Hidden TB subpopulations reveal missed rifampicin resistance

Tuberculosis remains a global threat, and modern genetic tools such as whole genome sequencing (WGS) are increasingly used to diagnose infections and predict drug resistance. But clinicians and public health teams have sometimes seen a mismatch: a laboratory’s genotypic prediction from WGS can say a Mycobacterium tuberculosis sample is susceptible, while phenotypic drug susceptibility testing shows resistance. That mismatch can delay correct treatment and complicate surveillance. Philip W. Fowler led a study to investigate whether small resistant subpopulations within clinical samples, and the presence of compensatory mutations, can explain these discrepancies for the key antibiotic rifampicin. The research analyzed a very large data set of 35,538 clinical M. tuberculosis samples to test how different analysis choices affect the sensitivity and specificity of WGS-based rifampicin resistance calls. By comparing standard approaches to ones that explicitly look for minority resistant variants and consider compensatory mutations, the team aimed to pinpoint how resistance is acquired and why some resistant infections are missed by routine genetic tests.

The study tested how requiring different fractions of sequencing reads to call a resistance-associated variant affects performance. When the minimum fraction of reads needed to identify a resistant variant was lowered from 0.90 to 0.05, sensitivity for detecting rifampicin resistance increased from 94.3% to 96.4%, while specificity did not show a significant decline. Allowing compensatory mutations to indicate resistance reduced the false negative rate further. The investigators also compared samples that were uniformly resistant with those that contained resistant subpopulations and found that samples with resistant subpopulations were less likely to carry compensatory mutations than homogeneous resistant samples. Additional analyses of the heterogenous samples revealed distinct clusters with differing amounts of within-sample genetic diversity. These patterns are consistent with different mechanisms of resistance acquisition, including within-host evolution and secondary infections, and the genetic diversity observed suggested that at least 28% of rifampicin resistance cases may arise from secondary infections.

These findings have practical implications for how WGS is used in clinical labs and surveillance programs. A substantial fraction of false negative WGS calls for rifampicin resistance can be explained by resistant subpopulations that are masked when analysis requires very high read fractions to call variants. Relaxing that threshold and incorporating the presence of compensatory mutations into interpretation can reveal resistance that would otherwise be missed, helping clinicians select effective therapy sooner. Recognizing within-sample diversity also offers a window into whether resistance likely evolved inside a patient or was acquired from another infection, information that matters for infection control. The suggestion that roughly 28% of rifampicin resistance may come from secondary infections highlights the need for robust contact tracing and public health measures alongside improved laboratory detection to curb the spread of resistant tuberculosis.