Drug resistance mutations and heteroresistance detected using the GenoType MTBDRplus assay and their implication for treatment outcomes in patients from Mumbai, India

Background Only 5% of the estimated global multidrug resistant TB (MDRTB) load is currently detected. Endemic Mumbai with increasing MDR would benefit from the introduction of molecular methods to detect resistance. Methods The GenoType MTBDRplus assay was used to determine mutations associated with isoniazid and rifampicin resistance and their correlation with treatment outcomes. It was performed on a convenience sample comprising 88 onset and 67 fifth month isolates for which phenotypic drug susceptibility testing (DST) was determined by the Buddemeyer technique for an earlier study. Simultaneous presence of wild type and mutant bands was referred to as "mixed patterns" (heteroresistance). Results Phenotypically 41 isolates were sensitive; 11 isoniazid, 2 rifampicin, 2 pyrazinamide and 5 ethambutol monoresistant; 16 polyresistant and 78 MDR. The agreement between both methods was excellent (kappa = 0.72-0.92). Of 22 rifampicin resistant onset isolates, the predominant rpoB mutations were the singular lack of WT8 (n = 8) and mixed D516V patterns (n = 9). Of the 64 rifampicin resistant fifth month isolates, the most frequent mutations were in WT8 (n = 31) with a further 9 showing the S531L mutation. Mixed patterns were seen in 22 (34%) isolates, most frequently for the D516V mutation (n = 21). Of the 22 onset and 35 fifth month katG mutants, 13 and 12 respectively showed the S315T1 mutation with loss of the WT. Mixed patterns involving both S315T1 and S315T2 were seen in 9 and 23 isolates respectively. Seventeen of 23 and 23/35 inhA mutant onset and fifth month isolates showed mixed A16G profiles. Additionally, 10 fifth month isolates lacked WT2. Five onset and 6 fifth month isolates had both katG and inhA mutations. An association was noted between only katG but not only inhA resistance and poor outcome (p = 0.037); and additional resistance to ethambutol (p = 0.0033). More fifth month than onset isolates had mixed profiles for at least 1 gene (p = 0.000001). Conclusions The use of the assay to rapidly diagnose MDR could guide simultaneous first- and second-line DST, and reduce the delay in administering appropriate regimens. Furthermore, detection of heteroresistance could prevent inaccurate "cured" treatment outcomes documented through smear microscopy and permit more sensitive detection of neonascent resistance.


Background
Tuberculosis (TB), a leading cause of death globally, with increasing rates of drug resistance is of concern. Timely diagnosis and treatment are the key elements of the effort to combat TB and reduce transmission by rendering infectious cases non-infectious.
Only 5% of the World Health Organization (WHO) estimated global multidrug resistant TB (MDRTB) case load of 440,000 is currently detected [1,2]. Detection by conventional drug susceptibility testing (DST) requires considerable resources of infrastructure and trained personnel. The WHO recommends the use of MGIT960 and line probe assays (LPAs) towards quicker MDR detection [3] since phenotypic DST takes 4 to 6 weeks from the receipt of clinical samples.
Commercial and in-house systems for the rapid detection of rifampicin (RIF) resistant Mycobacterium tuberculosis (M. tb) take 5-8 hrs from the time of sample collection [4,5]. The GenoType MTBDRplus is superior in that it detects mutations associated with both rifampicin and isoniazid resistance unlike the INNO-LiPA Rif.TB (Innogenetics, Belgium) which detects mutations only to the former. Unlike RIF resistance, in which 95% of isolates have mutations within an 81-bp region of the rpoB gene encoding the RNA polymerase β subunit [6], isoniazid (INH) resistance has been associated with mutations in several genes [7,8]. Furthermore, since the technique is polymerase chain reaction (PCR) based, it allows detection of low levels of resistant bacteria amidst a predominantly susceptible population, providing a more accurate representation of the susceptibility of the infecting bacteria [9,10]. Different mutations lead to varying degrees of resistance and influence bacterial ability to multiply [11]. Studies have reported the potential use of sophisticated techniques such as sequencing to detect drug resistance mutations which can serve as epidemiological markers, since the relative frequency of alleles associated with resistance varies geographically [12][13][14]. The application of the GenoType MTBDRplus assay has been reported in other high burden settings such as Russia, South Africa and China, but has not yet been reported from our setting [10,15,16].
Despite guidelines that advocate DST for patients failing any treatment regimen [17], it is only performed in 0.5% of notified previously treated TB cases [1]. The endemic setting of Mumbai with reports of increasing levels of MDRTB [18][19][20][21] and a high case load would benefit from the introduction of molecular methods to detect resistance, overcoming the drawbacks of culture methods.
This study was therefore undertaken to evaluate the MTBDRplus for detection of MDR, defined as resistance to at least INH and RIF, in pulmonary TB (PTB) patients in Mumbai. The dual objectives thus encompassed determination of the nature and frequency of mutations associated with resistance and correlations, if any, between the type of mutation and treatment outcome of the patient. Additionally, the assay enabled the detection of heteroresistance to both INH and RIF.

Location and patient selection
This study was carried out on isolates from sputum samples collected from April 2004 to September 2007, for an epidemiological project on MDRTB transmission in Mumbai [18]. Samples were collected at onset and fifth month of treatment from patients registered with the DOTS (Directly Observed Therapy Short Course) Centers of the RNTCP (Revised National Tuberculosis Control Programme) in 4 centrally located wards, characterized by a high sputum positive case load and covering a population of 3 million. Patients were sputum smear positive, within the 15-70 yr age group, and commenced on 2 months of isoniazid, rifampicin, pyrazinamide and ethambutol followed by 4 months of isoniazid and rifampicin, each phase thrice weekly. To minimize bias towards MDR, patients with no history of TB or antitubercular therapy, and those with no interruption of treatment or interruption for < 2 weeks, were included after informed consent. Clearance for the parent study was obtained from the Foundation for Medical Research (FMR) Institutional Ethics Committee (20.07.2001/01).
A total of 681 onset and 345 fifth month samples were collected for the parent study. A convenience subset comprising of 88 onset and 67 fifth month samples covering different susceptibility profiles was used for this analysis.
MDR was defined as resistance to at least INH and RIF. Other cases were categorized as: sensitive -absence of resistance to any drug, monoresistance -resistance to only 1 drug and polyresistance -resistance to two or three drugs excluding the INH-RIF combination.
External quality control of phenotypic DST Ten percent of the isolates from the parent study were sent single blinded to the Swedish Institute for Infectious Disease Control, Stockholm, a WHO/International Union Against Tuberculosis and Lung Disease (IUATLD) supranational reference laboratory for external quality assurance by the Bactec 460 method. Kappa scores showed excellent agreement (range 0.76-0.77), for INH and RIF [18].
The GenoType MTBDRplus assay (Hain Lifesciences, Germany) DNA for the assay was extracted from bacilli grown on LJ media by the cetyl trimethyl ammonium bromide (CTAB) and chloroform isoamyl alcohol method [25]. The assay was performed according to the manufacturer's instructions. The test is based on DNA strip technology comprising multiplex PCR amplification and reverse hybridization. In general for the 3 loci, a pattern comprising only WT bands was interpreted as sensitive. Resistance was interpreted as: i) absence of 1/more WT bands ii) presence of mutant bands with or iii) without the simultaneous absence of the complementary WT. The simultaneous presence of WT and corresponding mutant bands was referred to as a mixed pattern.

Data analysis
Patterns of RIF and INH resistance in MDRTB and non MDRTB isolates were analysed. The ability of RIF resistance alone to predict MDR was also investigated. Results obtained through phenotypic DST and the GenoType MTBDRplus were compared for RIF, INH and MDR and kappa scores were generated for concordance. The interpretations were: < 0.45 = Poor, 0.45 -0.70 = Fair and > 0.70 = Excellent [26]. The Chi square test using EpiInfo 2002 was used to determine association between parameters and a p-value of ≤ 0.05 was considered significant.

Comparison of the Buddemeyer and GenoType MTBDRplus assays
Excellent agreement was obtained between the 2 methods for the onset, fifth month and combined isolates for INH and RIF as represented in Table 1.

Mutations identified using the GenoType MTBDRplus assay
The frequency of mutations detected in the 3 genes is described in Table 3.

RpoB
Twenty-two onset and 64 fifth month isolates showed resistance. For the onset isolates, the most frequent mutation was the absence of only WT8 (n = 8). A further 4 isolates showed the corresponding S531L mutation. Mixed patterns for D516V were seen in 9 isolates. Of the fifth month isolates, the most frequent mutations were in codons 530-533 (n = 31), wherein 26 showed the absence of WT8 alone; or in combination with absence of WT6, 7 (n = 2); WT7 (n = 2) or WT1, 6 (n = 1). A further 9 isolates showed the corresponding S531L mutation. Only 1 isolate showed a H526Y mutation. Mixed patterns were seen in 22/64 isolates (34%), the most frequent for the D516V mutation (n = 21). For those WT bands with a corresponding mutant band (WT 3, 4, 7 and 8), loss of WT only was seen in 9 onset and 40 fifth month isolates.

KatG
Resistance was detected in 22 onset and 35 fifth month isolates. Lack of the wild type with the presence of the S315T1 mutation were seen in 13 onset and 12 fifth month isolates, of which 9 and 7 respectively also showed the S315T2. Mixed patterns involving both mutations were seen in 9 and 23 isolates respectively. Amongst onset isolates, an association was noted between katG mutations and resistance to ethambutol (c 2 = 8.63, p = 0.0033). Loss of WT only was seen in 3 onset and 10 fifth month isolates. Analysed collectively, 40 onset and 64 fifth month isolates showed INH resistance. Of these, 17 onset (43%) and 29 fifth month (45%) isolates showed resistance only through katG while in both groups an equivalent proportion [45% (18 onset and 29 fifth month)] were resistant only through inhA. The remaining 5 onset and 6 fifth month isolates had mutations in both genes.
Overall, 46 onset and only 1 fifth month isolate were sensitive to INH and RIF. Of 18 onset and 4 fifth month isolates resistant to only INH, 2 onset isolates had dual mutations in katG and inhA, while the remainder had mutations in either 1. No fifth month isolates had such dual mutations. Two isolates each among the onset and fifth month isolates were resistant to only RIF (Table 3).
Of the 82 MDR isolates (20 onset and 62 fifth month), the predominant patterns involved absence of WT8 with either S315T1/2 or Δ-8 mutations (Table 3). No association was found between a particular mutation and a monoresistant or MDR profile.
An analysis of the profiles for all 3 genes revealed that a significant number of fifth month isolates had mixed profiles in comparison to onset isolates (c 2 = 23.75, p = 0.000001). For katG, an association (p = 0.0012) was noted between a clean resistant profile (as against a mixed profile) and poor outcome (death or treatment failure). Similarly an association was noted between only katG resistance (as against only inhA resistance) and poor outcome (c 2 = 4.37, p = 0.037). Cure was associated with A16G (p = 0.02) and Δ-8 (p = 0.015) in comparison to a C15T mutation in the inhA.

Discussion
To the best of our knowledge this is the first study from this setting to investigate genotypic profiles in the rpoB, katG and inhA regions associated with DR using the MTBDRplus assay. Techniques which detect MDR mutations in new cases at onset or during therapy would enable rapid identification of MDR and facilitate the modification of regimens with improvement to programme practices.
Overall, the concordance between the methods for INH and RIF ranged from 86-97% and 94-99% respectively. Fifteen discordant results were obtained for INH resistance, of which 5 were phenotypically resistant but genotypically sensitive, possibly because the mutations lie outside the regions covered by the probes. Ten discordant results were Buddemeyer sensitive but MTBDRplus resistant, wherein 5 were inhA mutants, known to be associated with low level resistance [27]. Of the remainder, 4 were onset isolates with katG mutations, of which 2 showed a mixed profile. This may explain the phenotypic sensitivity despite the katG mutations being associated with high level resistance [27]. The remainder 2 profiles showed the katG S315T1 mutation but were INH sensitive by the Buddemeyer. For one of these isolates, the interpretation of the drug susceptibility could have altered from sensitivity to resistance on extended incubation. For the second isolate, there was no indication of resistance and the discordance could reflect a technical anomaly in the phenotypic assay, or an "adaptation" by the strain which has prevented phenotypic expression of the katG mutation.
Of 5 isolates that were MTBDRplus RIF resistant but Buddemeyer sensitive, 4 lacked the WT8 band, including 3 which showed the MUT3 (S531L). Though associated with high level resistance, its detection at onset could Only RIF (n = 4) imply a smaller proportion of resistant bacteria. Low level resistance which may remain undetected despite conventional DST has been previously reported [28,29]. Heteroresistance, reflecting the slow evolution of bacteria from a sensitive to resistant profile, is not uncommon in M. tb [30]. However our study is probably the first to report relatively high levels of heteroresistance in all 3 genes in an endemic setting. Two explanations are offered for this finding. Firstly, heteroresistance could have arisen due to transmission of both susceptible and resistant bacterial populations from drug resistant patients to previously untreated cases. An endemic setting like Mumbai would be prone to the prevalence of MDR strains, and hence new cases, even at onset, are likely to harbour resistant bacteria at proportions genotypically detectable but phenotypically undetectable, as suggested by the detection of mixed profiles in our onset isolates [31]. Studies on heteroresistance have shown that phenotypic DST results corresponded to the mutated, i.e. resistant, organism [31]. Secondly, the presence of exclusively sensitive bacteria at onset which gradually develop resistance during therapy, with incomplete elimination of the sensitive population by fifth month, would result in phenotypic resistance but with both forms remaining detectable genotypically. This possibility has been explained through mathematical modelling of the scenario in which MDR bacilli arise from a completely sensitive original infection [32]. The second scenario may explain the higher occurrence of mixed patterns among our fifth month isolates in comparison to the onset isolates. The probability of any bias towards detection of heteroresistance has been reduced by excluding patients with various other likely contributory factors such as prior treatment and defaulting. DNA fingerprinting techniques such as Mycobacterial Interspersed Repetitive Units Variable Number Tandem repeats (MIRU VNTR) would help gauge whether multiple infections have contributed to the heteroresistance detected.
The detection of heteroresistance seems to support our finding of an association between a clean resistant profile for the katG and a poor outcome, since the presence of exclusively resistant strains is more likely to result in non responsiveness to treatment. Identification of hetereoresistance can be used to probe outcomes of smear examination based "cure" since microscopy may not be sensitive enough to detect a small focus of bacteria which have evolved from sensitivity to resistance during therapy.
Mutations in katG 315 may be favoured because they decrease INH activation without abolishing catalase-peroxidase activity, reflecting its low fitness cost [11,33,[39][40][41]. The detection of equal proportions of katG and inhA mutations indicate that in our setting, inhA may also have a low fitness cost. This may be a consequence of the extended occurrence of MDR allowing for acquisition of compensatory mutations such as in the ahpC [42]. Despite their equivalent levels, katG but not inhA mutations, were associated with treatment failure.
Our data identified associations between katG 315 mutations and ethambutol resistance as well as poor outcome. It has been reported that the iniBAC promoter is induced by cell wall biosynthesis inhibitors such as isoniazid and ethambutol [43]. Overexpression of the iniA gene confers a tolerance-like phenotype to INH and ETB. Furthermore, the iniA gene product is an essential component of an MDR-like pump [44]. Resistance to INH through katG mutations might influence response to other drugs, allowing development of resistance to ethambutol [45] and streptomycin [46], leading to poor treatment response.
Of the 86 RIF resistant isolates, 70 showed a mutation in the 530-533 region of the rpoB. However only 23 (27%) showed the specific S531L mutation as compared to other studies reporting 46-79% of strains with this mutation [10,27,36,37]. The low fitness cost of rpoB S531L [47] may account for its high frequency in these regions (i.e. South Africa, France, Spain) though its occurrence has also been reported to be as low as 30-31% in India and Hungary [12,48]. The lower proportion of the S531L compared to the D516V mutation indicates that the latter is also not associated with a fitness cost, at least in our setting.
Our study failed to find any association between a particular mutation and the occurrence of monoresistance or MDR. However, other studies have reported a significantly higher level of katG and S531L mutations in MDR isolates compared to INH or RIF monoresistant isolates respectively [10,38]. It is likely that this difference is due to the relatively low occurrence of S531L and the equivalent proportions of katG and inhA mutations in our cohort.
The value of RIF as a surrogate MDR marker has been documented [49] and further corroborated in our study. Despite their advantages, genotypic methods do not always identify phenotypically resistant strains [38], highlighting the limitations of molecular testing and need for supplementation with culture or additional probes. Additionally resistance can be inferred from the absence of a wild type signal alone, without confirmation by a mutant probe signal (in 47% of our isolates) and may be due to a mutation in a region not associated with resistance [29]. Such susceptible isolates would be called resistant leading to the unnecessary removal of RIF and/ or INH from therapy. Moreover since a proportion of INH resistance, particularly in monoresistant isolates, could be due to resistance determinants other than katG S315T and inhA C15T, these isolates would also be indicated as susceptible. This highlights the need for the interpretation of genotypic data in conjunction with patient clinical status and the determination of mutations specific to certain geographical locales.

Conclusions
Despite the limitations of the MTBDRplus assay, its rapidity and sensitivity still favour its implementation as an initial screen for MDRTB. The presence of resistance mutations and the occurrence of heteroresistance could not only guide the need for simultaneously initiating first-and second-line DST, but also significantly reduce the delay in administering an appropriate regimen to an MDR (or extensively drug resistant) TB case.