Comparative Efficacy of the Novel Diarylquinoline TBAJ-587 and Bedaquiline against a Resistant Rv0678 Mutant in a Mouse Model of Tuberculosis

Since its conditional approval in 2012, bedaquiline (BDQ) has been a valuable tool for treatment of drug-resistant tuberculosis. More recently, a novel short-course regimen combining BDQ with pretomanid and linezolid won approval to treat highly drug-resistant tuberculosis.

associated with 50% of its maximal effect (EC 50 ) in MDR-TB patients receiving the registered dose, indicating that higher BDQ doses would achieve greater efficacy (15). However, potential safety concerns, including QTc prolongation, have limited enthusiasm for testing higher BDQ doses (18). Evidence also continues to emerge that the efficacy of BDQ-containing regimens can be compromised by inactivating mutations in the Rv0678 gene, which encodes a negative transcriptional regulator of the mmpL5-mmpS5 transporter in Mycobacterium tuberculosis (19)(20)(21)(22). Although mutations in Rv0678 are associated with relatively small reductions in susceptibility to BDQ and clofazimine, they are readily selected by BDQ and/or clofazimine treatment in mouse models of TB and also have been selected during clinical use of BDQ, including in the Nix-TB trial (14). Coupled with evidence that Rv0678 variants with reduced BDQ susceptibility have been isolated from MDR-TB patients without known prior exposure to BDQ or clofazimine (19,23,24), these reports raise concern that emergence of Rv0678 variants could undermine the promising clinical efficacy of BDQ-containing regimens.
Development of diarylquinoline drugs with superior potency could mitigate the aforementioned concerns associated with BDQ. For example, TBAJ-587 has greater in vitro potency than BDQ and reduced cardiovascular liability (25). Although mutation of Rv0678 reduces the in vitro activity of TBAJ-587 to a similar degree as BDQ, TBAJ-587 remains more potent than BDQ against such mutants and therefore may be more effective at killing them or preventing their selection during treatment (26).
In the current experiment, we evaluated the dose-dependent effects of BDQ (B) and TBAJ-587 (S) against an Rv0678 loss-of-function mutant compared to the wild-type H37Rv parent to determine the impact of such mutations on the activity of BDQ and its contribution to the efficacy of the BPaL and BPaMZ regimens. The results indicate that replacing BDQ in these regimens with TBAJ-587 could increase their efficacy against wild-type M. tuberculosis as well as Rv0678 mutants and reduce the emergence of resistance to diarylquinolines and companion drugs.

RESULTS
Bacterial strains and mouse infection model. The experimental schemes indicating the regimens used against the wild-type strain, M. tuberculosis H37Rv, and an isogenic mutant with an IS6110 insertion in the Rv0678 gene are given in Tables S1 and S2 in the supplemental material, respectively. Using the broth macrodilution method in 7H9 media, the MICs of BDQ and TBAJ-587 against the H37Rv strain were 0.0625 and 0.016, respectively, while against the Rv0678 mutant, the MICs were 0.5 and 0.0625, respectively.
One day after high-dose aerosol infection of BALB/c mice with either strain, mean (6 standard deviation [SD]) lung CFU counts were 4.11 6 0.06 log 10 for H37Rv and 4.19 6 0.08 for the Rv0678 mutant. Mean CFU counts increased over the following 2 weeks to 7.79 6 0.11 and 7.75 6 0.10, respectively, when treatment began.
Response to treatment. (i) Mice infected with wild-type M. tuberculosis H37Rv. As shown in Fig. 1A and Table S3, after 1 month of treatment, BDQ at 25 mg/kg alone reduced the mean lung burden by 3.29 log 10 compared to day 0 (D0) and was more active than the combination of PMD and LZD. When added to PMD and LZD, BDQ reduced the mean CFU count by an additional 2.88 log 10 (i.e., BPaL versus PaL). The combination with PMD, MXF, and PZA was much more active than PMD plus LZD, and the addition of BDQ reduced the mean CFU count by an additional 1.77 log 10 (i.e., BPaMZ versus PaMZ). Treatment with TBAJ-587 at 25 and 50 mg/kg resulted in reductions of lung CFU burden that were more than 1.5 log 10 greater than observed with BDQ alone (P , 0.0001). There was no statistically significant difference between TBAJ-587 at 25 mg/kg than at 50 mg/kg. When combined with PMD and LZD, the effect of TBAJ-587 again reduced the mean CFU burden by .1.5 log 10 compared to BDQ. The S 25 PaL regimen was significantly more active than B 25 PaL (P = 0.0002) and not significantly less active than S 50 PaL (P = 0.0954). Similarly, in combination with PMD, MXF, and PZA, TBAJ-587 reduced the mean CFU count by approximately 1.5 log 10 more than BDQ did. The S 25 PaMZ regimen was significantly more active than B 25 PaMZ (P = 0.0001) and not significantly different from S 50 PaMZ (P = 0.9049). The response to monotherapy was also evaluated at month 2 ( Fig. 1B) and showed that both diarylquinolines continued to reduce the lung CFU burden but that TBAJ-587 continued to kill at a higher rate than BDQ.
(ii) Mice infected with M. tuberculosis H37Rv with an Rv0678 mutation. As shown in Fig. 2A and Table S4, dose-dependent reductions in lung CFU were observed with both BDQ (at 12.5, 25, and 50 mg/kg) and TBAJ-587 (at 25 and 50 mg/kg) monotherapy. Both diarylquinolines (DARQs) were significantly less active against the mutant than against the wild-type H37Rv strain, but a similar difference in potency was observed between BDQ and TBAJ-587 against each strain. BDQ at 25 mg/kg reduced the mean CFU count by only 0.56 log 10 compared to D0. TBAJ-587 at 25 mg/kg reduced the mean CFU count by 2 log 10 and was significantly more active than BDQ at any dose (P , 0.0001) but less active (P = 0.0269) than TBAJ-587 at 50 mg/kg. PMD, in combination with either LZD or MXF and PZA, had similar activity against each strain. Both BDQ and TBA-587 significantly (P , 0.0001) increased the activity of the PaL and PaMZ combinations at month 1 (M1). As expected, both BDQ and TBAJ-587 contributed a smaller effect size in combination with PaL against the Rv0678 mutant than that  Selection of drug-resistant mutants. (i) Mice infected with wild-type M. tuberculosis H37Rv. At D0, the mean frequencies of CFU able to grow on agar containing 0.06 and 0.25 mg/ml of BDQ were 4.2 Â 10 25 and 1.8 Â 10 25 , respectively. Plates containing 0.06 mg/ml BDQ were used to quantify the resistant subpopulation at subsequent time points, based on prior evidence of its effectiveness for quantifying mutants with low-level resistance to BDQ (27). All mice infected with M. tuberculosis H37Rv and treated with BDQ monotherapy at 25 mg/kg for 1 or 2 months demonstrated selective amplification of BDQ-resistant CFU, which represented approximately 3% and 18% of the recovered CFU after 1 and 2 months, respectively (Table 1). However, it should be noted that the total number of CFU isolated on BDQ-containing plates decreased from D0 to M1 to M2 (Table S5), indicating that BDQ treatment reduced the size of the subpopulation of CFU able to grow on BDQ but not nearly as fast as it reduced the number of fully susceptible CFU. Monotherapy with TBAJ-587 at either 25 or 50 mg/kg prevented the selection of spontaneous BDQ-resistant mutants, with the exception of a single colony recovered from one of five mice after 2 months of TBAJ-587 monotherapy at 25 mg/kg (representing an estimated one-third of all CFU recovered). As expected, the BPaL and BPaMZ regimens resulted in less selection of BDQ-resistant CFU compared to BDQ monotherapy. Only three mice and one mouse receiving BPaL and BPaMZ, respectively, for 1 month harbored BDQ-resistant CFU. Replacing BDQ with TBAJ-587 at either 25 or 50 mg/kg in combination with PaL or PaMZ prevented the selection of spontaneous BDQ-resistant mutants. As expected, each diarylquinoline prevented selection of PMD-resistant mutants when combined with PaL.
(ii) Mice infected with M. tuberculosis H37Rv with an Rv0678 mutation. Among mice infected with the Rv0678 mutant (Table 2), the selection of CFU with a higher level of BDQ resistance (as quantified by growth on agar containing BDQ at 1 mg/ml) was rare, being observed as a very low percentage of recovered CFU in only 2 of 5 mice (average frequency of 1.4 Â 10 26 CFU) treated with PaL for 1 month, 1 of 3 surviving mice (frequency of 1.9 Â 10 27 CFU) treated with BDQ 12.5 mg/kg for 1 month, and 1 of 5 mice (frequency of 2.7 Â 10 25 CFU) treated with B 25 PaL for 1 month. No CFU resistant to BDQ at 1 mg/ml were recovered from mice treated with TBAJ-587 alone or in combination for 1 to 2 months or from mice treated with BDQ alone or in combination for 2 months. Selection of PMD-resistant mutants was more evident among mice infected with the Rv0678 mutant than those infected with the wild-type H37Rv and was associated with receipt of BDQ rather than TBAJ-587. PMD-resistant mutants were recovered from 10 of 10 mice receiving BPaL for 1 to 2 months, with similar CFU counts at both time points (Table S6), and comprised over 6% of the bacterial population after 2 months of treatment. In contrast, among 20 mice receiving SPaL for 1 to 2 months, only 1 mouse harbored PMD-resistant CFU at a frequency of 0.36%, indicating that TBAJ-587 more effectively eradicated double mutants with resistance to both BDQ (via the baseline Rv0678 mutation) and PMD (via spontaneous resistance mutations).

DISCUSSION
The BPaL regimen is now an important fully oral, short-course treatment option for XDR-TB and difficult-to-treat MDR-TB (1,14). BPaMZ has demonstrated potential to be an even shorter-course regimen for selected MDR-TB and possibly DS-TB (15). Therefore, it is of great concern that Rv0678 variants with reduced susceptibility to BDQ are reported with increasing frequency among patients treated with BDQ or clofazimine, including at least 1 participant treated with BPaL in the Nix-TB trial who relapsed with an Rv0678 mutant (14). In addition, several studies have reported Rv0678 variants among baseline MDR-TB isolates obtained prior to any known BDQ or clofazimine exposure (19,23,24), suggesting that they may be enriched among MDR-TB isolates due to selection by other factors. The results presented here confirm that loss-offunction mutations in Rv0678 significantly reduce the efficacy of BDQ in vivo (19) and show for the first time that they also significantly reduced the efficacy of the novel BPaL regimen. Mice infected with the Rv0678 mutant required 2 months of BPaL treatment to reach the same CFU count as observed in wild-type infected mice treated for 1 month. Moreover, spontaneous BDQ-resistant mutants were selectively amplified by BPaL treatment in wild-type-infected mice, approaching or surpassing the 1% threshold commonly used to define resistance by the agar proportion method in 3 of 5 mice. Interestingly, BDQ resistance was also observed in one of five mice treated with BPaMZ for 1 month. Therefore, although the likelihood of BDQ resistance emerging during treatment is undoubtedly higher when BDQ is combined with less effective companion drugs, it should be recognized that BDQ produces strong selective pressure favoring amplification of spontaneous BDQ-resistant mutants even in these highly active regimens.
Inadvertent treatment of patients infected with an Rv0678 mutant, perhaps even as a heteroresistant subpopulation, could lead to selection of additional mutations  TBAJ-587 against a Bedaquiline-Resistant Rv0678 Mutant Antimicrobial Agents and Chemotherapy conferring resistance to companion drugs. In the present study, dual BDQ and PMD resistance emerged in all 5 mice infected with the Rv0678 mutant and treated with BPaL for 2 months. In stark contrast, no PMD-resistant mutants were isolated from wildtype-infected mice treated with BPaL. These results raise concerns that inadvertent treatment of patients infected with an Rv0678 mutant with BPaL could lead to a dangerous new form of multidrug resistance defined by resistance to BDQ and PMD, which would likely extend to delamanid (28,29). Surprisingly, the BPaMZ regimen, like the PaMZ regimen, had similar bactericidal effects against both the wild-type and mutant infections, indicating that the contribution of BDQ to the efficacy of BPaMZ was not affected by the Rv0678 mutation. While further study is needed to confirm this observation and explore its potential mechanism, it is conceivable that PZA reduces the function of the MmpL5/MmpS5 transporter through disruption of membrane potential (30) or has other synergies with BDQ that enable bactericidal effects at lower intrabacillary BDQ concentrations.
In the present study, replacement of BDQ with the more potent diarylquinoline TBAJ-587 improved the bactericidal activity of the BPaL and BPaMZ regimens against the wild-type H37Rv strain, indicating its potential to shorten the duration of treatment needed to prevent relapse (31). The present work also demonstrates that TBAJ-587 is more effective than BDQ against an isogenic Rv0678 mutant. Although loss of Rv0678 function causes a similar shift in susceptibility to BDQ and TBAJ-587, TBAJ-587 retains superior potency. At 25 mg/kg/day, BDQ loses most of its bactericidal activity against the Rv0678 mutant, whereas TBAJ-587 exhibits bactericidal activity (i.e., 2-log kill over 1 month) as monotherapy. Indeed, SPaL was practically as effective against the mutant as BPaL was against the wild-type H37Rv strain. Interestingly, replacing BDQ with TBAJ-587 had the smallest apparent benefit in the BPaMZ regimen against the Rv0678 mutant.
As a function of its superior activity against Rv0678 mutants, TBAJ-587 prevented the emergence of new drug resistance more effectively than BDQ. In mice infected with the wild-type H37Rv strain, TBAJ-587, both alone and in combination with PaL or PaMZ, nearly abolished the selective amplification of spontaneous BDQ-resistant CFU. Whereas BDQ resistance emerged in 100%, 60%, and 20% of mice treated with BDQ alone, BPaL, and BPaMZ, respectively, it was observed in only 5% of mice treated with TBAJ-587 alone and in none of the 10 mice each receiving SPaL or SPaMZ. Taken together, these results indicate that, in addition to improving efficacy, replacing BDQ with TBAJ-587 would make regimens like BPaL and BPaMZ more robust to the emergence of diarylquinoline resistance. Importantly, we did not observe selection of higher level, or "second-step," BDQ resistance despite treating the Rv0678 mutant infection with BDQ or TBAJ-587 alone. This is likely a function of the low frequency of viable spontaneous atpE mutants and their fitness costs observed in vivo (27). Although atpE mutations have been identified in a small number of BDQ-resistant clinical isolates to date (32,33), Rv0678 mutations have been far more prevalent. Therefore, overcoming Rv0678-mediated resistance with more potent diarylquinolines like TBAJ-587 could greatly extend the utility and longevity of this important new class of drugs.
Use of TBAJ-587 in place of BDQ in the BPaL regimen in mice infected with the Rv0678 mutant also largely prevented the selection of spontaneous PMD-resistant double mutants. Following 2 months of treatment, all five BPaL-treated mice harbored dual BDQ/PMD-resistant isolates, compared with zero of 10 SPaL-treated mice. In fact, BPaL and PaL resulted in similar absolute numbers of dual BDQ/PMDresistant mutants at the end of 2 months of treatment, indicating that BDQ did not contribute to faster elimination of spontaneous PMD-resistant mutants in the Rv0678 mutant background. Thus, the benefits of replacing BDQ with TBAJ-587 in the BPaL regimen are expected to include the prevention of PMD resistance (i.e., a new form of multidrug resistance to diarylquinolines and nitroimidazoles) when infection with an Rv0678 mutant is inadvertently treated with a regimen combining a diarylquinoline with PaL.

MATERIALS AND METHODS
Bacterial strains. The laboratory strain, M. tuberculosis H37Rv, and a spontaneous bedaquiline-resistant mutant with an IS6110 insertion in Rv0678 at amino acid 116/nucleotide 349 (aa116/nt349) were used in this study. The Rv0678 mutant was previously identified as BDQ-8 when it was isolated from an untreated mouse infected by H37Rv (27). The Rv0678 mutation and the absence of other mutations in genes associated with drug resistance were confirmed by whole-genome sequencing. The MICs of BDQ and TBAJ-587 against both strains were determined by the broth macrodilution method in complete 7H9 broth using polystyrene tubes.
Infection model. Female BALB/c mice, 6 weeks old, were aerosol infected with ;4 log 10 CFU of each M. tuberculosis strain from a log-phase culture with an optical density at 600 nm (OD 600 ) of ;0.8 on day 214 (D214). Treatment started 2 weeks later (D0). Mice were sacrificed for lung CFU counts on D213 and D0 to determine the number of CFU implanted and the number present at the start of treatment, respectively.
Antibiotic treatment. Mice were treated as indicated in Tables S1 and S2 in the supplemental material at the following doses (mg/kg): 12.5, 25, or 50 mg/kg BDQ, 25 or 50 mg/kg TBAJ-587, 100 mg/kg PMD, 100 mg/kg MXF, 100 mg/kg LZD, and 150 mg/kg PZA once daily 5 days per week for 1 or 2 months, with the exception of BDQ at 50 mg/kg that was given twice daily at 25 mg/kg. BDQ and TBAJ-587 were formulated in 20% hydroxypropyl-b-cyclodextrin solution acidified with 1.5% 1N HCl. PMD was prepared in the CM-2 formulation as previously described (34). LZD was prepared in 0.5% methylcellulose. MXF and PZA were prepared in water. Except as noted for BDQ at 50 mg/kg, the diarylquinolines and PMD were administered in a single gavage in the morning, and LZD, MXF, and PZA were administered in the afternoon.
Evaluation of drug efficacy. Efficacy determinations were based on lung CFU counts after 1 month and 2 months of treatment. At each time point, lungs were removed aseptically and homogenized in 2.5 ml phosphate-buffered saline (PBS). Lung homogenates were plated in serial dilutions on 0.4% charcoal-supplemented 7H11 agar with 2Â selective antibiotics, i.e., to compensate for charcoal absorption of these drugs. Final concentrations for these antibiotics were 20 mg/ml cycloheximide, 100 mg/ml carbenicillin, 50 mg/ml polymyxin, and 40 mg/ml trimethoprim.
Evaluation of resistance selection. At each time point, lung homogenates from mice infected with the wild-type H37Rv strain were plated in parallel on drug-free 7H11 plates and on the same plates supplemented with 0.06 mg/ml of BDQ. Although this concentration is below the recommended critical concentration of 0.25 mg/ml for clinical use, it enables more effective recovery of Rv0678 and pepQ mutants with small shifts in drug susceptibility (i.e., 2 to 8 times increases in MIC) from lung homogenates of mice, especially when mice are on treatment with BDQ, which accumulates in lungs and is carried over onto plates, increasing the effective concentration of BDQ, as previously described (27,35). Aliquots of lung homogenates from mice infected with the Rv0678 mutant were plated on agar containing a higher BDQ concentration (1 mg/ml) to evaluate for atpE-mediated resistance (27) and on agar containing 2 mg/ ml of PMD.
Statistical analysis. Differences between regimens were assessed by one-way analysis of variance (ANOVA) with Dunnett's multiple-comparison correction using GraphPad Prism version 8.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 0.1 MB.