Of 28 isolates tested, one was reported by Xpert Ultra as “No R detected”, eight yielded an “R Indeterminate” result, and 19 were identified as RR (Table 1).
The reported RS isolate had a combo of Ins431(agCCAc) and Pro439Leu (ccg > ctg), and yielded somewhat deviating Tms for both probes 1 and 2 (68.3 and 72.9°C respectively), but still resulted in a wild type interpretation for both probes (See Additional file 1). Pro439Leu is classified as Ass-w-R-int, while the Ins431(agCCAc) insertion is not listed in the WHO catalogue, but would be Ass-w-R-Int since the isolate was phenotypically resistant on LJ medium. Hence, this case is considered a false-RS result by Xpert Ultra.
From the 8 isolates with an “R Indeterminate” result, six were singleton SNPs that each occurred twice: Gln432Leu, Gln432Pro and Lys446Gln. For codon-432 mutants no Tm was obtained for probe 1 despite successful amplification, while for Lys446Gln the amplification of probe 3 failed (See Additional file 1). The 432 mutants are classified as Ass-w-R, while Lys446Gln is Ass-w-R-Int. All six were RR on LJ medium. Nevertheless, Xpert Ultra failed to identify them as RR, which confirms previous observations of not detecting some codon-432 mutants (3). The Gln432Lys and Gln432Glu mutations were correctly identified as RR by Ultra.
Also for the combo Met434Thr/His445Asp an “R Indeterminate” result was reported, despite the fact that it correctly identified His445Asp as resistant (Tm of 71.9°C for probe 3; See Additional file 1). While we would have expected an impact from Met434Thr on the Tm of probe 1, this was not the case. Rather, probe 2 did not have a successful Tm despite successful amplification (Ct 17.8). This is in line with observations from Ng and colleagues, where codon-434 mutations only impacted probe 2 and not probe 1 (6). However, in our case no Tm was generated. This isolate was resistant on LJ and MGIT. Xpert Ultra failed to identify this combo as RR.
As for the last “R Indeterminate” case, the no-frameshift deletion at the edge of the targeted region (Del427-428(ACCAGC) prohibited amplification of probe 1, while the other probes had cycle threshold (Ct) values between 17.1 and 20.3 (See Additional file 1). Remarkably, this isolate was found phenotypically RS in MGIT (0.5 µg/ml) and on LJ with a minimal inhibitory concentration (MIC)99 of 20 µg/ml and MIC100 of 40 µg/ml. Hence, it seems that this deletion does not cause RR, and at most could be classified as “uncertain significance” or even “Not associated with resistance”. Hence, the obtained Xpert Ultra “R Indeterminate” seems acceptable until more certainty around the relevance of this insertion arises.
Among the 19 rpoB variants reported as RR by Xpert Ultra, three isolates appeared phenotypically RS by both LJ (MIC ≤ 10 µg/ml on LJ medium) and MGIT (0.5 µg/ml) testing: one with Ser428Arg and two with Asp435Ala. The Ultra-RR result was based on the predicted Tm from probe 1 or probe 2 respectively. In the WHO mutation catalogue, Ser428Arg has been observed in 5 RR isolates, while not being present among 24433 RS isolates. Overall, it is known that elusive rpoB mutations can be miss-classified as RS by rapid or imprecise pDST (7). Few pDST data for Ser428Arg mutants are publicly available though. El Maraachi and colleagues reported a MIC of 5 µg/ml in MGIT (8), justifying its classification as Ass-w-R-Int. More data would be needed to investigate the impact of these mutations.
Overall, our results corroborate the observations of Omar and colleagues regarding the “Indeterminate” results for SNPs at codon 432, while we add Lys446Gln as additional “Indeterminate” result and Pro439Leu as a false-RS result. Furthermore, we document other uncommon SNPs and indels across the rpoB gene that are mostly correctly identified as RR by Xpert ultra (V3). Missing RR at diagnosis may delay timely initiation of appropriate treatment with the risk of worse treatment outcome and continued RR-TB transmission, and from a public health perspective even diagnostics-driven selective advantage for less common mutations, as was probably the case for the rpoB_Ile491Phe mutation (9).
On the other hand, we identified two Ass-w-R-Int rpoB SNPs (Asp435Ala, Ser428Arg) that do not seem to cause phenotypic RR on their own. If these mutations are truly phenotypic/clinical RS, the Xpert Ultra generates a false-RR result, with potential unnecessary initiation of multi-drug resistant treatment. The associated risk for the patient may be limited if a safe, short multi-drug resistant TB regimen can be offered.
It must be noted that most of these mutations are uncommon. Their reported frequency as singletons in the WHO mutation catalogue varied from not being reported (n = 9 different mutations within the Ultra target region), to being reported between two and 706 times, compared to 6256 occurrences for the most common Ser450Leu mutation (5). Hence the impact of missed or non-reported RR with delayed treatment adaptation on one hand, or overtreatment with second-line drugs for erroneous RR calling for the investigated mutations will be limited on the global scale. Nevertheless geographical variations in mutation prevalence may occur, and may lead to discordant results for rifampicin-susceptibility testing, complicating patient management. Despite these minor shortcomings, we still fully support the use of rapid molecular testing for detection of (RR-TB, acknowledging that none of the tests is perfect.