Whole genome sequencing reveals large deletions and other loss of function mutations in Mycobacterium tuberculosis drug resistance genes.

Drug resistance in Mycobacterium tuberculosis, the causative agent of tuberculosis disease, arises from genetic mutations in genes coding for drug-targets or drug-converting enzymes. SNPs linked to drug resistance have been extensively studied and form the basis of molecular diagnostics and sequencing-based resistance profiling. However, alternative forms of functional variation such as large deletions and other loss of function (LOF) mutations have received much less attention, but if incorporated into diagnostics they are likely to improve their predictive performance. Our work aimed to characterize the contribution of LOF mutations found in 42 established drug resistance genes linked to 19 anti-tuberculous drugs across 32689 sequenced clinical isolates. The analysed LOF mutations included large deletions (n=586), frameshifts (n=4764) and premature stop codons (n=826). We found LOF mutations in genes strongly linked to pyrazinamide (pncA), isoniazid (katG), capreomycin (tlyA), streptomycin (e.g. gid) and ethionamide (ethA, mshA) (P<10-5), but also in some loci linked to drugs where relatively less phenotypic data is available [e.g. cycloserine, delaminid, bedaquiline, para-aminosalicylic acid (PAS), and clofazimine]. This study reports that large deletions (median size 1115 bp) account for a significant portion of resistance variants found for PAS (+7.1% of phenotypic resistance percentage explained), pyrazinamide (+3.5%) and streptomycin (+2.6%) drugs, and can be used to improve the prediction of cryptic resistance. Overall, our work highlights the importance of including LOF mutations (e.g. large deletions) in predicting genotypic drug resistance, thereby informing tuberculosis infection control and clinical decision-making.