High genetic diversity among Mycobacterium tuberculosis complex strains from Sierra Leone

Pathogens of the Mycobacterium tuberculosis complex (MTBC) such as M. tuberculosis and M. africanum are the causative agents of tuberculosis (TB) and remain one of the leading infectious killers worldwide. According to the latest WHO report, there have been approx. 8.8 million new TB-cases worldwide in year 2005 and 1.6 million people died of TB [1]. Furthermore, treatment and control of TB is increasingly complicated by the emergence of drug resistant or even multidrug resistant (MDR, resistance to at least isoniazid [INH] and rifampin [RMP]) strains [2]. Levels of MDR-TB among new patients have reached 14% in some areas of the world making successful treatment very difficult [2]. Among the regions with a high incidence of TB, Sub Saharan Africa is particularly affected with approximately 2 million new cases every year [1, 3]. Furthermore, the TB epidemic in Africa is reinforced by the mutual interaction of TB and HIV as HIV increases the risk of reactivation of latent TB as well as the rapid progression to active TB after infection [4].

Besides this dramatic situation in several Sub Saharan countries, data on the diversity of MTBC strains causing the epidemic in this area are only sparsely available. However, recent reports indicate that the genetic heterogeneity of MTBC strains is higher than previously anticipated and may influence host pathogen interaction, immunogenicity, transmissibility, development of drug resistance, and the performance of diagnostic tests [5–8]. Starting with the introduction of the first effective strain genotyping technique, IS6110 DNA fingerprinting [9], a number of further molecular methods such as spoligotyping [10], investigation of chromosomal deletions (regions of difference, RD), and most recently mycobacterial interspersed repetitive unit-variable number of tandem repeats (MIRU-VNTR) analysis have been established to analyze the diversity and population structure of clinical MTBC strains [6, 10–12]. These studies confirmed a phylogeographical population structure of the MTBC. In addition to the classical species M. tuberculosis, M. africanum, M. microti, M. caprae, M. bovis and M. prototuberculosis (including M. canetti), several well defined sub lineages were described within every species such as Haarlem or Beijing genotype among M. tuberculosis and West African-1 and West African-2 among M. africanum [6, 11, 13]. Based on the new tools for the efficient analysis of the MTBC population structure, an increasing number of investigations mounted evidence that the bacterial genomic background indeed can influence disease related characteristics [8, 14]. This might result e.g. in an enhanced capability to spread or an enhanced propensity to gain drug resistance when certain genotypes were considered and can influence diagnostic test such as the newly developed interferon-gamma-based T cell assays [5, 8, 14].

In contrast to Europe and USA, where the majority of human TB cases is caused by M. tuberculosis, a significant number of TB cases in West Africa has been reported to be caused by M. africanum strains which can be further subdivided into the two main lineages West African I and West African II [6, 15]. While some studies suggest a lower virulence and infection frequency of M. africanum which might result in a longitudinal extinction of this species [6, 16, 17], others still indicated higher levels of M. africanum infection prevalence and did not find significant differences in clinical presentation between M. africanum and M. tuberculosis [18, 19]. It appears to be a geographical gradient in the distribution of both M. africanum lineages with West African-1 dominating in Cameroon and West African-2 dominating in Gambia. However only a limited amount of data based on molecular classification of clinical isolates is available.

To further investigate this question, we have performed a retrospective study on the population structure of clinical MTBC strains obtained from patients registered for re-treatment in Sierra Leone in years 2003/2004 with a special focus on the prevalence of M. africanum. Sierra Leone is one of the high burden countries in West Africa with a TB incidence of 475/100.000 and TB mortality of 111/100.000 in year 2005 [1]. A study from 1993 reported a significant number of M. africanum infections in that region and was used to investigate the longitudinal trend of M. africanum prevalence [20]. In addition to molecular typing, drug susceptibility testing was carried out to monitor the level of MDR-TB among previously treated patients in the study area.

In the study period, MTBC cultures were grown from specimens of 103 patients registered for re-treatment between March 2003 and June 2004 in the Western Area and Kenema districts in Sierra Leone. Chromosomal DNA was successfully isolated from 97 strains which were used for further analysis. Six strains were not viable at the time when the DNA isolation was carried out. The main reason for registration for re-treatment among the 97 patient were relapse (n = 41) and treatment after interruption (n = 38). Twelve patients were treatment failure and for six no data are available.

To investigate the population structure of the 97 MTBC strains, molecular typing with IS6110 DNA fingerprinting, spoligotyping and analysis of 24 MIRU-VNTR loci was performed. For all strains, clear cut IS6110 DNA fingerprint and spoligotyping profiles were obtained. MIRU-VNTR typing was also successful for all strains with the exception that VNTR 2163b could not be amplified for one strain and an exceptional large fragment was obtained for VNTR 0577 for one isolate.

All genotyping data were digitized and analyzed with the Bionumerics software. A dendrogram based on the similarity of MIRU-VNTR typing data revealed a surprising high population diversity comprising several groups which can be mainly allocated to MTBC genotypes described previously (Fig. 1; for MIRU copy numbers and SpolDB4 types see Additional file 1). Genotype identification was carried out according to SpolDB4 [13] and by using the freely assessable MIRU-VNTRplus web database [23], which comprises genotyping data of a reference strain collection of validly described MTBC genotypes.

To further confirm the genotype classification, we used the MIRU-VNTR data to calculate a Minimum Spanning Tree (MST), which uses a maximum parsimony algorithm to investigate phylogenetic relationships and to identify clonal complexes within a population. The MST strikingly confirmed the genotype classification according to UPGMA tree based analysis (Fig. 1) and to comparison with the MIRU-VNTRplus web database [23]. All lineages suspected from dendrogram-based analysis were also detected as clonal complexes in the MST (Fig. 2) including the newly described genotypes Sierra Leone-1 and -2, thus confirming the existence of previously undefined phylogenetic lineages in Sierra Leone. The MST analysis also helped to assign strains with unclear classification to known genotypes as some of these were clearly grouped in the Haarlem complex and others to genotypes Sierra Leone-1 or -2 (Fig. 2). Overall, there is a very high concordance between genotype classification based on spoligotyping data according to SpolDB4 [13] and to that based on MIRU-VNTR data. However, MIRU-VNTR typing is clearly superior to spoligotyping as it allows a clear genotype identification also for strains of heterogeneous (ill defined) spoligotype clades such as the T-family [13]. Furthermore, compared to spoligotyping that is based on the analysis of one locus only it is more reliable to measure genetic distance since it based on 24 independent loci scattered around the genome [12].

To further investigate the deep phylogenetic branching we investigated sequence polymorphisms within the polymorphic pks1/15 locus. Interestingly, nearly 70% of all strains investigated (n = 66, Table 2) showed a 7 bp deletion that is typical for a major MTBC lineage so called Euro-American lineage according to Gagneux et al. [6]. The wild type pks1/15 sequence was obtained for strains of genotypes Beijing, EAI, and West African-1 while all West African-2 strains had a 6 bp deletion which has also been described before [24]. These data indicate that strains of the Euro-American lineages might be the major cause of TB in West Africa, however, with a large diversification in several well defined phylogenetic lineages as detected by MIRU-VNTR typing (Table 2).

If compared with other studies, it is striking that the M. tuberculosis Cameroon lineage, which represents 40% of TB cases in Cameroon [17] and 30% in Burkina Faso [25], was not prevalent in Sierra Leone where it represents only 4% of all strains investigated. This indicates a geographical influence on the distribution of MTBC lineages in West Africa. Possible reasons, e.g. differences in host-pathogen interaction [6], are not yet clear and represent an interesting field for future research.

Due to the limited amount of available patient information and the small study sample, we could not perform a deeper analysis of possible associations between disease characteristics and infection with certain genotypes. However, a striking observation made is that out of 4 Beijing strains three are MDR (Table 2). This is in accordance with higher rates of MDR among Beijing strains reported from several parts of the world especially Eastern Europe [14, 26]. Interestingly, the level of MDR was also unequally distributed among strains of other lineages (Table 2). However, due to the small numbers further studies are necessary to investigate this finding.

To analyze the overall discriminatory power of IS6110 DNA fingerprinting plus spoligotyping compared to 24 loci MIRU-VNTR typing plus spoligotyping in the study collection (spoligotyping was not considered alone since it is known that its discriminatory power is too low), we have determined the cluster rates obtained with the two different combinations. IS6110 DNA fingerprinting plus spoligotyping resulted in 79 distinct typing patterns and grouped 30 strains (31% of all) in 12 different clusters ranging in size from 2 to 5 strains (data not shown). The combination of MIRU-VNTR and spoligotyping had a slightly higher discriminatory power with 85 distinct types reducing the number of clusters to 10 with 22 strains (23% of all).

If the correlation between the two typing schemes is considered, there were a total of eight IS6110 DNA fingerprint/spoligotyping clusters that were split by MIRU-VNTR typing (data not shown) with confirmed changes in at least two loci making a real epidemiological linkage rather unlikely [27]. The highest number of MIRU-VNTR differences (12 loci out of 24 had a variation) occurred in a cluster with four EAI genotype strains with just one IS6110 band and an identical spoligotype. However, this is not surprising as in this case IS6110 DNA has no discriminatory power which can not sufficiently be increased by additional use of just spoligotyping in the case of EAI strains. Out of the 10 MIRU-VNTR/spoligotyping clusters, five were split up by differences in IS6110 DNA fingerprint patterns (one up to three bands, data not shown). The combination IS6110 DNA fingerprinting with MIRU-VNTR typing yielded the best discrimination of the strains analyzed resulting in 91 different types and only 6 clusters with 12 strains (13%).

These data suggest that the discriminatory power of both typing schemes is not optimal in this high incidence setting. This might be explained by our sampling procedure (short study period/not population based) which made it difficult to detect real transmission chains. However, it is noteworthy that the epidemiological significance of the discordant cases of clustering, especially for those involving slight changes in IS6110 DNA fingerprint patterns, remains unknown given the absence of epidemiological and contact tracing data for individual patients. Furthermore, the discriminatory power of molecular typing techniques can be significantly influenced by the composition of the MTBC population investigated. It has already been shown that IS6110 DNA fingerprinting is not discriminative in strains with lower band numbers as also observed for strains of the EAI genotype in this study [28]. In a diverse population of MTBC strains such as in Sierra Leone, this might result in a lower correlation between different typing methods if identical molecular typing patterns were considered.

However, the overall concordance between both typing systems is high if similarity rather than identity is considered. The close relationship of strains in IS6110 DNA fingerprinting/spoligotyping and MIRU-VNTR typing/spoligotyping cluster, respectively, is clearly supported by the fact that the MIRU-VNTR changes of strains in fingerprint clusters and changes of IS6110 DNA banding patterns of strains in MIRU-VNTR clusters are mainly small, respectively.

Our study confirmed a high level of drug resistance in patients registered for re-treatment in the Western Area and Kenema districts in Sierra Leone. This demonstrates the need for implementing routine drug susceptibility testing for all culture positive cases and improved public health measures to improve the control of the spread of drug resistant and MDR strains in Sierra Leone.

MIRU-VNTR typing proved to be an excellent tool for analyses of the population structure of MTBC strains. It allows the application of automated tools for genotype identification (such as the MIRU-VNTRplus database [23]) and is well suited for the identification of new phylogenetic lineages in a study population. Together with spoligotyping, its discriminatory power is superior to the combination of spoligotyping and IS6110 DNA fingerprinting, although the highest discriminatory power was achieved by the combination of all three methods.

The population structure of MTBC strains from Sierra Leone was found to be surprisingly divers with a total of 11 different genotypes/phylogenetic lineages. With 24% of all strains investigated, M. africanum still represent a significant number of cases which appears to be not lower compared to that reported from a study performed in years 1992/1993 in Sierra Leone [20]. However, the majority of strains belonged to the M. tuberculosis European American lineage [6], which can be further subdivided in seven sub lineages of which two have not been described before. The consequences of this intriguing high MTBC population diversity for the TB epidemic in this high incidence setting as well as for the application of new diagnostic tests or treatment strategies are not yet clear and need to be urgently addressed in further studies.