Population structure of drug-resistant Mycobacterium tuberculosis in Central Asia

Background Drug-resistant tuberculosis (TB) is a major public health concern threathing the success of TB control efforts, and this is particularily problematic in Central Asia. Here, we present the first analysis of the population structure of Mycobacterium tuberculosis complex isolates in the Central Asian republics Uzbekistan, Tajikistan, and Kyrgyzstan. Methods The study set consisted of 607 isolates with 235 from Uzbekistan, 206 from Tajikistan, and 166 from Kyrgyzstan. 24-loci MIRU-VNTR (Mycobacterial Interspersed Repetitive Units - Variable Number of Tandem Repeats) typing and spoligotyping were combined for genotyping. In addition, phenotypic drug suceptibility was performed. Results The population structure mainly comprises strains of the Beijing lineage (411/607). 349 of the 411 Beijing isolates formed clusters, compared to only 33 of the 196 isolates from other clades. Beijing 94–32 (n = 145) and 100–32 (n = 70) formed the largest clusters. Beijing isolates were more frequently multidrug-resistant, pre-extensively resistant (pre-XDR)- or XDR-TB than other genotypes. Conclusions Beijing clusters 94–32 and 100–32 are the dominant MTB genotypes in Central Asia. The relative size of 100–32 compared to previous studies in Kazakhstan and its unequal geographic distribution support the hypothesis of its more recent emergence in Central Asia. The data also demonstrate that clonal spread of resistant TB strains, particularly of the Beijing lineage, is a root of the so far uncontroled MDR-TB epidemic in Central Asia.


Background
Drug-resistant tuberculosis (TB) is a public health concern threathing the success of TB control efforts. This is particularily problematic in the Central Asian Republics (CARs) Uzbekistan, Tajikistan, and Kyrgyzstan where multidrug-resistant (MDR)-TB rates reach 24%, 22%, 27%, and 63%, 45% and 60% among new and previously treated cases, respectively [1].
Little is known about the population structure of drug resistant Myocbacterium tuberculosis (MTB) strains in Central Asia. Available reports are limited by either outdated genotyping techniques, sampling of special populations and/or small sample sizes [2][3][4]. Here, we provide a snapshot of the population structure of MTB strains based on World Health Organization resistance surveys performed in Kyrgyzstan, Tajikistan, and Uzbekistan representing two thirds of the Central Asian population.
For genotyping we combined 24-loci MIRU-VNTR typing (Mycobacterial Interspersed Repetitive Units -Variable Number of Tandem Repeats) with spoligotyping. MIRU-VNTR provides high-resolution discrimination of strains for epidemiological studies and phylogenetic classification [5]. MIRU-VNTR data allow also to generate parsimonious phylogenetic networks such as minimumspanning trees (MST) [6]. These results were combined with phenotypic drug susceptibility data to determine the presence of highly resistant dominant clones.

Methods
Seven hundred and seven clinical MTB isolates (306 from Uzbekistan, 216 from Tajikistan, and 185 from Kyrgyzstan) were collected in the framework of crosssectional national drug resistance surveys (DRS) [7,8] and included in the study. The study population was biased towards resistant strains since only isolates that required retesting for quality control purposes as per the WHO DRS protocol which was valid at the time were included [8]. This approach artificially increased the proportion of MDR-,pre-extensively drug-resistant (pre-XDR) and XDR-TB strains. Ethical approvals for the DRSs were granted by the respective national ethical committees.
DNA was extracted by heat lysis of bacteria from solid culture on LJ medium. 24-loci MIRU-VNTR genotyping and spoligotyping was performed as previously described [5,9]. Genotypes were identified on the MIRU-VNTRplus web database (www.miruvntrplus.org) [5]. Spoligotyping data were used to confirm strain relationships and for genotype classification. Typing data was analysed with Bionumerics (v7.5; Applied Maths, Sint-Martens-Latem, Belgium). MST analysis based on MIRU-VNTR data was performed using the categorical coefficient. For each 24-loci MIRU-VNTR pattern a unique multiple loci VNTR analysis (MLVA) MtbC15-9 haplotype was assigned by using the MIRU-VNTRplus nomenclature [10]. A cluster was defined as a minimum of two isolates harboring identical genotyping patterns from different patients. Patients with mixed molecular typing patterns were excluded. Drug susceptibility testing was performed by the resazurin microtiter assay [11].

Results
Twenty four-loci MIRU-VNTR typing and spoligotyping were successfully performed for 607 isolates, while 76 isolates (62 from Uzbekistan, seven from Tajikistan, and seven from Kyrgyzstan) yielded indeterminate results. For 23 isolates (9 from Uzbekistan, three from Tajikistan, and 11 from Kyrgyzstan), mixed genotyping patterns were observed. One isolate from Kyrgyzstan had incomplete phenotypic data.

Discussion
High prevalence of Beijing strains has previously been reported from Central Asia [4]. The star-like shape of the lineage 2/Beijing population in Fig. 1 with two central clones surrounded by layers of single and multi-locus variants is a typical pattern for emerging and expanding MTB populations [13]. The two largest clusters 94-32 and 100-32 were found in all three Central Asian countries and represented more than a third of all study isolates indicating strong population expansion across borders. Our findings confirm the results of a global Beijing population study which reported that 94-32 and 100-32 were predominantly prominent in Uzbekistan, Russia and Eastern European countries and had particularly high clustering rates [13]. The smaller size of 100-32 and its unequal distribution to the three CARs suggest however its more recent arrival in Central Asia and a shorter regional expansion time than of 94-32. This hypothesis is further supported by Mokrousov [14] and Skibe et al. [15] who have already reported 100-32 as minor group (< 4%) of Beijing isolates in Central Asia while it reached proportions of over 10% and over 20% in Eastern Europe and Russia, respectively, and while Skibe et al. observed 94-32 as dominant cluster in Kazakhstan. The extent of cross-border spread and locally restricted transmisison networks can however not be ultimately resolved with MIRU-VNTR cluster data and require further investigation with higher resolution.
High percentages of MDR among Beijing strains has also been reported from other Eastern European countries, like Georgia, Kazakhstan, and Uzbekistan [4,15,16]. Higher proportions of MDR-and XDR-TB were also observed within the 100-32 genotype by Merker and colleagues [13]. The authors analyzed the full genomes of MDR-TB strains belonging to MLVA-VNTR haplotypes 94-32 and 100-32 and showed that the maximal divergences were 17 and 23 SNPs, respectively, within genomes of those clusters [13].
Our  Fig. 2 Proportions of isolates based on drug resistance patterns. The number of isolates with a given genotype/MLVA MtbC15-9 haplotype is given above the corresponding column the ongoing MDR-TB epidemic in Central Asia. MIRU-VNTR is a powerful tool for investigating population structures and transmission dynamics, however, advances in the field of whole genome sequencing have resulted in molecular methods that can give better resolution and should ideally be used to elucidate specific transmission pathways. Analyzing the full genomes of this or an unbiased and thus even more representative study population could provide an even deeper insight into the MTB population structure in CAR.

Conclusions
Clusters 94-32 and 100-32 are the dominant genotypes driving population expansion of Beijing strains in the CARs Uzbekistan, Tajikistan, and Kyrgyzstan. The data demonstrate that spread of resistant TB strains, particularly of the Beijing lineage and its clusters 94-32 and 100-32, is an important root of the ongoing MDR-TB epidemic in Central Asia. Measures to combat this epidemic need to be improved.