Molecular methods for tuberculosis trials: time for whole-genome sequencing?

The fi rst genetic sequencing of Mycobacterium tuberculosis was a momentous achievement that required years of painstaking eff ort and substantial funding. It is remarkable that 15 years later, advances in laboratory techniques and informatics have enabled whole-genome sequencing to be done for hundreds of M tuberculosis isolates, making this information available for epidemiological studies. The increasing availability of whole-genome sequencing has raised questions about the interpretation of this new method, as well as its public health and clinical applications and utility. In the Lancet Respiratory Medicine, Josephine Bryant and colleagues report a study comparing molecular typing by mycobacterial interspersed repetitive unitvariable number of tandem repeat (MIRU-VNTR) versus whole-genome sequencing for isolates of M tuberculosis from 47 consecutive participants in a phase 3 trial of tuberculosis treatment. These participants had recurrent positive cultures after at least 17 weeks of treatment; the two molecular methods were used to determine whether these recurrences were a result of relapse or re-infection. None of these patients had treatment failure, none acquired drug resistance, and 11 were HIV-positive. Five recurrences were single isolated positive cultures followed by repeated negative cultures without treatment; four of these were attributed to lab contamination. On the basis of whole-genome sequencing of the remaining 42 pairs, 33 were judged to be relapses, three were considered re-infection, four had evidence of mixed infection initially, of which one of the two initial strains relapsed, two had single strains initially but mixed infection with two strains was identifi ed at the time of recurrence implying that both relapse and re-infection had occurred. Several fi ndings of this study are worthy of comment. Of the 42 participants with clinically signifi cant disease recurrence, at most fi ve (12%) could be considered to be a result of re-infection (including two with mixed infection at the time of recurrence), whereas relapse accounted for 37 recurrences (including the four patients with evidence of mixed infection initially). This proportion of re-infections is substantially lower than reports from other studies from South Africa, in which more than half of those with recurrent tuberculosis had evidence of re-infection. However, in those studies, most patients were HIV positive with advanced immune suppression, whereas in the Bryant study only 22% had HIV and patients with advanced immune suppression were excluded from the parent trial. For patients with HIV—as long as it is well controlled—the primary determinant of long-term outcomes seems to be the adequacy of their tuberculosis treatment. Although whole-genome sequencing is the new test being compared with MIRU-VNTR, it is still generally assumed to be the new gold standard, in view of the wealth of detailed data provided by analysis of single nucleotide polymorphisms (SNPs). MIRU-VNTR and whole-genome sequencing were concordant in the three patients judged to have re-infection, and of the 33 paired isolates that had few or no SNP diff erences by wholegenome sequencing, 27 also had identical MIRU-VNTR at all 24 loci. As with many clinical and epidemiological comparisons, the discordant cases are the most informative. Of the six discordant results deemed relapses on the basis of whole-genome sequencing, fi ve (83%) were discordant by a single MIRU-VNTR locus, meaning