A retrospective cohort study was conducted using a broadly representative sample of
Mtb isolates of varying resistance profiles from patients across South Africa. A BDQ naïve group was used to determine the wild-type distribution using all available RR-TB isolates from national drug resistance survey (NICD, 2016 ). The Clinical Laboratory Standards Institute (CLSI) recommends a sample size of > 300 isolates for ECV determination (CLSI, 2008 ). As the number was below what was required, additional isolates were included from routine drug resistance surveillance programs (GERMS-SA, 2015 ) in the country. The provincial distribution of isolates is shown in Fig. S1. A BDQ exposed group was used for comparison. These included patients on BDQ therapy having an elevated MIC on baseline testing on any method tested. The study was approved by the Human Research Ethics Committee of the University of the Witwatersrand, Johannesburg, South Africa under R14/49 and M160667.
The ECV study methods were designed in accordance with the CLSI M23-A3 guidance document: “Development of
in vitro susceptibility criteria and quality control parameters” (CLSI, 2008 ).
Mtb complex isolates, confirmed with the TBcID antigen test (Becton Dickinson, USA) and purity tested, we recovered on Middlebrook 7H10 agar and MGIT for MIC testing on M10A/BMD and MGIT respectively. BDQ MIC was performed by the M10A and BMD methods as previously described (Lounis et al., 2016 (
link), Kaniga et al., 2016 (
link)). CFZ MICs were performed on BMD only following the same methodology as was done for the BDQ MIC on BMD. Custom-made microtiter plates were prepared by Thermo Fisher Scientific (Oakwood Village, Ohio, USA). The MIC for the M10A and BMD methods was defined as the lowest concentration of the drug-containing plate or well respectively, with no visual growth. The batch results were valid only if the H37Rv control fell within the published QC range (Lounis et al., 2016 (
link)). The BACTEC™ MGIT™ 960 DST methods were followed as previously described (Torrea et al., 2015 (
link), Rusch-Gerdes et al., 2006 (
link)) with slight modification to allow specific MICs to be tested for BDQ (Supplementary Information Box 1). The EpiCentre TBExist software (Becton Dickinson, USA) was used for interpretation of MIC for this method, and the incubation period extended from the recommended 13 to 28 days, adjusted for slow growing drug resistant isolates. The MGIT MIC was defined as the lowest concentration of drug-containing tube reported having a Growth Unit < 100.
WGS was performed using the MiSeq (Illumina, UK). Library preparation was performed using the Nextera-XT library preparation kit (Illumina, UK) and sequencing performed using the 2 × 300 bp MiSeq cartridge v.3 (Illumina, UK) with a target of 30 ×-50 × paired coverage (~ 80–100X coverage). CLC Genomics Workbench 8.5.1 was used to detect RAVs within the genes
atpE,
pepQ, Rv1979c and
Rv0678 using Reference mapping against the annotated reference genome H37Rv (NC00962.3) and the quality-based variant analysis tools.
The epidemiological cut-off value (ECV) was defined as the upper limit of the MIC value that separates the wild-type from the non-wild-type population (Schon et al., 2009 (
link), Turnidge and Paterson, 2007 (
link)). An ECV of 95% (ECV
95) was deemed susceptible (S) and an ECV between 95%–99.9% (ECV
99.9) was deemed intermediate (I). For the range of dilutions used in the study, the frequency and cumulative frequency of MIC distribution were calculated for all three methods. The mode for each method was inspected against H37Rv QC range and if it differed by more than one dilution, the set was excluded (EUCAST, 2017 ) as this would imply variability in testing and skew the ECV. Histograms of the MIC distribution by each method were generated and the wild-type ECV estimated by iterative non-linear regression on expanding subsets (Turnidge et al., 2006 (
link)) using the ECOFF finder tool (Turnidge and Paterson, 2007 (
link)) and additional visual inspection. In addition, MIC ranges, MIC
90 and MIC
95 tables were generated. The MICs and associated ECVs were evaluated against the putative genes reported to encode BDQ resistance as well the MIC and RAV data of isolates from patients on BDQ based regimens. Lastly, the BDQ and CFZ MICs were cross tabulated to assess cross resistance and a Pearson's correlation coefficient determined.
Ismail N.A., Omar S.V., Joseph L., Govender N., Blows L., Ismail F., Koornhof H., Dreyer A.W., Kaniga K, & Ndjeka N. (2018). Defining Bedaquiline Susceptibility, Resistance, Cross-Resistance and Associated Genetic Determinants: A Retrospective Cohort Study. EBioMedicine, 28, 136-142.
