ABSTRACT
Listeria monocytogenes is a food-borne bacterial pathogen that causes a wide spectrum of diseases, such as meningitis, septicemia, abortion, and gastroenteritis, in humans and animals. Among the 13 L. monocytogenes serovars described, invasive disease is mostly associated with serovar 4b strains. To investigate the genetic diversity of L. monocytogenes strains with different virulence potentials, we partially sequenced an epidemic serovar 4b strain and compared it with the complete sequence of the nonepidemic L. monocytogenes EGDe serovar 1/2a strain. We identified an unexpected genetic divergence between the two strains, as about 8% of the sequences were serovar 4b specific. These sequences included seven genes coding for surface proteins, two of which belong to the internalin family, and three genes coding for transcriptional regulators, all of which might be important in different steps of the infectious process. Based on the sequence information, we then characterized the gene content of 113 Listeria strains by using a newly designed Listeria array containing the “flexible” part of the sequenced Listeria genomes. Hybridization results showed that all of the previously identified virulence factors of L. monocytogenes were present in the 93 L. monocytogenes strains tested. However, distinct patterns of the presence or absence of other genes were identified among the different L. monocytogenes serovars and Listeria species. These results allow new insights into the evolution of L. monocytogenes, suggesting that early divergence of the ancestral L. monocytogenes serovar 1/2c strains from the serovar 1/2b strains led to two major phylogenetic lineages, one of them including the serogroup 4 strains, which branched off the serovar 1/2b ancestral lineage, leading (mostly by gene loss) to the species Listeria innocua. The identification of 30 L. monocytogenes-specific and several serovar-specific marker genes, such as three L. monocytogenes serovar 4b-specific surface protein-coding genes, should prove powerful for the rapid tracing of listeriosis outbreaks, but it also represents a fundamental basis for the functional study of virulence differences between L. monocytogenes strains.
The current revolution in microbial investigations due to the sequencing of complete genomes has revealed insights into the genetic structures of a number of bacterial species. However, it has become evident that the genome sequence of one strain is not entirely representative of other members of the species and that the broad spectrum of physiological and virulence properties of bacterial pathogens mirrors the existence of different subsets of genes enabling different lifestyles. Besides whole-genome comparisons (1, 27), micro- and macroarray techniques are extensively used to study inter- and intraspecies diversity (16, 26). For example, microarrays and related techniques have been employed to study the genetic differences among members of the Mycobacterium tuberculosis complex (2, 7), Helicobacter pylori strains (30), Staphylococcus aureus strains (17), and Vibrio cholerae strains (14) and between Yersinia pestis and Yersinia pseudotuberculosis (23). The results underscore the fact that although considerable diversity is present among different bacterial isolates of the same species, clonal expansion of highly virulent subpopulations of a bacterial pathogen may exist.
Listeria monocytogenes, an intracellular pathogen, is the causative agent of serious epidemic and sporadic food-borne listeriosis. The clinical features of listeriosis include meningitis, meningoencephalitis, septicemia, abortion, perinatal infections, and gastroenteritis (34). Although rare when compared to other food-borne diseases, a significant feature of listeriosis is the high lethality rate (about 30%), which makes L. monocytogenes an important human pathogen. L. monocytogenes has the capacity to adapt and survive under extreme conditions, allowing it to ubiquitously exist in the environment and to survive and proliferate under conditions that exist within the food chain.
However, not all strains of L. monocytogenes seem to be equally capable of causing disease in humans. Isolates from 4 (1/2a, 1/2c, 1/2b, and 4b) of the 13 serovars identified within this species are responsible for over 98% of reported human listeriosis cases (24). Furthermore, all major food-borne outbreaks of listeriosis, as well as the majority of sporadic cases, have been caused by serovar 4b strains, suggesting that strains of this serovar may possess unique virulence properties. A number of different typing and population genetic studies suggested that different genetic divisions or lineages, which correlate with serovars, exist within the species L. monocytogenes (3, 6, 20, 28). Hereafter we will designate serovar 1/2a, 1/2c, and 3c strains as lineage I strains; serovar 4b, 1/2b, and 3b strains as lineage II strains; and serovar 4a and 4c strains as lineage III strains. Genetic analyses using multilocus sequence typing of virulence-associated genes, restriction fragment length polymorphism analysis, and ribotyping suggested that epidemic strains are mostly found in lineage II and that sporadic strains are found in lineages I and II, while lineage III strains are extremely rare and are mostly animal pathogens (25, 36). However, these methods are unable to further characterize the genetic basis for this observed variability.
Recently, the complete genome sequences of L. monocytogenes strain EGDe and Listeria innocua strain CLIP 11262 were determined (19). A comparison of these sequences revealed 10.5 and 14% specific sequences for each strain, respectively (19). The L. monocytogenes strain that was sequenced is from serovar 1/2a and belongs to lineage I. Therefore, it was important to investigate differences in gene content between lineage I and lineage II isolates and/or between different serovars.
In order to address questions regarding epidemiological and evolutionary relationships between pathogenic and nonpathogenic Listeria species and to define characteristics of particularly successful clonal pathovariants for causing disease, we partially sequenced an epidemic isolate of L. monocytogenes serovar 4b. Based on the comparison of the three Listeria sequences, we constructed specific arrays that were used to characterize 113 Listeria strains. The correlation of genomic, phylogenetic, and epidemiological properties of the strains allowed us to identify lineage-specific marker genes and to propose new evolutionary relationships. The results open new avenues for the development of rapid typing tools as well as for functional analysis of species- and serovar-specific genes to understand their roles in pathogenicity.
MATERIALS AND METHODS
Bacterial strains.The Listeria strains used in this study were selected from the culture collection of the National Reference Center for Listeria, Institut Pasteur, Paris, France, and from the production environment of different food plants. The strains were selected to represent all serovars of L. monocytogenes as well as disease- and non-disease-related isolates. A total of 93 L. monocytogenes strains from 12 different serovars isolated from humans (sporadic and epidemic cases), foods, animals, and the environment, as well as 20 representative strains of the different species of the genus Listeria (eight L. innocua strains, six Listeria ivanovii strains, two Listeria welshimeri strains, two Listeria seeligeri strains, and two Listeria grayi strains), were studied (Table 1). This set included eight epidemic strains from five outbreaks, 15 isolates from sporadic cases, and 19 strains from food plants. Strains were routinely grown overnight at 37°C without agitation in TPB broth (GIBCO).
Strains used in the DNA-DNA macroarray hybridization analysis
Sequencing, assembly, and sequence analysis.Cloning, sequencing, assembly, and annotation of L. monocytogenes CLIP 80489 were done as described by Glaser et al. (19). Essentially, 13,234 shotgun sequences were assembled into 1,020 contigs. For annotation, the program CAAT-box (18) was used.
Probe selection, primer design, PCR amplification, and array construction.The L. monocytogenes EGDe-specific probes spotted onto the Listeria array corresponded to the 270 genes defined previously as being specific for L. monocytogenes EGDe with respect to L. innocua CLIP 11262 (19). As some genes were too small to allow amplification of a PCR product of optimal size, the final array contained 262 EGDe genes. The 94 probes that were specific for L. innocua CLIP 11262 relative to L. monocytogenes EGDe were also selected from the previously defined list of 149 genes specific for L. innocua CLIP 11262 (19). For genomic regions containing several genes, only representatives were chosen for the array. In order to identify L. monocytogenes CLIP 80459-specific sequences, we used the program Cross-match (http:/bozeman.mbt.washington.edu/ ). With this approach, 141 sequence fragments, ranging from 33 to 3,025 bp, were identified as missing from both L. monocytogenes EGDe and L. innocua CLIP 11262. For probe design, only fragments longer than 1 kb were taken into account, allowing us finally to select 53 sequences that were specific for L. monocytogenes CLIP 80459. Primers were designed by use of a modified version of Primer 3 software (CAAT-box [18]) to amplify a specific fragment of 300 to 600 bp for each gene (melting temperatures were 55 to 65°C) (Eurogentec). Amplification reactions were performed in a 100-μl reaction volume containing 10 to 20 ng of chromosomal DNA. The concentration and size of each PCR product were verified on agarose gels. For array preparation, nylon membranes (Qfilter; Genetix) were soaked in TE solution (10 mM Tris [pH 7], 1 mM EDTA [pH 7.6]). Spot blots of PCR products and controls were performed by a Qpix robot (Genetix). Following spot deposition, membranes were fixed for 15 min in 0.5 M NaOH-1.5 M NaCl, washed briefly in distilled water, and stored wet at −20°C until use.
Hybridization.Genomic DNA was extracted by using a Qiagen DNeasy kit and was radiolabeled by using a random priming DNA labeling kit (Roche). Labeling was performed with 500 ng of genomic DNA and 50 μCi of 33P-labeled dCTP (Amersham). Labeled DNA was purified through Sephadex G-50 (Roche) or Qiaquick (Qiagen) minicolumns. High-density arrays were wetted in 2× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate) and prehybridized for 1 h in 10 ml of a solution containing 5× SSPE (1× SSPE is 0.18 M NaCl, 10 mM NaH2PO4, and 1 mM EDTA [pH 7.7]), 4% sodium dodecyl sulfate, 1× Denhardt's solution (50× Denhardt's solution is 1% Ficoll, 1% polyvinylpyrrolidone, and 1% bovine serum albumin), and 1 mg of denatured salmon sperm DNA. Hybridization was performed overnight at 60°C. Membranes were washed twice at room temperature and twice at 60°C in 0.5% SSPE-0.2% sodium dodecyl sulfate. Arrays were then sealed in polypropylene bags and exposed to a phosphorimager screen (Molecular Dynamics) for 24 h.
Verification of the specificity and quality of the macroarray.Fifteen percent of all PCR products were randomly chosen and sequenced. All 64 sequences corresponded to the expected PCR products. The membrane was then hybridized with chromosomal DNAs isolated from the three Listeria strains used to amplify the probes (L. monocytogenes EGDe, L. innocua CLIP 11262, and L. monocytogenes CLIP 80459) to test the quality and the correct spotting of probes.
Data analysis.For scanning, a model 445SI phosphorimager (Molecular Dynamics) was used. ArrayVision software (Imaging Research) was used for quantification of hybridization intensities and for normalization of data. For each spot, the hybridization intensity value was normalized by dividing it by the average of all significant intensity values on each filter (see supplemental tables at http://www.pasteur.fr/recherche/unites/gmp/sitegmp/gmp_projects.html ). For ratio calculation, a reference array, which was built by combining the average normalized data from three replicate hybridizations of the genomic DNAs of L. monocytogenes EGDe, L. innocua CLIP 11262, and L. monocytogenes CLIP 80459 to the corresponding spots on the array (http://www.pasteur.fr/recherche/unites/gmp/sitegmp/gmp_projects.html ), was used. In order to define the cutoff ratio for the presence of a gene, we analyzed the results for L. monocytogenes EGDe genes hybridized with L. innocua chromosomal DNA. The threshold for the presence of a gene was defined as >0.3. This corresponds to a DNA similarity of ≥80%, which was verified by sequence comparisons of these genes for both genomes. The data were then converted into binary scores (for ratios of >0.3, a gene was scored as present [1], and for ratios of <0.3, a gene was scored as absent [0]) (http://www.pasteur.fr/recherche/unites/gmp/sitegmp/gmp_projects.html ).The binary data were analyzed by hierarchical clustering with the program J-Express (13), by neighbor joining with the program MVSP 3.1 (Kovach Computing Services), and by intensive expert-based data mining with Excel spreadsheets.
RESULTS
Genome diversity among an epidemic L. monocytogenes strain, L. monocytogenes EGDe, and L. innocua and macroarray conception.The partial sequence of the epidemic L. monocytogenes serovar 4b strain CLIP 80459 (lineage II) was assembled into 1,020 contigs that encompassed 2.3 Mb of the genome (estimated complete genome, 2.9 Mb) and was compared to the genome sequence of L. monocytogenes EGDe (serovar 1/2a; lineage I). About 8% of the CLIP 80459 sequences were missing from EGDe. Thus, the genetic diversity between the two L. monocytogenes isolates seems to be close to that between L. monocytogenes EGDe and L. innocua (10.5%), which belong to two different species. One hundred forty-one serovar 4b-specific fragments, ranging from 33 to 3,025 bp, were identified. An analysis of these fragments identified partial or complete genes with homologies to genes encoding cell proteins, e.g., 7 surface proteins with an LPXTG motif, 6 transport proteins, 3 transcriptional regulators, 1 phosphotransferase system (PTS) components and 26 unknown proteins (seesupplemental material at http://www.pasteur.fr/recherche/unites/gmp/sitegmp/biodiversitylist.html ), suggestingthat epidemic L. monocytogenes strains differ substantially in gene content from the L. monocytogenes EGDe strain.
To extend the comparison from the three sequenced strains to a large collection of strains, we designed high-density membranes that were mainly focused on genes specific for each sequenced Listeria strain. This approach was chosen to increase the discriminatory power of the array. The membrane contained 409 probes, including 262 that were specific for L. monocytogenes EGDe relative to L. innocua CLIP 11262 and all its virulence genes, 94 that were specific for L. innocua CLIP 11262 relative to L. monocytogenes EGDe, and 53 that were specific for L. monocytogenes CLIP 80459. This membrane was used to analyze 113 Listeria isolates. Probes and primary hybridization results are available online as supplemental material (http://www.pasteur.fr/recherche/unites/gmp/sitegmp/biodiversitylist.html ).
Strain diversity and overall gene distribution.Based on the macroarray hybridization data predicting the presence or absence of the studied genes, we built bifurcating trees illustrating possible phylogenetic relationships between the different Listeria species by using the neighbor-joining method. Important points of gene conservation within each species of the genus Listeria and within distinct groups of L. monocytogenes strains were identified. The analysis grouped all strains without exception according to their species. Thus, the Listeria array allows accurate species identification, although the probes were defined from only two Listeria species. Each species was defined by a combination of genes that were specifically present or absent.
For the species L. monocytogenes, we identified 30 marker genes, including 18 that were present in all 93 L. monocytogenes strains tested (Table 2, group I) and absent from all other isolates of the remaining Listeria species and 12 that were present in all L. monocytogenes strains except the five serovar 4a and 4c isolates tested (Table 2, group II). Because serovar 4a and 4c strains are very rare and do not cause typical human listeriosis (35), thereafter these strains were not considered in the analysis of species-specific marker genes, but they will be discussed separately below. The 30 markers comprised the well-known virulence genes (plcA, plcB, and actA), 7 surface protein-coding genes (inlA, inlB, inlH, inlE, lmo0333, lmo0835, and lmo2821), 1 soluble internalin-coding gene (lmo0549), 3 genes for transcriptional regulators, and 11 genes for proteins of unknown function. For the species L. innocua, we identified four markers (lin0739, lin0803, lin2741, and lin2918) that were consistently present in all L. innocua strains tested. However, 29 of the 94 L. innocua genes spotted onto the membrane were detected only in L. innocua isolates, suggesting them to be species specific.
L. monocytogenes-specific marker genes
Since the macroarray did not contain probes specific for the other Listeria species, no specific markers for these species could be defined. However, orthologs of about a third of the EGDe genes and about a quarter of the CLIP 80459 or L. innocua genes were identified in at least one of the L. ivanovii, L. seeligeri, or L. welshimeri strains. L. grayi was found to be the most distantly related species. From the 409 probes, only 12 (6 from EGDe, 3 from L. innocua, and 3 from strain CLIP 80459) hybridized with DNA from L. grayi. These probes corresponded to genes present in all Listeria species, such as lmo1136, coding for an LPXTG protein, and genes coding for proteins of unknown function.
Subgrouping within the species L. monocytogenes.In addition to the neighbor-joining method, hierarchical clustering (J-Express) was used for the identification of specific gene clusters. Analysis of the 93 L. monocytogenes strains defined three lineages (I, II, and III) and distinguished two subdivisions within each lineage (Fig. 1). For each lineage and subgroup, specific markers were identified.
Listeria genetic diversity. Red and black areas denote the presence and absence of genes, respectively. (A) Dendrogram showing estimates of genomic relationships of the 113 strains constructed by hierarchical cluster analysis with the program J-Express. Phylogenetic lineages and subgroups are indicated. (B) Enlargements representing the groups of lineage-specific genes whose numbers are indicated to the right. I, lineage I (serovars 1/2a, 1/2c, 3a, and 3c); II, lineage II (serovars 4b, 4d, 4e, 1/2b, and 3b); III, lineage III (serovars 4a and 4c); I.1, serovars 1/2a and 3a; I.2, serovars 1/2c and 3c; II.1, serovars 4b, 4d, and 4e; II.2, serovars 1/2b and 3b.
Nineteen genes were associated specifically with lineage I (Table 3, group A). Twelve of these genes clustered in two regions (lmo0734-lmo0739 and lmo1968-lmo1974) coding for proteins putatively involved in sugar metabolism. Furthermore, a two-component regulatory system (lmo1060 and lmo1061), an ABC transporter complex (lmo1062 and lmo1063), and a gene coding for a surface protein containing an LPXTG motif (lmo0171) were lineage I specific. Surprisingly, the bvr locus (bvrABC) (4) was present only in isolates of lineage I and in the two serovar 4c strains. Eight genes allowed for the subdivision of lineage I. They were present in lineage I.2 (serovars 1/2c and 3c) but were generally absent from lineage I.1 (serovars 1/2a and 3a) (Table 3, group B).
L. monocytogenes lineage-specific marker genes
Five of the 53 serovar 4b-specific genes were markers for lineage II (Table 3, group C). Two code for transcriptional regulators, and three code for surface proteins containing an LPXTG anchor. Because serovar 4b strains are mainly responsible for human listeriosis, it is particular interesting to identify markers for serovar 4b strains or for the subgroups 4b, 4d, and 4e (lineage II.1). One such specific marker was ORF0799, coding for an unknown protein. ORF2372 (encoding a putative teichoic acid protein precursor C) and ORF2110 (encoding a putative secreted protein) were present only in serovar 4b and in two and four, respectively, of the six L. ivanovii strains. Furthermore, 35 of the 53 serovar 4b genes spotted on the membrane were conserved in all 4b strains, suggesting their implication in characteristic features of serovar 4b strains.
For lineage III strains (serovars 4a and 4c), no specific genes were identified, as our macroarray did not contain representative sequences of this lineage. However, lineage III was characterized by the absence of over 37% (96 genes) of the EGDe genes that were spotted on the membrane. Thirteen genes, clustered in eight different chromosomal regions, were specifically absent from lineage III strains. They code for surface proteins (lmo1666 and lmo0835), the arginine metabolic pathway proteins (lmo0036-lmo0041), and proteins of unknown function. Strains of serovar 4a (lineage III.1) were distinguished from those of serovar 4c (lineage III.2) by the lack of an additional 20 genes, 7 of which code for cell surface proteins (inlC, inlEHG, lmo0333, lmo0549, and lmo2821). These genes were also absent from all L. innocua strains tested.
Distribution of known virulence genes.The virulence gene cluster of L. monocytogenes comprises prfA, plcA, hly, mpl, actA, and plcB. Since these genes are a prerequisite for the virulence of L. monocytogenes, differences in virulence among different isolates could be due to the absence of one or more of these genes. However, macroarray hybridization showed that all 93 L. monocytogenes isolates contained this virulence gene cluster. The above-mentioned genes have been reported also to be present in L. ivanovii and L. seeligeri. We detected hybridization signals for the hly, mpl, and prfA genes, whereas the plcA, actA, and plcB genes were either absent or did not give a signal due to a high divergence of the corresponding gene orthologs. Indeed, the sequence similarity of plcA, plcB, and actA from different L. ivanovii and L. seeligeri strains compared to L. monocytogenes EGDe does not exceed 60%.
Several other genes of L. monocytogenes have been implicated in adhesion and internalization. Among those, the best studied are inlA and inlB. Our analysis revealed the presence of these two genes in all L. monocytogenes strains tested, confirming their species specificity. The uhpT gene (10) and the bsh gene (12) were identified in all isolates of the three hemolytic Listeria species (L. monocytogenes, L. ivanovii, and L. seeligeri).
High diversity level of surface proteins within the species L. monocytogenes.Fifty-five genes coding for putative surface proteins (http://genolist.pasteur.fr/ListiList/ ) belonging to the three sequenced Listeria genomes were spotted onto the array. We identified two groups of genes. The first group comprises 25 genes specific for the species L. monocytogenes, including inlAB, the inlGHE cluster, inlF, and a number of surface proteins of unknown function (Table 4). Two of the genes for surface proteins (lmo0171 and lmo2026) are lineage I specific and three (ORF2568, ORF2017a, and ORF0029) are lineage II specific. inlG seems to be specifically absent from all lineage II and serovar 4a strains. None of the serovar 4b surface protein-coding genes was identified in L. monocytogenes serovar 1/2c and 3a strains. For L. innocua, we identified two specific surface protein-coding genes (lin0739 and lin0803). The second group comprises surface protein-coding genes that are heterogeneously distributed among the different Listeria isolates and species (Table 4).
Distribution of cell surface proteins
For corroboration of the hybridization results, six of the surface protein genes specific for L. monocytogenes (inlA, inlB, inlE, inlG, inlH, and inlF) and two of the surface protein genes found within several or all Listeria species (lmo0550 and lmo1289) were amplified by PCR from one reference strain for each serovar and one for each species. The PCR amplifications confirmed the array results (data not shown).
Carbohydrate metabolism and PTSs.The distribution patterns of genes for 12 PTS permeases and of 14 genes encoding proteins predicted to be implicated in sugar metabolism and degradation were similar to that of surface protein-coding genes; all of these genes were highly conserved in lineage I strains and most were lacking in the other Listeria species. Except for two PTS genes (lmo2733 and lmo2782) and three carbohydrate metabolism genes (lmo2143, lmo2735, and lmo2781), all other genes in this group were missing from the L. monocytogenes serovar 4a strains.
Cell wall proteins—two subdivisions within teichoic acid biosynthesis genes.Despite the fact that the majority of genes grouped the 93 L. monocytogenes strains according to previously defined lineages which correlate mainly with flagellar antigen combinations (serovars), we identified 13 genes implicated in cell wall biosynthesis that divide the L. monocytogenes strains into two groups according to their somatic antigens (serogroup 4 and serogroups 1/2, 3, and 7). These genes code for teichoic acid biosynthesis proteins and were detected within the strains of serogroups 1/2, 3, and 7, but were absent from strains of serogroup 4. This finding is in agreement with previous studies that have identified two distinct structural types of teichoic acid within L. monocytogenes, for which the first type was found in strains of serogroups 1/2, 3, and 7 and the second was found in strains of serogroup 4 (15). This suggests that these genes may be implicated in the synthesis of the specific teichoic acid type. Nine of these genes, located within a 19-kb region of the L. monocytogenes chromosome (lmo1076-lmo1077, lmo1080 to lmo1084, lmo1088, and lmo1091), were also shared with the L. seeligeri strains, which are of serogroup 1/2, suggesting that L. seeligeri has a teichoic acid type similar to that of L. monocytogenes serogroup 1/2.
Similarly, one (lin1073) of two L. innocua genes that are implicated in teichoic acid biosynthesis was uniquely shared with the L. monocytogenes strains of serogroup 4 and L. welshimeri. This suggests an implication in the specificity of the cell wall type of serogroup 4 strains of L. monocytogenes, which is more closely related to that of serogroup 6 of L. innocua than to that of L. monocytogenes serogroup 1/2 strains (15).
Variable genomic regions and analysis of junction sequences.Several L. monocytogenes EGDe gene clusters were missing only from L. monocytogenes serovar 4a strains but were present in all other L. monocytogenes strains. Two regions were absent from L. monocytogenes serovar 4a and L. innocua but were present in L. ivanovii, L. seeligeri, and L. welshimeri. They seemed good candidates for use as evolutionary markers. To further analyze them, we sequenced the junction regions of the putative deletion sites from six isolates of L. innocua and L. monocytogenes serovar 4a.
Analysis of the region lmo2671-lmo2672 revealed the existence of three deletion events, two of which were located in the coding sequence of lmo2672, resulting in the deletion of two internal fragments, of 621 and 35 bp. The third deletion was 355 bp and was located downstream of the 5′ end of the coding sequence of lmo2671. All three junction sequences were identical, containing either an insertion of TTGCATT, an insertion of A, or no insertion (see supplemental material at http://www.pasteur.fr/recherche/unites/gmp/sitegmp/biodiversitylist.html ). From the analysis of the junction sequence of the region of lmo2771 to lmo2773, we obtained the same result for all the strains studied, with identical junction sites sequenced 38 bp downstream of the 3′ end of lmo2770. The junction site had an insertion of the sequence TTATTTAAG replacing genes lmo2771 to lmo2773 (http://www.pasteur.fr/recherche/unites/gmp/sitegmp/biodiversitylist.html ).
The third region investigated (lmo1030 to lmo1036) was absent from L. innocua and L. monocytogenes serovar 4a strains and was present in L. ivanovii. The analysis of the junction region identified a minor sequence variation (insertion of TCA in L. innocua and of AT in L. monocytogenes serovar 4a) at the deletion site (http://www.pasteur.fr/recherche/unites/gmp/sitegmp/biodiversitylist.html ).
Furthermore, the inlGHE cluster, which is missing from L. monocytogenes 4a and all other Listeria sp., was analyzed. We identified again an identical sequence for the five strains sequenced, suggesting that a single deletion event had occurred in a common ancestor of L. monocytogenes serovar 4a and L. innocua (http://www.pasteur.fr/recherche/unites/gmp/sitegmp/biodiversitylist.html ).
DISCUSSION
Overall genetic diversity.One of the first criteria by which the L. monocytogenes species could be subdivided was the varying antigenic properties of diverse subpopulations. A scheme developed by Seeliger and Höhne described 13 serovars (32). Based on somatic antigens, L. monocytogenes isolates were divided mainly into serogroups 1/2 and 4, and based on flagellar antigen combinations, each of these serogroups was subdivided into serovars 1/2a, -b, and -c and 4b or some other less common serovars (32). Today, this scheme is still in use, and numerous studies have identified correlations between certain phenotypic or genetic features and specific serovars (3, 5, 20, 28). The combined analysis of the genome sequences of L. monocytogenes serovar 1/2a and L. innocua (19), the partial genome sequence of L. monocytogenes serovar 4b, and the macroarray hybridization of 113 Listeria DNAs undertaken for this study substantiated this classification at the genomic level.
One of the most striking observations of this study was the magnitude of divergence that exists within the species L. monocytogenes. We found that the genetic divergence between lineage I and lineage II of L. monocytogenes was nearly equally important (about 8%) as the interspecies difference between the sequenced L. monocytogenes EGDe serovar 1/2a strain and L. innocua (10%). These results are in line with a previous report (22) which identified 39 specific gene fragments for the epidemic L. monocytogenes strain F.4565 compared to L. monocytogenes EGDe by use of a subtractive hybridization method. This is of particular importance since strains of serovar 4b mainly represent epidemic L. monocytogenes strains and are isolated from severe invasive human cases more frequently than strains of other serovars, such as serovar 1/2a. Apart from the important divergence between the two lineages of L. monocytogenes, the macroarray results identified remarkable genomic conservation within the major lineages and subgroups (Fig. 1) but variations between the different subgroups. These results seem to mirror the evolution within the genus Listeria.
Specific features.An important gene family in L. monocytogenes encodes surface proteins (8, 19). The macroarray hybridization and the analysis of the partial L. monocytogenes serovar 4b sequence indicated that a group of surface protein-encoding genes which includes all known internalin genes (inlA, inlB, inlG, inlH, inlE, inlC, and inlF) is highly specific for the species L. monocytogenes. Furthermore, each subgroup of L. monocytogenes is characterized by a specific set of surface proteins. Finally, a third group of surface protein-coding genes is distributed quite heterogeneously among the different Listeria species. Interestingly, in the rarely isolated L. monocytogenes serovar 4a strains, which are mostly animal pathogens, 13 of the 25 L. monocytogenes-specific surface proteins, including all internalins except inlAB, were missing. The lack of these proteins may be related to the lower disease potential of these strains for humans. The fact that different subgroups of L. monocytogenes strains contain different sets of surface proteins may also reflect their different potentials to cause disease or to multiply in different niches. The elucidation of the functions of the different surface proteins and the putative strain-specific characteristics that they confer will be one of the challenging questions of the future and may give additional insights into our understanding of the tropism of L. monocytogenes toward different cell types.
Proteins implicated in sugar transport and metabolism, in particular PTSs, form another important gene family in Listeria species (19). An analysis of the distribution of these genes again underscored the genetic divergence of the different subgroups of L. monocytogenes, as each lineage was characterized by a specific set of PTS permeases. Most PTS genes present in serovar 1/2a, 1/2c, 3a, and 3c strains were missing from the serovar 4b and 1/2b strains. The finding that the bvrABC locus, a β-glucoside-specific PTS system previously described as being implicated in virulence gene expression (4), was absent from all L. monocytogenes strains of lineage II (serovars 4b, 4d, 4e, 7, 1/2b, and 3b) was surprising. Because regulation of the PrfA regulon by β-glucosides also takes place in lineage II strains, it can be assumed that another PTS system fulfills the functions of the bvrABC proteins. The finding that one of the PTS permeases identified in the sequence of the L. monocytogenes serovar 4b strain was present in all strains in which the bvrABC locus was lacking might be consistent with this hypothesis.
Marker genes for L. monocytogenes and each lineage.We identified 30 markers for the species L. monocytogenes (Table 2), as well as markers of each subpopulation within the species L. monocytogenes (Table 3). One of our major questions was whether the pronounced differences in virulence among different subgroups of strains can be explained by different gene contents. Scanning of our results for the presence or absence of known virulence genes (inlAB, prfA, plcA, hly, mpl, actA, plcB, uhpT, and bsh) revealed that they are present in all L. monocytogenes strains tested. However, an analysis of the correlation between epidemiological data, the origins of the strains, and the genomic profiles clustered the L. monocytogenes serovar 4b strains isolated from epidemics and from incriminated food sources in a group separate from the other environmental, food, and animal isolates (http://www.pasteur.fr/recherche/unites/gmp/sitegmp/biodiversitylist.html ). Thus, disease-related L. monocytogenes isolates seem to be characterized by a particular combination of genes, and the Listeria array, combined with the knowledge of the marker genes identified in this study, should prove to be a powerful tool for identifying these strains (M. Doumith, C. Buchrieser, P. Glaser, C. Jacquet, and P. Martin, unpublished data). The oligonucleotides used to amplify marker genes are available online as supplemental material (http://www.pasteur.fr/recherche/unites/gmp/sitegmp/biodiversitylist.html ).
Evolutionary genomics of Listeria.As shown in Fig. 1, the combined use of bioinformatics and macroarray results for 113 Listeria strains generated a large data set, from which a detailed analysis allows us to group strains according to shared genetic profiles. In addition to conclusions concerning genetics, epidemiology, and the virulence of Listeria strains, these data also allow us to hypothesize how the different Listeria species and phylogenetic lineages may have evolved. Several studies, using analysis of 16S and 23S rRNA (11, 31), PCR-based DNA fingerprinting techniques (34), or virulence locus and genome comparisons (9, 19, 34), indicated a phylogenetically close relationship between L. monocytogenes and L. innocua and suggested that L. innocua lost the virulence locus by deletion. Most interestingly, we identified several other regions missing from L. monocytogenes serovar 4a strains which were also missing from L. innocua, such as the inlGHE gene cluster. Sequence analysis of the different junction regions identified identical sequences among L. monocytogenes serovar 4a and the L. innocua strains, suggesting single deletion events. The presence of these genes in the other Listeria species suggests that they were part of the genome of a common ancestor and that L. innocua evolved by successive gene loss from an ancestor of L. monocytogenes serogroup 4 strains. This hypothesis is also substantiated by the similar teichoic acid structures of L. monocytogenes serogroup 4 and L. innocua strains (15) and by the structural and functional similarity of the cell wall anchor of the autolysin Ami of L. monocytogenes serogroup 4 and L. innocua, but divergence between Ami of L. monocytogenes serovar 1/2 and that of serovar 4 (E. Milohanic, R. Jonquières, P. Glaser, P. Dehoux, C. Jacquet, P. Berche, P. Cossart, J.-L. Gaillard, submitted for publication). Further evidence for this close relationship also comes from the flagellar antigen structures of L. monocytogenes serogroup 4 and L. innocua, which are the same for these two groups but are different from that of serovar 1/2a and 1/2c strains (32). Based on our analysis and the literature, we suggest an alternative model of evolution within the L. monocytogenes-L. innocua branch. The separation into phylogenetic lineages is based on the divergence of serovar 1/2c and serovar 1/2b strains from a common ancestor (Fig. 2). Later in its evolution, the serovar 1/2b branch gained genes, such as gtcA, which conferred serogroup-specific expression of TA-associated serotype-specific antigens (29) and evolved into serogroup 4 and later on into the species L. innocua, mainly by successive gene loss.
Evolutionary scheme of the different lineages and serovars of L. monocytogenes. The scheme is derived from a hypothesis based on the presence and absence of genes and their correlation with antigenic characteristics.
Conclusions.Taken together, the results of DNA-DNA hybridization of a specific Listeria array containing genes of three different Listeria isolates showed that L. monocytogenes strains differ substantially in gene content. These differences are most pronounced in surface proteins and proteins for sugar metabolism, which are most likely to confer traits that provide selective advantages in the environment and the infected host. Our results further provide an explanation for why previous studies have found an association between various characteristics of L. monocytogenes and serovars. We showed that this association is due to evolutionary differentiation. To date, the microbiological surveillance of listeriosis, a disease that causes the death of at least 400 to 500 persons per year in Europe and North America, is based on subtyping by serotyping and pulsed-field gel electrophoresis (21). The precise characterization of L. monocytogenes is essential for following long-term trends in sporadic cases as well as for the detection of clusters of cases and epidemics and the identification of their common source. As such, the selective markers for the different subpopulations are an essential contribution for the construction of rapid, accurate identification and subtyping methods, which should be powerful tools applicable in health institutions and the food industry. Finally, the identification of genes that are consistently absent or present in epidemic-associated L. monocytogenes strains now opens the way for mutational and functional analysis of these genes in order to decipher the molecular basis for the increased pathogenic potential of certain L. monocytogenes strains.
ACKNOWLEDGMENTS
This work received financial support from the European Commission contract REALIS (QLG2-CT-1999-00932) and the Pasteur Institut (PTR 6 and GPH9). M. Doumith holds a postdoctoral position from the Pasteur Institut (PTR 6).
We thank SOREDAB for providing L. monocytogenes strains, Rachel Purcell and Elisabeth Couvé for technical assistance, Claude Parsot for critical reading of the manuscript, and Roland Brosch for fruitful discussions and ideas. P. Cossart is an international scholar from the Howards Hughes Medical Institute.
FOOTNOTES
- Received 12 September 2003.
- Returned for modification 20 October 2003.
- Accepted 22 October 2003.
- Copyright © 2004 American Society for Microbiology
