INTRODUCTION

Top Abstract Introduction Materials and Methods Results and Discussion References

Genetic fingerprinting of clinical isolates of Mycobacterium tuberculosis using the consensus IS6110 restriction fragment length polymorphism (RFLP) method (19), is of great value in studying the epidemiology of tuberculosis (14). However, systematic fingerprinting of all bacterial isolates by IS6110 RFLP analysis remains cumbersome in large epidemiological studies, particularly in developing countries. Previously, a PCR-based spoligotyping method for diagnosis and epidemiology of tuberculosis was proposed as an alternative to hybridization-based fingerprinting methods (10). However, as spoligotyping used alone overestimates the number of epidemiological links, it was suggested that it should be used in association with another rapid fingerprinting technique (7). Consequently, spoligotyping was used in association with double-repetitive-element PCR (DRE-PCR) (4), and was found to be a suitable alternative to IS6110 RFLP analysis (16).

Recently, thanks to the full genome sequencing of M. tuberculosis H37Rv (2), a total of 11 loci containing variable numbers of tandem DNA repeats (VNTR) were localized, out of which 5 comprised multiple polymorphic tandem repeats (MPTR) with substantial sequence variation and 6 represented exact tandem repeats (ETRs) (i.e., ETR-A to -F) that contained multiple alleles (5). Although VNTR fingerprinting to study the polymorphism of M. tuberculosis was initially attempted using all 11 VNTR loci (5), a recent multicenter study suggested that only 5 VNTR loci (ETR-A to -E) are sufficiently discriminant to be retained for further investigations (11). Out of these five VNTR loci, ETR-A was initially identified as the promoter region of the katG gene (8), and ETR-D and ETR-E were found to be identical to the Mycobacterial interspersed repetitive units (MIRUs) described recently (18; P. Supply, personal communication). Each ETR locus has multiple alleles, and recently a combined analysis identified 22 distinct allele profiles in 25 wild-type strains of the M. tuberculosis complex (5). These allele profiles were reproducible and stable, as demonstrated by analysis of multiple isolates of reference strains obtained from different laboratories, and consequently VNTR fingerprinting was proposed as a rapid typing method for strain differentiation and evolutionary studies of mycobacteria (5). Following this first study, VNTR typing was recently used to characterize clinical isolates of Mycobacterium africanum from West Africa (6). Because of very limited studies relating to VNTR fingerprinting (5, 6, 11) and the fact that its potential as a second-line test in association with spoligotyping has not yet been evaluated for epidemiological studies of tuberculosis, it was desirable to compare the discriminatory power of VNTR typing used alone and in association with spoligotyping in an insular geographic model and to compare the results obtained with those obtained by spoligotyping and DRE-PCR.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results and Discussion References

All the clinical isolates of M. tuberculosis used in this study were obtained between January 1998 and June 1999 from clinical specimens of patients residing in the French Caribbeans. The isolates were cultured and identified locally at the mycobacterial reference laboratory of the Pasteur Institute of Guadeloupe using classical mycobacteriological procedures (3) and were grown as fresh Löwenstein-Jensen slants at 37°C prior to experiments. The DNAs were prepared using the cetyltrimethylammonium bromide method as described previously (20). The quality and quantity of DNAs were checked both by agarose gels and UV spectrophotometry. DNAs were stored in TE (10 mM Tris-1 mM EDTA, pH 7.4) buffer at 4°C. IS6110 fingerprinting was performed using the internationally agreed-on methodology (19). Labeling and detection were performed using direct ECL kits (Amersham, Little Chalfont Buckinghamshire, United Kingdom). Gel-Compar (Applied Maths, Kortrijk, Belgium) and/or Taxotron (Institut Pasteur, Paris, France) softwares were used to calculate the molecular weights of hybridizing bands and to compare the isolates. Strains were defined as being clustered when patterns were identical or differed by one band. The results obtained were always checked by visual inspection of the gels.

The DRE-PCR was essentially performed according to a previously published protocol (4), modified recently to increase the number of bands and the discriminative power of the DRE-PCR technique (12) by adding 6% dimethyl sulfoxide in each reaction tube and an extension step of 3 min instead of 1 min. Spoligotyping was performed as previously described using homemade membranes (10). The results were documented under the form of a binary code according to the results of hybridization (positive or negative result) for each spacer oligonucleotide probe (n = 43), and entered in an Excel spreadsheet file or into a Recognizer file (Taxotron, Institut Pasteur, Paris) for phylogenetic reconstruction.

VNTR typing was essentially performed as described previously (5), with slight modifications. Briefly, PCR was performed in a total volume of 60 µl containing 6 µl of 10× recombinant Taq buffer (AP-Biotech, Uppsala, Sweden), 2 mM MgCl₂, an 80 nM concentration of each primer, a 500 µM concentration of each of the four desoxynucleoside triphosphates, 6 µl of dimethyl sulfoxide, 1 U of recombinant Taq (AP-Biotech), and 50 to 200 ng of DNA sample. Two drops of paraffin oil was added to each tube. An initial denaturation of 7 min at 94°C was followed by 35 cycles of denaturation at 94°C for 30 s, annealing at 60°C for 1 min, and extension at 72°C for 2 min, and this was followed by a final extension step at 72°C for 10 min. An aliquot (15 to 25 µl) from the reaction tubes was run on a 3% Metaphor gel (FMC Bioproducts, Rockland, Maine). Molecular weight standards (100-bp ladder or PhiX-HaeIII; AP-Biotech) were run every 4 to 5 lanes. The molecular weight determination of PCR fragments was performed using the Taxotron software on images digitized using the Video-Copy system (Bioprobe, Montreuil, France). Once the length of the PCR fragments was precisely calculated, the number of copies for each ETR was deduced according to a previously published scheme (5, 11) and documented as a five-digit number representing allele profiles for ETR-A to ETR-E.

The discriminatory power of each of the typing methods was calculated using the Hunter-Gaston index (HGI) as previously reported (9).

	RESULTS AND DISCUSSION

Top Abstract Introduction Materials and Methods Results and Discussion References

A collection of 66 individual patient isolates of M. tuberculosis, isolated between January 1998 and June 1999, was typed by spoligotyping, DRE-PCR, and VNTR fingerprinting. Using VNTR typing alone, a total of 48 (73%) isolates were grouped in 12 clusters (A to L) comprising 2 to 24 strains, as compared to a total of 50 (75%) isolates grouped in 18 spoligotyping-defined clusters (Fig. 1 and Table 1). Although the detailed results for DRE-PCR alone are not illustrated in Table 1, this method grouped a total of 25 (38%) isolates in 11 clusters. These observations underlined that contrary to the initial study (5), VNTR typing was poorly discriminant as a first-line fingerprinting method in our setting; e.g., the VNTR-defined cluster A (allele designation, 32333) accounted for one-third (n = 24) of the isolates in our study. Cluster A could be further subdivided into 10 distinct subtypes containing one to six isolates upon spoligotyping (Table 1). However, one interesting observation was the fact that VNTR typing used as a second-line test was able to further subdivide the two ubiquitous spoligotypes, namely, patterns 50 and 53 (15), into two and four VNTR-defined subtypes. Last but not least, VNTR typing used as a second-line test gave results comparable with DRE-PCR, except for three clusters (spoligotype profile numbers 14, 17, and 20) comprising seven isolates, that could be further subdivided using DRE-PCR but not VNTR typing. However, this discrimination was based only on a single additional band revealed upon DRE-PCR, probably indicating a phylogenetical relatedness of these subtypes.

View larger version (30K): [in this window] [in a new window]   FIG. 1.   Dendrogram constructed by the unweighted pair group method using arithmetic averages and similarity index of 66 clinical isolates fingerprinted by the VNTR typing method. The VNTR results were entered into a Recognizer file of the Taxotron software, and the 1 Jaccard index was calculated for each pairwise comparison of strains. Letters A to L highlight individual clusters. The five-digit numbers in column I represent strain designations. The five-digit numbers in column II represent VNTR allele designations. The numbers in column III are spoligotypes patterns (1 to 137). Spoligotype patterns 1 to 69 have been recently published (15), whereas patterns 70 to 137 are available from the authors upon request.

A calculation of the discriminative index of each of the methods used alone or in combination was performed using the HGI (9), and the results obtained are summarized in Table 2. When used alone, VNTR was the least-discriminatory method (HGI = 0.863). Although VNTR was significantly more discriminatory when used in association with spoligotyping (HGI = 0.982), the combination of spoligotyping and DRE-PCR (HGI = 0.992) was still the most efficient among rapid, PCR-based methodologies, giving results comparable to IS6110 RFLP analysis. Indeed, it is generally admitted that a typing system must have an index of at least 90% to be considered as an efficient test for epidemiology (9), and consequently, VNTR typing alone did not fulfill these requirements in our setting.

Out of the collection of 66 isolates studied by the three PCR-based methods in this study, 27 (41%) isolates were randomly selected for confirmatory IS6110 RFLP fingerprinting, as this latter remains a "gold standard" in the molecular epidemiology of tuberculosis (11, 19, 20). A total of 19 distinct profiles were obtained by IS6110 RFLP analysis, and 13 out of 27 (48%) isolates were grouped in five clusters containing two to four isolates (Fig. 2). As shown in Fig. 2, a perfect correlation was found between IS6110 RFLP analysis and PCR-based methodologies for isolates containing more than five copies of IS6110; however, as expected, IS6110 RFLP analysis was less discriminatory for M. tuberculosis isolates containing fewer than five copies of the IS6110 element (11). For the isolates containing five or fewer copies of IS6110 (Fig. 2 [two isolates with spoligotype 91, and 3 isolates with spoligotype 14]), DRE-PCR was more discriminant than IS6110 RFLP analysis for spoligotype 14 isolates, whereas VNTR typing was more discriminant for spoligotype 91 isolates (Table 1; Fig. 2). For the three isolates of spoligotype 14, the DRE-PCR based subdivision was independently confirmed by available polymorphic GC-rich sequence (PGRS)-RFLP data (results not shown). It may therefore be concluded that both spoligotyping with DRE-PCR and spoligotyping with VNTR typing allow the definition of epidemiologically relevant clusters comparable to IS6110 RFLP analysis, a fact also corroborated by relative HGIs (Table 2).

View larger version (65K): [in this window] [in a new window]   FIG. 2.   Dendrogram constructed by the unweighted pair group method using arithmetic averages and the corresponding lanes for 27 clinical isolates of M. tuberculosis typed by IS6110 RFLP analysis. The results were analyzed using the Gel-Compar software. The strains were also typed in parallel using spoligotyping, VNTR typing, and DRE-PCR (results not shown). (A) Spoligotyping patterns; (B) strain designation; (C) VNTR allele designation.

A recent multicenter study showed that the fingerprinting of M. tuberculosis isolates by IS6110 RFLP analysis led to a clustering similar to that obtained by other genetic markers such as VNTR or those based on the polymorphism of the direct repeat locus, such as spoligotyping (11). This strong mutual association of unlinked genetic markers suggests a clonal population structure of circulating M. tuberculosis strains (11). This study (11) further showed that both spoligotyping and VNTR fingerprinting were reproducible methods (respectively, 94 and 97% reproducibility), whereas DRE-PCR was poorly reproducible (58%). As the authors pointed out (11), this might have been due to some events of arbitrary priming in performing the DRE-PCR technique. In our hands, DRE-PCR and VNTR typing have always been performed on N-cetyl-N,N,N-trimethyl ammonium bromide-purified DNA with fixed DNA concentrations within a range of 50 to 200 ng, which certainly helps increase the reproducibility of DRE-PCR. Furthermore, a computer-assisted interpretation of the bands further facilitates the comparison of isolates. Although VNTR analysis, which requires 5 PCRs per sample, remains more cumbersome to perform than DRE-PCR, the results obtained are much more easily interpreted due to their numerical format. However, one should be extremely careful to avoid potential sources of error when analyzing the VNTR results based on five different loci. A future robotization of PCR steps and automatic sizing of PCR products on sequencing gels using fluorescent primers may facilitate the use of VNTR fingerprinting as a routine rapid typing method, eventually as an adjunct to spoligotyping or other PCR-based methods such as ligation-mediated PCR, which was recently proposed as a first choice alternative to spoligotyping (1, 13).

Concerning evolutionary genomics, M. tuberculosis is remarkably homogeneous at the molecular level (17), and transposition, homologous recombination, and replication slippage are the driving forces of M. tuberculosis genome evolution (11). Although the relative speed of VNTR typing as a molecular clock versus IS6110 RFLP and direct repeat locus analysis remains to be studied, we also used the VNTR results of this investigation in parallel for a phylogenetical analysis (results not shown). The trees obtained were compared to those generated using spoligotyping, as shown recently (15), and this preliminary analysis showed a very good congruence between the trees generated using these two distinct PCR-based method, suggesting that VNTR typing may also be very useful for the construction of detailed evolutionary trees of M. tuberculosis complex. A detailed evaluation of the phylogenetic potential of VNTR typing for studying M. tuberculosis genome evolution is currently under investigation by analyzing M. tuberculosis profiles versus M. africanum and Mycobacterium bovis profiles.