6-Cyano Analogues of Bedaquiline as Less Lipophilic and Potentially Safer Diarylquinolines for Tuberculosis.

Bedaquiline (1) is a new drug for tuberculosis and the first of the diarylquinoline class. It demonstrates excellent efficacy against TB but induces phospholipidosis at high doses, has a long terminal elimination half-life (due to its high lipophilicity), and exhibits potent hERG channel inhibition, resulting in clinical QTc interval prolongation. A number of structural ring A analogues of bedaquiline have been prepared and evaluated for their anti-M.tb activity (MIC90), with a view to their possible application as less lipophilic second generation compounds. It was previously observed that a range of 6-substituted analogues of 1 demonstrated a positive correlation between potency (MIC90) toward M.tb and drug lipophilicity. Contrary to this trend, we discovered, by virtue of a clogP/M.tb score, that a 6-cyano (CN) substituent provides a substantial reduction in lipophilicity with only modest effects on MIC values, suggesting this substituent as a useful tool in the search for effective and safer analogues of 1.

Novel drugs that can reduce the treatment time for tuberculosis (TB) are vital, particularly in cases of multi- and extensively drug resistant tuberculosis (MDR-TB and XDR-TB).[1] Ideally, new TB drugs are effective against drug-resistant and drug-sensitive TB, well tolerated, suitable for once daily oral dosing, and compatible with antiretroviral therapies for individuals coinfected with HIV. After several decades without the approval of a new class of drug for TB, the diarylquinoline (DARQ) bedaquiline (TMC207, Sirturo, Janssen Pharmaceuticals; 1, Figure ) was approved by the US Food and Drug Administration in December 2012 for use in pulmonary multidrug resistant (MDR) TB. Bedaquiline has a novel mechanism of action, through inhibition of the mycobacterial ATP synthase enzyme.[2] Improved outcomes were seen when bedaquiline was added to standard therapy regimens for MDR-TB in a Phase II registration trial.[3] Other multidrug trials are in progress, with positive results being reported for a bedaquiline/pretomanid/pyrazinamide (BPaZ) combination therapy phase IIa trial.[4,5] Bedaquiline shows inhibition of the hERG cardiac potassium channel, with the concomitant risk of QTc prolongation.[6] This raises concerns about potential interactions with other drugs that also prolong the QTc interval (fluoroquinolones, clofazimine) in MDR-TB patients.[7]

Figure 1

Structure of bedaquiline (1).

It is also very lipophilic (measured log P 7.25), which may contribute to its induction of phospholipidosis, seen at high doses in preclinical models.[8] Its high lipophilicity may also contribute to bedaquiline’s long terminal elimination half-life,[9] which may lead to tissue overproportional accumulation at high doses or with daily dosing. Due to these pharmacokinetic properties, bedaquiline is currently dosed three times per week, following a period of once daily loading. Additionally, due to the possibility of tissue overproportional accumulation, efficacy has not been thoroughly explored at higher doses.[10] These observations suggest that less lipophilic analogues of bedaquiline would be of potential interest, to reduce the potential for tissue overproportional accumulation and hence to increase suitability for once daily dosing.

Bedaquiline emerged from a whole-cell screen of 70,000 library compounds against the nonpathogenic M.smegmatis strain of TB,[11] where the racemic mixture (comprising four diastereomers) was shown to have useful activity against both M.smegmatis and M.tuberculosis (M.tb), with the R,S enantiomer being the most potent. A structure–activity relationship (SAR) study of about 200 analogues of 1 (as mixtures of RR,SS or RS,SR diastereomers) against M.smegmatis showed a rank order correlation between M.smegmatis and M.tb, with the latter about 10-fold more resistant. The SAR study[11] showed that the dimethylaminoethyl side chain was near optimal for activity, with weaker bases being less effective. This is consistent with later crystallographic studies[12] of 1 bound to its major target (the c subunit of the ATP synthase Fo moiety), where the dimethylaminoethyl unit making a H-bond to Glu65 in the ion-binding site of the enzyme, anchoring the rest of the molecule to make multiple additional hydrophobic contacts. The study also evaluated eight analogues of 1 with differing substituents at the 6-position of the methoxyquinoline ring, including compounds 1, 2, 5, and 6 in Table below (mostly as RS,SR diastereomer mixtures). The authors noted that while substituents generally improved potency over the unsubstituted parent (2), there seemed to be little electronic effect, with the IC90s of the 6-substituted compounds within a 2-fold range compared with the lead compound (1).[11]

Table 1

6-Substituted Quinoline Analogues of Bedaquiline

clogP calculated by ChemDraw Ultra v13.0 (CambridgeSoft).

MIC90 (mg/mL); minimum inhibitory concentration for 90% inhibition of growth of M.tb strain H37Rv, determined under aerobic (MABA)[20] conditions. Each value is the mean of at least two independent determinations.

IC50 (μM).

Data for R,S enantiomer.

In the present Letter, we expand the range of 6-substituents in this series, across a number of modified B and C ring scaffolds, seeking more polar alternatives to Br that provide analogues of bedaquiline with similar potency of M.tb inhibition.

The 6-Br compounds in Tables and 2 were primarily synthesized following a route described previously,[11] from appropriate benzylquinoline A/B units and 3-(dimethylamino)-1-phenylpropan-1-one C/D units (Scheme ). While there have been two reported asymmetric syntheses of bedaquiline,[13,14] these syntheses are lengthy and nonconvergent, calling for stepwise installation of the B, C, and D units, and were not suitable for a medicinal chemistry SAR program. Our synthetic efforts utilized some expedient synthetic routes to a range of bedaquiline analogues, mainly by employing some common intermediates (e.g., functionalized A/B units where X = Br, I). Tables and 2 report a vast number of DARQ analogues with combinations of various A/B and C/D units, and their building blocks are detailed in the Supporting Information. However, we highlight in Scheme the key reactions for the preparation of a subset of the compounds most relevant to this study.

Table 2

Comparison of Different 6-Quinoline Substituents on Modified B/C Scaffolds

#	B-ring substituent	C-ring substituent	X	clogPa	MIC90b	clog P/M.tb scorec
1	H	1-naphthyl	Br	7.25	0.09
27	H	3-F	Br	6.22	0.23
28			CN	4.86	0.69	3.0
29			C≡CH	5.60	0.36	4.8
30			NMeSO2Ph	6.39	4.1d	–0.04
31			NMeSO2NMe2	4.20	5d	0.42
32	3-F	3-OCF3	Br	7.25	0.25
33			CN	5.89	0.47	6.2
34			N(CH2CH2)2O	6.22	0.85	1.7
35			Cl	7.10	0.09	–0.94
36	2-F, 3-OMe	3-F	Br	6.22	0.10
37			CN	4.87	0.18	17
38			Xe	5.04	0.51d	2.9
39			N(CH2CH2)2S	6.03	0.19d	2.1
40			N(CH2CH2)2SO2	4.23	>5d	0.41
41			N(CH2CH2)2SO	4.31	>5d	0.39
42			SMe	6.03	0.13d	6.3
43			SO2Me	4.11	2.3d	0.96
44	2-F, 3-OMe	3-OCF3	Br	7.11	0.09
45			CN	5.75	0.26	8.0
46			Xe	5.92	0.77	1.8
47			N(CH2CH2)2O	6.08	0.14	21
48			N(CH2CH2)2NH	6.07	0.66	1.8
49			Npiperidyl	7.46	2.3	–0.16
50			F	6.39	0.21	6.0
51			OCF3	7.60	1.1	–0.51
52	2-F, 3-OMe	3-OMe	Br	6.00	0.10
53			CN	4.64	0.09	–130.0f
54			Xe	4.81	0.87d	1.5
55	2-F, 3-OMe	3-Cl	Br	6.79	0.07
56			CN	5.44	0.13	23
57			Xe	5.61	0.31d	4.9
58	2-F, 3-OMe	3-Me	Br	6.58	0.04
59			CN	5.22	0.09	30
60			Xe	5.39	1.1d	1.1
61	2,3-diOMe	3-F	Br	5.48	0.04
62			CN	4.12	0.17	11
63			Xe	4.29	0.66	1.9
64			N(CH2CH2)2S	5.28	0.28d	0.83
65			N(CH2CH2)2SO2	3.49	>5d	0.40
66			N(CH2CH2)2SO	3.57	4.8d	0.40
67			NMeSO2NMe2	3.45	1.13d	1.9
68	2,3-diOMe	2,3-diOMe	Br	4.99	0.20
69			CN	3.64	0.34	9.6
70			N(CH2CH2)2O	3.97	2.5	0.44
71			Cl	4.84	0.25	3.0
72	2,3-diOMe	3-OCF3	Br	6.36	0.09
73			CN	5.01	0.21	11
74			Xe	5.18	0.75	1.8
75			N(CH2CH2)2O	5.34	0.68	1.7

clogP calculated by ChemDraw v.13.0 (CambridgeSoft).

MIC90 (in μg/mL) for inhibition of M.tb.

clog P/M.tb score = clogP(Br) – clogP(Xsub)/MIC90(Xsub) – MIC90(Br).

Data for R,S enantiomer.

For X, see compound 26 in Table .

This value was not included as a data point for Figure , as the CN analogue was more potent than the Br, producing a negative score.

Scheme 1

Syntheses of a Representative Subset of Mannich Bases and Diarylquinoline Analogues

Reagents and conditions: (a) (i) HN(iOPr)2 or TMP, n-BuLi, THF, −40 °C, 0.25 h; (ii) 103–132, THF, −78 °C, 1.5 h; (iii) 133–139, THF, −78 °C, 4 h; (b) acetophenone, CH2O, Me2NH·HCl, c.HCl, EtOH, 90 °C, 18 h; (c) P(o-tol)3, Zn, Zn(CN)2, Pd2(dba)3, DMF, 50 °C, 5–18 h.

Figure 2

Mean lipophilicity/M.tb activity score of most suitable X substituents (cf. with X = Br).

The majority of final DARQ compounds (including all 6-Br in Table ) was synthesized via condensation of the appropriate A/B unit and C/D unit (Scheme ). C/D units were prepared in one step via Mannich reaction of appropriate acetophenones. The DARQ product was formed in one step as a racemic mixture of four diastereomers, and the desired RS,SR diastereomer was isolated by supercritical fluid HPLC at BioDuro LLC (Beijing). A wide range of yields were observed for the key condensation reaction, even when the A-ring substituent remained constant (e.g., X = Br, 20–77%; A = spiromorpholine, 16–75%). Moreover, the yield appeared to also be dependent on B-ring substituents, with the 2-F, 3-OMe substituent seemingly preferred over its 3-F and 2,3-diOMe counterparts.

Cyano DARQs (28, 33, 37, 45, 53, 56, and 59) were prepared from their corresponding bromides (27, 32, 36, 44, 52, 55, and 58) via a Pd-catalyzed cyanation.[15] Cyanation conditions were optimized using various palladium sources and ligands. The purity of tris(dibenzylideneacetone)dipalladium(0) was variable from several commercial sources and was repurified before use.[16] The order of addition of reagents was also crucial, with addition of cyanide source (zinc cyanide) to a preheated mixture (50 °C) of other reagents critical for high yields and complete conversion to products.[15]

For substituents other than cyano, common intermediates of A/B units where X = Br, I allowed the introduction of amine or sulfamide substituents to the 6-position of the A-ring via Buchwald coupling (107–110, 121–127, and 131–132) or Ullmann type reaction (113 and 128), respectively (see Supporting Information). Other reactions such as Suzuki coupling with the 6-bromo A/B unit afforded a 6-cyclopropyl derivative (112).

Alternatively, DARQ compounds were directly functionalized at the 6-position of the A-ring. An amino substituent was accessed via hydrolysis of imine (22), whereas silanes were reduced to form alkynes (9 and 29) or further down to ethyl substituents (7). Thio-based DARQs were oxidized to sulfoxides (41 and 66) or sulfones (40, 43, and 65) using m-chloroperoxybenzoic acid (m-CPBA) or N-methylmorpholine N-oxide (NMO), respectively.

As the preparation of bedaquiline (1) has been reported previously, its synthesis has not been described here. The syntheses of compounds 2, 4, 6, 10–14, 16, 20, and 21 were conducted by Janssen Pharmaceutica (Belgium) previously[17−19] and so are not reported in the Supporting Information.

Table shows the structures and physicochemical and biological properties of bedaquiline (1) and 25 analogues bearing a wide variety of different 6-substituents. For the majority of the compounds, MIC90 values were determined for inhibition of M.tb (strain H37Rv) under aerobic conditions (MABA assay[20]). The majority of the compounds were evaluated as the RS,SR diastereomers, but a few (23, 25, 26; noted) were available only as the pure R,S enantiomer.

Representative examples (as pure R,S-enantiomers) were also evaluated for their ability to inhibit potassium ion through the hERG potassium ion channel.[21,22] While some compounds showed less potent hERG inhibition, there was no significant correlation seen between any 6-substituent properties and hERG inhibitory potency.

Calculations show a modest correlation of lower MIC90 with higher overall lipophilicity (measured as 6-substituent π values) (eq ) but not with substituent electronic properties. The latter suggests there is little 6-substituent interaction with the enzyme active site, consistent with the crystal structure of 1 bound to the c subunit of M.phlei.[12]

No correlation was seen between 6-substituent properties and hERG IC50 values.

In Table we extend these studies on the suitability of different 6-substituents to ten sets of compounds containing a variety of other substituents in the B and C units of the bedaquiline structure. The aim was to seek more polar 6-substituents that would contribute to lowering overall drug lipophilicity and potentially hERG inhibition while (in spite of the overall trend shown by eq ) not adversely affecting potency (MIC90) against M.tb. We have measured this by calculating the expression [clog P/M.tb score] for each compound (eq )

This is the ratio of the difference in overall lipophilicity over the difference in MIC90 value for each compound, averaged over all the compounds containing that substituent. The more positive the value of the ratio for a particular substituent, the better it fulfills the desired role. Figure shows that, by this measure, the polar but strongly electron-withdrawing CN substituent is the most preferable of the 6-substituents studied, with the highest average clog P/M.tb score of 13 (albeit ranging widely from 3 to 30). Across the 10 individual Br/CN sets in Table , the CN compounds had an average MIC of 0.25 μM (only 2-fold greater than the average 0.12 μM MIC for the corresponding Br compounds), but with an average clogP 1.5 log units lower than that of their Br counterparts (4.9 versus 6.4). Thus, CN is suggested as an accessible and stable 6-substituent across a range of analogues to substantially lower lipophilicity with minimal effects on MIC90.

Table also reveals preliminary SAR information from variations of B/C units. With X = Br or CN (the preferred 6-substituents), a comparison of the MIC90 across different B-units suggests that the disubstituted B-units (2-F, 3-OMe, and 2,3-diOMe) may be more favorable than the monosubstituted ones, showing similar potency as 1. Changing from a bicyclic naphthalene C-unit to a 3-substituted phenyl ring (as well as a 2,3-diOMe phenyl) were found to be tolerated. A few examples of these B/C scaffolds with a 6-CN substituent afforded comparable potencies to 1, which warrant further investigation into other combinations of B/C units with a 6-CN quinoline scaffold that may further lower both lipophilicity and MIC90.

The results of Table suggest that, for a range of 6-substituted analogues of 1, there is a positive correlation between potency (MIC90) toward M.tb and drug lipophilicity, as has been observed previously. Despite this, in Table , we show that a 6-cyano (CN) substituent provides a substantial reduction in lipophilicity with only modest effects on MIC90 values, by determining the ratio of the difference in overall lipophilicity over the difference in MIC90 value for compounds. This is a valuable new substituent to use in the search for effective but less lipophilic and potentially safer analogues of 1.