Performance Evaluation of the BACTEC MGIT 960 System for Rifampin Drug-Susceptibility Testing of Mycobacterium tuberculosis Using the Current WHO Critical Concentration

ABSTRACT

The World Health Organization recently lowered the rifampin (RIF) critical concentration (CC) for drug-susceptibility testing (DST) of Mycobacterium tuberculosis complex (MTBC) using the mycobacterial growth indicator tube (MGIT) 960 system. Here, we evaluated the diagnostic performance of the MGIT system with the revised CC for determining MTBC RIF resistance with 303 clinical MTBC isolates, including 122 isolates with rpoB mutations, of which 32 had single borderline-resistance mutations, and 181 wild-type rpoB isolates. The phenotypic RIF resistance was determined via the absolute concentration method (AC) and via MGIT using both previous (1 mg/L) and revised (0.5 mg/L) CCs for the latter method. The diagnostic accuracy of each phenotypic DST (pDST) was assessed based on rpoB genotyping as the reference standard. The overall sensitivity of the AC was 95.1% (95% confidence interval [CI], 89.6 to 98.2%), while the MGIT results with previous and revised CCs were 82.0% (95% CI 74.0 to 88.3%) and 83.6% (95% CI 75.8 to 89.7%), respectively. The 32 MTBC isolates with single borderline-resistance mutations showed a wide range of MICs, and sensitivity was not significantly increased by reducing the MGIT CC. All 181 wild-type rpoB isolates were RIF-susceptible in the AC and with MGIT using the previous CC, whereas 1 isolate was misclassified as RIF-resistant with the revised CC. Our results demonstrate that the overall diagnostic performances of the MGIT DST with the revised RIF CC and previous CC were comparable. A further large-scale study is required to demonstrate the optimal RIF CC for MGIT.

INTRODUCTION

Multidrug-resistant or rifampin (RIF)-resistant tuberculosis (MDR/RR-TB) is a public health threat because of its prolonged duration, toxicity, and unsatisfactory outcomes of MDR/RR-TB treatment (1–3). MDR/RR-TB occurs due to acquired resistance caused by spontaneous chromosomal mutations and selective pressure during disease treatment or to primary resistance in which a patient is infected with MDR/RR-TB bacteria (4). Therefore, to successfully reduce the development and spread of MDR/RR-TB, the standard drug regimen should be appropriately prescribed and administered to patients, and direct transmission of MDR/RR-TB should be prevented through early diagnosis and prompt treatment (5).

Drug-susceptibility testing (DST) is essential for diagnosing MDR/RR-TB and selecting antibiotic regimens for treatment. Growth-based phenotypic DST (pDST) is currently the gold standard method used to detect drug resistance (6). Traditionally, pDST has relied on testing at a single, critical concentration (CC) for each anti-TB drug (7). Multiple pDST methods based on solid media, including the proportion method and absolute concentration (AC) method, or based on liquid media, such as the Bactec Mycobacterial Growth Indicator Tube (MGIT) 960, have been widely used (6). MGIT contributes significantly to reducing the turnaround time of pDST by providing results within 4 to 13 days (6). However, previous studies have indicated that Mycobacterium tuberculosis complex (MTBC) isolates with borderline-resistance rpoB mutations are often misclassified as susceptible by pDST, especially with the MGIT (8–11).

Recently, the World Health Organization (WHO) lowered the RIF CC for 7H10 and MGIT testing from 1 mg/L to 0.5 mg/L, based on data from a systematic review, to classify RIF resistance more accurately (7). Although several previous reports investigated the diagnostic accuracy of MGIT DST based on a revised RIF CC, large-scale validation studies are scarce (10, 12–18). This study aimed to evaluate the diagnostic performance of the MGIT system with the revised RIF CC for determining RIF resistance in MTBC.

MATERIALS AND METHODS

MTBC isolates.

This double-center study was conducted at Samsung Medical Center, a tertiary-care hospital in Seoul, Korea, and at the Korean Institute of Tuberculosis (KIT), a WHO-designated supranational reference laboratory. The study was approved by the Institutional Review Board of Samsung Medical Center, Seoul, Korea (approval number: 2021-04-168) and involved 303 clinical MTBC isolates, including 122 isolates with rpoB mutations and 181 wild-type rpoB isolates. Among the 122 MTBC isolates with rpoB mutations, 32 had single borderline resistance mutations (Fig. 1).

FIG 1.

Distribution of rpoB mutations stratified by pDST using MGIT (0.5 and 1.0 mg/L RIF) and the AC method. Results for pDST (R, resistant; S, susceptible) are indicated on the left in areas shaded gray. The types of single and double/triple mutations found in isolates used in this study are shown, respectively, above and below the central row, which shows the wild-type amino acid sequence at several positions of the RRDR, along with M. tuberculosis and Escherichia coli numbering. For single mutations, the number of isolates with the given mutation is shown in parentheses after the mutation. For double and triple mutations, which are joined by horizontal lines, the number of isolates is given in the rightmost mutation box.

pDST.

The RIF resistance of MTBC isolates was determined via the AC method on Löwenstein-Jensen (LJ) medium and via the Bactec MGIT 960 system (Becton Dickinson, Sparks) using both previous (1 mg/L) and revised (0.5 mg/L) CCs for the latter method. For the AC method, each MTBC isolate was inoculated on drug-free LJ medium as the control and on LJ medium containing 40.0 mg/L RIF. Cultures were incubated at 37°C for 4 weeks. RIF resistance was defined as ≥20 colonies grown on RIF-containing media or colony counts on RIF-containing plates that were ≥1% of the counts on control plates. For MGIT testing, we used the Bactec MGIT 960 system according to the manufacturer’s instructions with the previous CC (1 mg/L) (19). Briefly, a lyophilized drug vial (Bactec MGIT 960 SIRE kit; BD Biosciences) was reconstituted with 4 mL of sterile distilled/deionized water to make a stock solution. For MGIT with the revised CC (0.5 mg/L), the RIF vial was reconstituted in 8 mL of sterile distilled/deionized water. Thereafter, 0.1 mL of lyophilized drugs, 0.8 mL of supplements, and 0.5 mL of a positive broth culture were added to the growth indicator tubes. The control tube did not contain any drug. All the cultures were incubated in the MGIT 960 system.

RIF MICs were determined using the 7H9 broth microdilution method according to the Clinical and Laboratory Standards Institute (CLSI) (20, 21). Briefly, isolated colonies were harvested from LJ medium, and the inoculum was adjusted to a 100-fold dilution of a 1.0 McFarland standard after vigorous mixing and transferred to Middlebrook 7H9 medium with oleic acid–albumin–dextrose–catalase (OADC) enrichment (Becton Dickinson). This inoculum was added to each well, with RIF concentrations ranging from 0.0625–8.0 mg/L. Assay plates were incubated at 37°C and interpreted after 2 weeks of incubation.

rpoB gene sequencing.

Genomic DNA was extracted from MTBC isolates using the boiling method. A 700-bp rpoB gene fragment containing the RIF-resistance determining region (RRDR) was amplified using the primers previously described (22). Positive and negative controls were used for all PCRs. Sanger sequencing of the amplified products was carried out with an ABI 3730xl Genetic Analyzer (Applied Biosystems) according to the manufacturer’s instructions. Sequencing results were analyzed using Sequencher software (Gene Codes); sequences were aligned with the corresponding reference strain (M. tuberculosis H37Rv; GenBank accession no. NC_000962.3); results were reported according to the M. tuberculosis amino acid numbering system; and mutations within codons 341 to 560 were identified. Seven mutations, including Leu430Pro, Asp435Tyr, His445Asn, His445Leu, His445Ser, Leu452Pro, and Ile491Phe, were defined as borderline resistance mutations based on the recommendation of the WHO Technical Expert Group (TEG) (7).

For RRDR wild-type isolates with MICs >0.5 mg/L in the 7H9 broth microdilution method, the entire rpoB gene was sequenced to exclude the possibility of mutations outside of the RRDR, using previously described primers (23).

Data analysis.

The sensitivities and specificities of each pDST were assessed based on rpoB gene sequencing as the reference standard. Statistical analyses were performed using MedCalc Statistical Software version 20.118 (MedCalc Software Ltd). McNemar's test was used to compare diagnostic accuracy of the MGIT DST for previous versus revised CC, and a P-value <0.05 was considered significant.

RESULTS

The rpoB gene sequencing revealed that 181 isolates were wild-type with no mutations. The remaining 122 isolates had rpoB mutations, with single mutations in 110 isolates and double or triple mutations in 12 isolates (Fig. 1). The most frequent mutations that conferred high-level RIF resistance resulted in amino acid substitutions Asp435Val and Ser450Leu (26 and 25 isolates, respectively), followed by 7 other substitutions of His445Tyr (n = 16), His445Asp (n = 8), His445Arg (n = 2), Ser450Trp (n = 2), Gln432Pro (n = 1), Gln432Lys (n = 1), and Asp435Gly (n = 1). Among the seven mutations classified as borderline resistance mutations (7), 6 were identified in this study, and the most frequent mutation was Asp435Tyr (n = 29), followed by His445Leu (n = 3), His445Asn (n = 3), Leu430Pro (n = 2), Leu452Pro (n = 2), and Ile491Phe (n = 2).

Testing of all 303 clinical isolates by the 7H9 broth microdilution method resulted in 288 RIF MIC values, with 15 isolates, including 6 wild-type and 9 mutant isolates, failing to grow. A total of 175 wild-type isolates had MICs ranging from ≤0.0625 to 1.0 mg/L (Fig. 2). Among 113 mutant isolates, 102 (90.3%) had MICs of ≥2.0 mg/L, ranging from 2.0 mg/L to >8.0 mg/L. All isolates with high-level resistance mutations had MICs of ≥4.0 mg/L, whereas the subgroup with single borderline resistance mutations had a wide range of MIC values, from ≤0.0625 to >8.0 mg/L.

FIG 2.

Distribution of RIF MICs via the 7H9 broth microdilution method stratified by type of rpoB mutation. Isolates harboring at least one high-confidence resistance mutation were included in the ‘high-confidence resistance’ group. All isolates harboring mutations other than high-confidence resistance or single borderline resistance mutations were placed in the ‘others’ group.

Overall sensitivities of the pDST method for detecting RIF resistance are shown in Table 1. For the AC method, 2 isolates with Asp435Tyr, 2 isolates with Ile491Phe, 1 with Pro439Leu, and 1 with Asp545Glu were identified as RIF-susceptible, yielding a sensitivity of 95.1% (95% confidence interval [CI], 89.6 to 98.2%). On the other hand, an additional 16 isolates with borderline resistance mutations, including 14 Asp435Tyr, 1 Leu430Pro, and 1 His445Leu, were misclassified as RIF-susceptible in MGIT with the previous CC (1 mg/L), yielding a sensitivity of 82.0% (95% CI, 74.0 to 88.3%). The revised CC (0.5 mg/L) for MGIT reduced the misclassification by 2 isolates, which had borderline resistance mutations of Asp435Tyr and His445Leu, resulting in a sensitivity of 83.6% (95% CI, 75.8 to 89.7%). All isolates with high-level resistance mutations were concordantly resistant to RIF in all pDSTs (Fig. 1). All 181 wild-type isolates were RIF-susceptible in the AC method and in the MGIT test with the previous CC, while 1 isolate was misclassified as RIF-resistant using the revised CC in the MGIT, yielding a specificity of 99.5% (95% CI, 97.0 to 100%) (Table 1).

TABLE 1.

Diagnostic performance of pDSTs for detection of RIF resistance in MTBC isolates

pDST (RIF CC)	Result	rpoB sequencing		Sensitivity (95% CI)	Specificity (95% CI)	Accuracy (95% CI)
		Mutation detected	Mutation not detected
AC on LJ (40 mg/L)	Resistant	116	0	95.1%	100%	98.0%
	Susceptible	6	181	(89.6–98.2%)	(98.0–100%)	(95.7–99.3%)
MGIT (1.0 mg/L)	Resistant	100	0	82.0%	100%	92.7%
	Susceptible	22	181	(74.0–88.3%)	(98.0–100%)	(89.2–95.4%)
MGIT (0.5 mg/L)	Resistant	102	1	83.6%	99.5%	93.1%
	Susceptible	20	180	(75.8–89.7%)	(97.0–100%)	(89.6–95.7%)

For the subgroup of 32 isolates with single borderline resistance mutations, the diagnostic sensitivities of each pDST method are shown in Table 2. The AC method showed higher sensitivity compared to the MGIT DST. The revised CC for MGIT increased the sensitivity from 37.5% (21.1 to 56.3%) to 43.8% (26.4 to 62.3%), although statistical significance was not achieved (P = 0.50). A total of 18 out of 32 isolates with single borderline resistance mutations were RIF-susceptible by the revised CC for MGIT; as summarized in Table 3, these isolates had lower MIC values with a wide range from ≤0.0625 to >8.0 mg/L.

TABLE 2.

Diagnostic sensitivities of pDSTs for detection of RIF resistance in MTBC isolates with single borderline resistance mutations

pDST (CC)	Sensitivity (95% CI)
AC on LJ (40 mg/L)	28/32, 87.5% (71.0–96.5%)
MGIT (1.0 mg/L)	12/32, 37.5% (21.1–56.3%)
MGIT (0.5 mg/L)	14/32, 43.8% (26.4–62.3%)

TABLE 3.

Distribution of 7H9 broth microdilution MIC values in MTBC isolates with single borderline resistance mutations

Mutation type	RIF pDST (CC, mg/L)			MIC (mg/L)
	AC (29)	MGIT (1)	MGIT (0.5)	≤0.0625	0.125	0.25	0.5	1.0	2.0	4.0	8.0	>8.0	NAa	Total
Leu430Pro	Rb	Sc	S	0	0	0	0	0	0	1	0	0	0	1
Asp435Tyr	S	S	S	0	0	1	0	0	0	1	0	0	0	2
	R	S	S	0	0	0	3	4	3	2	0	1	0	13
	R	S	R	0	0	0	0	0	1	0	0	0	0	1
	R	R	R	0	0	0	0	0	0	0	3	7	0	10
His445Leu	R	S	R	0	0	0	0	0	1	0	0	0	0	1
	R	R	R	0	0	0	0	0	0	1	1	0	0	2
Ile491Phe	S	S	S	1	0	0	0	0	0	0	0	0	1	2
Total				1	0	1	3	4	5	5	4	8	1	32

^aNA, not available.

^bR, resistant.

^cS, susceptible.

Three isolates showed discordant results in the MGIT DST between the previous and revised CCs (Table 4). Two of these isolates with borderline resistance mutations showed low-level RIF resistance with an MIC of 2.0 mg/L, whereas the other genotypically wild-type isolate was susceptible with an MIC of 1.0 mg/L.

TABLE 4.

Discordant MGIT DST results between CCs of 0.5 mg/L and 1.0 mg/L of RIF

Sample ID	RIF pDST (CC, mg/L)
	AC (40)	MGIT (1)	MGIT (0.5)	MIC (interpretation)a	rpoB sequencing	Type of isolate
SMC_036	Sb	S	Rc	1.0 mg/L (S)	Genotypically wild-type	INHd-R (pre-treatment)
KMRC-028	R	S	R	2.0 mg/L (R)	His445Leu	MDRe
KMRC-077	R	S	R	2.0 mg/L (R)	Asp435Tyr	XDRf

^aResults of the 7H9 broth microdilution method. MIC values of ≥2.0 mg/L were interpreted as resistant according to CLSI criteria (20).

^bS, susceptible.

^cR, resistant.

^dINH, isoniazid.

^eMDR, multidrug-resistant.

^fXDR, extensively drug-resistant.

DISCUSSION

Since the WHO endorsed 1.0 mg/L as the RIF CC for the MGIT 960 system in 2008, that CC has been used for more than 10 years (24). However, previous studies have shown that this RIF CC (1.0 mg/L) incorrectly classified strains with borderline resistance mutations (8–11, 25). A TEG meeting was convened by the WHO in 2020 to assess results of a systematic review of published literature on isoniazid and RIF CCs for pDST. They revealed that one reason for the discordant results for borderline resistance mutations was that the CCs for MGIT and 7H10 had been set too high for the phenotypical wild-type (pWT) population. Therefore, the WHO lowered the RIF CCs to the tentative epidemiological cutoff value (ECOFF) of 0.5 mg/L (7). However, it is of note that this would only reduce, not eliminate, the discordance between genotype and phenotype, more so for MGIT than for other DSTs, for 2 reasons (7, 26). First, it was unclear whether the revised MGIT CC was at the upper end of the pWT MIC distributions, and it may still be above the true ECOFF. Second, the MIC distributions of MTBC harboring borderline resistance mutations usually overlap with those of pWT isolates because of the shorter incubation period for MGIT DST (10). In the systematic review by the WHO, the sensitivity of MGIT DST with the revised CC was estimated to be increased from 26% to 47% by reducing the CC to 0.5 mg/L; and the specificity was estimated to be 99.8% (522/523) (7). Our study demonstrated that only one isolate was misclassified as RIF-resistant with the revised CC. However, the sensitivity of the MGIT DST was lower than that of the AC method and was not significantly increased by reducing the CC to 0.5 mg/L; only 2 additional isolates out of 32 isolates with single borderline resistance mutations were correctly classified as RIF-resistant using the revised CC. Additionally, we reaffirmed that the misclassification of borderline resistance mutations as susceptible is primarily a problem of testing with the MGIT system rather than with the AC method, which uses LJ medium (10, 26).

Torrea et al. previously evaluated the ability of MGIT DST to detect RIF borderline resistance-conferring rpoB mutations with 38 mutant strains (10). They found a significant increase in the ability of MGIT DST to detect borderline-resistance mutations at an extended incubation period of 15 days with a CC of 0.5 mg/L. However, extension of the incubation period cannot be easily adopted in routine clinical practice. Torrea et al. (10) also demonstrated that a CC of 0.125 mg/L with a standard incubation time dramatically increased the sensitivity; however, 2 of 13 pWT strains were misclassified as RIF-resistant. Recently, Wang et al. reevaluated the RIF breakpoint concentration using 61 MTBC isolates with 5 different borderline resistance mutations and 40 RIF-susceptible pWT isolates (18). They determined MICs of MTBC isolates via MGIT system ranging from 0.031 to 16 mg/L and demonstrated that a large proportion of the isolates with borderline-resistance mutations would not be detected using the revised CC, which yielded a sensitivity of 49.2% for detection of RIF resistance, consistent with our observations. Of particular note, their study suggested that the CC for MGIT DST should be reduced to 0.125 mg/L to significantly increase the sensitivity (18). Although their study showed that all 40 RIF-susceptible isolates were correctly detected at their suggested CC (18), more MGIT MIC data for pWT isolates are needed to investigate whether the CC can be further lowered without resulting in a high rate of false resistance in RIF-susceptible isolates (7).

The WHO has recently reaffirmed that any mutations in the RRDR should be assumed to confer RIF resistance (6, 7). Specifically, borderline-resistance mutations should be regarded as clinically relevant for the current dose of RIF and should be treated with a second-line drug regimen (7, 27). Cases classified as RIF-resistant by genotypic DSTs (gDSTs) and susceptible by pDSTs can be associated with borderline-resistance mutations. In this study, we confirmed that MTBC isolates with borderline resistance mutations had lower MIC values, which ranged widely from ≤0.0625 to >8.0 mg/L using the 7H9 broth microdilution method. Indeed, the revised CC for MGIT misclassified a majority of isolates with borderline resistance mutations as RIF-susceptible, and the revision of CC for MGIT did not significantly reduce the misclassification of borderline resistance mutations. It is noteworthy that gDSTs are designated as the reference standard for borderline-resistance mutations. Therefore, a combination of pDST and gDST is still advisable for MTBC DST to accurately detect borderline resistance mutations.

Although our study comprises the largest number of MTBC isolates for evaluating the diagnostic performance of the MGIT DST with the revised RIF CC, this study has some limitations. First, not all the known mutations associated with borderline resistance were represented. Isolates with His445Ser substitutions were not included since they are rarely isolated in South Korea (25, 28). In addition, small numbers of isolates in our study had His445Asn and Leu452Pro substitutions; however, they were excluded from subgroup analysis for isolates with single borderline resistance mutations since they had multiple mutations. Second, rpoB mutations were identified by partial (700 bp) rpoB gene sequencing, and although all 7 borderline resistance mutations were covered, rare mutations outside RRDR, such as Val170Phe, might not be detected. To compensate for this limitation, entire rpoB gene sequencing was performed for five isolates, which were genotypically wild-type by partial rpoB sequencing and showed MIC values of >0.5 mg/L, to exclude the possibility of mutations outside of the RRDR; however, no mutations were found in the entire rpoB gene sequence. Third, we only tested the previous and revised CCs for MGIT DST, and other RIF concentrations may warrant testing.

In conclusion, our results demonstrate that the overall diagnostic performance of the MGIT DST based on the revised 0.5 mg/L RIF CC was comparable to the performance with the previous CC. Moreover, for the subgroup with single borderline resistance mutations, the sensitivity was not significantly increased by reducing the CC by one dilution. A further large-scale study with a broader RIF concentration range for MGIT DST and isolates with more diverse mutations is required to demonstrate the optimal MGIT RIF CC.