|Home | About | Journals | Submit | Contact Us | Français|
We describe an improved multiple-locus variable-number tandem-repeat (VNTR) analysis (MLVA) scheme for genotyping Staphylococcus aureus. We compare its performance to those of multilocus sequence typing (MLST) and spa typing in a survey of 309 strains. This collection includes 87 epidemic methicillin-resistant S. aureus (MRSA) strains of the Harmony collection, 75 clinical strains representing the major MLST clonal complexes (CCs) (50 methicillin-sensitive S. aureus [MSSA] and 25 MRSA), 135 nasal carriage strains (133 MSSA and 2 MRSA), and 13 published S. aureus genome sequences. The results show excellent concordance between the techniques' results and demonstrate that the discriminatory power of MLVA is higher than those of both MLST and spa typing. Two hundred forty-two genotypes are discriminated with 14 VNTR loci (diversity index, 0.9965; 95% confidence interval, 0.9947 to 0.9984). Using a cutoff value of 45%, 21 clusters are observed, corresponding to the CCs previously defined by MLST. The variability of the different tandem repeats allows epidemiological studies, as well as follow-up of the evolution of CCs and the identification of potential ancestors. The 14 loci can conveniently be analyzed in two steps, based upon a first-line simplified assay comprising a subset of 10 loci (panel 1) and a second subset of 4 loci (panel 2) that provides higher resolution when needed. In conclusion, the MLVA scheme proposed here, in combination with available on-line genotyping databases (including http://mlva.u-psud.fr/), multiplexing, and automatic sizing, can provide a basis for almost-real-time large-scale population monitoring of S. aureus.
Staphylococcus aureus is a pathogen of worldwide clinical significance. For this reason, it is the subject of intensive investigations in terms of virulence and drug resistance phenotypes and, also, population genetics. Although the latter is not of significant use for short-term patient care, it is essential for understanding the emergence and spread of new phenotypes. For instance, it was initially considered most likely that methicillin (meticillin)-resistant variants were appearing only rarely through the acquisition of a mobile DNA region designated staphylococcal cassette chromosome mec (SCCmec) and that these variants were spreading efficiently worldwide (34). However, the most recent population genetics investigations suggested instead that SCCmec was acquired hundreds of times independently worldwide and that, as a rule, the geographic spread of these resistant strains was limited (28, 36). This knowledge could be produced in the past 10 years due to sequence-based approaches, mainly multilocus sequence typing (MLST) analysis, in which approximately 3 kb of coding genome sequence (or 1/1,000 of the whole genome) are scanned for polymorphism. MLST has allowed the creation of shared and high-quality databases which can be easily queried over the Internet, and this has proved to be highly valuable (7). However, as the sequences used in MLST schemes evolve slowly and are highly conserved, the resolution provided by MLST is too low for the investigation of recent evolution and, above all, for short-term epidemiological studies. The sequencing of much larger portions of the genome to increase resolution can only be used in dedicated research projects analyzing a limited number of strains (28). Presently, pulsed-field gel electrophoresis (PFGE) remains the most discriminatory technique for S. aureus typing, but it allows the constitution of shared databases only at the national level and is not appropriate for population studies (1, 35). There is, consequently, still a need for a technology as discriminatory as PFGE and as portable as MLST at a low cost.
Tandemly repeated sequences provide a very valuable source of polymorphism. Multilocus variable-number tandem-repeat (VNTR) analysis (MLVA) is now used in genotyping several bacterial species (26, 48). MLVA typing relies upon a basic and widespread methodology, the measurement of the length of DNA fragments. It is not a “pattern”-producing method, even when run on agarose gels. The genotype, in the form of a string of numbers corresponding to the number of repeats at each locus, is highly portable and can be readily incorporated in large databases (13).
VNTRs were proposed years ago to genotype S. aureus isolates, first with a tandem repeat (TR) adjacent to the coagulase gene coa (12) and later with a single TR present in the S. aureus protein A (spa) gene (9). Subsequently, new TRs present in individual genes were analyzed simultaneously to produce patterns, or “fingerprints” (8, 37). A second level of TR-associated polymorphisms due to repeat unit variations was exploited by sequencing TR alleles. The spa gene, providing a high level of information, is most frequently used. It allows the relatively correct assignment of isolates to MLST-defined clonal complexes (CCs) (17, 22, 41), with some occasional exceptions. The resulting data have a limited phylogenetic value (28). In order to increase the discriminatory power and phylogenetic content, an assay called double-locus sequence typing (DLST) in which a second TR locus (clfB) is included has recently been developed (23).
MLVA stricto sensu, in which a repeat copy number is deduced for each locus, was first applied to S. aureus genotyping by Hardy et al., using seven members of a class of repeated elements called staphylococcal interspersed repeat units (SIRUs) (15, 16). Eleven additional TRs were later identified, and different combinations were used to improve the assay (11). Ikawaty et al. recently described an MLVA scheme with six SIRUs (20) that showed a higher discriminatory power than MLST and spa typing. However, the clustering only identified three large clusters of MLVA types and the correlation with MLST CCs was partial. In a recent work, Schouls et al. confirmed that MLVA with as few as eight VNTRs provided clustering similar to that of spa typing and PFGE, but they did not demonstrate agreement between MLST complexes and MLVA complexes with their assay (40).
Although MLVA appears to have the potential to provide a technique for short-term epidemiological studies that is fast and reliable in comparison to other techniques, there is still no consensus on the set of VNTR markers to be used for an efficient genotyping protocol and, more generally, for the potential use of MLVA for S. aureus typing. Different TRs vary at different rates, and homoplasy levels at individual VNTRs may be high. Due to intraspecies genetic variability in S. aureus, some primer combinations fail to amplify a significant fraction of the strains. The loci and primers to be used need to be carefully selected. In the present study, we have investigated new VNTRs, as well as previously described ones, and we propose an MLVA genotyping scheme, the MLVA-14 assay, made up of two complementary panels totaling 14 markers, which provides an easy and highly informative genotyping assay with a strong phylogenetic content. For this purpose, we analyzed three diverse S. aureus strain collections which were previously characterized in detail with both MLST and spa typing.
Eighty-seven strains from the Harmony collection (1) were provided by Alex van Belkum. Twenty-five methicillin-resistant S. aureus (MRSA) carriage strains were isolated in the Maastricht University Medical Center (MUMC) in The Netherlands between 2002 and 2006 and comprise five strains each from MLST CC5, CC8, CC22, CC30, and CC45 (29). Fifty methicillin-sensitive S. aureus (MSSA) nasal carriage strains were isolated from patients attending their general practitioner in The Netherlands during 2005 and comprise five strains each from MLST CC5, CC7, CC8, CC12, CC15, CC22, CC25, CC30, CC45, and CC51 (6). These 75 strains were previously spa typed (30).
One hundred thirty-five nasal carriage strains, two of which were MRSA, were isolated from newly employed hospital personnel during their first medical checkup at a tertiary care hospital in Lausanne, Switzerland, and have been genotyped using MLST, spa typing, amplified fragment-length polymorphism, and DLST (39). Reference strain Mu50 was purchased from the Centre de Ressources Biologiques de l'Institut Pasteur (CRBIP). The Ridom nomenclature was used to describe the organization of the spa repeats (17).
Oligonucleotide primers targeting the 5′ and 3′ flanking regions of the selected loci and matching the sequenced genomes of strains COL, MRSA476, MW2, N315, NCTC8325, JH1, JH9, Newman, USA300 (FPR3757), and USA300 (TCH1516) were used for amplification. Some mismatches existed with the genomes of strains MRSA252 and RF122. DNA was extracted by using a DNeasy tissue kit (Qiagen) after treatment of bacteria with lysostaphin (Sigma) at a concentration of 1 mg/ml. PCRs were performed in 15-μl volumes containing 2 ng DNA, 1× PCR buffer, 1.5 mM MgCl2, 1 U Taq DNA polymerase (Qiagen, Courtaboeuf, France), 200 μM of each deoxynucleoside triphosphate, and 0.3 μM of each flanking primer (Eurogentec, Angers, France). Amplification was performed with a PTC 200 thermocycler (Bio-Rad, Marnes-la-Coquette, France) using the following conditions: initial denaturation cycle for 5 min at 94°C, 35 cycles of denaturation for 30 s at 94°C, annealing for 30 s at various temperatures (indicated in Table Table1),1), and elongation for 45 s at 72°C plus a final elongation step for 10 min at 72°C. Three microliters of PCR products was separated in a 2% agarose gel (Eurogentec, Angers, France). Electrophoresis was performed in 20-cm-long gels made in 0.5× Tris-borate-EDTA buffer (Sigma), run at 8 V/cm. In each run, the PCR product from reference strain Mu50 was included at least once. The 100-bp DNA size marker was from MBI Fermentas (Euromedex, Souffelweyersheim, France). The gels were stained after the run in 0.5 μg/ml ethidium bromide for 15 to 30 min and then rinsed with water and photographed under UV illumination (Vilber-Lourmat, Marne la Vallée, France). To prevent carryover contamination, the different steps of the procedure were performed in separate rooms with dedicated materials.
In the first phase of the study, the size of the amplicons was measured with the assistance of BioNumerics 5.1 (Applied-Maths, Sint-Martens-Latem, Belgium), and the number of repeats was deduced using the Mu50 genome sequence as a reference. Thereafter, the size of the amplicons was directly estimated by eye before import and conversion into a character data set in BioNumerics.
The repeat lengths and numbers of repeat units in the different sequenced genomes were determined by using the Microorganisms Tandem Repeats Database (http://minisatellites.u-psud.fr) (4, 13, 25). For each locus, the size of the PCR product, S, in the first sequenced genome (as predicted by in silico analysis of reference strain Mu50 with the primer pair used here), the size of the repeat unit, U, and the corresponding number of repeat units, N, are indicated in Table Table1.1. Amplification of DNA from reference strain Mu50 produced amplicons of the expected size. The exact copy number for each allele was calculated as follows: S was subtracted from the estimated allele size, and the result was divided by U, added to N, and rounded up to the nearest integer that was distant by less than 0.2. Alleles which could not be rounded up following this rule were double-checked and eventually sequenced to confirm the existence of intermediate alleles (0.5) and to establish the reason for this intermediate size (which may result from small deletions in the flanking sequence).
The polymorphism index of individual or combined VNTR loci was calculated using the Hunter-Gaston diversity index (19), an application of Simpson's index of diversity (43). Confidence intervals (CI) were calculated as described by Grundmann et al. (14). The results of using the MLVA-14 assay, with 14 VNTRs in two panels, to genotype a strain are expressed as its allelic profile, corresponding to the number of repeats at each VNTR in the order Sa0122 (spa), Sa0266 (coa), Sa0311, Sa0704, Sa1132, Sa1194, Sa1291 (SIRU13), Sa1729, Sa1866, and Sa2039 (panel 1) and Sa0906, Sa1213, Sa1425, and Sa1756 (SIRU15) (panel 2) (Table (Table1).1). The genotype of Mu50, deduced from its genomic sequence, is 10 6 3 4 6 7 4 5 3 3 (panel 1) 3 5 4 2 (panel 2). The categorical coefficient (also called Hamming's distance) and the unweighted pair group method with arithmetic mean clustering method were run within BioNumerics. A cutoff value of 45% similarity was applied to define clusters. It corresponds to differences at a maximum of 3 VNTRs out of 14. This is still empirical since there is no precise knowledge about the VNTR evolutionary mechanism and speed, but it seems to correctly define clusters when compared to those defined by other genotyping methods. The minimum spanning tree was produced in BioNumerics, allowing the creation of missing links. The circle size is proportional to the number of isolates. A logarithmic scale was used when drawing branches.
The MLVA profiles and allele size ranges are available for comparison in the MLVAbank for Bacterial Genotyping (http://mlva.u-psud.fr/ [Staphylococcus aureus database]).
In a preliminary study performed in 2003, 14 VNTRs were selected by comparing the available sequenced genomes of strains Mu50, N315, MRSA252, NCTC8325, MW2, and MSSA476 using the strain comparison tool developed by Denoeud and Vergnaud (4) and available at http://minisatellites.u-psud.fr/ (Microorganism Tandem Repeats database and Strain Comparison pages) (Table (Table1).1). They were tested on a collection from the Pasteur Institute given to us by Nevine El Sohl and previously typed by PFGE. These initial results showed that MLVA could efficiently cluster strains with similar pulsotypes (31). We then investigated the informativeness and potential use of additional VNTRs described in other studies. Most of the published primers did not perfectly match the genome of the strains analyzed or they were predicted by in silico analysis of thirteen sequenced genomes to amplify more than one locus. This is the case, for instance, for the sdr locus in which the three highly informative TRs present in genes sdrC, sdrD, and sdrE could not be amplified independently. In addition to the spa locus (SIRU21), only SIRU01, SIRU13, and SIRU15 (15) and SAV920 and SAV1078 (10) were retained and tested on a larger collection of strains (data not shown). For SAV920 and SIRU01, lack of amplification was observed in about 10% of strains, and therefore, these markers were not kept for a first-line MLVA scheme. The selected set of loci comprises 14 VNTRs that are present in all of the sequenced genomes, 6 of which correspond to S. aureus repeat (STAR) elements (Sa0311, Sa0906, Sa1213, Sa1425, Sa1729, and Sa1866), a family of intergenic elements whose copy numbers vary from 13 to 21 in individual strains (3). The VNTR set contains four additional intergenic TRs (Sa1756 [SIRU15], Sa0704, Sa1194, and Sa1291 [SIRU13]) and four TRs located inside the coding regions of the spa, coa, SAV1078, and SAV1738 genes (corresponding to Sa0122, Sa0266, Sa1132, and Sa1866, respectively). The MLVA scheme was run with individual PCRs and agarose gel electrophoresis of amplicons in this study, as shown in Fig. Fig.11 for a subset of VNTRs. The size of the amplicons can be easily estimated by eye on agarose gels. For markers Sa0906 and Sa1213, the absence of amplification was frequently observed when using primers localized about 20 bp on each side of the TRs. Therefore, primers inside the two genes flanking the STAR element were selected, allowing correct amplification in all the strains. A large number of alleles could be obtained for Sa0906, reflecting the complexity of the locus as confirmed by examination of the sequence (see “The structure of Sa0906” below).
This MLVA assay is perfectly reproducible, as attested by the repeated use of Mu50 DNA as control, which always gave the same result. In addition, the stability of the VNTRs in cultured bacteria is demonstrated by the use of different batches of Mu50 DNA and by the fact that the observed amplicon sizes are identical to those predicted by the genome sequence.
To assess the informativeness of MLVA compared to that of MLST and spa typing, we genotyped three complementary and well-referenced strain collections previously analyzed by these two techniques and belonging to the major MLST CCs. Overall, the efficiency of PCR amplification was excellent. Only in three instances could no amplification be obtained, which may be due to the absence or the mutation of the target of one of the primers: Sa0311 failed to amplify two MLST sequence type 398 (ST398) samples from nasal carriage, and Sa1291 failed to amplify the ST8 sample NL33. The data were used to perform a clustering analysis based upon the categorical distance coefficient and unweighted pair group method with arithmetic mean clustering method. With a cutoff value of 45%, 21 clusters were defined, two of which corresponded to a single isolate (2 strains of nasal carriage, Laus167 of ST50 and Laus325 of ST78) (see dendrogram in Fig. S1 in the supplemental material). These clusters correspond to CCs defined by MLST and by spa typing together with the algorithm Based Upon Repeat Pattern. Figure Figure22 shows a minimum spanning tree which produces a more condensed representation of the clustering and suggests relationships between the clusters. An almost perfect correlation between the results of the different techniques was observed, with the exception of a few isolates which did not cluster by MLVA with strains of the same ST or spa type. Strain NL33, spa typed as t701 that is associated with CC8, is grouped with CC7, and Laus356, ST45 and spa type t1081, is grouped with CC12, both with long branches. In general, a much higher diversity was found with MLVA than with MLST, particularly inside the major CCs, due to polymorphism at one or several VNTRs (Fig. (Fig.3).3). For example, the groups of 19 ST8 and t008 isolates, 16 ST5 and t002 isolates, and 18 ST30 and t012 isolates were each resolved into 12 MLVA genotypes. Thirteen ST45 t015 isolates were resolved into seven MLVA genotypes (44). The diversity index of the MLVA-14 assay is 0.9965 (95% CI, 0.9947 to 0.9984) (243 genotypes). In comparison, the diversity index of the MLST assay with the same isolates (52 different STs) is 0.9314 (95% CI, 0.9198 to 0.9431) and the diversity of the spa assay (127 spa types) is 0.9802 (95% CI, 0.9755 to 0.9849).
In order to test whether the complete set of 14 VNTRs was necessary to get a good resolution, clustering was performed with different combinations of markers, and we found that very satisfying clustering was already obtained with 10 VNTRs (panel 1), excluding those occasionally showing very small alleles (Sa0906, Sa1123, Sa1425, and Sa1756). The only inconsistency observed with panel 1 is the clustering of an ST121 (CC51) isolate with CC45 strains. Therefore, we propose to use panel 1 to assign the isolates to a cluster and four additional markers (panel 2) for more informativeness. Panel 1 discriminates 215 genotypes with a diversity index of 0.9946 (95% CI, 0.9925 to 0.9967).
As explained above, VNTR Sa0906 is a TR present in a STAR element, which shows a complex organization requiring the use of primers in the flanking coding regions for efficient PCR amplification (SAV0834 and SAV0835) (Fig. (Fig.4).4). A large number of alleles was observed, and in order to better assess their nature, these alleles were sequenced. The basic organization of a STAR element is a group of three sequences, or “boxes,” in the order B, C, A that are 46 bp, 57 bp, and 97 bp long, respectively (24). Box C can be tandemly repeated (Fig. (Fig.4,4, cat1). The shortest allele, corresponding to a basic structure, was observed in only four isolates from the present collection of strains (NL33 and a group of nasal carriage isolates, Laus253, Laus369, and Laus292), all localized in out-group positions. In isolates of CC45, the presence of a 21-bp duplication (Fig. (Fig.4,4, cat2) was observed and different alleles were seen to possess one to four C sequences. CC5 and CC8 isolates had an identical structure, with an insertion within the box C repeats of the STAR element and variations in copy number on both sides (Fig. (Fig.4,4, cat3). In several other CCs (and in strain RF122), it seemed that a deletion had removed part of this insertion (Fig. (Fig.4,4, cat4). An artificial convention is proposed so that all the different alleles and deletion combinations can be coded as in the case of an ordinary TR variation (described in the support website at http://minisatellites.u-psud.fr/MLVAnet/index.php?&largeur-1680).
We demonstrate in this study that the results of MLVA genotyping of S. aureus are highly informative and congruent with those of two other widely used techniques, MLST and spa typing. MLVA provided better resolution and was more congruent with the results of MLST than with those of spa typing. VNTRs were previously used to genotype S. aureus in a method called multilocus variable-number tandem-repeat fingerprinting (MLVF) to reflect the fact that the data are not interpreted in terms of repeat copy number at each locus but simply used to compare multiple banding patterns as is done with PFGE (21, 27, 38, 45). The discriminatory power of MLVF is similar to that of PFGE, but the profiles are not easily comparable between laboratories and this precludes the constitution of international databases. The reproducibility of MLVF might be also limited by the performance of 10-plex PCRs. Other PCR-based genotyping techniques, such as repetitive sequence-based PCR, are satisfactory for determining strain relatedness, but their informativeness is low and the data are not suitable for population studies (47). Although MLVA is a gel-based method (using either traditional gels or sophisticated capillary electrophoresis equipment), it relies on coding the results as a string of numbers. This, in contrast to all fingerprinting methods, including PFGE and MLVF, makes the resulting profiles highly reproducible and portable, like MLST and spa typing.
In the present MLVA scheme, markers which could be easily analyzed using agarose gel electrophoresis were favored (i.e., with a size difference between two alleles that is greater than 5% of the amplicon size across the whole allele size range). As such, the assay can be performed in a laboratory equipped with simple molecular biology equipment. In addition, preliminary experiments showed that the assay can be easily adapted to capillary electrophoresis. For instance, six markers could be analyzed simultaneously on a Beckman CEQ8000 apparatus in a multiplex PCR, using three fluorophores and selecting VNTRs according to the allele size range (see Fig. S2 in the supplemental material). Further development of this procedure should allow the analysis of the complete set of markers in two to three PCRs.
In addition to informativeness and portability, the cost issue is also very important when choosing a genotyping technique, particularly when thousands of isolates need to be investigated. Based on the cost of consumables only, the cost of in-house MLST typing for one sample (seven PCR amplifications and seven sequencing reactions) was estimated to be $30, with an initial start-up cost of $100,000 to $300,000 for equipment. When sequencing is performed outside the laboratory, it amounts at the very least to $50 per strain for sequencing costs alone. The cost for consumables alone for MLVA with 14 VNTRs amounts to $10 per strain whether the analysis is done with agarose gels or capillary electrophoresis, although the later requires a considerably higher start-up investment and expensive maintenance, whereas the former requires more hands-on time.
In recent years, MLST has provided a clear view of the S. aureus population and its evolution, with the definition of a number of well-defined CCs that are evolving, largely independently from each other, to such an extent that some authors suggest giving species status to each CC (46).
The collection of strains analyzed here covers the two groups or clades identified by MLST analysis (2, 18). This is illustrated in Fig. Fig.5,5, using a minimum spanning tree representation and allowing the prediction of missing links. The phylogenetic relationships observed by using MLVA data also separate the isolates into two main groups (Fig. (Fig.2).2). The two clades are not as sharply defined as those found by using MLST data, but the main relationships between CCs are similar using either MLVA or MLST. MLVA clearly allows the identification of CCs and the emergence of new families, as shown here with ST239 emerging from ST8, for example (33).
In the nasal carriage isolates, the clustering produced by MLVA is very similar to that identified by eBURST analysis of MLST data (39). In this previous work, ST942 and ST707 were grouped by DLST, although, when grouped by MLST, they differed by more than two mutations. This was also the case for ST291 and ST398. The MLVA-14 assay, which investigates 14 loci throughout the genome, also groups these isolates, suggesting a closely related origin of at least some part of their genome. Their spa types are compatible with a recent common origin. In the previous study mentioned above, amplified fragment-length polymorphism analysis placed Laus356, an ST45 isolate with a spa type characteristic for CC45, with a long branch next to the main CC45 cluster. By MLVA, this isolate is positioned with a long branch near the CC12 cluster.
Among the 14 selected VNTRs, 6 are present in STAR elements. The polymorphism of box C in STAR elements is used in a typing protocol based on RFLP and PCR (32) and contributes significantly to the present MLVA-14 assay. Additional polymorphism is provided by different rearrangements, including the insertion or deletion of sequences which can be further described upon sequencing. In particular, marker Sa0906 shows at least four structures which appear to be of phylogenetic value. Indeed, MLVA clustering places certain strains in an ancestral position to specific clusters, for example, NL33, a CC8 isolate from The Netherlands.
In the present collection, strains belonging to CC7, CC15, and CC25 were all MSSA, whereas both MSSA and MRSA isolates were found intermixed in all the other clusters. CC8 and CC5 isolates are mainly MRSA, as previously observed. This is in agreement with the CCs observed in typical MRSA and MSSA lineages worldwide (reviewed in reference 5).
We describe in CC45 an unusual polymorphism at locus Sa0906, as well as in loci Sa1213 and Sa1425, with no or only a fraction of one repeat (data not shown). In addition, this CC also possesses only one copy of the repeat for loci Sa1132, Sa1291, Sa2039, and Sa1756 (SIRU15). Among these VNTRs, only locus Sa1132 is within a hypothetical protein-coding gene. In addition, CC45 displays the highest level of spa polymorphism, apparently through loss of motifs, whereas the other VNTRs show relatively less diversity (44). Therefore, we believe that in this CC, there is for some reason an important level of recombination leading to deletion. We are presently exploring the basis for this phenomenon. At the moment, there is no complete sequenced genome available for a CC45 strain.
There is a need to expand investigations of pathogenic bacterial populations on a worldwide scale. The currently available data come from very few countries (36), and the main typing method (MLST) has an insufficient discriminatory power for epidemiological and shorter-term evolutionary studies, which is well exemplified by the study of Nubel et al. (28). Single-nucleotide polymorphism typing might provide an interesting alternative to MLST. However, in a highly clonal species like S. aureus, the use of single-nucleotide polymorphisms is an inherently biased approach which will fail to explore CCs not previously identified by other means. For epidemiological studies, MLVA entails costs and time investment similar to those of spa typing while providing considerably higher resolution. MLVA appears to be adapted to large-scale investigations and, in addition, might give some insight into the effects of DNA plasticity and recombination through the analysis of the TR mechanism of instability. In contrast to MLVF, in which VNTRs are used to produce a pattern, MLVA investigates each locus independently. The MLVA-14 assay described here, comprising 10 loci in panel 1 and 4 loci in panel 2, is highly discriminatory, cost efficient, reproducible, and portable. The assay can be used in automated or more manual protocols, as is best adapted to local conditions. If necessary, additional VNTRs described in the literature (Table (Table1)1) could be added by taking advantage of the numerous available genome sequences to design new primer sets able to amplify most if not all strains. In addition to the present collection of strains, the MLVA-14 assay has been applied to about 300 isolates from French patients with cystic fibrosis, allowing a follow-up during chronic infection (H. Vu Thien, K. Hormigos, G. Corbineau, B. Fauroux, H. Carvol, D. Moissenet, G. Vergnaud, and C. Pourcel, unpublished data).
We are grateful to Alex van Belkum and René te Witt for the generous provision of the Harmony strain collection.
This work was supported by CNRS, Université Paris-Sud, and the Association Vaincre la Mucoviscidose.
Published ahead of print on 26 August 2009.
†Supplemental material for this article may be found at http://jcm.asm.org/.