Yersinia massiliensis sp. nov., isolated from fresh water

At the time of writing, the genus Yersinia consists of 12 species, including Yersinia pestis (the plague agent), Yersinia enterocolitica and Yersinia pseudotuberculosis (the two species responsible for a complex clinical condition known as yersiniosis) (Perry & Fetherston, 1997; Bottone, 1999; Carniel & Mollaret, 1990), and nine Y. enterocolitica-like species, i.e. Yersinia frederiksenii, Yersinia intermedia, Yersinia kristensenii, Yersinia bercovieri, Yersinia mollaretii, Yersinia rohdei, Yersinia ruckeri, Yersinia aldovae (Sulakvelidze, 2000) and Yersinia aleksiciae (Sprague & Neubauer, 2005). Except for Y. pestis, Yersinia species are ubiquitous and have been found in soil, fresh water sources, animals, foods and sewage (Bercovier & Mollaret, 1984; Bottone, 1997; Sulakvelidze, 2000). They have been occasionally isolated from clinical specimens (Gray, 1995) and virulence factors have been detected in several isolates of these species (Robins-Browne et al., 1991; Sulakvelidze et al., 1999; Sulakvelidze, 2000). Herein we have characterized two freshwater Yersinia isolates by extensive phenotypic and molecular methods and found them to exhibit unique characteristics, suggesting that they are representative of a hitherto undescribed Yersinia species.

Isolate 50640^T was isolated in a dialysis unit from the hospital water distribution system in Marseilles, France, and isolate 823 came from a well in another part of southern France. Water specimens were filtered through a 0.45 μm nitrocellulose filter membrane (Millipore) and incubated at 30 °C for 5 days on cefsulodin-irgasan-novobiocin agar (Becton Dickinson) (Schiemann, 1979). Strains Y. mollaretii CCUG 26331^T, Y. bercovieri CCUG 26329^T, Y. frederiksenii CCUG 11293^T, Y. aldovae CCUG 18770^T, Y. intermedia CCUG 11292^T, Y. kristensenii CCUG 11294^T, Y. rohdei CCUG 38833^T, Y. ruckeri CCUG 14190^T, Y. enterocolitica CCUG 11291^T, Y. enterocolitica CCUG 12369^T, Y. enterocolitica CCUG 7758 and Y. enterocolitica CCUG 8239A were obtained from the Culture Collection, University of Göteborg, Sweden, and investigated in parallel. Motility was checked by microscopic observation. Morphological properties were determined by Gram staining. The size and ultrastructure of the cells were determined by transmission electron microscopy. Colonies were harvested from trypticase soy medium (bioMérieux), pelleted by centrifugation and prefixed for 1 h at room temperature in 2.5 % glutaraldehyde in 0.1 M cacodylate buffer (pH 7.2) containing 0.1 M sucrose. After washing overnight with the same buffer, the bacteria were fixed for 1 h at room temperature with 1 % osmium tetroxide in 0.1 M cacodylate buffer, dehydrated through increasing concentrations (25–100 %) of ethanol, and then embedded in Epon 812. Thin sections were cut and post-stained with a saturated solution of methanol uranyl acetate and lead citrate in water before examination on a Morgagni M 268D electron microscope (FEI France) at an operating voltage of 60 kV. Oxidase activity was detected by using a dimethyl-p-phenylenediamine oxalate disk (Sanofi Diagnostic Pasteur). Catalase activity was detected by emulsifying a colony in 3 % hydrogen peroxide and checking for the presence of microscopic bubbles. Mannitol-fermenting colonies were identified by using an API 20E strip (bioMérieux), a commercial identification system for Enterobacteriaceae (Neubauer et al., 1998), according to the instructions of the manufacturer with the exception of incubation at 28 °C (Sharma et al., 1990). API 20E strips were read after 24, 48 and 72 h incubation under a highly humidified atmosphere, and tentative identifications were obtained using the Analytical Profile Index, fourth edition (Archer et al., 1987; Sharma et al., 1990). API Coryne, Api NH and API 20A (bioMérieux) were used for further determination of pyrazinamidase, lipase, aesculin and salicin. Isolates were serotyped by slide agglutination with antisera against Y. enterocolitica O : 3, O : 8 and O : 9 (Denka Seiken). Every test was done in triplicate on three separate days to ensure the reproducibility of the results.

After 24 h culture on MacConkey agar, DNA extraction, purification, degradation and G+C content determination by HPLC were performed as described by Mesbah et al. (1989), except that a Waters 625 LC system with a Waters 486 Tenable Absorbance Detector and a Water 746 Data Module (Millipore) were used. Three determinations were done. DNA extracted from colonies by using Qiagen columns (QIAamp tissue kit) was used as a template for PCR amplification and sequencing of the 16S rRNA gene using primers fD1/rp2 (Weisburg et al., 1991). Four additional housekeeping genes, including rpoB, hsp60, gyrB and sodA, were amplified and sequenced. rpoB gene amplification and sequencing was performed as described by Mollet et al. (1997). Because of limited sequence availability of the remaining three genes for some of the Yersinia species in GenBank, a 683 bp fragment of the hsp60 gene was amplified and sequenced using primers hsp60-498F (5′-GGAAAAAGTCGGTAAAGAAGGCG-3′) and hsp60-1181R (5′-GCTTTCTTCTCTTTCATTTCAACTTC-3′) designed after alignment of the hsp60 gene of Y. enterocolitica Y-108c (GenBank accession no. X68526), Y. pseudotuberculosis IP32953 (NC006155, locus_tag YPTB0405) and Y. pestis CO92 (NC003143, locus_tag YPO0351). A 662 bp gyrB gene fragment was amplified and sequenced by using primers gyrB-303F (5′-CGGTAAATTTGACGATAACTCC-3′) and gyrB-965R (5′-ACAGCAATCAAGCCTTCACGGG-3′) designed after alignment of the gyrB gene of Y. enterocolitica ATCC 2610 (AB084022), Y. pseudotuberculosis IP32953 (NC006155, locus_tag YPTB3940) and Y. pestis CO92 (NC003143, locus_tag YPO4094). A 624 bp sodA gene fragment was amplified and sequenced by using the primer pair sodA-1F (5′-ATGAGTTAYTCACTGCCAT-3′) and sodA-624R (5′-TTACTTRGCTTGAGCGAA-3′) designed after alignment of the sodA gene of Y. enterocolitica WA-314 (X96852), Y. pseudotuberculosis (NC006155, locus_tag YPTB3925) and Y. pestis CO92 (NC003143, locus_tag YPO4061). Products of sequencing reactions were recorded with an ABI Prism 3130 xl DNA sequencer following the standard protocol of the supplier (Perkin Elmer Applied Biosystems). The percentage similarity between the sequences was determined using the clustal w program supported by the PBIL website (). For phylogenetic analysis, sequences were trimmed in order to start and finish at the same nucleotide position for all the strains under study. Multisequence alignment was performed by using the clustal_x program, version 1.81 in the phylip software package (Thompson et al., 1994). Phylogenetic trees were obtained from DNA sequences by using the neighbour-joining method with Kimura's two-parameter distance correction model with 1,000 bootstrap replications in the mega version 3.1 software package (Kumar et al., 2004). Phylogenetic trees based upon each gene sequence and one based on the combined gene sequences were then compared using the neighbour-joining and maximum-parsimony methods.

Isolates 50640^T and 823 were shown to be Gram-negative bacilli and were motile at 28 °C. The rods were 2.23 μm long and 0.85 μm wide, with one to two flagella of 6.75 μm (Fig. 1⇓). Biochemical characteristics that differentiate isolates 50640^T and 823 from closely related Yersinia species are summarized in Table 1⇓. Both isolates were catalase-positive and oxidase-negative. An API 20E strip inoculated with isolate 50640^T or 823 yielded a biochemical pattern indistinguishable from that of Y. enterocolitica coded as 1154723 in the API 20E system. Parallel inoculation of Y. enterocolitica CCUG 12369^T, Y. frederiksenii CCUG 11293^T and Y. pseudotuberculosis IP 32953 yielded codes of 1154723, 1354733 and 1014113, respectively, allowing very good identification of these species. However, the codes 1114523 for Y. bercovieri and 1114523 for Y. mollaretii identified these strains as Y. enterocolitica according to the API 20E coding system. Isolates 50640^T and 823 differed from Y. mollaretii CCUG 26331^T and Y. bercovieri CCUG 26329^T by exhibiting positive indole and inositol tests, and from Y. frederiksenii CCUG 11293^T by the lack of l-rhamnose fermentation and citrate utilization. After 48–72 h incubation, isolates 50640^T and 823 exhibited citrate utilization and weak acetoin production, resulting in a profile number of 1354723 that identified them as Y. enterocolitica (52 % confidence) or Y. frederiksenii (48 % confidence). The two isolates were positive for salicin, aesculin and pyrazinamidase, but did not react with any of the tested Yersinia typing sera.

The DNA G+C content was 48.7±0.6 mol%. The 1461 bp 16S rRNA gene sequences of isolates 50640^T and 823 were identical and shared 98.7 % similarity with Y. bercovieri, 98.5 % with Y. mollaretii and 98.4 % with Y. frederiksenii corresponding to 17, 20 and 21 bp differences, respectively (Table 2). The 16S rRNA gene sequence did not discriminate Y. enterocolitica CCUG 11291^T, Y. enterocolitica CCUG 12369^T and Y. enterocolitica CCUG 7758. As for the hsp60, gyrB, sodA and rpoB gene sequences, isolates 50640^T and 823 shared 93.5, 90.4, 92.4 and 96.6 % similarity with the respective homologous sequences of Y. mollaretii, 93.0, 90.1, 89.1 and 96.3 % with Y. bercovieri and 93.2, 89.8, 88.9 and 95.3 % with Y. frederiksenii (Table 2⇓). Comparison of sequences derived from isolates 50640^T and 823 yielded one point mutation in gyrB, two in sodA, four in hsp60 and five in rpoB.

Phylogenetic trees were constructed based upon each gene sequence, 16S rRNA (1461 bp), hsp60 (635 bp), gyrB (619 bp), sodA (624 bp) and rpoB (1049 bp), and by combining the four housekeeping gene sequences (about 3000 nt) determined for 11 Yersinia species in addition to isolates 50640^T and 823 (Figs 2⇓ and 3⇓). Bootstrap values >75 % in the 16S rRNA gene tree occurred for only 7/12 (58.3 %) nodes, too low to provide much confidence (Fig. 2⇓). Likewise, hsp60-, gyrB-, sodA- and rpoB-based phylogenetic trees showed low bootstrap values with only 7/13 (51.8 %), 5/12 (41.7 %), 4/13 (30.8 %) and 8/13 (61.5 %) nodes >75 %, respectively (see supplementary Figs S1, S2, S3 and S4 available with the online version of this paper). In contrast, the combination of the four housekeeping gene sequences improved the bootstrap values with 12/13 (92.3 %) nodes >75 % (P<0.05) (Fig. 3⇓). This analysis also revealed that isolates 50640^T and 823 formed a unique line within the genus Yersinia. A bootstrap value of 93 % in the neighbour-joining tree supported the fork separating isolates 50640^T and 823 from Y. mollaretii and Y. bercovieri, giving a high level of confidence to this analysis (Fig. 3⇓). Parsimony and maximum-likelihood methods confirmed that isolates 50640^T and 823 formed a lineage that was clearly different from closely related species and quite distant from other recognized Yersinia species. In most, but not all cases, the four housekeeping genes and the 16S rRNA gene sequence trees correlated well with each other, and agreed with a separate species status for isolates 50640^T and 823 (Figs 2⇓ and 3⇓).

Two isolates, 50640^T and 823, of biochemically atypical Yersinia were recovered from haemodialysis unit water and well water. An API 20E strip misidentified these isolates as Y. enterocolitica. However, the two isolates were positive for salicin, aesculin, pyrazinamidase, lipase, xylose, trehalose, indole and the Voges–Proskauer test, a pattern in agreement with the recent description of some Y. enterocolitica-like species (Hallanvuo et al., 2006). We observed a 1.3 % divergence in the 16S rRNA gene sequence of isolates 50640^T and 823 with that of Y. bercovieri, 1.5 % with that of Y. mollaretii and 1.6 % with that of Y. frederiksenii. These levels were below the 1.3 % threshold proposed by Stackebrandt & Ebers (2006) to delineate bacterial species by using 16S rRNA gene sequencing, thus firmly suggesting that the isolates herein studied belong to a new species in agreement with their unique phenotypic profile.

We further characterized isolates 50640^T and 823 by sequencing four housekeeping genes to fulfil the recommendations of the ad-hoc committee for the re-evaluation of the definition of the bacterial species (Stackebrandt et al., 2002) and as recently suggested for Yersinia isolates speciation (Kotetishvili et al., 2005). These four housekeeping genes are spaced well apart on the genome and could potentially reflect the evolution of the whole genome (Thomson et al., 2006). The topologies of the four housekeeping-based phylogenetic trees and the 16S rRNA gene-based phylogenetic tree were quite similar. However, the four housekeeping genes were more consistent with the biochemical species designation than the 16S rRNA gene analysis, i.e. Y. mollaretii and Y. bercovieri, which exhibited a similar biochemical profile, were in the same branch in the four housekeeping gene trees (Fig. 3⇑) whereas they were in separate branches in the 16S rRNA gene tree (Fig. 2⇑). The four housekeeping genes yielded bootstrap values higher than those derived from the 16S rRNA gene. These data indicate that the 16S rRNA gene sequence alone cannot resolve the phylogenetic relationships between all currently recognized Yersinia species.

By combining the four housekeeping genes studied herein, we found that Y. mollaretii and Y. bercovieri on one hand, and Y. aldovae and Y. intermedia on the other were closely related. In the rpoB, gyrB, hsp60 and sodA gene sequences, similarity levels between Y. mollaretii and Y. bercovieri were 97.9, 92.5, 94.9 and 91.3 %, and 96.9, 91.7, 93.5 and 92.9 % between Y. aldovae and Y. intermedia, respectively (Table 2). Isolates 50640^T and 823 showed less similarity with respect to the four housekeeping genes with all other Yersinia species (Table 2). Although there are no validated cut-off values for Yersinia species delineation, we observed that similarity values of <98.0 % for rpoB, <92.5 % for gyrB, <95.0 % for hsp60 and <93.0 % for sodA effectively delineated the currently recognized species in the Yersinia genus. These values supported previous DNA–DNA hybridization data for the genus (Bottone, 1997). As for the rpoB gene, these cut-off values agree with the similarity value (97.6 % over 1049 bp) that delineates two other enteric species, Klebsiella variicola and Klebsiella pneumoniae; these species share 99.3 % 16S rRNA gene sequence similarity (Rosenblueth et al., 2004). The same strategy, based on this short rpoB gene portion, has been used to delineate Klebsiella singaporensis from K. pneumoniae, two species which share 99.3 % 16S rRNA gene sequence similarity and 97.5 % rpoB gene sequence similarity (Li et al., 2004). Based on a 512 bp portion of the rpoB gene sequence reported for Enterobacteriaceae species (Mollet et al., 1997), these authors found 97.7–98.8 % similarity among representative species of three genetic groups, i.e. Klebsiella, Escherichia and Salmonella. We found that the interspecies similarity in this region for the 11 Yersinia species examined in this study varied from 93.5 % (Y. pestis and Y. enterocolitica) to 98.8 % (Y. kristensenii and Y. enterocolitica). Also, our data showed that isolates 50640^T and 823 shared 96.6 % similarity with Y. mollaretii, 96.2 % with Y. bercovieri, and 95.2 % with Y. frederiksenii. This analysis further supported that isolates 50640^T and 823 formed a novel species within the genus Yersinia.

Comparative 16S rRNA gene sequence analysis revealed that isolates 50640^T and 823 shared 100 % sequence similarity with non-typable Yersinia isolate 11A recovered from raw milk samples in Queensland, Australia (Ibrahim et al., 1997). The dendrogram of relationships based on the 16S rRNA gene sequence indicated that the atypical isolate 11A formed an individual line of descent distinct from Y. frederiksenii (Ibrahim et al., 1997). Although 16S rRNA gene sequence identity does not necessarily indicate close relatedness, this Yersinia isolate also exhibited a similar biochemical phenotype to isolates 50640^T and 823. It yielded only 41 % DNA–DNA relatedness to the Y. frederiksenii type strain (Ibrahim et al., 1997) and has been tentatively assigned to Y. frederiksenii hybridization group 2 (unnamed genospecies 2). Isolate 11A, however, was not available for this study and the question of its taxonomic status remains open, pending further molecular comparisons with isolates 50640^T and 823.

Original phenotypic traits and molecular analysis, including sequence analyses of the 16S rRNA, hsp60, sodA, gyrB and rpoB genes, identified isolates 50640^T and 823 as belonging to a novel Yersinia species, for which the name Yersinia massiliensis is proposed.