Retrospective Analysis of False-Positive and Disputed Rifampin Resistance Xpert MTB/RIF Assay Results in Clinical Samples from a Referral Hospital in Hunan, China

Concerns about the specificity of the Xpert MTB/RIF (Xpert) assay have arisen, as false-positive errors in the determination of Mycobacterium tuberculosis complex (MTBC) infection and rifampin (RIF) resistance in clinical practice have been reported. Here, we investigated 33 cases where patients were determined to be RIF susceptible using the Bactec MGIT 960 (MGIT) culture system but RIF resistant using the Xpert assay.

ABSTRACT

Concerns about the specificity of the Xpert MTB/RIF (Xpert) assay have arisen, as false-positive errors in the determination of Mycobacterium tuberculosis complex (MTBC) infection and rifampin (RIF) resistance in clinical practice have been reported. Here, we investigated 33 cases where patients were determined to be RIF susceptible using the Bactec MGIT 960 (MGIT) culture system but RIF resistant using the Xpert assay. Isolates from two of these patients were found not to have any mutations in the rifampin resistance determining region (RRDR) region of rpoB and had good treatment outcomes with first-line antituberculosis (anti-TB) drugs. The remaining 31 patients included 5 new cases and 26 previously treated patients. A large number of well-documented disputed mutations, including Leu511Pro, Asp516Tyr, His526Asn, His526Leu, His526Cys, and Leu533Pro, were detected, and mutations, including a 508 to 509 deletion and His526Gly, were described here as disputed mutations for the first time. Twenty-one (81%) of the 26 previously treated patients had poor treatment outcomes, and isolates from 19 (90%) of these 21 patients were resistant to isoniazid (INH) as determined using the MGIT culture system. Twenty-seven of the 31 isolates with disputed rpoB mutations were phenotypically resistant to INH, 21 (78%) being predicted by GenoType MTBDRplus to have a high level of INH resistance. Most (77.4%) of the isolates with disputed mutations were of the Beijing lineage. These findings have implications for the interpretation of false-positive and disputed rifampin resistance Xpert MTB/RIF results in clinical samples and provide guidance on how clinicians should manage patients carrying isolates with disputed rpoB mutations.

INTRODUCTION

Although not a complete surrogate for multidrug-resistant tuberculosis (MDR-TB), rifampin (RIF) resistance is the most important indicator of MDR-TB, particularly in settings where levels of drug resistance are low (1). A patient with tuberculosis (TB) that is resistant to RIF is defined as having RIF-resistant TB (RR-TB), whether detected using phenotypic or genotypic methods and with or without resistance to other anti-TB drugs (2). RR-TB requires treatment with second-line drugs according to guidelines issued by the World Health Organization (WHO) (3). In 2017, 160,684 cases of MDR/RR-TB were notified worldwide, up from 153,119 in 2016 (4).

The growing number of RR-TB cases notified may be attributed to the widespread use of Xpert MTB/RIF (Cepheid, Sunnyvale, CA, USA), recommended by the WHO in 2010. Xpert MTB/RIF is an automated cartridge-based assay designed to simultaneously detect Mycobacterium tuberculosis and resistance to RIF directly in clinical specimens using heminested real-time PCR (5) which targets the 81-bp rifampin resistance determining region (RRDR) of the rpoB gene. As mutations in the RRDR occur in 95% to 98% of all RIF-resistant strains (1, 6), Xpert MTB/RIF provides highly accurate RIF resistance detection, as has been demonstrated through extensive large-scale studies (7–10). In 2017, the WHO also recommended the use of the Xpert MTB/RIF Ultra assay (Xpert Ultra), which incorporates two different multicopy amplification targets (IS6110 and IS1081) and uses improved assay chemistry and cartridge design, to improve assay sensitivity in the detection of M. tuberculosis.

However, concerns about Xpert’s specificity have been rising, as previous studies have reported false-positive errors in M. tuberculosis (11, 12) and RIF resistance (13–15) detection in clinical practice. In some cases, false-positive RIF resistance results have been attributed to silent mutations within the RRDR of rpoB (13, 14), or to a large cycle difference in the cycle threshold (C_T) value between any two probes (defined as ΔC_T maximum values) (13, 15). In other cases, however, some discrepancies in RIF resistance Xpert results are attributed to disputed rpoB mutations (16, 17), such as Leu511Pro, Asp516Tyr, His526Leu, His526Asn, His526Cys, and Leu533Pro. Disputed rpoB mutations often cause “borderline” or subcritical levels of RIF resistance, which are easily missed in phenotypic susceptibility assays, particularly using the automated MGIT 960 system (BD, Sparks, MD, USA) (18–20).

Previous studies on the clinical significance of disputed rpoB mutations have claimed that disputed rpoB mutations could phenotypically conceal multidrug-resistant M. tuberculosis and challenge the use of phenotypic drug susceptibility testing (DST), the gold standard, for RIF resistance (21–23). However, the significance of these disputed mutations is still debatable, as we have little knowledge of the consequences of infection with strains harboring such mutations (24). In addition, in light of these findings, it is unclear whether and to what extent Xpert MTB/RIF might outperform phenotypic DST methods for detecting RIF resistance. WHO intends to continue to collect and evaluate data on this issue (25).

In this study, our aim was to analyze false-positive and disputed rifampin resistance results obtained with the Xpert MTB/RIF assay in clinical samples so as to provide greater understanding on the issue of discordance between genotypic resistance but phenotypic sensitivity for RIF as well as insight into the relationship between rare disputed mutations in the RRDR of rpoB and treatment outcomes of the patients infected with strains harboring such mutations.

MATERIALS AND METHODS

Setting.

This study was conducted at Hunan Chest Hospital in Hunan province, China. Hunan province, located in central-southern China, has a population of approximately 72 million people (26). Between 2005 and 2009, the average annual TB incidence in Hunan province was 111.75 per 100,000 in male individuals and 43.44 per 100,000 in female individuals (27). Between 2009 and 2010, the prevalences of isoniazid (INH) and rifampin (RIF) resistance were 35.7% and 26.9%, respectively (28). Hunan Chest Hospital is a TB referral hospital that provides diagnostic and treatment services for patients with chest and lung diseases, including TB, MDR-TB, and extensively drug-resistant TB (XDR-TB), referred from throughout Hunan province (26). Phenotypic DST based on solid and liquid culture and molecular tests such as GenoType MTBDRplus (Hain Lifescience GmbH, Nehren, Germany) and Xpert MTB/RIF are performed routinely.

Patients.

Medical records of 4,575 patients with DST results obtained using both the mycobacterial growth incubator tube (MGIT) culture system (BD, Sparks, MD, USA) and Xpert MTB/RIF between September 2014 and September 2016 were reviewed (Fig. 1). Medical records included patient registration category (new case or previously treated case), laboratory examination results (mainly Xpert MTB/RIF and MGIT culture results), and treatment outcome with first-line anti-TB drugs (previously treated cases only). Patients who initially registered as new cases were contacted by phone in June 2017 to ascertain their health status. Patients were included only if they were diagnosed with RIF-resistant TB by Xpert MTB/RIF but were RIF susceptible by MGIT culture. Thirty-five patients met the requirements for inclusion in the study, and 33 patients were included (isolates from the other two patients could not be recovered). Of the 33 patients, 15% (n = 5) were female, the median age of the patients was 44 years (range, 17 to 70 years), and 79% (n = 26) were previously treated cases according to WHO definitions (29). This study was approved by the Hunan Chest Hospital Ethics Committee and the need for written informed consent from the patients was waived.

FIG 1.

Study outline. RIF, rifampin; RRDR, rifampin resistance determining region.

Diagnostic tests.

All patients included had been suspected to have TB or MDR-TB and had submitted sputum specimens for Xpert MTB/RIF and MGIT culture analysis in the clinical laboratory of Hunan Chest Hospital. In each case, one sputum sample was analyzed by Xpert MTB/RIF (version 4.0), and another sputum sample was submitted for MGIT culture. Patients who were M. tuberculosis complex (MTBC) culture positive then underwent phenotypic DST for the four first-line drugs using the MGIT culture system. To confirm RIF resistance, isolates recovered from frozen suspensions stored in a −80°C freezer were tested by genotypic DST using GenoType MTBDRplus, and phenotypic DST using the MGIT culture system was repeated. All the above methods were performed according to the manufacturers’ instructions, as described briefly in previous studies (30, 31).

Storage and recovery of MTBC strains.

MTBC isolates obtained by culture were stored at −80°C for further analysis. The storage medium contained tryptone and glycerol and was prepared according to previously described protocols (32). To recover the isolates, the frozen suspensions were thawed at room temperature, and then an aliquot of the thawed suspension was inoculated on Löwenstein-Jensen (LJ) medium and incubated at 37°C.

MIC determination.

MIC determination was performed for recovered isolates using the modified resazurin microtiter assay (REMA) described by Palomino et al. (33). Briefly, 100 µl of 7H9 broth containing 10% oleic acid-albumin-dextrose-catalase (OADC) was dispensed in each well of a sterile flat-bottom 96-well plate, and serial 2-fold dilutions of each drug were prepared directly in the plate. Final RIF concentrations were 0.125 to 128 µg/ml (for high-level RIF resistance tests) or 0.001 to 1.024 µg/ml (for low-level RIF resistance tests). The last column of wells on the plate was used as a RIF negative control. One hundred microliters of inoculum was added to each well except the A- and H-line perimeter wells, in which sterile water was added to avoid evaporation during incubation. A sterile control was also included for each isolate. The plate was covered, sealed with Parafilm, and incubated at 37°C. After 7 days of incubation, 20 µl of resazurin solution (filter sterilized, 10 mg/ml concentration) was added to each well, and the plate was incubated for 24 to 48 h. A change in color from blue to pink indicated the growth of bacteria, and the MIC was defined as the lowest concentration of drug that prevented this change in color.

DNA extraction.

Genomic DNA was extracted from LJ cultures. Briefly, colonies of each strain were scraped from LJ culture medium and resuspended in 500 µl of 0.9% saline water. After heat inactivation at 95°C for 20 min, the cell suspension was incubated in an ultrasonic bath for 15 min. Cells were then centrifuged at 13,000 × g for 5 min, and the supernatant containing chromosomal DNA was transferred to a fresh tube. The DNA solution was stored at −20°C for further analysis, including GenoType MTBDRplus testing, rpoB sequencing, and mycobacterial interspersed repetitive-unit–variable number tandem-repeat (MIRU-VNTR) typing.

rpoB sequencing.

The rpoB gene containing the 81-bp RRDR was amplified from genomic DNA using a pair of specific primers (rpoB-f, 5′-CTTGCACGAGGGTCAGACCA-3′ and rpoB-r, 5′-ATCTCGTCGCTAACCACGCC-3′), which yielded a product of 543 bp. PCR conditions were as follows: 94°C denaturation for 5 min, followed by 30 cycles of 94°C denaturation for 1 min, annealing at 60°C for 1 min, and extension at 72°C for 1 min, with a final extension cycle at 72°C for 7 min. The PCR product was purified using a DNA gel extraction kit (Beyotime, Shanghai, China). The purified product was directly sequenced using a 3730xl DNA sequencer (Applied Biosystems, Foster City, CA, USA) by Beijing Ruibio Biotech Co., Ltd. Sequences were analyzed using BLAST (Basic Local Alignment Search Tool, Bethesda, MD, USA) by comparing with the MTB H37Rv sequence (GenBank accession no. NC_000962.3).

MIRU-VNTR.

The 12-locus MIRU-VNTR typing method was performed using previously reported primers (34). PCRs for all MIRU-VNTR loci were performed in a reaction volume of 20 µl containing 10 µl 2× Taq mixture, 2 µl genomic DNA template, and 0.5 µmol of each primer set. The PCR amplification program was 94°C for 3 min, followed by 30 cycles at 94°C for 1 min, 58°C for 1 min, and 72°C for 1 min, with a final extension at 72°C for 5 min. Five microliters of the amplicon was run on a 1.5% agarose gel, with a 100-bp ladder being run once every ten lanes. Analysis of PCR fragment size and assignment of the various VNTR alleles was performed using Quantity One (version 4.6.2) software (Bio-Rad Laboratories). The corresponding 12-locus MIRU-VNTR genotypes were analyzed on the MIRU-VNTRplus database, as described previously, to identify lineages and construct a minimum spanning tree (MST) (35, 36).

RESULTS

Repeat phenotypic DST and mutations in rpoB.

To ensure the reliability of previous phenotypic DST results, we repeated the MGIT culture-based DST for the 33 isolates that were successfully recovered. Twenty-eight isolates had consistent RIF susceptible results, but only 5 isolates were RIF resistant (Fig. 1). Sequencing of the RRDR region of rpoB to determine if mutations associated with RIF resistance were present in the 33 isolates showed that 2 of the 33 isolates did not harbor any RRDR mutations, while the remaining 31 isolates each harbored one mutation type, four mutation types occurring at position 526 based on the Escherichia coli rpoB codon numbering system (His526Asn, His526Leu, His526Arg, and His526Cys). None of the sequencing profiles showed underlying peaks indicating heteroresistance.

Laboratory investigations and clinical features of 33 cases.

In the two TB cases where isolates did not have mutations in the RRDR, Xpert assays showed delayed C_Ts, resulting in ΔC_T maximum values exceeding 4.0 cycles (a predefined ΔC_T maximum cutoff value indicating RIF resistance in the 4th version of the Xpert software). One of these cases (TB02) was retested by submitting an additional sputum sample for Xpert assay testing; a RIF susceptible result with a ΔC_T maximum value of 2.2 cycles was obtained. Both isolates were RIF and INH susceptible in the GenoType MTBDRplus test and showed MICs of 0.016 and 0.032 μg/ml (Table 1), respectively, that were much lower than the critical concentration (1.0 μg/ml) used in the MGIT culture system. In addition, both TB patients were registered as new cases and had good treatment outcomes with first-line anti-TB drugs (Table 1).

TABLE 1.

Results of laboratory investigations and clinical features of two cases with a wild-type rpoB RRDR region

Case	Laboratory resultsa			Clinical features
	Xpert	MGIT	MIC (μg/ml)	Patient classification	Treatment outcome
TB01	Low, RIF-R	RIF-S, INH-S	0.032	New case	Cured
TB02	Very low, RIF-R	RIF-S, INH-S	0.016	New case	Cured

^aRIF, rifampin; INH, isoniazid; S, susceptible; R, resistant.

Table 2 shows data for five TB patients whose isolates were confirmed to have phenotypic RIF resistance. Four of these patients were confirmed to be MDR-TB and one to be RIF monoresistant TB based on DST results from the MGIT culture system. Results obtained using Xpert were also confirmed by GenoType MTBDRplus, except for one case with an undetermined RIF resistance result due to weak staining of WT8 (see Table S1 in the supplemental material). Four of the isolates showed elevated MICs ranging from 2.0 μg/ml to 64 μg/ml. MIC determination for one isolate failed due to poor growth in the control well (MIC determination was repeated three times). Four patients in this group were registered as retreatment cases and one patient as a new case. Three previously treated patients (TB03, TB05, and TB07) with isolates harboring disputed mutations had poor treatment outcomes (e.g., treatment failure or relapse) using first-line anti-TB drugs, including RIF. The treatment outcome for the fourth previously treated patient (TB06) was unknown due to loss to follow-up.

TABLE 2.

Mutation types, results of laboratory investigations, and clinical features of cases confirmed to be phenotypically RIF-R by repeating MGIT testing

Case	Mutation type	Laboratory resultsa			Clinical features
		MGIT	GenoType MTBDRplus	MIC (μg/ml)	Patient classification	Treatment outcomes
TB03	His526Arg	RIF-R, INH-S	RIF-R, INH-S	64	Treatment after failure	Treatment failed
TB04	His526Leu	RIF-R, INH-R	RIF-R, INH-R	4	New case	Not evaluatedb
TB05	His526Leu	RIF-R, INH-R	RIF-R, INH-R	4	Relapse	Treatment completed
TB06	His526Leu	RIF-R, INH-R	RIF-R, INH-S	NA	Treatment after loss to follow-up	Lost to follow-up
TB07	Leu533Pro	RIF-R, INH-R	Indeterminate RIF resistancec , INH-R	2	Treatment after failured	Treatment failed

^aNA, not available; RIF, rifampin; INH, isoniazid; S, susceptible; R, resistant.

^bTreatment outcome was not evaluated as patient could not be contacted.

^cRIF resistance was indeterminate due to weak staining of WT8 in the GenoType MTBDRplus test.

^dPatients who were previously treated for TB and whose treatment failed at the end of their most recent course of treatment.

Repeating the MGIT tests confirmed the previous phenotypic RIF susceptible status of 26 isolates (Table 3). The MICs of these isolates (0.016 to 2.0 μg/ml) were lower than those of the five isolates described above. Missense mutations in rpoB were detected in all 26 isolates by sequencing, confirming the results of the Xpert assay. In addition, GenoType MTBDRplus detected RIF resistance in 19 of the 26 isolates and indeterminate RIF resistance in the remaining seven isolates due to weak staining of WT8 (Table S1). In this group, 3 were registered as new cases, 22 (85%) as retreatment cases (relapse, n = 11; retreatment after treatment failure, n = 7; and treatment after loss to follow-up, n = 4), and 1 patient had an uncertain registration category due to a lack of medical records. The three new patients were started on first-line treatment when they registered at the hospital and were then transferred to second-line treatment after RIF resistance was detected by the Xpert assay. Two of the new cases were cured using second-line anti-TB drugs, and the treatment outcome of the other new patient was unknown. Eighteen of the 22 retreatment cases showed a poor treatment outcome (e.g., treatment failure or relapse) using first-line anti-TB drugs, including RIF, and the remaining four patients were lost to follow up.

TABLE 3.

Mutation types, results of laboratory investigations and clinical features of cases confirmed to be phenotypically RIF-S by repeating MGIT testing

Case	Mutation type	Laboratory resultsa			Clinical features
		MGIT	Genotype MTBDRplus	MIC (μg/ml)	Patient classification	Treatment outcomes
TB08	508 to 509 deleted	RIF-S, INH-R	RIF-R, INH-R	0.128	Relapse	Treatment completed
TB09	Leu511Pro	RIF-S, INH-R	RIF-R, INH-R	0.032	Relapse	Treatment completed
TB10	Leu511Pro	RIF-S, INH-R	RIF-R, INH-R	0.016	Relapse	Treatment completed
TB11	Leu511Pro	RIF-S, INH-R	RIF-R, INH-R	0.032	Relapse	Cured
TB12	Leu511Pro	RIF-S, INH-R	RIF-R, INH-R	0.512	Relapse	Treatment completed
TB13	Asp516Tyr	RIF-S, INH-R	RIF-R, INH-R	0.064	New case	Curedb
TB14	Asp516Tyr	RIF-S, INH-S	RIF-R, INH-S	0.128	Relapse	Cured
TB15	Ser522Leu	RIF-S, INH-R	RIF-R, INH-R	0.512	Treatment after failurec	Treatment failed
TB16	His526Asn	RIF-S, INH-R	RIF-R, INH-R	0.512	Relapse	Treatment completed
TB17	His526Asn	RIF-S, INH-R	RIF-R, INH-R	0.128	Relapse	Treatment completed
TB18	His526Asn	RIF-S, INH-S	RIF-R, INH-S	0.256	Treatment after loss to follow-up	Lost to follow-up
TB19	His526Leu	RIF-S, INH-R	RIF-R, INH-R	1.024	Treatment after failurec	Treatment failed
TB20	His526Leu	RIF-S, INH-R	RIF-R, INH-R	0.256	Treatment after failurec	Treatment failed
TB21	His526Leu	RIF-S, INH-R	RIF-R, INH-R	0.256	New case	Not evaluatedd
TB22	His526Leu	RIF-S, INH-R	RIF-R, INH-R	2.0	Treatment after loss to follow-up	Lost to follow-up
TB23	His526Gly	RIF-S, INH-R	RIF-R, INH-R	0.128	New case	Curedb
TB24	His526Gly	RIF-S, INH-R	RIF-R, INH-R	0.064	Treatment after failurec	Treatment failed
TB25	His526Cys	RIF-S, INH-R	RIF-R, INH-R	0.128	Treatment after failurec	Treatment failed
TB26	His526Cys	RIF-S, INH-R	RIF-R, INH-R	0.512	Treatment after loss to follow-up	Lost to follow-up
TB27	Leu533Pro	RIF-S, INH-R	Indeterminate RIF resistancee , INH-R	0.512	Relapse	Treatment completed
TB28	Leu533Pro	RIF-S, INH-R	Indeterminate RIF resistancee , INH-S	0.512	Treatment after loss to follow-up	Lost to follow-up
TB29	Leu533Pro	RIF-S, INH-R	Indeterminate RIF resistancee , INH-R	1.024	Relapse	Treatment completed
TB30	Leu533Pro	RIF-S, INH-S	Indeterminate RIF resistancee , INH-R	0.256	NAf	NAg
TB31	Leu533Pro	RIF-S, INH-R	Indeterminate RIF resistancee , INH-R	0.256	Relapse	Treatment completed
TB32	Leu533Pro	RIF-S, INH-R	Indeterminate RIF resistancee , INH-R	0.256	Treatment after failurec	Treatment failed
TB33	Leu533Pro	RIF-S, INH-R	Indeterminate RIF resistancee , INH-R	0.512	Treatment after failurec	Treatment failed

^aRIF, rifampin; INH, isoniazid; S, susceptible; R, resistant.

^bPatients were cured using therapies based on second-line drugs.

^cPatients who were previously treated for TB and whose treatment failed at the end of their most recent course of treatment.

^dTreatment outcome was not evaluated as the patient could not be contacted.

^eRIF resistance was indeterminate due to weak staining of WT8 in the GenoType MTBDRplus test.

^fPatient treatment status was not assigned due to lack of medical records at registration.

^gTreatment outcome was not available due to lack of information on treatment outcome in medical records.

Disputed rpoB mutations and genotypes.

Twelve-locus MIRU-VNTR analysis was applied to all 33 isolates to investigate a possible association between disputed mutations and specific genotypes or lineages. A minimum spanning tree (MST) was constructed by online analysis of genotyping data (see Table S2) (35, 36), and VNTR genotyping divided the 33 isolates into 4 clusters and 11 unique types, with a clustering rate of 54.5% and a discriminatory index of 86.6%. Although isolates in each cluster showed identical MIRU-VNTR type patterns, the RRDR mutation types were diverse within each cluster. For example, four mutation types (His526Asn, His526Leu, His526Gly, and Leu533Pro) were found among the 11 isolates within the largest cluster (cluster I), while two mutation types (Leu511Pro and Leu533Pro) were found within each small cluster (cluster II and cluster IV), both of which comprised 3 isolates. All 4 clusters as well as another 8 unique types were grouped into a large clonal complex (CC) comprising 30 isolates. The remaining three unique MIRU-VNTR types were assigned as singletons. Among the 31 isolates harboring disputed rpoB mutations, the most prominent lineage was found to be Beijing (77.4%), followed by Ghana (6.5%) and NEW-1 (6.5%). Only three isolates (9.7%), grouped into cluster IV, were not assigned to a known phylogenetic lineage.

DISCUSSION

Prompt and accurate DST is crucial in the diagnosis, appropriate treatment, and control of drug-resistant tuberculosis (DR-TB). However, when genotypic and phenotypic DST are performed, discordant results invariably occur (37). In this study, we reviewed the discordance in results from the Xpert MTB/RIF assay and the Bactec MGIT 960 culture system used for diagnosis of RR-TB patients, focusing specifically on patients that were RIF susceptible by the MGIT system but RIF resistant by the Xpert assay.

Of the 33 patients investigated, two patients were wrongly diagnosed with RR-TB by the Xpert assay, with ΔC_T maximum values of greater than 4.0 cycles. False-positive RIF resistance caused by large ΔC_T maximum values has also been reported by Ocheretina et al. (13) and Van Rie et al. (15). To improve the specificity of the Xpert assay for detecting RIF resistance, the fourth-generation Xpert assay has adopted a modified ΔC_T maximum value of 4.0 cycles (13). Nevertheless, based on the findings from our study as well as those of Ocheretina et al. (13), the fourth-generation Xpert assay still produces false-positive RIF resistance results, especially in the case of patients with a “low” or “very low” level of M. tuberculosis detection.

Isolates from the remaining 31 patients studied collectively harbored the majority of the well-documented disputed mutations, such as Leu511Pro, Asp516Tyr, His526Asn, His526Leu, His526Cys, and Leu533Pro. These mutations are reported to often be missed by phenotypic DST, particularly when the MGIT 960 system is used (18, 19). Of note, His526Arg is not a well-documented mutation and was previously reported only by Abanda et al. (17), who showed that RIF resistance arising from this mutation could be detected by the MGIT 960 system. Surprisingly, all five patients confirmed to have RR-TB based on the results of the repeated MGIT tests also harbored disputed mutations (Table 1). We speculate that phenotypic resistance in these five isolates was not solely determined by these disputed mutations, and other genomic variants or transcriptional alterations may have caused low levels of RIF resistance (38, 39). The reason why initial phenotypic DST testing gave RIF susceptible results for these five patients is unclear but may have been due to laboratory error when performing the phenotypic DST (37), particularly in the case of the patient whose isolate harbored a His526Arg mutation and had a MIC of 64 μg/ml.

We also found that RIF resistance caused by mutation types such as deletion of 508 to 509 and His526Gly were also missed by the MGIT 960 system. To our knowledge, these mutation types were not previously reported as disputed mutations, possibly because previous studies showed that these two disputed mutations can yield phenotypic resistance to RIF in the LJ proportion method (39, 40), which has been reported to miss fewer disputed mutations than the MGIT system in the detection of RIF resistance (18).

We carefully evaluated the treatment outcomes of the patients investigated in our study. Whereas all new TB cases were treated at Hunan Chest Hospital, all retreatment cases given standard therapies with first-line drugs containing RIF were managed at peripheral hospitals or local sites of the Centers for Disease Control and Prevention of China (China CDC), where both phenotypic and genotypic DST were not available. Consistent with previous studies (21–23), treatment outcomes were poor for previously treated patients whose isolates harbored disputed mutations, including the disputed mutations such as deletion of 508 to 509 and His526Gly identified here for the first time. For example, of the 26 previously treated patients whose isolates harbored disputed mutations, 21 (81%) patients were registered as relapse and retreatment after treatment failure cases (Tables 2 and 3). Our study therefore indicates that disputed rpoB mutations may be of clinical significance.

Williamson et al. (21) and Ho et al. (22) suggested that disputed rpoB mutations may phenotypically conceal multidrug-resistant M. tuberculosis, and cases with isolates harboring such mutations are thus more likely to experience treatment failure or recurrent disease. Indeed, of the 21 patients studied here with poor treatment outcomes, MGIT culture results showed that 19 (90%) of the corresponding isolates were also resistant to isoniazid (INH) (Tables 2 and 3). With the exception of three isolates whose RIF resistance results were confirmed by repeated MGIT system culturing, most of the remaining 18 isolates, with RIF MICs that varied from 0.016 to 2.0 μg/ml, were likely detected as multidrug-resistant M. tuberculosis, given that the RIF critical concentration used in the MGIT culture system is set at 0.0625 μg/ml (41).

In total, 27 of the 31 isolates with disputed rpoB mutations were phenotypically resistant to INH. It is intriguing that these disputed rpoB mutations often occur in INH-resistant isolates, a finding that is consistent with previous reports (21, 22, 24). In our study, although the INH MICs of these 27 isolates were not determined, the INH resistance level can be predicted from GenoType MTBDRplus test results (see Table S1 in the supplemental material). Twenty-one (78%) isolates harbored Ser315Thr mutations within katG or had no copy of the katG locus, scenarios that likely result in high-level INH resistance (42).

Another important finding was that most (77.4%) of the isolates with disputed mutations were of the Beijing lineage, although a possible association between disputed mutations and a specific genotype has not been observed in other studies (22, 43). Of note, four of the five isolates with disputed mutations reported by Ho et al. (22) were of the Beijing lineage, and interestingly, three of the four Beijing lineage isolates were isolated from patients who were born in China. Based on the findings from Ho et al. and our study, further research on the prevalence of disputed rpoB mutations in China and the relationship between disputed mutations and the Beijing lineage is merited.

Our study has some limitations. First, we only focused on cases that were determined to be RIF susceptible by the MGIT system but RIF resistant by the Xpert assay, rather than on all cases that were Xpert RIF resistant. We were therefore unable to compare treatment outcomes of patients whose isolates harbored disputed mutations with those that did not harbor disputed mutations. Our finding that disputed rpoB mutations are of clinical significance would be more persuasive if the comparison between these two groups was performed. In addition, the prevalence of disputed mutations could have been investigated if all Xpert RIF-resistant isolates had been included. Second, although we did not preselect the patients based on their clinical outcomes, as was the case in the study by Van Deun et al. (23) who studied first-failure and relapse patients, we included some new patients carrying isolates with disputed mutations; however, as these patients were either lost to follow-up or were treated directly with second-line therapies, as results from Xpert assays indicated the presence of existing RIF resistance, information on treatment outcome for these patients based on first-line anti-TB therapies was not available. An additional limitation of this study is that the rate of patient relapse may be overestimated, as the possibility of reinfection was not ruled out by genotyping the isolates at initial diagnosis and recurrence.

In conclusion, our findings have implications for the interpretation of false-positive and disputed rifampin resistance results obtained using the Xpert MTB/RIF assay in clinical samples. Importantly, these findings inform clinicians on how they should manage patients carrying isolates with disputed rpoB mutations. Given that disputed rpoB mutations are associated with poor treatment outcomes (e.g., treatment failure and relapse), careful clinical assessment and use of additional therapies, including second-line drugs, remain essential for the management of patients with disputed rifampin resistance results.