, with the concomitant risk of delayed ventricular repolarization (QTc interval), 6 and this is a point to be considered in the planning of combination regimens with other TB drugs with similar effects (e.g. fluoroquinolones, clofazimine). 9 Thus bedaquiline analogues of comparable antibacterial activity but with lower clog P and less potent inhibition of the hERG potassium channel, but similar in anti-Mtb potency to bedaquiline, would be of substantial interest. We have previously 10 explored the effects of a range of more polar 6-substituents on the quinoline ring and showed that, on balance, a 6-CN group offered a significant reduction in overall lipophilicity over the standard 6-Br group (about 1.25 log P units) with the least deleterious effect on antibacterial potency.
In the current paper we explore the effects of replacing the phenyl B-ring unit of bedaquiline with heterocycles of differing lipophilicity. One example of such a heterocyclic analogue, the 3-furan 2 also with a smaller (phenyl) C-ring unit, (Fig. 1) has already been reported to show significant activity in M. smegmatis, with the separated racemic diastereomer pairs (configurations not assigned) having values of 1.57 and 0.06 mg/mL. 11 The bedaquiline analogues were synthesized from appropriate benzylquinoline A/B-units and 3-(dimethylamino)-1-arylpropan-1-one C/D-units, following a route described previously. 7 In Scheme 2, the 2-furyl analogues 15-21 were prepared by the same procedure as in Scheme 1, reacting 51 and 2-furylaldehye (65) to give alcohol 66, which was deoxygenated to give A/B-unit 67. Condensation of this with the appropriate Mannich bases (selected from 55a-55j) gave compounds 15-21. The 3-furyl analogues 22-28 were synthesized by preparation of the boronic acid 68 (from 51) and Suzuki coupling of this and 3-(bromomethyl)furan 69 to give A/B-unit 70, and subsequent condensation with the appropriate Mannich base as above (see Supplementary Material for further details).
Reaction of the A/B units prepared in Schemes 1 and 2 with in situ-prepared LDA, (lithium diisopropylamide) or LiTMP (lithium tetramethylpiperidide) at À78°C for 60-90 min) followed by addition of the appropriate Mannich bases 55a-j (À78°C, 3-4 h, then AcOH) under previously-reported 7,1 conditions gave required diarylquinolines 3-6, 8, 9 and 11-14 of Table 1, as a racemic mixture of two diastereomers (see Supplementary Material for further details). This mixture was purified by column chromatography to >95% purity in-house, and the desired RS, SR enantiomer was then separated from the mixture by preparative super-critical fluid HPLC at BioDuro LLC (Beijing). The coupling yields for classes E-H appear to be higher than classes A-D, although this comparison is complicated by the small sample sizes in classes B, C and D. Moreover, in some cases, the coupling reaction proceeded in high conversion, but due to difficulty in purifying the product from impurities the final yield was reduced. There is however, an apparent correlation between the electron density at the benzylic position of the A/B-unit and the yield of the reaction. The best coupling unit (E) is a 2-substituted furan which exerts a strong electron donating effect towards the benzylic position. While the second best unit (G) is a more electron withdrawing pyridine, the 2 or 4-substituted electron donating substituent contributes some electron density towards the benzylic position. Unit H has a 4-aza atom which renders the benzylic position less electron rich, despite a 2-methoxy substituent, and the average coupling yield is even lower. These observations suggest that the lithium anion formed at the benzylic position could be more nucleophilic with a higher electron donation from the B-ring, making the A/B-unit more reactive and favouring the coupling reaction.
The compounds were evaluated for their inhibition of growth (measured as MIC 90 values in mg/mL) against M.tb (strain H37Rv) under both replicating (MABA) 16 or non-replicating (LORA) 17 conditions. Under these conditions bedaquiline (1) is a potent inhibitor of both (MICs 0.05 and 0.08 mM respectively). In a previous structure-activity relationship (SAR) study of bedaquiline analogues 7 it was shown that electron-withdrawing groups, especially F or Cl at the 3-and 4-positions on the phenyl B-ring unit (Fig. 1)   heterocycles of varying lipophilicity (Table 1). A representative subset of compounds were also evaluated for a number of pharmacological properties, and compared against bedaquiline (1) ( Table 2). Compounds 3-10 explore the use of dimethylthiophene B-ring units. While these are slightly more lipophilic than benzene, the set of compounds prepared had clogP values lower than bedaquiline (between 6.33 and 4.24) and MABA MICs between 0.05 and 0.58 mg/mL. The exception was the much less lipophilic CN derivative 7 (clog P 4.24), which was less active. The 2,3-dimethylthiophene 9, bearing a more lipophilic 3-F C-unit substituent, had comparable activity to 1 and its cyano counterpart 10, with a much lower clog P (4.67) was equally potent. The 2-Me and 2-OMe thiophenes 12 and 13 were also effective inhibitors, despite clog P values of around 5.
Given that good activity had been reported for furan 2 (albeit in M. smegmatis), 11 we evaluated a series of both 2-and 3-furyl analogues bearing different C-unit substituents. For the whole dataset (49 compounds, including 1), despite a wide variation in structure and some MICs being determined using a slightly different protocol (see Table 1), there was a modest positive correlation between potency and lipophilicity (Eq. (1)), as was shown previously for both bedaquiline derivatives 10 and other classes of M.tb inhibitors. 18 log MIC ¼ À0:53c log P þ 1:96 The 2-furyl (15-21) and 3-furyl (22-28) compounds had a relatively narrow range of potencies (MICs 0.68-1.96 mg/mL), despite a wide lipophilicity range (clog Ps from 7.51 to 4.02). For this subgroup of similar compounds, high lipophilicity (clog P) correlated with higher potency (MIC) (Eq. (1)). Compounds 29-43 of Table 1 ii iii   Table 2 shows that, for both 1 and selected analogues, IC 50 values for cytotoxicity in mammalian cells (VERO green monkey kidney epithelial cells) were all >10 mg/mL, 19 and (in the majority of cases) the IC 50 values for inhibition of the common CYP3A4 metabolizing enzyme were !10 mM. The potency of 1 for inhibition of the hERG calcium channel (IC 50 0.37 mM) 20 is seen as a potential liability (cardiovascular toxicity) and changes that attenuated this effect would be beneficial. However, for the compounds tested there was only modest (at best 5-fold; compound 43) and unpredictable improvement in this parameter. Compound 1 has quite low clearance in human liver microsomes, as reflected in its Cl int and t1/2 values in Table 2, with all of the other analogues having significantly higher rates of clearance. Finally, compound 49 showed an oral bioavailability comparable to that of 1, despite its significantly lower lipophilicity. Indeed, all of the compounds in Table 2 had acceptable bioavailability.

Conclusions
This work, part of a study 10 exploring analogues of 1 with an altered profile of biological properties, has focused on the effects of replacing the phenyl B-ring unit of bedaquiline with heterocycles of differing lipophilicity, in compounds with a smaller C-ring unit; thiophene-, furan-and pyridyl-based B-ring units were studied. In addition to the expected broad positive correlation between lipophilicity and anti-TB potency (Eq. (1)), the 3-pyridyl and especially the 4-pyridyl compounds did seem to have an additional measure of potency, suggesting that further exploration of the B-unit region might be fruitful.