|Home | About | Journals | Submit | Contact Us | Français|
Populations of the food-borne pathogen Listeria monocytogenes are genetically structured into a small number of major clonal groups, some of which have been implicated in multiple outbreaks. The goal of this study was to develop and evaluate an optimized multilocus variable number of tandem repeat (VNTR) analysis (MLVA) subtyping scheme for strain discrimination and clonal group identification. We evaluated 18 VNTR loci and combined the 11 best ones into two multiplexed PCR assays (MLVA-11). A collection of 255 isolates representing the diversity of clonal groups within phylogenetic lineages I and II, including representatives of epidemic clones, were analyzed by MLVA-11, multilocus sequence typing (MLST), and pulsed-field gel electrophoresis (PFGE). MLVA-11 had less discriminatory power than PFGE, except for some clones, and was unable to distinguish some epidemiologically unrelated isolates. Yet it distinguished all major MLST clones and therefore constitutes a rapid method to identify epidemiologically relevant clonal groups. Given its high reproducibility and high throughput, MLVA represents a very attractive first-line screening method to alleviate the PFGE workload in outbreak investigations and listeriosis surveillance.
Listeriosis is a food-borne infection caused by the bacterium Listeria monocytogenes. Invasive forms of human listeriosis include septicemia, meningitis, and maternal-fetal infections (1). Listeriosis is associated with high hospitalization and fatality rates (almost 100% and 25 to 30%, respectively). Populations at risk include pregnant women, immunocompromised individuals, and the elderly. L. monocytogenes is widely present in the environment, including soil, water, vegetation, and silage, as well as in animals and animal-derived food, and can contaminate food in processing plants and retail establishments. L. monocytogenes is recognized as a public health issue and a serious challenge for the food industry, and this has led to the establishment of national surveillance systems in several countries. L. monocytogenes also stands out as a model system in the fields of microbiology, cell biology, and immunology and for the study of host-pathogen interactions (2–5).
L. monocytogenes strain characterization on the basis of serotyping and molecular typing methods is used for surveillance, epidemiological tracking, and outbreak investigation purposes (6, 7). Genetic variants of L. monocytogenes have diversified into four major phylogenetic lineages, with lineages 1 and 2 each containing multiple clonal groups of public health importance (8–14). As these groups appear to differ in virulence and epidemic potential (6, 15), it will be interesting to better define their epidemiological, clinical, and microbiological specificities. For this purpose, tools for the easy identification of clonal groups are needed to recognize such groups and determine their presence in a large variety of sources. Several typing methods are currently available for L. monocytogenes strains. Conventional serotyping (16) and its molecular proxy PCR serogrouping (17) discriminate major categories of strains that correlate strongly (albeit not totally) with lineages and clones (11, 12, 14), but these methods do not have the necessary discriminatory power in the context of outbreak investigations. Pulsed-field gel electrophoresis (PFGE) is established as the gold standard for L. monocytogenes strain subtyping and is widely used for listeriosis surveillance and outbreak investigations (18). Yet, PFGE presents several practical disadvantages, as it is time-consuming and requires stringent standardization for interlaboratory data comparison. Multilocus sequence typing (MLST) is a well-established reference method for global epidemiology and population biology (19, 20), as it renders interlaboratory genotype comparisons easy and unambiguous and as sequence data can be used to infer useful population genetic information such as amounts of genetic diversity, recombination rates, and strain phylogeny. MLST also provides backward compatibility with genome sequencing (21). However, MLST is neither rapid nor cheap and has limited discriminatory power within L. monocytogenes (12, 22). Given the current limitations of available methods for L. monocytogenes strain typing, a potentially useful complementary approach is multilocus variable number of tandem repeats (VNTR) analysis (MLVA). This method is largely used for epidemiological tracking of bacterial pathogens (23, 24) because it is relatively easy and cheap to implement and because it has remarkable discriminatory power in many, although not all, bacterial species. MLVA relies on the study of the variability of the number of tandem repeats at specific loci in bacterial genomes. MLVA schemes are constructed on the basis of an open choice of several VNTR chromosomal loci, and five different MLVA schemes have been developed almost simultaneously for L. monocytogenes strain typing (25–29). Subsequent use of MLVA in outbreak investigations and strain diversity studies has relied on these different schemes or combinations thereof (30–34).
The five previously proposed MLVA schemes differ in the number of VNTR loci that were included, ranging from only 3 (27) to 8 (29) and even 10 (25). These studies also differed in the number, diversity, and inclusion criteria of L. monocytogenes isolates used to evaluate strain typeability and discrimination. As strains of both lineage I and lineage II are frequent among sporadic human infections and can cause outbreaks (8, 11, 14, 35), an MLVA scheme should ideally be applicable to all strains of both lineages. Many of the primer pairs defined so far to amplify VNTR loci did not amplify a number of strains, which limits their discriminatory power and complicates the interpretation of differences among strains. Besides, it is important to calibrate the MLVA method against PFGE on the one hand and MLST on the other hand. Indeed, PFGE is the standard for epidemiological investigations, whereas MLST is well established as a population biology tool and provides a standard operational definition of clones as clonal complexes (CCs) (12, 36). Only one of the five proposed MLVA schemes was compared with PFGE data obtained by following the current standard protocol of dual restriction enzyme (ApaI and AscI) use (29), while a comparison with MLST has been performed in only one study so far (27). The added value of MLVA relative to PFGE and MLST thus clearly remains to be precisely defined.
The aims of our study were (i) to identify novel VNTR loci and evaluate them, as well as all previously described loci, for typeability of isolates representative of a broad range of genotypes of lineages I and II, including all major clones defined by MLST; (ii) to evaluate the ability of MLVA to identify MLST-defined clonal groups; and (iii) to evaluate and compare to PFGE, the added value of MLVA in terms of strain discrimination within clonal groups of particular interest.
A total of 255 isolates were included in this study (see Table S1 in the supplemental material). First, 217 L. monocytogenes isolates were selected from the National Reference Centre for Listeria (NRC-L) and from the World Health Organization Collaborative Centre for Listeria (WHO-CC-L) collections. This included 155 isolates previously characterized by us, i.e., 58 isolates from a study by Ragon et al. (12) and 97 isolates from a study by Chenal-Francisque et al. (36). We also included 38 isolates selected to represent additional epidemiologically unrelated isolates of CC3, CC4, CC5, and CC8, which were represented by only a few isolates in our previous studies. To assess the epidemiological concordance of MLVA, we also included 24 strains collected during a large outbreak that occurred in 1992 in France and was linked a ready-to-eat meat product, pork tongue in jelly (37). The 217 isolates described above were epidemiologically unrelated, except for some of the 24 human or food isolates corresponding to the period of the 1992 French outbreak (see Results). Overall, the 217 isolates were recovered from human infection cases (n = 130), food (n = 47), animal (n = 18), the environment (n = 9), and vegetation (n = 1) and 12 were of undocumented origin. They were collected from 31 countries across six continents and were isolated between 1933 and 2010.
Second, 38 International Life Sciences Institute (ILSI) strains were included in order to place these well-documented strains (38) within the diversity based on the Institut Pasteur MLST scheme (http://www.pasteur.fr/mlst) and thereby establish a correspondence between MLST clonal groups and previously defined clonal groupings. The ILSI strains (38) included 23 strains representative of past outbreaks defined as members of epidemic clones (ECs) ECI, ECII, ECIII, and ECIV and 15 strains from the ILSI diversity set (see Table S1 in the supplemental material). These isolates were recovered from human (n = 20), food (n = 10), animal (n = 7), and environmental (n = 1) sources.
All isolates were confirmed as L. monocytogenes with the API Listeria system (bioMérieux, Marcy l'Etoile, France) and the multiplexed PCR serogrouping method (17). Hemolytic activity was confirmed on blood agar plates.
Total genomic DNA was extracted with the Promega Wizard Genomic DNA purification kit (Promega, Madison, WI) according to the manufacturer's instructions. DNA samples diluted in water at 50 ng/μl were used as the template for PCR amplifications.
First, we included all 15 of the VNTR loci described previously (25–29). The VNTR loci common to more than one of the above studies were amplified with all of the different PCR primer sets defined previously (Table 1). Second, to identify additional VNTR loci, the tandem repeat site http://minisatellites.u-psud.fr (39), which uses the software Tandem Repeat Finder (40), was used for the genomes of strains EGDe and F2365, with a minimal unit length of 6 bp, a minimal copy number of two repeats, and an 80% sequence match. Three novel repeat regions (Lis-TR357, Lis-TR495, and Lis-TR1869) were identified. To optimize the primer design for these loci, as well as for previously described loci, alignment of flanking regions of the VNTR loci was performed on the basis of publicly available genomes. This optimization effort resulted in a total of 38 distinct PCR assays (Table 1) corresponding to 18 distinct loci (Fig. 1).
A preliminary screening was performed on a genetically diversified subset of isolates of major clonal groups of phylogenetic lineages 1 and 2. Each VNTR region was separately amplified by PCR with Isis DNA polymerase (MP Biomedicals, Santa Ana, CA.). The PCR run conditions included a denaturation step of 94°C for 4 min, followed by 35 cycles including denaturation at 94°C for 30 s, elongation at a temperature depending on the primer pair (Table 1) for 30 s, and an extension step of 72°C for 1 min. A final extension of 72°C for 10 min was performed. The finalized protocol consisted of two multiplexed PCR mixtures called M1 (containing six primer sets) and M2 (five primer sets). The Qiagen Multiplex PCR kit (Qiagen, Valencia, CA) was used. The concentrations of the primers were adjusted to obtain even intensities for all fragments, with the forward primers being fluorescently labeled at the 5′ end (Table 2). The same PCR cycle was used for M1 and M2 multiplex PCRs: 95°C for 15 min, followed by 35 cycles of 94°C for 30 s, 56°C for 90 s, and 72°C for 90 s, with a final extension at 72°C for 10 min.
The purified PCR products of single PCRs performed with unlabeled primers were sequenced to control the molecular origin of size variation. The DNA sequence for every distinct allele of all VNTR loci was determined. Nucleotide sequence data were edited and compared with BioNumerics version 6.6 (Applied Maths, Sint-Martens Latem, Belgium) and Multalin (http://multalin.toulouse.inra.fr/multalin/).
The L. monocytogenes EGDe strain was passaged on Columbia agar medium 65 times over a period of 4 months. At each passage, a sweep was taken and streaked onto the next Columbia agar medium tube. After every 10th passage and the last one, DNA from a bacterial sweep was extracted and the total DNA was subjected to the MLVA assay.
Each isolate was typed by PFGE according to PulseNet standardized procedures with the AscI and ApaI restriction enzymes (18). Data analysis was performed with BioNumerics version 6.5. ApaI and AscI PFGE types were defined as differing from other types by at least two bands for each individual enzyme.
PCR products obtained with fluorescently labeled primers (Table 2) were subjected to capillary electrophoresis on an ABI 3730XL DNA sequencer. The sizes of PCR fragments were determined with the Applied Biosystems GeneMapper v4.0 software. The number of tandem repeats was then calculated according to the PCR fragment size and the known length of the repeat unit of each VNTR locus. The number of tandem repeats was rounded to the closest whole integer. In accordance with this procedure, an allele number corresponding directly to the number of tandem repeats was assigned to each PCR fragment. Allele strings were then imported into a BioNumerics database. Because no data were available at locus Lm11 for most isolates of CC7 and CC8, we chose to exclude this VNTR locus from subsequent polymorphism calculations and from the determination of the overall number of MLVA types. Each unique combination of the 10 remaining allelic numbers was converted into a distinct MLVA type. PFGE patterns were compared with a tolerance parameter of 1% and an optimization parameter of 1%. Dendrogram analysis by the unweighted-pair group method using average linkages (UPGMA) and minimum spanning tree (MStree) construction were performed with BioNumerics.
Homoplasy along the branches of the MStree was defined as 1 − (K − 1)/M, where K is the number of alleles and M is the number of changes along the MStree. Hence, if each allele was generated by a single evolutionary event, the number of changes since the ancestral state would be K − 1 and the homoplasy index would be equal to zero. In contrast, if alleles are often changing by convergent evolution or reversion to an ancestral state, M would become much greater than K and the homoplasy index would increase toward 1.
To determine their diversity and whether they belong to major MLST clones, we analyzed the 255 isolates by MLST. Data for 155 isolates were derived from our previous studies (12, 36), and we performed MLST analysis of 100 additional isolates for the purposes of the present study. These 100 isolates included 38 strains from the reference ILSI collection, 24 isolates collected during the large 1992 outbreak in France (37), and 38 isolates representing CC3, CC4, CC5, and CC8 (see Materials and Methods). The seven MLST genes of the 100 isolates could be amplified, and their sequence types (STs) were determined. The 20 STs that were not previously described (ST326 or above in Table S1 in the supplemental material) were incorporated into the Institut Pasteur MLST database at www.pasteur.fr/mlst.
The 217 isolates from the NRC-L and WHO-CC-L included 110 STs, thus representing a genetically diverse population. The 38 ILSI strains represented 18 STs, 9 of which were unique to the ILSI strains. Figure 2 represents the joint analysis of the 255 isolates with the MLST diversity previously described by using the Institut Pasteur MLST scheme (12, 36, 41). For a tree locating the ILSI strains relative to the other strains in this study, see Fig. S1 in the supplemental material.
MStree analysis of the MLST data (Fig. 2) confirmed that the population of L. monocytogenes is dominated by a few numerically dominant CCs that we have named “major clones” (12, 36). In the present study, most lineage I strains (172 out of 176) belonged to CC1 (62 strains), CC2 (41 strains), CC3 (25 strains), CC4 (16 strains), CC5 (10 strains), or CC6 (18 strains). In lineage II, CC7 (22 strains), CC8 (13 strains), and CC9 (23 strains) were the most strongly represented, although 21 strains (27%) belonged to less frequent CCs or STs.
MLST analysis of the ILSI reference strains representing past outbreaks revealed that all of the outbreaks except three were caused by strains that belong to major clones as defined by MLST (see Fig. S1 in the supplemental material). The three exceptions were the 2000 North Carolina queso fresco outbreak caused by a strain that belong to ST558, which differs by three MLST genes from ST2, and the 1988 turkey frank case and the 2000 sliced turkey deli meat multistate outbreak caused by a strain that belongs to ST11. The latter two events were traced to the same food processing facility, and the corresponding strains were shown to be highly similar, which led to the definition of ECIII (6). ST1, the central genotype of MLST CC1, included strains of the 1986 Los Angeles Jalisco outbreak, the 1981 Nova Scotia coleslaw outbreak, and the 1983 to 1987 Switzerland Vacherin Mont d'Or outbreak. These strains were previously defined as belonging to ECI (6). ST2, the central genotype of CC2, comprised the reference strains of the Massachusetts pasteurized milk outbreak and the 1987 United Kingdom and Ireland pâté outbreak. These isolates were assigned to ECIa, which was later renamed ECIV (42). ST3, the central genotype of CC3, included the strain from the 1994 Illinois pasteurized milk chocolate outbreak. Finally, ST6, the central ST of CC6, included the 1988 to 1989 multistate hotdog outbreak and the 2002 multistate deli turkey outbreak. Strains involved in these two later outbreaks were assigned to ECII (6). Note that the 2008 Canadian outbreak was assigned to newly defined ECV, which belongs to MLST CC8 (43, 44). The ILSI strains from the diversity set were scattered across the MStree (see Fig. S1), consistent with the fact that they were selected to represent a diversity of ribotypes or PFGE types. However, two of these strains (FSL J2-064 and FSL J1-169) belong to ST5, the central genotype of major CC5, one strain (FSL C1-122) belongs to ST1, and two other strains (FSL C1-056 and FSL J1-094) represent major lineage 2 CC7 and CC9, respectively. Therefore, seven out of nine major MLST clones appear to be represented in the ILSI collection, with the exceptions of CC4 and CC8.
L. monocytogenes phylogenetic lineages I and II account for almost all cases of human listeriosis. To identify a panel of PCR primer pairs able to amplify VNTR loci from a diverse sample of isolates of these lineages, we selected a panel of strains representing a diversity of STs and CCs within lineages I and II, i.e., two strains each of CC1, CC2, CC3, CC4, CC5, CC7, and CC9 plus strain EGDe. A total of 38 primer sets corresponding to 18 distinct VNTR loci were investigated in this preliminary screening (Table 1; for convenience, we assigned the letters A to R to the 18 loci). Most of these primer pairs have been defined in five previous studies (25–29). In addition, we tested three novel loci (P, Q, and R; Table 1) identified herein. We noted that most of the VNTR loci were located in the first half of the chromosome of strain EGDe, especially in the first quarter (Fig. 1). We also redesigned primer pairs for the two previously identified loci, A and B (Table 1), which correspond to loci Lm3 and Lm32, respectively (29). Previously published procedures were followed for previously described primer sets. The technical results obtained with the 38 PCR assays are summarized in Table 1. For each of the A, B, C, E, F, G, H, I, and J loci, we found a PCR primer pair that gave optimal amplification results with all of the screening isolates, with a unique amplified fragment of the expected size. One PCR primer set was thus selected for each of these loci (Table 1; Fig. 1) as follows: LisTR881, locus A; Lis-TR1317, locus B; LMV1, locus C; LMV6, locus E; JLR1, locus F; JLR2, locus G; Lm11, locus H; LM-TR4, locus I; LMV9, locus J. For loci L and R, a single previously defined primer pair gave a unique expected band with all of the isolates screened. In contrast, none of five PCR primer sets defined for locus D were satisfactory, as some did not amplify all of the isolates tested or nonspecific PCR products were observed. Besides, this locus exhibits size variation due to insertions-deletions in the sequences flanking the tandem repeat region (see below), rendering fragment size variation difficult to interpret in terms of tandem repeat numbers. Similarly, a lack of PCR amplification was observed with some screening isolates with LisTR-495 (locus Q), LM-TR5 (locus N), and LM-TR6 (locus O). The latter is consistent with the initial report of PCR failure for locus LM-TR6 of many strains (28). For locus K, the two PCR primer sets also turned out to be unsatisfactory: TR3 gave nonspecific or no amplification, depending on the strain, and primer pair Lm8 yielded a unique PCR fragment but this fragment was same size for nearly all of the strains, indicating a lack of discriminatory power, as reported previously (29). Likewise, LM-TR2 (locus M) and Lis-TR357 (locus P) provided low discrimination. Loci D, K, M, N, O, P, and Q were thus eliminated from the assay optimization process. The 11 loci that were retained (Fig. 1; Tables 2 and and3)3) included one novel locus identified herein (Lis-TR1869), one locus (JLR4) uniquely described by Larsson (25), and nine loci previously identified by various authors. Among the latter, two novel primer pairs were used (Lis-TR881 for locus A and Lis-TR1317 for locus B). Notably, the 11 loci retained included seven out of eight loci selected for inclusion in the Sperry scheme (29) but only two out of six loci proposed by Murphy et al. (28) and one of the three loci used by Miya et al. (27).
The nucleotidic sequences of all of the VNTR loci of one representative isolate of each allele were determined. Sequence alignments confirmed that the expected target regions were amplified and that the size variation of the PCR fragments, as determined by capillary electrophoresis, was due to variation in the number of repeats. For VNTR locus LM-TR3, we observed DNA sequence variability in the flanking region, as noted previously (29). Independently of this observation, locus LM-TR3 was eliminated because of nonspecific PCR amplification. The size range of the fragments obtained at the 11 loci is shown in Table 3.
To test the repeatability of these assays, we analyzed 50 isolates in duplicate, and the results showed complete agreement of the deduced repeat number at all loci. In order to develop a rapid and simple MLVA assay, the 11 primer sets selected from the preliminary screening were distributed into two distinct groups (Table 2), and two separate multiplexed PCR assays called M1 (six VNTR loci) and M2 (five VNTR loci) were developed. To test for the stability of the VNTR loci during laboratory subculture, strain EGDe was subcultured 65 times and no repeat number variation was observed during these passages at any of the 11 VNTR loci.
The two multiplexed PCR tests were applied to the analysis of an expanded panel of 255 L. monocytogenes isolates. These isolates represented multiple isolates of major L. monocytogenes MLST clones and several other STs of lineages I and II (Fig. 2). Ten out of 11 loci showed nearly exhaustive amplification on the 255 isolates, with only 3 missing data points out of 2,550: locus JLR1 for FSL J2-063 (ST16, serotype 1/2a), locus Lis-TR1317 for FSL C1-115 (ST370, serotype 3a), and locus LMV6 for strain LM70290 (ST251, serotype 4b). All amplified PCR products were sized, and the corresponding tandem repeat number was determined, indicating nearly complete typeability with these 10 loci (Table 3). However, we observed no Lm11 PCR product for 38 isolates. Interestingly, these isolates all belonged to MLST CC7 or CC8; the only exception was ST376 (strain FSL J2-066, serotype 1/2a). Locus Lm11 could be amplified and sized for two isolates of CC7 but none of CC8. This result suggests a loss of this locus in isolates of CC8 and in the majority of the isolates of CC7. A search for the Lm11 locus in genomic sequences confirmed its absence from SLCC5850 (CC7) and from 08-5578 and 08-5923 (CC8). Regarding strains other than CC7 and CC8, locus Lm11 had allele 5 in most of the strains of lineage I and allele 4 in most of the lineage II strains. However, strains of CC4 (lineage I) had Lm11-4, and a subset of CC1 strains had allele Lm11-6, which was specific to these strains. Likewise, only strains of ST11 (ECIII) had allele Lm11-1. This locus is therefore very useful for the characterization of strains from these specific groups.
MLVA analysis with multiplexed PCR assays M1 (excluding Lm11) and M2 combined resolved the 255 isolates into 66 different MLVA types. In contrast, the M1 multiplexed PCR assay generated 44 MLVA M1 types, whereas the M2 multiplexed PCR assay distinguished 35 MLVA M2 types. Figure 2 illustrates the distribution of the distinct MLVA types onto the MLST diversity (note that the same color in the lineage I and II panels represents distinct MLVA types). It was remarkable that all of the major clones were characterized by specific MLVA patterns. In other words, no single MLVA type was shared by isolates belonging to different MLST clones. In addition, cluster analysis on the basis of MLVA data (Fig. 3) was highly consistent with the classification into MLST clones, as isolates of a single MLST clone clustered together in all cases, except for CC121.
Although less informative than when used together, the M1 and M2 multiplexed PCR assays also provided useful information when considered independently. By M2 PCR assay, most major clones showed clone-specific patterns that could prove useful for their identification. For example, CC1 isolates were characterized by M2 profile 8-2-21-2-11, within only four exceptions that had profile 8-2-21-2-12 because of variation at locus LMV9 (Fig. 3; see Table S1 in the supplemental material). Likewise, CC2 isolates had profile 7-1-21-3-12, with a single exception (7-1-21-3-11). In particular, whereas most of the clones showed several patterns, CC3 and CC7 showed only one M2 type that was specific to either CC. Cluster analysis of M2 patterns recovered all major clones as single branches, except CC9 (data not shown). In contrast, because of the higher variability of M1 multiplex PCR markers, cluster analysis of M1 data did not recover the MLST clones in general.
We observed different degrees of MLVA variation according to CC (Table 4). In particular, it was interesting that MLVA efficiently subdivided some of the major CCs, including their central ST (Fig. 2). Therefore, MLVA may represent a useful typing complement to MLST for isolates of specific clones. Table 4 shows the number of MLVA profiles and Simpson's diversity index per major clone. MLVA had more discriminatory power than MLST for CC4, CC5, CC6, and CC9. On the contrary, isolates of CC7, which represented 13 STs, were almost not discriminated at all by MLVA, with the exception of one strain at locus JLR2.
Subtyping of isolates within clones was mostly contributed by the M1 multiplex PCR assay, independently of the use of the M2 PCR assay (Table 4). For example, CC1 isolates were subdivided into four M1 types and the M2 PCR assay did not add to their discrimination; similar results were obtained for CC6 and CC8. However, in several cases, the use of multiplex M2 did improve the discrimination of isolates within clones (Table 4).
The 38 isolates from the ILSI collection, including the 23 isolates representing outbreaks, were divided into 22 MLVA types. The CC1 reference strains representing the 1981 Nova Scotia coleslaw outbreak and the 1983 to 1987 Switzerland Vacherin Mont d'Or outbreak were of the same MLVA type, which differed by a single marker (Lm11) from the 1986 to 1987 California Jalisco soft cheese outbreak (Fig. 3; see Table S1 in the supplemental material). CC2 strains from the 1983 Massachusetts milk outbreak and the 1987 to 1989 United Kingdom and Ireland outbreak shared the same pattern, except for variation at locus Lis-TR881 (Fig. 3; see Table S1). CC6 isolates from the two represented outbreaks could be distinguished by two loci, Lis-TR1869 and LMV6. Finally, the four isolates of CC11 (ECIII) from 1988 to 1989 and 2000 were all identical by MLVA. The CC1, CC2, CC6, and CC11 strains previously assigned to different ECs (ECI to ECIV) were clearly distinguished by MLVA, as were the isolates of the ILSI diversity set (Fig. 3; see Table S1).
VNTR loci can evolve following two alternative models. In the stepwise model, the number of repeats varies along lineages by the progressive addition or removal of single repeat units. In the saltational model, any allele has the same probability of evolving from any other allele, regardless of repeat number differences. To investigate the evolutionary mode of the 11 VNTR loci selected, we analyzed the 28 changes observed among closely related MLVA profiles, defined as profiles differing by a single allele out of 11. Out of 28 such changes, 22 (78.5%) corresponded to single-repeat differences; 3 changes involved two repeat differences, 2 changes involved three repeats, and 1 change (at locus LMV6) corresponded to six repeat differences. We conclude that L. monocytogenes VNTR markers evolve predominantly by the stepwise addition or deletion of a single repeat unit, as previously shown for other microorganisms (45).
To evaluate the degree of homoplasy, i.e., evolutionary convergence or reversion events, we compared the number of changes observed along the MStree deduced from MLVA data to the minimal number of changes that would be obtained without homoplasy. The homoplasy ratios ranged from 0.41 for JLR1 to 0.67 for JLR2 (Table 3). In other words, there were approximately twice as many changes as the number required to generate all of the distinct alleles, indicating a substantial degree of homoplasy.
PFGE analysis was performed with ApaI and AscI enzymes for all isolates. The 255 isolates displayed 123 combined AscI-ApaI PFGE types (see Table S1 in the supplemental material). To estimate the respective discriminatory powers of the MLVA and PFGE methods with only unrelated isolates (46), we excluded 9 of the 10 isolates that were shown to be identical by MLVA and PFGE and that were associated with the 1992 French outbreak (see below). On the basis of the 246 remaining isolates, Simpson's index of discrimination was 98.5% for ApaI-AscI PFGE (95% confidence interval, 0.980 to 0.990). In contrast, Simpson's index was only 93.6% (confidence interval [CI], 0.921 to 0.951) for MLVA based on the 10 loci (excluding Lm11). When taking into account locus Lm11, 69 distinct MLVA profiles were observed and Simpson's index increased to only 94.5% (CI, 0.932 to 0.959), still well below the discriminatory power of PFGE (see Table S1). As expected, several PFGE profiles were observed for isolates that shared the same MLVA type. However, in some cases, the reverse was observed: several MLVA types were found for isolates with the same PFGE profile. Interestingly, the relative discriminatory powers of MLVA and PFGE appeared to depend on the MLST clone (Table 4). The most notable difference between MLVA and PFGE was found for CC7 isolates, with Simpson indexes of 0.10 and 0.90, respectively. Isolates of this clone had only two distinct MLVA types, whereas they corresponded to 13 distinct PFGE types. Similarly, CC1 isolates showed only 4 MLVA types, in contrast to 28 PFGE profiles, whereas CC2 was subtyped into 5 MLVA and 19 PFGE patterns, and CC6 showed 6 MLVA and 11 PFGE patterns (Table 4). CC1, CC2, and CC6 include strains included in ECI, ECIV, and ECII, respectively. Therefore, they correspond to clonal groups that have caused multiple outbreaks. These results show that MLVA has limited discriminatory power in epidemiologically important clones. However, interestingly, MLVA had more discriminatory power than PFGE among isolates of CC4 (11 types versus 7; Simpson index of 0.91 versus 0.75) and CC9 (8 types versus 5; Simpson index of 0.78 versus 0.32). Therefore, MLVA may be useful for the subtyping of isolates that belong to these particular clones.
In order to evaluate the potential use of MLVA for outbreak investigations and to compare this method with PFGE within the context of an outbreak, we studied 24 isolates that had been collected during the 1992 French outbreak of listeriosis. These isolates included 13 human clinical isolates that spanned the July-to-October 1992 period, as well as 11 food isolates collected during epidemiological investigations of this outbreak. The 24 isolates displayed 13 distinct combined ApaI-AscI PFGE types and only three distinct MLVA patterns. The most frequent PFGE pattern (ApaI-AscI 38/41) was represented by 10 isolates, among which was the LM25703 strain, which was isolated from the incriminated source, pork tongue in jelly. These 10 isolates displayed MLVA type 0025, showing that MLVA did not exclude any isolate that was associated with the outbreak PFGE pattern. However, out of the 14 isolates with other PFGE types, 9 also had MLVA pattern 0025, whereas 4 had MLVA pattern 0022, differing by a single repeat unit at both JLR1 and Lis-TR1869), and one had MLVA pattern 0031, differing from MLVA pattern 0025 at 6 out of 11 loci (Fig. 3). The nine isolates with MLVA pattern 0025 but with a PFGE type that was distinct from ApaI-AscI 38/41 differed from the latter pattern by up to eight bands with AscI and up to five bands with ApaI. These results showed that MLVA did not distinguish from the source isolate several isolates that were clearly distinguished from the source isolate by PFGE.
L. monocytogenes strain subtyping is widely practiced in the context of human listeriosis surveillance and food safety control. Because the typing methods that are currently widely used, including serotyping, PFGE, and MLST, present a series of limitations, MLVA has attracted intense interest. MLVA fulfills a number of criteria recognized as important for the successful implementation of a typing method and the interpretation of its results (46). In particular, the rapidity, low cost, and easy implementation of MLVA are practical advantages that are largely recognized (23, 46). In the present study, we have developed an optimized MLVA scheme as a typing method for L. monocytogenes and compared it to two reference methods, MLST and PFGE. We have exhaustively tested the previously published VNTR primer sets and defined new ones against a genetically diverse set of L. monocytogenes strains. This allowed us to (i) develop a simple MLVA typing system, MLVA-11, which is based on two multiplex PCR assays combining 11 selected VNTR loci; (ii) analyze the variation at these 11 loci among a collection of 255 isolates; and (iii) compare the MLVA data obtained with PFGE and MLST data.
Previous MLVA schemes have been associated with a number of PCR failures for some VNTR loci, resulting in the incomplete characterization of some isolates (26, 28, 29, 33). Missing data lead to lower discrimination and loss of informative characters for phylogenetic placement and render the interpretation of differences among MLVA profiles problematic. Here, we optimized the selection of VNTR loci and the sequences of primers. We thus achieved nearly complete (99.9%) typeability for 10 markers, even when combining these assays into two multiplexed PCRs. As previously reported (29), we found null alleles at locus Lm11, but these PCR failures were restricted to isolates of CC7 and CC8, consistent with the absence of the corresponding locus in the genomes of representative strains of these two clones. However, we chose to keep this locus in the M1 multiplex assay because it provides some discrimination within CC1, one of the epidemiologically most important clones of L. monocytogenes. Besides, the inclusion of the Lm11 primers in the M1 multiplex PCR mixture did not negatively affect the assay for the other loci. As our sample covers a large breadth of the clonal diversity of lineages I and II, including all of the clinically most frequent serotypes and clonal groups, we demonstrate that the MLVA-11 system has broad applicability, an important characteristic of typing systems (46).
While this work was in progress, two studies were published in which distinct MLVA schemes were compared and combined (31, 33). Our MLVA scheme includes four loci that were not used by Li et al., while these authors' scheme has two loci that we chose not to include (LM-TR3 and LM-TR6). Chen et al. combined the three VNTR loci from Murphy et al. (28) with the three loci from Miya et al. (27). However, LM4b-TR1 (27) and LM-TR3 (28) correspond to the same locus and are therefore redundant. Besides, locus LM-TR3 was removed from our scheme as variation at this locus is difficult to interpret because of size variation in the regions flanking the repeat array (29).
Allele sequencing allowed us to confirm that size variation at the 11 selected loci is attributable to repeat number differences. We showed that MLVA loci are stable during laboratory subculture and found complete repeatability of allele coding by capillary electrophoresis separation. Elimination of loci with nonspecific amplification also facilitated the sizing of fragments and the determination of tandem repeat numbers. We also showed that the L. monocytogenes VNTR loci selected have a very strong tendency to evolve by the stepwise addition or removal of single repeat units. This observation indicates that the quantitative difference in the number of repeats, rather than simple allelic mismatch, can be taken into account to estimate strain relationships.
We compared MLVA-11 with the three widely used methods serotyping, PFGE, and MLST. While several MLST schemes have been published (9, 12, 47), only the Institut Pasteur scheme (12) provides a standardized nomenclature through a publicly accessible database and is used in a coordinated manner by multiple laboratories (www.pasteur.fr/mlst). With this MLST scheme, we characterized for the first time reference strains of several past outbreaks and show that previously defined ECs appear to correspond to MLST-defined major clones. Our MLVA assay successfully discriminated all MLST CCs, including those that correspond to ECI, ECII, ECIII, and ECIV. MLVA-11 thus appears as a very powerful method to identify these clones, as MLVA patterns are clone specific and as cluster analysis of MLVA patterns is strongly concordant with MLST clones. This good agreement can be explained by the fact that MLVA and MLST markers have similar levels of variation and by the rarity of recombination among L. monocytogenes strains. MLVA could therefore be used as a rapid identification tool for epidemiologically important clonal groups. For this purpose, it will be important to elaborate a more complete MLVA-MLST dictionary by mapping every novel MLVA pattern onto the MLST diversity. Along the same lines, one previous study compared MLVA with MLST data and found that MLVA could be useful for the recognition of three ECs of serotype 4b (27). It is interesting that the two multiplex assays of the MLVA-11 system may be used for different purposes. Whereas M2 multiplex alone clustered strains according to their clone and could therefore represent a useful rapid clone identification method by itself, in turn, M1 PCR was more useful to discriminate among isolates within clones.
One advantage of molecular typing methods, in contrast to serotyping, is that genetic markers may be used to estimate phylogenetic relationships among strains. MLVA is generally regarded as an unreliable phylogenetic method because of the high frequency of homoplasy. Such events can correspond either to reversion to ancestral states or to convergent evolution leading to the same allele by independent changes in different lineages (48, 49). We calculated that approximately half of the changes in our data set were evolutionary reversions or convergences. Therefore, MLVA data comprise a substantial degree of phylogenetic noise and must be interpreted with caution. As most VNTR loci are located within genes putatively encoding surface-exposed proteins (Table 2), it is possible that the number of repeats is subjected to selective pressures. In this context, it is remarkable that cluster analysis of the MLVA-11 profiles did recover the two main subdivisions corresponding to the two major phylogenetic lineages, I and II, and classified strains according to their MLST clone (Fig. 3). This result indicates that despite homoplasy, MLVA variation in L. monocytogenes does convey useful phylogenetic information. Several MLVA alleles are largely conserved within either lineage I or II; for example, Lis-TR1317 has predominantly allele 4 in lineage I and allele 3 in lineage II, while locus JLR4 has mostly alleles 8 and 4 in lineages I and II, respectively. This remarkable stability within lineages indicates that some VNTR markers diversify very slowly in L. monocytogenes. Another requirement for MLVA to convey phylogenetic information is that these loci must undergo restricted amounts of genetic recombination among L. monocytogenes strains, which is consistent with low recombination rates estimated on the basis of MLST and full-genome sequence analyses (12, 50). It is also important to use a relatively high number of VNTR markers, as the use of each multiplex individually did not recover phylogenetic placements as accurately (data not shown).
Listeriosis is a global public health issue that led to the implementation of surveillance systems in several countries in the European Union, the United States, and Canada (18). In this context, the high discriminatory power of PFGE made this method the de facto gold standard typing method. Our study indicates that MLVA has lower discriminatory power than PFGE based on enzymes ApaI and AscI, which is the current standard. Two previous evaluations of MLVA have concluded that MLVA had better discrimination than PFGE (26, 27). However, PFGE in these studies was performed with a single enzyme, either ApaI or AscI, and were thus based on a less discriminatory implementation of PFGE. Sperry et al. (29) compared MLVA data with PFGE data for 123 isolates and demonstrated a lower discrimination of MLVA than ApaI and AscI PFGE, a conclusion with which our findings fully agree. Our scheme includes seven of the eight loci included in the scheme of Sperry et al. and four additional loci. Surprisingly high discrimination was achieved by Li et al. (33) on the basis of nine MLVA loci, possibly because of the inclusion of solely unrelated strains, many of which were from food and the environment. From the above, we conclude that MLVA cannot be viewed as a replacement for PFGE when discrimination is a key requirement.
Interestingly, we showed that the discrimination of MLVA, relative to that of PFGE, is highly dependent on the clone. In particular, MLVA should be a useful addition to PFGE for the discrimination of strains that belong to CC4 or CC9. It is conceivable that MLVA loci evolve faster in some clonal lines, even though the reasons behind this heterogeneity are currently unknown. Alternatively, mobile elements including insertion sequences and phages may be less dynamic in some clones.
The high number of distinct MLST STs per clone and the Simpson index of discrimination of the MLST method (Table 4) should not be taken as evidence that MLST has higher discrimination power than MLVA. Indeed, strains with distinct STs were purposely included in this study on the basis of our previous MLST analyses (12, 36) in order to test the ability of MLVA to identify all of the variants within clones. This selection has inflated MLST's Simpson index artificially. However, as multiple isolates with the same ST or closely related ones were not discriminated by MLVA, it indicates that MLVA does not subtype MLST clones efficiently, with the exceptions of CC4 and CC9.
Concordance of typing results with epidemiological information is a desirable characteristic of a typing method (46). In our study, we used mostly independent isolates collected over wide temporal and geographic scales. Nevertheless, we found several groups of epidemiologically unrelated isolates with the same MLVA pattern and sometimes also the same PFGE pattern (see Table S1 in the supplemental material; PFGE16/54 and MLVA0010 [n = 9], PFGE27/11 and MLVA0055 [n = 6], PFGE53/14 and MLVA0048 [n = 7], and PFGE42/41 and MLVA0022 [n = 5]). These results show that identity of MLVA and PFGE profiles does not necessarily imply a direct epidemiological link. This is especially true for MLVA, and in our retrospective study of isolates from the French pork jelly outbreak, some of the isolates defined as unrelated by PFGE were not distinguished by MLVA.
MLVA has strong potential for interlaboratory standardization, as this method is highly reproducible and as data scoring into integer numerals provides unambiguous results. This is largely regarded as an advantage over PFGE, which can imply partly subjective decisions during band scoring. Standardization would provide benefits to international surveillance and population biology. However, MLVA standardization requires calibration of fragment sizing apparatuses and interlaboratory reproducibility needs to be carefully evaluated. For this purpose, we identified a set of 12 isolates that display distinct alleles at each of the 11 loci and together represent 80% of the distinct alleles found in this study (see Table S1 in the supplemental material). This MLVA reference set of strains is available upon request and should constitute a useful resource for interlaboratory calibrations.
MLVA was implemented in this study in the form of two multiplexed PCRs combining a total of 11 VNTR markers. The variation disclosed at these loci proved highly consistent with MLST data and was phylogenetically informative. These results show that MLVA could be used as a rapid identification method for MLST-defined clonal groups, including those corresponding to so-called ECI to ECIV. Because it has lower discriminatory power, MLVA cannot replace PFGE in outbreak investigations. However, given its simplicity, low cost, high throughput, and rapid time to results (around 8 h), MLVA could represent a useful screening method to alleviate the PFGE workload. Within the context of an outbreak, MLVA could advantageously fill the gap between the throughput needed to characterize a high number of isolates in a short period of time and the high discrimination level needed for informed epidemiological decisions. MLVA may also represent a suitable first-line assay for listeriosis surveillance, with PFGE efforts being focused on common MLVA genotypes. A two-step MLVA-PFGE strategy could significantly lighten the workload and would position MLVA as an important new tool in listeriosis surveillance.
We thank Martin Wiedmann for providing strains from the ILSI collection and Marie Ragon and Alban Le Monnier for contributing to the initiation of this work. We acknowledge Nathalie Tessaud-Rita and Arnaud Bertel for technical help.
This work was supported financially by the Institut Pasteur, INSERM, and the Institut de Veille Sanitaire (Saint Maurice, France).
Published ahead of print 10 April 2013
Supplemental material for this article may be found at http://dx.doi.org/10.1128/JCM.00606-13.