High resolution, on-line identification of strains from the Mycobacterium tuberculosis complex based on tandem repeat typing

The precise identification of bacterial pathogens at the strain level is essential for epidemiological purposes. Consequently, constant efforts are undertaken to develop easy to use, low cost and standardized methods which can eventually be applied routinely in a clinical laboratory. Newer developments are usually genetic methods based on PCR (Polymerase Chain Reaction) to type variations directly at the DNA level. The development of polymorphic markers is now further facilitated by the availability of whole genome sequences for bacterial genomes. Recently, it has been shown that tandem repeat (usually called minisatellites or VNTRs for Variable Number of Tandem Repeats) loci provide a source of very informative markers not only in humans where some are still in use for identification purposes (paternity analyses, forensics) but also in bacteria. Tandem repeats are easily identified from genome sequence data, the typing of tandem repeat length is relatively straight forward, and the resulting data can be easily coded and exchanged between laboratories independently of the technology used to measure PCR fragment sizes. Furthermore, the resolution of tandem repeats typing is cumulative, i.e. the inclusion of more markers in the typing assay can, when necessary, increase the identification resolution. However, the density of tandem repeats in bacterial genomes varies from species to species, and not all tandem repeats are polymorphic [1]. In addition, some tandem repeats are so unstable that they have no or little long-term epidemiological value [2]. This indicates that for each species under consideration, tandem repeats must be evaluated using representative collections of strains before they can be used. Tandem repeats for bacterial identification have already proved their utility for the typing of the highly monomorphic pathogens Bacillus anthracis, Yersinia pestis, [1] and M. tuberculosis. In this last case, the value of tandem repeat based identification was recognised very early [3]. The so-called DR (direct repeat) locus is a relatively large tandem repeat locus of unknown biological significance. The motif is 72 bp long, one half is highly conserved, whereas the other half (called the spacer element) is highly diverged. The spoligotyping method [4] takes advantage of these internal variations to distinguish the hundreds of different alleles at this locus, which have been reported in the M. tuberculosis complex among the thousands of strains typed so far [5]. Although it is quite powerful, with many advantages, spoligotyping suffers from a lack of resolution compared to the current gold-standard in M. tuberculosis genetic identification, IS6110 typing [6]. IS6110 typing is an RFLP (Restriction Fragment Length Polymorphism) method using the mobile element IS6110 as a probe. Strains with a low-copy number of IS6110 elements (such as most M. bovis strains) are poorly resolved by this method. The so-called PGRS (polymorphic GC-rich sequence) method is an other RFLP approach in which the probe used is a GC-rich tandem repeat. The polymorphisms which are scored at multiple loci simultaneously on the Southern blot are variations in the tandem repeats length (and not internal variations at a single locus as assayed by spoligotyping). The profiles generated are very informative, but in comparison with IS6110 typing, PGRS results are more difficult to score, because the intensity of the bands are highly variable (alleles with a small tandem array yield a lower hybridisation signal) [6]. Both PGRS and IS6110 typing are hindered by the requirement for relatively large amounts of high quality DNA which is an issue for slow-growing mycobacteria.

This collection of markers should already provide a typing resolution comparable to the current reference methods. Given that not all tandem repeats present in M. tuberculosis have been evaluated for polymorphism, it is likely that the typing resolution of minisatellites could further be improved. Eventually, normalisation work will have to be done in order to promote the use of tandem repeats. A number of the loci analysed are known under different names in different studies, (for instance, ETRD [7] is also known as MIRU4 in [10]; and VNTR 0580 in [11]) and the coding (number of motifs in an allele) of alleles can also be different in different studies, for reasons explained in [11]. This is due in part to the fact that the number of repeats is not necessarily an integer value (Table 1). Furthermore, because the repeats in an array are not necessarily exact repeats, there can be ambiguities in the definition of the first and last base pair of the array. Finally, in addition to length variations due to the addition or deletion of an exact number of units, microdeletions or insertions within some repeat units are sometimes observed (MIRU4 is one such instance [12]).

One purpose of the present report is to contribute to the development of Multiple Loci VNTR Analysis (MVLA) through the evaluation of new markers and the setting up of an on-line identification tool for the M. tuberculosis complex which can be queried very easily with the user's personal data. In the present report, we first take advantage of the availability of genome sequence from two M. tuberculosis strains to complement the current collection of polymorphic tandem repeat markers. We identified in silico tandem repeats showing a different length in the two strains using the previously described tandem repeat database http://minisatellites.u-psud.fr[1]. Thirteen loci with a different predicted length in the two genomes and which have not been previously investigated have been tested for polymorphism and ease of typing.

Eight among the 13 polymorphic loci were used together with 13 among the previously described markers to genotype a collection of different M. tuberculosis complex strains. The data produced clusters the strains as suggested by morphological observations and biochemical analyses. The resulting data can be queried from a dedicated web page http://bacterial-genotyping.igmors.u-psud.fr/bnserver.

The list of 40 markers given in Table 1 is close to representing the complete collection of tandem repeats of interest for MLVA typing in M. tuberculosis. It includes all loci with a different predicted size in H37Rv and CDC1551 and which are amenable to routine PCR typing. Nine additional loci which have been quoted in published reports are also included even if they do not fulfill this criteria. Clustering analysis (Figure 1) shows that the two strains CDC1551 and H37Rv (Figure 1) are relatively distant within the M. tuberculosis species. This would predict that tandem repeats of identical size in the two strains are likely to be poorly informative across the complex. However, this appears not to be absolutely true, since for instance, ETR-E (H37Rv_3192_53 bp) happens to have the same size in H37Rv, CDC1551 and even AF2122/97 (Table 1) in spite of its very high polymorphism index (0.69, Table 1). Consequently, the few additional loci, not explored here, which are of equal size in H37Rv and CDC1551, but differ with the predicted size for M. bovis strain AF2122/97 might also prove to be of interest.

As can be seen in Table 1, most repeat units are more than 50 bp long and allele sizes rarely exceed 1000 bp. As a result, the precision which can be achieved by ordinary agarose gel electrophoresis is sufficient to estimate the number of units in an allele. The selection of 21 markers proposed here was tested specifically in order to be easily assayed using this low-cost technological approach. Although a database system is necessary to efficiently manage a genotyping project with a high number of markers and strains, the identification of up to a few strains per day in a clinical setting for instance requires no sophisticated equipment nor costly consumables. Genotypes can be scored by visual analysis of the gel images, and a subset of the collection of available markers can be chosen for routine identification purposes. The data can then be analysed using the site described in Figure 2.

The role of tandem repeats in the M. tuberculosis genome is largely unknown. Twenty-one of the loci listed in Table 1 have repeat units which are a multiple of three base-pairs. The majority (fifteen) falls within putative genes, often of unknown function, such as the PPE family of genes [19]. The most remarkable instance is probably PPE34 at position 2163–2165 of the genome (Rv1917c in http://genolist.pasteur.fr/TubercuList/) which contains three minisatellites [20] (Table 1, Qub11a, Qub11b, ETR-A).

The present study includes 17 M. africanum strains. All strains have been identified as such independently, based on morphological features of the colonies grown on Lowenstein-Jensen medium, and biochemical analyses. M. africanum has long since been recognized as showing an extensive phenotypic heterogeneity [21], suggesting that M. africanum could display a phenotypic continuum between M. tuberculosis and M. bovis. This was recently supported by the study of deletion events distinguishing the H37Rv M. tuberculosis strain and the BCG M. bovis strain [22] and suggesting that M. bovis is the most recent member of the M. tuberculosis complex. The analysis of deletion events in the M. africanum strains investigated showed that West African strains fall into two groups, clearly distinguished from the M. tuberculosis strains. In contrast, no deletion event distinguished East African M. africanum strains from M. tuberculosis strains. The present study includes three Africanum type II strains (positive nitrate reductase test). All three originate from East Africa (Djibouti). Although the MLVA analysis presented here does confirm that they are very close to M. tuberculosis strains, they are clearly distinct, at least within the collection of strains evaluated. Interestingly, they appear to be closest to the Beijing type of M. tuberculosis strains (Figure 1, Group III, strains percy7, percy27 and percy91).

In its present form, the database should be considered as preliminary. More strains must be typed in order to provide a continuous and robust coverage of the M. tuberculosis complex, and the clustering analysis presented in Figure 1 should be considered as provisional. If the MLVA approach is considered to be of use by the community, and given that the associated data is highly portable, then it should be relatively easy, through collaborative efforts, to significantly expand the available data. It is hoped that this data will constitute an easy-to-use high-resolution classification resource which will then help address medical and epidemiological issues regarding the M. tuberculosis complex.