Multilocus Sequence Analysis of Close Relatives Vibrio anguillarum and Vibrio ordalii

Vibriosis, i.e., systemic infection with either Vibrio anguillarum or Vibrio ordalii (1), probably constitutes the most significant bacterial disease of wild and farmed fish in temperate marine environments worldwide. While 23 different serotypes of V. anguillarum have been identified (2), by far the majority of fish-pathogenic isolates belong to serotypes O1 and O2. In addition, strains not corresponding to known serotypes are commonly identified, mainly from environmental sources (3). Despite considerable phenotypic differences, V. ordalii and V. anguillarum are genetically very closely related (4); V. ordalii was initially characterized as V. anguillarum biotype 2, prior to its description as an independent species (1).

Vibriosis was previously a significant problem in salmonid farming in the North Atlantic area, but its impact in this industry has been drastically reduced due to the development of effective vaccination procedures (3). Despite vaccination, vibriosis caused by V. anguillarum and, to a lesser degree, V. ordalii remains a major problem in the relatively novel marine fish (e.g., cod and cleaner fish) farming industries in Norway (5), Ireland (6), and Scotland (7). The disease also remains a significant problem in farmed fish in Europe and Asia (8). In order to facilitate vaccine development and better understanding of the epidemiology of vibriosis, more information on the population structure is needed. Despite a considerable number of studies focusing on phenotypic (9, 10), serotypic (11–14), and genetic (15) variations among strains of fish-pathogenic V. anguillarum, few studies have performed detailed taxonomic analyses of both V. ordalii and V. anguillarum. None of the previous genus-wide studies addressing the phylogeny of vibrios (16–20) included significant numbers of V. anguillarum/V. ordalii isolates.

Multilocus sequence analysis (MLSA), relying on essentially unequivocal sequence information from several genetic loci, addresses most of the shortfalls associated with PCR-based DNA fingerprinting or single-gene phylogenetic studies. MLSA has the advantage of being portable, reproducible, and taxonomically sound, allowing the precise identification and classification of Vibrio isolates. Therefore, the main aim of the current study was to perform MLSA with a large number of V. anguillarum and V. ordalii isolates representing different geographical locations and fish species, to improve our knowledge of the population structure within and between these closely related fish-pathogenic species.

Partial sequences were obtained for eight protein-encoding genes (atpA, ftsZ, gapA, gyrB, mreB, topA, rpoA, and pyrH) from 103 V. anguillarum isolates and 13 isolates previously identified as V. ordalii. A codon-based Z-test of selection indicated that all genes were under purifying selection and thereby suitable for MLSA (Table 3). The Pacific (American and Chilean) V. ordalii isolates, all isolated from fish, clustered separately and apart from the main V. anguillarum clade in the maximum likelihood (ML) phylogenetic tree reconstruction (Fig. 1, far right cluster in red oval). While the Norwegian fish-pathogenic isolates putatively identified as Vibrio ordalii appeared to represent an earlier branch of the same lineage (Fig. 1, left cluster in red oval), the Pacific environmental (putative) V. ordalii isolates were spread around the main V. anguillarum cluster (Fig. 1, green oval), with the exception of FF167, which was closer to V. anguillarum but was placed basally within the fish-pathogenic V. ordalii lineage (Fig. 1, rightmost blue oval). Obvious associations between the host species and the major V. anguillarum serotype clusters could be seen. The serotype O1 clade (Fig. 1, yellow dot) was dominated by Norwegian, Danish, and American salmon and trout isolates but also included isolates from Spain (turbot), Japan (ayu), and Australia (rainbow trout), as well as the well-characterized V. anguillarum 775 (4, 33). The serotype O2b clade (Fig. 1, blue dot) was the most uniform; all 25 isolates originated from gadoid fish from Norway, Scotland, or Ireland, with 23 of them appearing clonal. The O2a/O2aII/O2b/O2c clade (Fig. 1, red dot) consisted of isolates from Norway, Denmark, and the United Kingdom that were identical in all eight sequenced loci. This clade also contained the full-genome-sequenced isolate RV22 (described as serotype O2b) from turbot (Scophthalmus maximus) (from Spain), a single non-O1/O2a/O2b Turkish isolate, and a single O2a biotype II isolate (defined as such due to the ability to produce lysine decarboxylase). The last of the four major clusters, i.e., the O2/O2a biotype II clade, included 10 somewhat diverse isolates, including a single serotype O2a isolate from Norwegian salmon.

The MLSA of the 116 isolates identified 44 HTs. eBURST clustering of the 44 HTs (with stringent group definition, i.e., 7 of 8 shared allele types) divided them into 5 CCs and 27 singletons (see Fig. S1 in the supplemental material). Seven singletons (HT1, HT3, HT14, HT22, HT30, HT31, and HT41) were represented by more than one isolate, with the largest (HT3) containing 18 isolates. The majority of the isolates (11 isolates) within HT3 originated from Norwegian turbot, salmonids, and gadoids, while some of the remaining 7 isolates originated from European seabass, brown trout, and a sediment sample (Turkey, United Kingdom, and Denmark). Three of the five identified CCs contained more than 10 isolates. The predominant clonal complex, CC-HT4, included 7 HTs, consisting of 22 isolates, with the predicted founder being HT4. Eight of the 16 isolates in HT4 were isolated from salmonids (Norway, Denmark, Sweden, and the United States), while the remainder were isolated from other fish species (not including Gadidae) in other countries (Turkey, Spain, Japan, Australia, and China). The second major clonal complex, CC-HT2, included 3 HTs and 25 isolates, with HT2 being its predicted founder. In comparison with CC-HT4, CC-HT2 was a much more homogeneous group, with all isolates originating from Norwegian, Irish, and Scottish Gadidae. The third clonal complex, CC-HT28, was made up of 3 HTs containing 10 isolates, with HT29 as the predicted founder. This group was also homogeneous, with 9 of 10 isolates originating from Norwegian gadoids and a single isolate originating from Norwegian Atlantic salmon. The two minor groups, CC-HT20 and CC-HT24, included only 2 HTs each, with 2 or 3 isolates, respectively, and neither had a predicted founder.

The genetic diversity (H) in the total data set (5,208 bp from each of 116 isolates) was estimated to be 0.767 ± 0.026 (Table 4). The standardized index of association (I^S_A) for the whole population was estimated to be 0.506, which differed significantly from zero (P < 0.05); therefore, the null hypothesis of linkage equilibrium was rejected. When overrepresentation of particular haplotypes was avoided by restricting the analysis to single representatives of the 44 HTs, the I^S_A value decreased to 0.281 (P < 0.05). When the analysis was further limited to include only CC founders plus all singletons (27 isolates), the I^S_A value was 0.097, closer to linkage equilibrium, but it remained significantly different from zero (P < 0.05).

When results were sorted by geographical origin, the greatest variation in haplotypes per number of isolates, i.e., 12 HTs/20 isolates (H = 0.815 ± 0.022), was found in the Pacific group (United States, Chile, Japan, and China) (Table 4). When results were sorted by isolation source, the greatest variations in haplotypes per number of isolates, i.e., 9 HTs/11 isolates (H = 0.909 ± 0.030), were found among the environmental isolates (including seawater, sediment, and rotifers) and in the nongadoid nonsalmonid group, with 11 HTs/19 isolates (H = 0.551 ± 0.067).

The haplotype network construction revealed that fish disease-related isolates of Vibrio ordalii from Chile and the United States and the isolates from Norway that shared several phenotypic traits with V. ordalii represented distinct populations (HT24, HT25, HT30, and HT31) (Fig. 2). Furthermore, the five environmental isolates from the United States previously considered to represent V. ordalii (HT41, HT42, HT43, and HT44) represented disparate isolates. V. anguillarum isolates from the United States (HT4, HT39, and HT40) grouped with isolates from other countries, including five closely related isolates from Turkey (HT15, HT16, HT17, HT18, and HT19). Three isolates from Greece also constituted a distinct group (HT20 and HT21).

The PHI test in SplitsTree4 identified statistically significant evidence of recombination among isolates included in the present study. The reticulate structure (see Fig. S2 in the supplemental material) is consistent with such recombination events in the evolutionary history of all studied isolates.

The intergenomic distances and DNA-DNA hybridization (DDH) estimates between V. anguillarum 775 and other full-genome-sequenced V. anguillarum strains (four strains) and putative V. ordalii strains, as well as the V. ordalii type strain (six strains), were calculated (Table 5). Of the V. anguillarum strains studied, RV2 displayed the greatest intergenomic distance to V. anguillarum 775 (distance, 0.0171). V. ordalii NCIMB 2167^T (clinical isolate) showed the greatest intergenomic distance to V. anguillarum 775 (distance, 0.0428). The intergenomic distances to the five environmental putative V. ordalii isolates were less (distances, 0.0136 to 0.0234) and similar to the intergenomic distances to bona fide V. anguillarum strains (distances, 0.0012 to 0.0171). The calculated DDH value for V. anguillarum 775 versus V. ordalii NCIMB 2167^T was 65.4 ± 2.9% (species delineation norms are defined as DDH values of >70%) (34). It is noteworthy that the calculated DDH values for the only environmental putative V. ordalii isolate, FF167, which groups with clinical isolates from fish, versus V. ordalii NCIMB 2167^T and V. anguillarum 775 were 69.3 ± 3.0% and 80.0 ± 2.8%, respectively.

The present study represents a multigene analysis of a substantial number of V. anguillarum and V. ordalii isolates. Examination of phylogenetic trees and networks (Fig. 1 and 2; also see Fig. S2 in the supplemental material) based on concatenated sequences revealed a high degree of sequence conservation among pathogenic V. anguillarum serotype O1 isolates from around the world. Norwegian V. anguillarum isolates showed relatively great diversity (13 haplotypes), but Danish isolates showed the greatest diversity of the study, compared to the number of isolates (12 haplotypes) (Fig. 2). Interestingly, most of the Turkish (non-O1/O2a/O2b) isolates were either identical or highly similar in sequence to serotype O1 isolates. Only one distinct serotype/haplotype-host relationship was identified, i.e., serotype O2b was isolated exclusively from gadoid fish from diverse geographical origins in the North Atlantic area. Twenty-three of the 25 serotype O2b isolates that were isolated from gadoids between 1979 and 2008 possessed identical sequences for all eight loci. Our results showed that many V. anguillarum isolates collected internationally grouped together in three main clonal complexes (see Fig. S1 in the supplemental material). The distribution of haplotypes versus serotypes (see Fig. S2) is consistent with relatively frequent horizontal gene transfer within this group of bacteria. For example, HT3 isolates occurred in four different O2 serotypes (O2a, O2aII, O2b, and O2c), while the O2a serotype occurred in three more distantly related haplotypes (HT1, HT26, and HT38), in addition to HT3. According to Jedani et al. (35), V. anguillarum serotype O2 isolates have several intact insertion sequences that may have the potential to move or shuffle genes, thus creating new O-antigen/capsule specificities within Vibrionaceae.

MLSA-based DNA similarities for all eight genes in all isolates were between 97.5 and 99.0%, with the exception of ftsZ (≥95%), and were well above the 95% threshold for Vibrio species differentiation proposed by Sawabe et al. (36). However, the calculated DDH value for V. ordalii NCIMB 2167^T versus V. anguillarum 775 of 65.4%, which agrees well with the original DDH estimates of 53% to 67% reported by Schiewe et al. and Thompson et al. (1, 37), clearly suggests that these two strains represent separate bacterial species. The GGDC 2.0 software calculations are independent of genome size, provide results directly comparable to wet-laboratory DDH results, and have been shown to correlate better with DDH results than the more commonly used average nucleotide identity (ANI) method (34).

Naka et al. (4) identified a significantly smaller genome in V. ordalii NCIMB 2167^T, compared to V. anguillarum 775, which suggests that physical and ecological differences exist between these two species/strains and that fish-pathogenic V. ordalii may be evolving toward an endosymbiotic lifestyle. The sequences retrieved from public databases for five putative V. ordalii isolates originating from seawater revealed intermediate-size genomes, genetically considerably closer (by MLSA and DDH analysis) to V. anguillarum 775 than the fish-pathogenic V. ordalii isolates. It is not clear from the available literature whether these environmental isolates are phenotypically consistent with V. ordalii (38, 39). With the exception of isolate FF167, the isolates show no relationship to the V. ordalii reference strain but are spread inside the V. anguillarum clade (Fig. 1), which clearly indicates that these isolates represent V. anguillarum. The precise taxonomic situation for the environmental isolate FF167 and the Atlantic cod pathogens NVI 5286 and NVI 5918 is unclear, as the latter two isolates display several phenotypic traits consistent with V. ordalii and appear very closely related to V. anguillarum by ML analysis but are situated at the base of the V. ordalii lineage. Intergenomic distances and in silico DDH values of 80.0% and 69.3% for FF167 versus V. anguillarum 775 and V. ordalii NCIMB 2167^T, respectively, indicate that putative V. ordalii strain FF167 is closer to V. anguillarum 775 than V. ordalii NCIMB 2167^T. It should be noted, however, that the distance to V. anguillarum type strain NCIMB 6 could not be calculated, as there is no full-genome sequence available for that isolate. Interestingly, FF167 has the smallest genome of the five environmental isolates and is also the only one to possess a full complement of biofilm-associated syp genes (data not shown), which were previously suggested to be species specific and present in V. ordalii NCIMB 2167^T but not in V. anguillarum 775 (4). These genes may represent genetic markers for phenotypic and epidemiological separation of V. ordalii linage members from V. anguillarum isolates.

In conclusion, the described MLSA largely resolved V. anguillarum isolates according to serotype. V. anguillarum serotype O1 and O2b isolates associated with disease in fish displayed a high degree of sequence conservation, as expected. Taken together, the phylogenetic relationships revealed by MLSA, DNA similarity values, and DDH estimates do not justify incorporation of V. ordalii within the species V. anguillarum. Further study of larger numbers of North Atlantic fish-pathogenic isolates that share several phenotypic traits with V. ordalii are warranted.

We thank Dawn Austin (Heriot-Watt University, Edinburgh, Scotland) and Harry Birkbeck (University of Glasgow, Scotland) for the donation of isolates.

The research was financially supported by the Norwegian Research Council (grant 184634). We are grateful for the support of Istanbul University, which financed a research period in Norway for S.K. P.M.M. and F.L.T. thank CNPq, CAPES, and FAPERJ for support.