In the current study we evaluated the performance of the molecular assay (HAIN Lifescience GmbH GenoType® MTBDRplus) for rapid detection of resistance to the most important anti-TB drugs (RIF and INH) on sputum samples in the region with middle TB incidence and high prevalence of MDRTB in Russian Federation (Samara oblast). Until now, within the Russian Federation, there was only limited evidence of applicability of rapid molecular techniques (biochips) for detection of mutations conferring drug resistance directly in clinical specimens ; molecular assays for diagnosis of anti-TB drug resistance are not widely available in the Russian Federation, a high TB burden country .
Substantial reduction in the time to diagnose drug-resistant TB, the earlier commencement of appropriate therapy and the potential to prevent transmission of drug-resistant strains constitutes the major advantages of these methods. Recent studies demonstrated the feasibility of the MDRTBPlus assay as an effective tool for MDRTB screening in a high TB, and high MDRTB incidence, region and good concordance with phenotypic DST results [11, 18, 19]. However, rapid DST on clinical samples using molecular tools has (or potentially may have) a number of drawbacks, generally related to the low concentration of bacilli and possible presence of various types of Mycobacteria (eg sensitive and resistant ones) in the sputum specimen. The former issue, potentially leading to a problems with an assay sensitivity, has been addressed by incorporation of two stages (nested) PCR or increasing of number of PCR cycles (INNO-LiPA Rif.TB and GenoType® MTBDRplus respectively), but the sensitivity was still low in smear-negative culture-positive samples [11, 18, 28]. In addition, increase in assay sensitivity achieved by a large number of PCR cycles may lead to increased sensitivity to bacterial and/or amplicon contamination. The sensitivity of the assay may also be affected by a suboptimal selection of DNA probes and targets for the population studies because prevalence of mutations associated with RIF and INH resistance vary in different geographical regions [5, 29, 30].
In our collection of strains the spectrum of mutations associated with the resistance to RIF and INH (dominance of single mutations in the codon 531 of the rpoB gene and codon 315 of the katG gene) was similar or close to previously reported on larger populations in Samara and other regions of Russian Federation [20, 31–33]. The proportion of drug resistant strains in which no mutations were detected was low (3.8% and 2.7% for RIF and INH resistance) suggesting that the set of the DNA probes used in the GenoType® MTBDRplus assay covers most of the mutations prevailing in the Russian Federation. Previously reported associations between the above mutations and Beijing strains [20, 34] suggest that the assay may be potentially useful in other areas outside Russia with a high prevalence of Beijing family isolates (Eastern Europe, China and South-East Asia).
The proportion of interpretable GenoType® MTBDRplus assay results in our study, i.e. sensitivity of the assay for detection of TB bacilli (91.7%) was high but slightly lower compared to that reported previously (96.8% and 98.6% in  and  respectively), and, contrary to previous publications, lower readability rates were clearly associated with lower AFB grading in smear microscopy results. The performance of the assay on sputum samples with low concentrations of TB bacilli could probably be enhanced through the use of alternative methods of sputum treatment, potentially involving concentration of microorganisms prior to DNA extraction. An increased number of PCR cycles (40 cycles for sputum analysis compared to 30 cycles for cultures) does not resolve this problem completely and increases the sensitivity of the assay to amplicon contamination.
Overall sensitivity of the GenoType® MTBDRplus assay for detection of RIF, INH and multidrug resistance was high at 96.2%, 97.4%, and 97.1% respectively, which is similar to previously reported results in studies from South Africa, Germany, and Italy [11, 18, 19] supporting the use of this assay for MDRTB screening. Specificity and negative predictive values were 90.7%, 83.3%, and 88.9% for rifampicin, isolniazid, and MDR respectively suggesting that the molecular assay slightly overestimates drug resistance as defined by phenotypic DST on cultures derived from relevant sputum specimens. We believe that these discrepant results (10 disagreements in total), as well as "double patterns" in a proportion of strips (25.8%) could be explained by a "heteroresistance", i.e simultaneous presence of both drug resistant and sensitive TB bacilli in clinical samples.
We hypothesized that in these cases initial sputum samples contained mixtures of resistant and sensitive bacilli and, whilst mutant genotypes were recognized by the molecular assay (therefore masking sensitive genotypes) the sensitive bacilli might have grown preferentially on liquid media giving "sensitive" phenotypic DST results.
Heteroresistance, initially reported by Rinder et al. , is an important factor which can affect the accuracy and reliability of drug susceptibility testing using clinical specimens, because phenotypic results after isolation of pure cultures may not be representative of the initial mixture of Mycobacteria in the sputum. This usually underestimated phenomenon complicates interpretation of diagnostic assay results and may have been a reason for discordant results and "double patterns" (positive hybridization with mutant and wild type probes) on GenoType® MTBDR and GenoType® MTBDRplus membranes in our and recent studies [15, 18]. We assume that heteroresistance is more likely to occur in high TB incidence areas and in cultures isolated from chronic patients as they have more opportunity to become infected with various populations of Mycobacteria ; therefore "double patterns" were more common in our collection of samples obtained in high TB drug resistance region. Heteroresistance should be further investigated by molecular fingerprinting of clinical specimens (eg PCR-based VNTR typing) and detection of mutations in artificial "spiked" sputum specimens to establish mechanisms of preferential recognition of certain genotypes present in the specimen.