Isoniazid Resistance in Mycobacterium tuberculosis Is a Heterogeneous Phenotype Composed of Overlapping MIC Distributions with Different Underlying Resistance Mechanisms

MIC testing using the Bactec mycobacteria growth indicator tube system 960 of 70 phylogenetically diverse, isoniazid-resistant clinical strains of Mycobacterium tuberculosis revealed a complex pattern of overlapping MIC distributions. Whole-genome sequencing explained most of the levels of resistance observed.

terium tuberculosis H37Rv ATCC 27294 as the control strain. Four serial 2-fold dilutions were prepared from an INH stock solution to provide a final test range of 0.016 to 0.25 g/ml for susceptible strains and H37Rv, 0.25 to 4 g/ml for inhA promoter mutant isolates with or without a concurrent inhA coding mutation, 1 to 16 g/ml for S315T/N mutant isolates, and 4 to 64 g/ml for isolates with double mutations in katG and the inhA promoter. Whole-genome sequencing (WGS) was carried out with the Nextera-XT DNA kit to prepare paired-end libraries of 150-bp read lengths for Illumina sequencing. Data analysis and single-nucleotide polymorphism calling were performed using the MTBseq pipeline (9) (for additional details, see Supplementary methods in the supplemental material).
The six susceptible controls had INH MICs of 0.03 to 0.06 g/ml ( Fig. 1 and Table S2 in the supplemental material). In contrast, resistant strains displayed a series of overlapping MIC distributions. Strains that had only mutations that are interrogated by the WHO-endorsed genotypic AST assays (i.e., the Hain GenoType MTBDRplus version 2 and Nipro NTMϩMDRTB version 2) resulted in three MIC distributions (10), i.e., strains that had only an inhA promoter change or a mutation at codon 315 of katG had nonoverlapping MIC distributions of 0.25 to 2 and 4 to 16 g/ml, respectively, and strains with both mutations displayed MICs of 8 to 64 g/ml. The variation in these distributions was likely largely due to the normal variation in MIC testing (i.e., even in the same laboratory, a variation of plus or minus one dilution is inevitable, which is further exacerbated by the variation in testing between laboratories).
The precise level of resistance cannot be predicted by using the Hain and Nipro assays alone because mutations that are not interrogated by these assays can increase the MICs. To some extent, this can be overcome by using WGS data, provided that known mutations with predictable effects are identified (i.e., the level of resistance could not be fully explained even with WGS). For example, some but not all strains with loss-of-function mutations in katG had MICs of Ͼ64 g/ml (Fig. 1). Moreover, a C deletion 34 nucleotides upstream of the main transcriptional start site of inhA likely accounted for the unusually high MIC of 64 g/ml for the katG S315N mutant (11). In contrast, another mutation upstream of inhA at codon 203 of fabG1, which is known to result in the overexpression of inhA by creating an alternative promoter, did not appear to increase the MIC above the level explained by the katG S315T mutation in the strain in question (i.e., 8 g/ml) (12). Similarly, there was an almost complete overlap between FIG 1 Isoniazid MIC results stratified by known or likely resistance mutations in the coding region of katG or inhA or mutations that result in the overexpression of inhA. All of the latter mutations are upstream of inhA, but "promoter" is used to highlight mutations in the Ϫ16 to Ϫ8 region upstream of the transcriptional start site of the fabG1-inhA operon, which can be detected by the WHO-endorsed Hain GenoType MTBDRplus version 2 and Nipro NTMϩMDRTB version 2 assays (all mutations interrogated by these assays are shown in bold in the key of the plot [25]). gWT, genotypically wild-type strain (i.e., strain without known resistance mutations); LOF, loss-of-function mutation (i.e., insertion, deletion, or nonsense mutation).
the MIC distributions of strains that harbored only inhA promoter mutations and those that had additional inhA coding mutations at codon 21, 94, or 194 (i.e., 0.25 to 2 g/ml versus 0.5 to 2 g/ml).
This study was conducted in a single center, so we could not quantify the effect of laboratory-to-laboratory variation. Data from additional laboratories are needed to define robust quality-control ranges/targets to evaluate whether the current CC of 0.1 g/ml corresponds to the epidemiological cutoff and to define the lower and upper ends of the various resistance mechanisms more accurately (3,7,13). Moreover, breakpoints cannot be set based on MIC data alone (5,14,15). Nevertheless, the MIC distributions in this study have implications for defining low-level resistance. The current clinical breakpoint of CLSI to define low-level resistance (i.e., 0.4 g/ml, which corresponds to 0.5 g/ml using our dilution series) does not correspond to the upper end of the MIC distribution of inhA promoter mutants. Instead, the upper end of the MIC distribution of a mechanism has to be considered when assessing whether it is treatable with either the standard or an elevated dose of INH (16)(17)(18). For strains with only inhA promoter mutations, this target concentration would be 1 or 2 g/ml (i.e., at least 10 times higher than the current CC) (3,7). Should pharmacokinetic/pharmacodynamic, drug penetration, and clinical outcome data confirm that this target is achievable, 1 or 2 g/ml may be adopted instead of 0.4 g/ml (14,15). This would avoid splitting the MIC distribution of inhA promoter mutants and would, consequently, reduce or eliminate their misclassification as high-level resistant because of the technical variation in AST, as is the case with the current clinical breakpoint of CLSI.
One argument against setting a clinical breakpoint at 1 or 2 g/ml might be that it would result in the misclassification of strains with both inhA promoter and coding mutations as low-level resistant, as stressed in a previous consensus statement (7,19). However, two aspects should be borne in mind. First, only 3% (95% confidence interval, 1% to 6%) of strains with inhA promoter mutations in the Ϫ16 to Ϫ8 region that do not have katG mutations have additional inhA coding mutations based on recent WHO population-level surveillance data from seven countries (20). Therefore, in most settings, misclassifications of double mutants would be rare compared with the increased ability to detect inhA promoter mutants with a higher clinical breakpoint. Second, the effect of these coding mutations on the INH MIC and thus clinical outcome is likely modest at worst, but more MIC data are needed for the mutations at different inhA codons (21)(22)(23)(24). Nonetheless, it might be advisable for countries that conduct routine WGS to err on the side of caution by classifying these double mutants as high-level resistant until clinical data to the contrary are available, although, in practice, it would be challenging to conduct a sufficiently powered study to address this question because these mutations are so rare. The funders had no role in study design, data collection, interpretation, or the decision to submit the work for publication. C.U.K. is a consultant for the WHO Regional Office for Europe, QuantuMDx Group Ltd., and the Foundation for Innovative New Diagnostics, which involves work for Cepheid Inc., Hain Lifescience, and WHO. C.U.K. is an advisor to GenoScreen. The Bill & Melinda Gates Foundation, Janssen Pharmaceutica, and PerkinElmer covered C.U.K.'s travel and accommodation to present at meetings. The Global Alliance for TB Drug Development Inc. and Otsuka Novel Products GmbH have supplied C.U.K. with antibiotics for in vitro research. C.U.K. is collaborating with YD Diagnostics.