Comparison of Three Commercial Molecular Assays for Detection of Rifampin and Isoniazid Resistance among Mycobacterium tuberculosis Isolates in a High-HIV-Prevalence Setting

Abstract

In a head-to-head comparison of the MTBDRplus version 2.0 (Hain Lifescience), the Xpert MTB/RIF (Cepheid), and the Anyplex MTB/NTM (Seegene) assays, we demonstrated equal sensitivity (59/61; 96.7%) and specificity (53/54; 98.1%) for detecting rifampin resistance with further analysis of discordances. The Xpert assay does not detect isoniazid resistance while the Anyplex assay showed high false positivity.

TEXT

There are a limited number of commercial molecular assays available for the rapid detection of drug-resistant tuberculosis (DR-TB), and currently only two are endorsed by the WHO for this purpose: the GenoType MTBDRplus version 1.0 (Hain Lifescience GmbH, Germany) and the Xpert MTB/RIF (Cepheid, USA) (1–3). The MTBDRplus has been further optimized, and this new version (2.0) can now be performed on both smear-positive and smear-negative clinical specimens as well as cultured isolates, according to the manufacturer (4, 5).

The Anyplex MTB/NTM (Seegene, South Korea) assay has not undergone WHO review for use in detecting DR-TB but is widely used. It is a multiplex real-time PCR assay capable of distinguishing between Mycobacterium tuberculosis and nontuberculosis mycobacteria (NTM) while allowing for the amplification of drug resistance-related gene (rpoB, katG, and inhA) sequences simultaneously (6, 7).

In this study, we evaluated the performance characteristics of these three molecular assays for the detection of DR-TB by performing a head-to-head comparison of these technologies.

This was a retrospective laboratory-based evaluation study. Cultured isolates were collected from the TB laboratory, Diagnostic Division of the Department of Medical Microbiology, Tshwane Academic Division, National Health Laboratory Services (NHLS). The results from the MTBDRplus version 1.0 assay, performed as part of the routine laboratory testing formed the basis of stratification into 50 multidrug-resistant (MDR) isolates, 30 fully susceptible isolates, 20 monoresistant (10 rifampin and 10 isoniazid resistant) isolates, and 20 isolates with undefined mutations. The 20 isolates with undefined mutations, having a wild-type missing with no corresponding mutation band on the MTBDRplus version 1.0 assay, were further characterized by means of Sanger sequencing.

These isolates were collected consecutively to make up a total of 120, which was calculated to provide the required sample size for the comparative evaluation. Four isolates from this sample subset were excluded as they were duplicates and another had poor banding repeatedly on the MTBDRplus version 2.0 assay. A total of 115/120 (96%) cultured isolates were analyzable on all three systems.

The MTBDRplus version 2.0, the Xpert MTB/RIF (GXP) (G3 cartridge), and the Anyplex MTB/NTM (Anyplex) assays were performed according to the manufacturer's instructions. For the GXP, 1 ml of cultured isolate was used, and the sample reagent buffer was added in a 2:1 ratio. The extracted DNA templates of isolates with discordant results and isolates with undefined mutations based on MTBDRplus version 1.0 were amplified utilizing the primers as previously described for rpoB, katG, and inhA (8). Cycle sequencing utilizing the BigDye Terminator cycle sequencing ready reaction kit (ABI Prism; Applied Biosystems, USA) was performed according to the manufacturer's instructions. The sequence of the PCR-amplified DNA was then determined utilizing the ABI 3500 XL (Applied Biosystems) genetic analyzer. Single nucleotide polymorphisms and indels were detected using the CLC Genomics Workbench (Qiagen, Germany). All variants were evaluated using the Dream DB database, and high-confidence mutations were regarded as resistant (9). For the comparison of the three molecular assays, we used as a molecular gold standard the original MTBDRplus version 1.0 assay mutation results as well as sequencing results as mentioned before with the sequencing result regarded as final.

Of the 115 isolates evaluated, 61 were rifampin resistant and 59 were isoniazid resistant based on the molecular gold standard. All three of the assays had exactly the same sensitivity (59/61; 96.7%) and specificity (53/54; 98.1%) for the detection of rifampin resistance. However, the discordances were not the same and are shown in Table 1.

TABLE 1.

Resolution of discordant rifampin results by sequencing of the rpoB region^a

Isolate	Result for:			Sequencing
	MTBDRplus version 2.0	Anyplex	GXP
40	S	R	S	Poor sequencing
41b	R	S	R	Insertion TTC 432−433
92c	R	S	S	Wild type
99d	R	S	R	SNP 1862 C>T His>Tyr 445His526Tyr
87e	S	R	S	SNP 1184 T>C Leu>Pro 452Leu533Pro
117f	S	R	S	SNP 1184 T>C Leu>Pro 452Leu533Pro

^aR, resistant; S, susceptible; WT, wild type.

^bIsolate 41: MTBDRplus version 2.0 mutation in codon 510−519 (rpoB WT 3 missing), no mutation band, GXP probe B not binding.

^cIsolate 92: MTBDRplus version 2.0 S531L mutation, no wild type missing, possibly mixed infection.

^dIsolate 99: MTBDRplus version 2.0 mutation in codon 526−529 (rpoB WT 7 missing) and H526Y mutation, GXP probe D not binding.

^eIsolate 87: Repeat on MTBDRplus version 2.0 (later version) mutation in codon 530−533 (rpoB WT 8 missing), repeat on GXP G4 cartridge probe E not binding.

^fIsolate 117: Repeat on MTBDRplus version 2.0 (later version) mutation in codon 530−533 (rpoB WT 8 missing), repeat on GXP G4 cartridge probe E not binding.

Two cases of rifampin resistance were not detected by either the MTBDRplus version 2.0 or the GXP assay; both of these cases had L533P mutations in the rifampin resistance-determining region upon sequencing. When evaluations were repeated using the most recent release of the version 2.0 assay as well as the GXP (G4 cartridge) assay, these mutations were detected by the MTBDRplus version 2.0 assay in codon 530 to 533 of the rpoB region (missing wild-type 8 band with no corresponding mutation band) and by the GXP (G4 cartridge) assay (probe E not binding). This mutation has been associated with discordant genotypic and phenotypic susceptibility results, notably when the widely employed broth-based phenotypic susceptibility testing methods are utilized; hence, the L533P mutation was thought to be a silent mutation, with no impact on the efficacy of rifampin (10), and has also been shown to be missed by the G3 cartridge (11, 12). Upon a retrospective review of our laboratory records, these two isolates had tested as rifampin susceptible with the MGIT 960 system. However, recent data have demonstrated that this L533P mutation is associated with low-level resistance, which is clinically significant, and retesting using the latest versions of the assays has addressed this shortcoming (13–15).

One isolate was detected as falsely rifampin resistant by the MTBDRplus version 2.0 assay against the molecular gold standard, displaying a S531L mutation and wild-type banding pattern at codon 530 to 533, which suggests a mixed population. This isolate was recorded as rifampin resistant according to MGIT 960 phenotypic drug susceptibility testing performed as part of the routine laboratory testing, further supporting the findings for the line probe assay. However, no mutations were detected utilizing Sanger sequencing. Interestingly the GXP assay showed binding of all probes, but this may be a masking effect of a wild-type strain in a mixed population, which is a limitation of the assay design (16). Both the line probe assay and phenotypic susceptibility testing, on the other hand, are known to be able to detect mixed populations (17), and thus the specificity of the MTBDRplus version 2.0 assay may be falsely low in our study.

The Anyplex assay detected one case of false rifampin resistance and two cases of false rifampin susceptibility. Of the two falsely susceptible cases, one had a H526Y mutation and the other an insertion TTC at codon 514 on sequencing. The insertion TTC at codon 514 of the rpoB gene has been described in various studies (18–22) although at low frequency: 2.5% of rifampin-resistant isolates in one study and 4% in another (18, 23). This mutation has been associated with various levels of rifampin resistance with corresponding rifampin MICs of 16 to >256 μg/ml (20, 22). The second mutation that was not detected by the Anyplex assay was an H526Y mutation. This is a commonly occurring mutation associated with drug resistance and thus has important treatment implications if not detected (24). The case of false rifampin resistance could not be resolved by means of Sanger sequencing as this isolate displayed poor sequencing despite repeat attempts; however, the isolate remained wild type upon repeat testing with both the MTBDRplus version 2.0 and the GXP assays.

No comparison for the detection of isoniazid resistance was possible for the GXP assay as this does not form part of the test. The MTBDRplus version 2.0 and the Anyplex assays both had sensitivities of 100% for the detection of isoniazid resistance mutations at codon 315 for the katG gene and inhA promoter region. The Anyplex assay had poor specificity of 82.4% due to falsely reporting 10 isolates as having mutations present in the inhA and/or katG regions. These isolates were defined as wild type by the MTBDRplus version 1.0 assay. All 10 isolates with discordant isoniazid results were wild type on sequencing and also had all wild-type bands present on the MTBDRplus version 2.0 assay.

The three molecular assays under evaluation in this study performed well for the detection of rifampin resistance. The MTBDRplus version 2.0 and GXP (G3 cartridge) assays had good performances and the later versions of both addressed the shortcomings of the earlier versions. The MTBDRplus version 2.0 assay was superior to the Anyplex assay overall for the diagnosis of MDR-TB. The main drawback of the GXP assay is the lack of isoniazid susceptibility testing.

The strengths of this study are the large numbers of resistant isolates evaluated and the direct comparison between all three systems. The limitations of this study are the restrictive genotypic standard which is expected to miss 10 to 15% of isoniazid resistance (25), the lack of sequencing and phenotypic drug susceptibility data for all isolates, and the fact that the assays were not performed on direct samples. The Anyplex assay has the advantage of detecting NTM, but this was not assessed. In the South African context, this is of lesser concern but may be of importance in specialized populations.

Molecular assays can detect nonviable mycobacteria, which can lead to problems in interpretation by clinicians; however, they do offer faster clinically actionable results. The three assays showed overall good performance for the detection of DR-TB, with each having its own strengths and weaknesses.