Quality control and considering systematic MIC shifts are key when evaluating the role of mmpR5 (Rv0678) frameshifts in bedaquiline resistance

LETTER

The accurate interpretation of genotypic antimicrobial susceptibility testing (gAST) results for bedaquiline for Mycobacterium tuberculosis is crucial as it is a key drug in most 6- to 9-month all-oral regimens for rifampicin-resistant tuberculosis (1, 2). Unfortunately, a large spectrum of resistance mutations is possible in mmpR5 (Rv0678), the most important bedaquiline resistance mechanism that also confers cross-resistance with clofazimine. To improve the sensitivity of gAST, WHO endorsed an additional grading rule, whereby any loss-of-function in mmpR5 (i.e., full-gene deletion, frameshifts, mutations that abolish the start codon, and premature stop codons) should be interpreted as an interim resistance mutation, taking epistasis caused by loss-of-function mutations in mmpL5 into consideration where possible (3, 4).

WHO and others acknowledged that exceptions to these additional grading rules may exist and should be studied because they may overcall resistance (3, 5). Snobre et al. recently proposed that some mmpR5 frameshifts subjected to the additional grading rule might retain susceptibility to bedaquiline by producing functional MmpR5 through alternative reading frames (1). However, when reviewing the Becton Dickinson (BD) BACTEC Mycobacteria Growth Indicator Tube 960 (MGIT) MIC data used to support their hypothesis, we identified several factors that should be considered.

49% (33/68) of the frameshift MGIT MICs analyzed were from Peretokina et al., which includes data from Zimenkov et al. from the same laboratory, whereas the remaining results were from three separate laboratories (1, 6–10). Only one of the 33 frameshift MICs (3%, 95% confidence interval [CI] 0–16) from Peretokina et al. was above the current interim critical concentration (CC) of 1 µg/mL, whereas 31 of the remaining 35 frameshift MICs (89%, 95% CI 73–97) from the remaining studies were in the resistant range (1). This suggested that the MGIT MICs in Peretokina et al. were systematically low, which was acknowledged by the authors and is in line with the following observations (8).

First, although no formal quality control (QC) range for the H37Rv control strain currently exists for bedaquiline, the MGIT MICs from Peretokina et al. appeared to be shifted towards lower values compared with the data considered by WHO to set the current CC for MGIT (8, 11).

Second, the 148 MGIT MICs of isolates without mmpR5 or atpE mutations prior to bedaquiline treatment from Peretokina et al. were between ≤0.03 and 0.25 µg/mL with a mode at 0.125 µg/mL (Table S1), which was lower than the modes of 0.25 or 0.5 µg/mL observed in the MIC distributions used by WHO (8, 11). The equivalent mode for 7H11 was ≤0.03 µg/mL (Table S1) and thus again lower than 0.06 µg/mL in the WHO distributions (8, 11).

Third, 12 of the 15 isolates with MGIT MICs in the resistant range either had or were later found to have an atpE mutation (Table S2) (8). For 7H11, the same applied to seven of the eight MICs in the resistant range (Table S2). Given that atpE mutations typically confer larger MIC increases than mmpR5, this indicates systematic shifts towards lower MICs for both media (12).

Fourth, 17 of 33 isolates with mmpR5 frameshifts MICs that Snobre et al. included from Peretokina et al. arose during bedaquiline treatment of seven patients who had genotypically wild-type isolates at baseline. Notably, the detection of the frameshift usually correlated with an MIC increase in MGIT as well as 7H11 (Table S3) (8). This is evidence of positive selection and means that epistasis linked to mmpL5 can be excluded as a confounder (3, 13).

Systematic MIC shifts have been described for other anti-TB agents in the past and, depending on their magnitude and direction, are difficult to detect if only the CC is tested or MIC distributions are not scrutinized (11, 14, 15). Instead, the inclusion of a QC strain that yields an on-scale MIC is key to detect shifts towards lower or higher MICs (16). For example, a BD BACTEC PZA reagent quality issue was detected early by the California Department of Public Health because of abnormally high MICs for H37Rv with various lots of reagents, which coincided with an increased frequency of likely false-resistant phenotypic results for clinical strains without gAST support. BD’s user QC protocol, which recommends testing only the CC, initially did not reveal any problems (17). Later, false resistance for PZA was noted globally, ultimately resulting in a recall of affected reagent lots by BD (18). This underlines that all manufacturers of pAST devices for anti-TB agents should establish QC ranges/targets that should be published for internal as well as external QC. Indeed, EUCAST calls for setting QC ranges/targets for the EUCAST reference MIC method as well as any other method that is calibrated against the reference method (19).

In summary, the in vitro evidence for susceptibility supporting the hypothesis by Snobre et al. was largely from Peretokina et al., which should have been excluded from the analysis because of the likely systematic MIC shift that resulted in false-susceptible results using the current CC (1, 8). Given the clinical implications of their hypothesis, additional data sets with appropriate QC results should be interrogated to study this important question.

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/aac.00266-25.

Supplemental tables. aac.00266-25-s0001.xlsx.

Tables S1 to S3.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.