Evaluation of Genotype MTBDRplus and MTBDRsl Assays for Rapid Detection of Drug Resistance in Extensively Drug-Resistant Mycobacterium tuberculosis Isolates in Pakistan

Pakistan ranks 5th among the world's highest tuberculosis (TB) burden countries alongside the 6th among countries with the highest burden of drug-resistant TB, including multi-drug resistant (MDR)-TB. Methods for rapid and reliable drug susceptibility testing (DST) are prerequisite for the prompt institution of effective anti-TB treatment. The aim of this study was to evaluate the efficiency of Genotype MTBDRplus and MTBDRsl assays for the detection of MDR and (pre-) extensively drug-resistant (XDR-TB) isolates in Pakistan. The study included 47 pre-XDR and 6 XDR-TB isolates, recovered from 53 patients from Pakistan. Conventional DST was performed using the standard 1% proportion method on the Löwenstein-Jensen medium. For molecular determination of drug resistance, GenoType MTBDRplus and GenoType MTBDRsl assays (Hain Lifescience, Germany) were used. To evaluate discrepancies between conventional and molecular DST results, mutation profiling was performed by amplifying and sequencing seven genetic loci, i.e., katG, inhA, and mabA-inhA promoter, rpoB, gyrA, embB, rrs. The sensitivity of Genotype MTBDRplus was 71.7% for isoniazid (INH) and 79.2% for rifampicin (RIF). Sequence analysis revealed non-synonymous mutations in 93.3 and 27.3% of isolates phenotypically resistant to INH and RIF, respectively, albeit susceptible when tested by GenoType MTBDRplus. GenoType MTBDRsl had a sensitivity of 73.6, 64.7, 20, 25, and 100% for the detection of fluoroquinolones, ethambutol, kanamycin, amikacin, and capreomycin resistance, respectively. Upon sequencing, mutations were detected in 20, 77.8%, and all isolates phenotypically resistant to aminoglycosides, ethambutol, and fluoroquinolones, respectively, yet declared as susceptible with GenoType MTBDRsl. Low sensitivities seriously impede the large-scale application of the Genotype MTBDRplus and MTBDRsl assays. Unless further optimized, the currently available line-probe assays should rather be auxiliary to the conventional, phenotype-based methods in the detection of MDR- and XDR-TB in Pakistan.


INTRODUCTION
Tuberculosis (TB) remains an inglorious leader among infectious diseases in mortality, with its annual toll of 1.5 million lives worldwide (Ullah et al., 2016). Although the global TB incidence has been on a downward trend since 2000, the emergence and persistence of drug-resistant (DR) tubercle bacilli strains, particularly those multi-drug resistant (MDR), defined as resistant to at least isoniazid (INH) and rifampicin (RIF), and extensively drug-resistant (XDR), defined as MDR with additional resistance to a fluoroquinolone (FQ) and a secondline injectable drug (SLID), have substantially undermined efforts to control and eliminate the disease. According to a most recent World Health Organization (WHO) report, every seventeenth TB patient expels MDR bacilli. One in sixteen of such patients expel strains of XDR phenotype (World Health Organization, 2017). For almost two decades, Pakistan, with a population of 193 million people, continues to be on the top of the list high TB-burden countries (HBCs), in terms of total TB caseload (TB incidence rate, 268 per 100,000 population) and in terms of DR-TB, including MDR-TB (MDR-TB incidence rate, 14 per 100,000 population), globally (World Health Organization, 2017). Studies on the prevalence of DR-TB, including MDR-TB and XDR-TB in Pakistan are quite scarce and fragmentary (Ali et al., 2011(Ali et al., , 2015Farooqi et al., 2012;Javaid et al., 2016). Also, data collected from national surveys on DR-TB and reported to WHO have to be cautiously treated. The true prevalence of DR-TB is thought to be underestimated, mostly due to a limited number of facilities offering drug susceptibility testing (DST) and poor availability of modern and advanced technologies, allowing for fast and reliable drug resistance profiling (Domínguez et al., 2016).
Conventional DST is laborious, time consuming and requires growth of mycobacteria either on solid or liquid media . Identification of the XDR phenotype is particularly protracted as it is a two-step procedure that first involves testing against first-line anti-TB drugs, and then against second-line agents, once multidrug resistance is disclosed (Abebe et al., 2011). Rapid and reliable DST results are prerequisite for the prompt implementation of effective treatment, and reducing the risk of acquired resistance (Hillemann et al., 2009). Over the last decade, diagnosing DR-TB has been greatly improved and expedited with the introduction of various molecularbased DST technologies (Thumamo et al., 2012). They all fall into two major categories that is probe-based and sequencebased methods. The former are represented by Cepheid's GeneXpert MTB/RIF assay (Cepheid, USA) with molecular beacon probes (Ioannidis et al., 2011) and Hain's MTBDRplus and MTBDRsl assays (Hain Life Sciences, Germany) with line probes (Nathavitharana et al., 2017). Whereas molecular beacon-based assays use real-time PCR, dual-labeled probes that form a quenched, stem-loop structure in native state and fluoresce upon hybridization to the target nucleotide sequence (Lawn and Nicol, 2011), line probe assays involve PCR and reverse hybridization with specific oligonucleotide probed fixed to a nitrocellulose strip in parallel lines (World Health Organization, 2008). Sequence-based methods include pyrosequencing (Molina-Moya et al., 2017), Sanger sequencing (Schleusener et al., 2017), and next-generation sequencing (Jagielski et al., 2016).
The implementation of routine sequencing to track drugresistance profiles has already become a part of a strategy aiming at TB elimination in England, and toward similar plans in other high-income countries Satta et al., 2018). However, DNA sequencing-based approaches remains prohibitevely expensive and complex for routine use in low-and middle-income countries, especially those heavily burdened with TB (Votintseva et al., 2017).
In December 2010 WHO approved the GeneXpert MTB/RIF test for detection of DR strains in high-burden, resource-limited countries (World Health Organization, 2013). An important drawback of the system is that it detects resistance to RIF but not to other anti-TB drugs such as INH, FQs, and SLIDs. Furthermore, whereas PCR inhibition or unsuccessful DNA extraction may increase the chances of false-negative results, false positive results might occur in previously treated patients and having mixed TB/non-tuberculous mycobacteria (NTM) infection (Liu et al., 2017).
The Hain probe assays use reverse hybridization technology to detect mutations associated with resistance of tubercle bacilli to both first-and second-line anti-TB drugs. MTBDRplus allows for identification of INH and RIF resistance by disclosing mutations in the katG, inhA, and rpoB genes, while MTBDRsl detects resistance to FQs, ethambutol (EMB), aminoglycosides (kanamycin, KAN; amikacin, AMK; viomycin, VIO), and cyclic peptide (capreomycin, CAP) by finding mutations in three different loci, that is gyrA, embB, and rrs, respectively (Bai et al., 2016).
Based on two recent meta-analyses, including over 50 studies, the overall sensitivities of the two Genotype assays varied widely from 96% for RIF to 44% for KAN, while the specificities ranged from 99% for INH, RIF, AMK, and KAN to 79% for EMB (Feng et al., 2013;Bai et al., 2016). Still, conventional, culture-based DST is considered a gold standard in mycobacteriology.
The aim of the study was to evaluate the efficiency of Genotype MTBDRplus and MTBDRsl assays in the detection of drug resistance in the context of conventional DST profiling and results of PCR-sequencing targeted at selected drug resistanceassociated loci.

Isolates
The study sample was selected out of a pool of 3056 Mycobacterium tuberculosis isolates, recovered from as many patients, originally diagnosed as having pulmonary TB and originated from seven major cities of Punjab province of Pakistan from January 2013 to June 2015. MDR-TB, based on the conventional DST results, was initially identified in 362 of the cases ( Figure S1). These cases were further reviewed to select those where (pre-)XDR-TB was confirmed. Thereby, the final study group of 53 patients was achieved ( Table 1).
pre-XDR phenotype was defined as MDR with either resistance to ofloxacin (OFX) or any of the SLIDs (i.e., AMK, KAN or CAP). XDR-TB was defined as MDR-TB with additional resistance to OFX and one of the SLIDs.
Primary isolation, culturing, and species identification were performed with standard mycobacteriological methods in Provincial TB Reference Laboratory, Institute of Public Health, Lahore. Briefly, clinical samples were decontaminated with 1% N-acetyl-L-cysteine/NaOH and centrifuged. Each sample was then inoculated onto Löwenstein-Jensen and MGIT (Becton Dickinson, Franklin Lakes, NJ, USA) media. Identification was performed using the BACTEC NAP TB Differentiation Test Kit (Becton Dickinson, USA), growth in para-nitrobenzoic acidcontaining media, nitrate reduction, and niacin accumulation (Koneman et al., 1983).
This study was carried out in accordance with the recommendations of ethical policy of the University of Punjab.

Drug Susceptibility Testing
Conventional DST was performed using the standard 1% proportion method on the Löwenstein-Jensen (L-J) medium, with the M. tuberculosis H37Rv strain as a quality control, following the WHO recommendations (Van Embden et al., 1993). The critical concentrations for specific drugs were as follows: INH, 0.2 mg/L; RIF, 40 mg/L; EMB, 2 mg/L; STR, 4 mg/L; KAN, 30 mg/L; AMK, 30 mg/L; CAP, 40 mg/L; and OFX, 4 mg/L.
For molecular determination of drug resistance, GenoType MTBDRplus and GenoType MTBDRsl assays (Hain Lifescience, Germany) were used. The principle of these tests was essentially described elsewhere (Anek-vorapong et al., 2010). Both assays were performed and interpreted in accordance with the manufacturer's instructions. A test result was considered valid, only if all control bands appeared correctly. An isolate was declared resistant if at least one wild type probe was absent or if any mutant probe was present. If all wild type probes were present and all mutant probes were absent, an isolate was recognized as susceptible.

DNA Extraction
Genomic DNA from M. tuberculosis isolates, grown on L-J medium, was extracted using the cetyl-trimethyl ammonium bromide (CTAB) method, as described previously (Jagielski et al., 2015).

PCR and Sequencing
To evaluate discrepancies between conventional and molecular (GenoType MTBDRplus and MTBDRsl) DST results, selected fragments of seven genetic loci, containing the hot spots for mutations associated with resistance of tubercle bacilli to INH (katG, inhA, and mabA-inhA promoter), RIF (rpoB), EMB (embB), aminoglycosides (rrs), and FQs (gyrA) were PCRamplified and sequenced. The oligonucleotide primers used for PCR are described in Table S1. Amplification reactions were set up and performed according to the manufacturer's instructions (TopTaq DNA polymerase, Qiagen, Germany). Purified PCR amplicons (Clean-up kit, A&A Biotechnology, Poland) were sequenced in both directions using the same primers as for PCR amplification. Mutations were detected using Clone Manager software (v. 8.0, Scientific & Educational Software, USA) by comparing the obtained sequences with the M. tuberculosis reference strain H37Rv sequences of respective loci, deposited in the GenBank database (National Center for Biotechnology Information; http://www.ncbi.nlm.nih. gov/). An isolate was declared resistant if at least one nonsynonymous mutation was detected in a resistance-associated locus.

Statistical Analysis
Data were analyzed with the IBM SPSS software (version 22.0, USA). To analyze differences between categorical variables, χ 2test was used. A value of P < 0.05 was considered statistically significant.

Diagnostic Performance
To assess the diagnostic performance of the line-probe assays, sensitivities and positive predictive values (PPV) were calculated for all drugs covered by the GenoType MTBDRplus and MTBDRsl assays. Since the study sample included no INH-, RIF-, FQ-susceptible isolates, specificities and negative predictive values (NPV) were calculated only for KAN, AMK, CAP, and EMB of GenoType MTBDRsl.

Conventional DST
All (

Performance of Genotype MTBDRplus
The results of DST with the GenoType MTBDRplus assay are shown in Table 2, while representative patterns obtained with the assay are depicted in Figure 1.
Concordance between DST results obtained with the MTBDRplus assay and conventional methods is shown in Table 3. The sensitivity of the MTBDRplus assay was calculated at 79.2% and 71.7% for the detection of RIF and INH resistance, respectively.

Performance of Genotype MTBDRsl
The results of DST with the GenoType MTBDRsl are shown in Table 2. Detailed patterns of drug resistance are given in Table S2.

DISCUSSION
Genotype MTBDRplus and MTBDRsl have become increasingly popular in mycobacteriology laboratories serving as fast molecular assays for detection of drug resistance. However, studies exploring the diagnostic performance of these two assays in HBCs are seriously lacking. This work evaluates the usefulness of Genotype MTBDRplus and MTBDRsl for the detection of drug resistance in M. tuberculosis isolates in Pakistan. It is also the first study from Pakistan investigating drug resistance profiles of tubercle bacilli (MDR-TB isolates) with a threepronged approach, that is conventional, culture-based method, line-probe, hybridization assays, and PCR sequencing.
Among isolates in which Genotype MTBDRplus assay failed to detect INH resistance, all but one had a missense mutation in codon 463 of the katG gene, as evidenced by PCR-sequencing. This codon is not covered by any of the probe of the Genotype MTBDRplus assay. Mutations at codon 463 of the katG gene have been identified in both INH-resistant (29-59.2%) and INH-susceptible (32-63%) strains, suggesting that they not directly associated with INH resistance (Van Doorn et al., 2001;Arjomandzadegan et al., 2011;Torres et al., 2015). One isolate carried a katG S315N mutation, which had previously been correlated with INH-resistant phenotype (Wei et al., 2003). Since the Genotype MTBDRplus test strip has no probe specific for this mutation, it cannot be detected with the assay. Still, almost a fourth (23.8%) of INH-resistant isolates can harbor this mutation (Jin et al., 2012). Apart from codons 315 and 463, mutations in the katG gene were identified at two other codons i.e., 189 and 232. Only mutation in the latter codon (P232R) had previously been described in one INHresistant isolate by (Greif et al., 2012), with a frequency of 2.2% among INH-resistant isolates. The precise role of these mutations in the development of INH resistance needs further investigation.
In case of RIF, three out of 11 isolates falsely designated as RIFsusceptible with GenoType MTBDRplus had missense mutations in their rpoB gene sequences i.e., S531L, S531W (in regions covered with the assay), and I572F (region not covered with the assay). I572F mutation had previously been linked with RIF resistance (Siu et al., 2011).
All fourteen isolates falsely designated as FQ-susceptible upon GenoType MTBDRsl carried a S95T gyrA substitution, which is a natural polymorphism, not associated with FQ resistance (Bakuła et al., 2016). Five isolates carried gyrA alterations missed with Genotype MTBDRsl (i.e., A90V, S91P, D94A, D94G, or D94Y). According to the literature, they have all been associated with FQ resistance.
Fourteen out of 18 isolates in which Genotype MTBDRsl failed to detect EMB resistance had a change in the embB gene, as revealed by sequence analysis. The most frequent were amino acid changes in codons 497 and 406. Mutations at both these codons were shown to increase resistance to EMB (Safi et al., 2010). Additionally, mutations in five other codons of the embB gene were identified and four of them (except in codon 284) have been described as conferring resistance to EMB (Ali et al., 2015).
Among five isolates phenotypically resistant to SLIDs but missed by GenoType MTBDRsl, only one, with resistance to KAN, carried a mutation (C517T) in the rrs gene, localized in highly mutable region known as the 530 loop (Jagielski et al., 2014). Mutations in this region are associated with resistance to STR but not to other aminoglycosides (Bakuła et al., 2016).
According to this study, both GenoType MTBDRplus and GenoType MTBDRsl, assays display an important level of inconsistency with conventional DST. When compared to the latter, the line-probe assays were unable to detect 20.7, 28.3, 26.4, 35.3, 80, and 75% of isolates resistant to RIF, INH, FQ, EMB, KAN, and AMK, respectively. These false negative results may be explained by two reasons. First is that mutations not covered by the probes might be more prevalent in Pakistan compared to other geographical locales. Inclusion of probes specific for mutations at two embB codons (Q497 and G406) would considerably increase the sensitivity of the MTBDRsl assay to detect EMB-resistant isolates from our sample (i.e., from 64.7 to 82.3%). Also, addition of other probes specific for mutations conferring resistance to INH (e.g., S315N, KatG) or RMP (e.g., I572F, RpoB) would improve the sensitivity of the MTBDRplus by approximately 2%, each (71.7 vs. 73.6% for INH; 79.2 vs. 81.1% for RMP). Poor outcomes of the GenoType MTBDRplus and MTBDRsl assays due to the lack of probes specific for certain mutations were described in previous studies (Huang et al., 2011;Jin et al., 2012;Maschmann et al., 2013). Second are possible technical errors in phenotypic and molecular assays. Factors such as handling procedures, incubation conditions, and end-point interpretation may influence the outcome of conventional DST (Schön et al., 2017). Whereas, losses on DNA extraction or inhibitors of amplification present in the specimens may affect the efficiency of line probe assays (Mäkinen et al., 2002;Padilla et al., 2004). In this study they missed 10 mutations detected by sequencing of the katG (S315T, one isolate, 1.9%), rpoB (S531L or S531W, two isolates, 3.8%), gyrA (A90V, S91P, D94A, D94G or D94Y, five isolates, 9.4%), and embB (M306V, one isolate, 1.9%) genes. Third, drug resistance may originate from mutations at other genes than those included in the probe-line assays (e.g., in ahpC, kasA, and ndh for INH) (Ferro et al., 2013) and this may also result in false negative results.
Finally, false positive results were noted for 50, 11.1, 14.5, and 9.7% of isolates declared as susceptible to EMB, AMK, KAN, and CAP, respectively, with conventional DST. This can be explained by probe mispriming (due to sample contamination or suboptimal procedure conditions) or the presence of synonymous mutations (Ajileye et al., 2017).
To conclude, for detecting drug resistance in TB cases, especially in high TB incidence countries, such as Pakistan, molecular approaches should still be a complement rather than a replacement to conventional DST. The knowledge on frequencies of drug-resistance conferring mutations in clinically and geographically diverse settings, should guide the inclusion of new specific probes in the test strips in the line-probe assays, such as Genotype MTBDR.

AUTHOR CONTRIBUTIONS
HJ executed the strain isolation and culturing, analyzed and interpreted the data, and wrote the manuscript. ZB analyzed and interpreted the data, and wrote the manuscript. MP executed the Genotype MTBDRplus and MTBDRsl analysis, and the sequence analysis. HH performed the species-level identification and drug susceptibility testing. ZT executed the administrative,the technical, and the material support. NJ critically revised the manuscript. TJ designed the study and supervised and wrote and critically revised the manuscript.