Implications of edge cases for antimicrobial susceptibility predictions for rifampin, delamanid, and pretomanid with the WHO Mycobacterium tuberculosis mutation catalog

ABSTRACT The WHO mutation catalog is the reference for genotypic antimicrobial susceptibility predictions (ASPs) for Mycobacterium tuberculosis . To increase sensitivity, WHO endorsed additional grading rules to classify mutations with insufficient experimental evidence. For rifampin, this includes any non-silent rpoB mutation in the rifampin resistance-determining region (RRDR). For delamanid and pretomanid, the rules include loss-of-function (LoF) variants in six resistance genes, including fbiC . Here, we describe three edge cases in which these additional grading rules overcalled resistance. The first case was a low-frequency rpoB variant in RRDR caused by a sequencing artifact. The second case involved a tandem repeat region starting at the end of fbiC . Because the H37Rv reference genome carries 2.7 repeats, deletion of one or two full repeats caused mapping errors, making these deletions appear as minority variants that trigger the fbiC LoF rule. However, assembly showed that these deletions did not affect the fbiC coding region because the remaining repeat substituted for the deleted repeats. The third case was a fbiC frameshift in lineage 4.6.1 Uganda genotype strains, caused by a 19 bp duplication upstream of the same tandem repeat region. Although this frameshift affects the fbiC coding region and meets the fbiC LoF rule, sequence assembly revealed only a modest impact on the end of fbiC . Although catalog performance remains high, WHO and others have acknowledged that, in specific situations, exceptions to its additional grading rules may result in unintended overcalling of resistance. Thus, systematic external quality control and communication of edge cases are essential. IMPORTANCE Antimicrobial susceptibility predictions (ASPs) for Mycobacterium tuberculosis using next-generation sequencing and the WHO mutation catalog are increasingly used for clinical decision-making. This avoids systematic false-resistant results of real-time PCR- or hybridization-based ASP assays caused by mutations that do not confer resistance (e.g., synonymous rifampin resistance-determining region [RRDR] mutations) and, potentially, subsequent inappropriate treatment adjustments. However, we demonstrate that false-resistant results occur even using the WHO mutation catalog for rifampin, delamanid, and pretomanid. Similar exceptions linked to the additional grading rules of the catalog are possible for other drugs, underlining the importance of monitoring the performance of the catalog continuously.