False detection of rifampicin resistance using Xpert® MTB/RIF Ultra assay due to an A451V mutation in Mycobacterium tuberculosis

Abstract

Background: In a 12 month period, three Irish-born adult cases with pulmonary TB were initially diagnosed by Xpert^® MTB/RIF Ultra assay, which detected a rifampicin resistance-conferring mutation prompting treatment as potential MDR cases.

Methods: Further laboratory investigations on the cultured isolates included GenoType MTBDRplus assay, phenotypic drug susceptibility tests using the BD BACTEC MGIT culture system and MIC broth microdilution tests. Sequencing of the rpoB gene was performed using Sanger sequencing and WGS.

Results: Phenotypic drug susceptibility tests determined the isolates to be rifampicin susceptible. Molecular investigations identified an A451V (codon 532) mutation in the Mycobacterium tuberculosis rpoB gene that has not previously been found to cause rifampicin resistance. Genome sequencing revealed that the three isolates' genomes differed by ≤5 SNPs, indicating a high likelihood of recent transmission events. Furthermore, a cluster of six related M. tuberculosis isolates from our in-house typing database showed four were highly related; all were rifampicin susceptible and lacked this mutation.

Conclusions: False detection of rifampicin resistance, albeit rare, should be considered possible with Xpert^® MTB/RIF Ultra assay, particularly in low TB incidence settings. Confirmatory sequencing methods should be performed to prevent the unnecessary use of second-line anti-tuberculous drugs.

Figure 1.

SNP-based analysis performed using the MTBseq v1.0.4 pipeline with minimum stringency threshold confirmed the frequency of the mutant allele at 58% in sample 1 with 88× coverage depth. The frequency of the mutation in samples 2 and 3 was 100% in all instances with coverage in that region 172× and 85×, respectively. A maximum likelihood tree was constructed with the RAxML GUI v2.0.0. The analysis parameters selected were ML + rapid bootstrap with 1000 reps and compute branch lengths selected. The resulting output file was visualized on the interactive tree of life website.