Diagnostic Implications of Inconsistent Results Obtained with the Xpert MTB/Rif Assay in Detection of Mycobacterium tuberculosis Isolates with an rpoB Mutation Associated with Low-Level Rifampin Resistance

Abstract

Xpert-MTB/Rif is one of the most frequently used molecular screening tests for multidrug-resistant tuberculosis worldwide. We report false-negative assay results in the presence of rpoB Leu533Pro, which is associated with low-level phenotypic rifampin resistance. Accurate and timely confirmation of rifampin susceptibility results obtained with Xpert-MTB/Rif is imperative.

TEXT

The accurate diagnosis of drug-resistant and multidrug-resistant (MDR) tuberculosis (TB) is imperative to initiate adequate treatment, to avoid transmission of the disease, and to prevent the development of further drug resistance. Because of its worldwide rollout and rapid implementation, the Xpert-MTB/Rif assay (Cepheid) has become one of the most frequently used molecular screening tests for TB and MDR TB in both resource-poor and resource-rich countries (1). Recently, a 67-year-old Swiss-born male patient was admitted to a secondary-care hospital in Switzerland with clinical and radiologic suspicion of pulmonary TB. Rapid testing by Xpert-MTB/Rif showed the presence of Mycobacterium tuberculosis complex in two sputum samples (Table 1; samples 1 and 2). In addition, these specimens showed indeterminate and definite rifampin (RMP) resistance, respectively (Table 1, samples 1 and 2). Since the patient had no history of TB and was from Switzerland (a low MDR TB incidence setting), based on previous reports of false RMP resistance assay results (2, 3), the clinician doubted the accuracy of the Xpert-MTB/Rif RMP results. Therefore, an additional four sputum samples were submitted to the Swiss National Reference Center for Mycobacteria for rapid confirmation before initiation of MDR TB therapy. Further patient testing revealed HIV positivity.

Table 1.

Xpert-MTB/Rif results and assay parameters of six sputum specimens tested at the primary laboratory (samples 1 and 2, gray shading) and at the Swiss National Reference Center for Mycobacteria (samples 3 to 6)

^aProbe E C_T and endpoint values for all six specimens are markedly different from those obtained with probes A to D, but the testing software (version 4.3) of the assay flagged RMP resistance in only two initial specimens with high and low bacterial loads, in contrast to other specimens with very low or medium bacterial loads, indicating an analytical problem with the software.

^bThe software uses a ΔC_T max cutoff of >5 for the automated detection of RMP resistance.

^cSPC, sample processing control.

The four additional sputum specimens (Table 1, samples 3 to 6) were acid-fast smear positive by the Ziehl-Neelsen method (4), and Xpert-MTB/Rif testing detected the presence of M. tuberculosis complex. Surprisingly, all of the samples were scored RMP susceptible by the molecular assay (Table 1). In order to resolve the discrepant Xpert-MTB/Rif results and to rapidly confirm or rule out MDR TB, direct rpoB sequencing of the 81-bp core region and additional molecular screening for isoniazid (INH) resistance were performed as described previously (3, 5) and revealed rpoB Leu(CTG)533Pro(CCG) and katG Ser(AGC)315Thr(ACC) mutations, respectively, in all four specimens. Growth detection and quantitative phenotypic drug susceptibility testing (DST) for first- and second-line antituberculosis drugs with the MGIT 960 system and EpiCenter software with the TB eXiST module (Becton, Dickinson Microbiology Systems, Sparks, MD) were performed as described earlier (6). Quantitative DST identified resistance to RMP at 0.5 μg/ml and susceptibility at 1.0, 4.0, and 20 μg/ml, and resistance to rifabutin at 0.1 μg/ml and susceptibility at 0.4 and 2.0 μg/ml. DST for INH showed resistance at 0.1 and 1.0 μg/ml and intermediate resistance at 3.0 and 10.0 μg/ml. No drug resistance was identified by conventional DST for other first- and second-line drugs. Direct molecular results were confirmed by DNA sequencing of the rpoB and katG genes of the culture isolates and gave concordant results.

Previously, in vitro experiments indicated that Xpert-MTB/Rif cannot detect Leu533Pro unless 100% of the DNA population was mutant (7). PCR followed by DNA sequencing directly from the sputum samples or cultures did not show double peaks (potential signs of heteroresistance) at the corresponding Leu533Pro mutant position in the sequence electropherograms. Therefore, we have no indications from molecular testing that a heteroresistant population was present in the patient. Moreover, the clinical information also argues against heteroresistance since resistance in patients with newly diagnosed TB (such as our case) is usually associated with drug resistance in a large proportion of the population compared to patients with acquired drug resistance who were previously treated for tuberculosis (8).

Detailed analysis of the Xpert-MTB/Rif assay parameters revealed that rpoB probe E (encompassing codon 533) was hybridizing significantly less than the other four probes and showed ΔC_T max values markedly lower than 5 (the Xpert-MTB/Rif software uses a ΔC_T max cutoff of >5 for the automated detection of RMP resistance) (9) but that did not result in RMP resistance identification by the Xpert software (version 4.3) (Table 1). The bacterial loads of the different specimens were variable, which may have resulted in variable probe E hybridization and lack of detection of mutation Leu533Pro (Table 1). These findings may warrant a thorough revision of the probe E sequence length and its hybridization and software detection parameters. Our report clearly shows that Xpert-MTB/Rif can produce false RMP susceptibility results in the presence of mutation Leu533Pro in clinical specimens with different bacterial loads. Since the bacterial loads of patients may vary in subsequent specimens, this nonreproducibility may generate serious diagnostic confusion and may significantly impact assay predictive values, especially in low-incidence settings.

RMP-resistant tuberculosis has a major impact on patient outcome, and removal of RMP is a very significant alteration of the standard treatment. Reliable molecular detection of rpoB mutations Leu533Pro, Leu511Pro, Asp516Tyr, and His526Leu has important diagnostic consequences since these mutations can be associated with treatment failure using standard regimens that include 600 mg/day RMP (2, 10, 11, 12). Such rpoB mutant strains can be missed by conventional DST since they are resistant to lower RMP concentrations that are not routinely tested in many laboratories (10). Although the rpoB Leu533Pro, Leu511Pro, Asp516Tyr, and His526Leu mutations are less common, their prevalence may reach 10 to 22% (13). However, it has also been shown that screening for RMP resistance alone is not always a reliable proxy for the presumptive diagnosis of MDR TB in either low- or high-prevalence settings, indicating the need for rapid INH susceptibility screening in addition to RMP susceptibility screening (14, 15).

In line with these observations, our results underline the fact that accurate and timely confirmation of RMP resistance or nonreproducible Xpert-MTB/Rif results and rapid molecular screening for INH resistance are imperative, especially in a low-incidence setting when clinical suspicion argues against MDR TB or in peripheral settings where staff may be less experienced in test interpretation, critical review of assay parameters, and troubleshooting. Correct and rapid identification of mutations can also provide valuable preliminary information on the level of drug resistance and can direct what additional nonroutine drug concentrations and second-line drugs can be included in subsequent conventional DST. Finally, the present report shows that a definitive diagnosis of MDR TB may not be based on a single test but should always be established by a thorough evaluation and verification of all clinical and laboratory findings.