Tracing Genomic Divergence of Vibrio Bacteria in the Harveyi Clade

To investigate the mechanisms underlying speciation in the genus Vibrio, we provided a well-assembled reference of genomes and performed systematic genomic comparisons among three evolutionarily closely related species. We resolved taxonomic ambiguities and identified genomic features separating the three species. Based on the study results, we propose a hypothesis explaining how species in the Harveyi clade of Vibrio bacteria diversified.

ABSTRACT

The mechanism of bacterial speciation remains a topic of tremendous interest. To understand the ecological and evolutionary mechanisms of speciation in Vibrio bacteria, we analyzed the genomic dissimilarities between three closely related species in the so-called Harveyi clade of the genus Vibrio, V. campbellii, V. jasicida, and V. hyugaensis. The analysis focused on strains isolated from diverse geographic locations over a long period of time. The results of phylogenetic analyses and calculations of average nucleotide identity (ANI) supported the classification of V. jasicida and V. hyugaensis into two species. These analyses also identified two well-supported clades in V. campbellii; however, strains from both clades were classified as members of the same species. Comparative analyses of the complete genome sequences of representative strains from the three species identified higher syntenic coverage between genomes of V. jasicida and V. hyugaensis than that between the genomes from the two V. campbellii clades. The results from comparative analyses of gene content between bacteria from the three species did not support the hypothesis that gene gain and/or loss contributed to their speciation. We also did not find support for the hypothesis that ecological diversification toward associations with marine animals contributed to the speciation of V. jasicida and V. hyugaensis. Overall, based on the results obtained in this study, we propose that speciation in Harveyi clade species is a result of stochastic diversification of local populations, which was influenced by multiple evolutionary processes, followed by extinction events.

IMPORTANCE To investigate the mechanisms underlying speciation in the genus Vibrio, we provided a well-assembled reference of genomes and performed systematic genomic comparisons among three evolutionarily closely related species. We resolved taxonomic ambiguities and identified genomic features separating the three species. Based on the study results, we propose a hypothesis explaining how species in the Harveyi clade of Vibrio bacteria diversified.

INTRODUCTION

The genus Vibrio (Bacteria, Gammaproteobacteria, family Vibrionaceae) has over 110 species with valid standings in the nomenclature (1, 2). Studies of bacteria in the genus are mostly driven by the presence of important human and marine animal pathogens (3). Studies of evolutionary processes within the genus revealed great diversity between species, frequent inter- and intraspecies gene transfer, and high recombination rates (2, 4, 5). However, important questions about Vibrio evolution remain unanswered, including how Vibrio species originated from within clusters of evolutionarily closely related strains.

Vibrio spp. were shown to undergo ecological differentiation, which could have triggered speciation in the bacteria (6, 7). The results from these studies were used to propose the hypothesis that speciation in the genus is propelled by ecological differentiation (8). This hypothesis so far have not been incorporated into the genus' taxonomy, which relies on a polyphasic species concept that uses results of phenotypic, genotypic, and phylogenetic analyses to delineate Vibrio strains into species (2). It is currently not clear how the two approaches to delineate Vibrio species, the ecotype concept and the taxonomic classification using polyphasic analysis, relate to each other and which one can be used to accurately describe the diversity of bacteria in the genus.

Our previous studies focused on characterizing the diversity of Vibrio bacteria in the so-called Harveyi clade. This clade consists of 10 evolutionarily closely related species with validly described names (2, 9). Bacteria in the clade are difficult to classify using polyphasic taxonomy and require the use of genomic data to accurately classify strains into species (9–12). Studies of Harveyi clade Vibrio bacteria allow a good opportunity to compare the delineation of bacteria based on their ecology with the classification using polyphasic taxonomy. Bacteria in the Harveyi clade are common in the coastal marine environments and are easily isolated and grown in laboratory media. As a result, representative strains from some species in the Harveyi clade have been isolated from all over the world over the last hundred years (13). Many of the strains are available for analysis, making it possible to study the clade's diversity over long temporal and spatial scales. Of particular interest is the diversity of three closely related species, Vibrio campbellii, Vibrio jasicida, and Vibrio hyugaensis.

V. campbellii has been isolated from a wide range of habitats, including from seawater samples and associations with marine animals (1), and it has been found to cause disease in marine fish and shrimp (14). Phylogenetic analyses recognized two clearly distinguishable clades within the species (4, 5). Half of the strains in one of the two clades were isolated from associations with animals, including pathogenic associations, while strains from the second clade were in all cases isolated directly from seawater samples. Genomic analysis revealed that strains from the first clade were enriched in Pfam domains with functions related to antibiotic transport and galactose metabolism (5). Despite the observed ecological and genomic differences, strains from both clades were classified as members of the same species. The classification was supported by genomic analysis, including the calculation of average nucleotide identity (ANI), and exhaustive phylogenetic analyses. Furthermore, bacteria in the two clades had higher frequencies of intraclade recombination than recombination with other Vibrio species (4).

The diversity observed in V. campbellii can be contrasted with that of V. hyugaensis and V. jasicida, which are evolutionarily closely related to V. campbellii (15, 16). All but one strain from the two species have been isolated directly from seawater samples. Strains of V. hyugaensis and V. jasicida have been isolated from the same geographical locations and, in some cases, from seawater samples collected on the same day (4, 15). The only V. jasicida strain isolated from associations with fish had no known pathogenic characteristics (16). These results suggest that the two species occupy similar ecological niches. Only one phenotypic feature, esculin hydrolysis, differentiates the species, but they could be distinguished using multilocus sequence analysis and ANI (15). However, unlike the two clades of V. campbellii, taxonomic analyses classify V. hyugaensis and V. jasicida as two different species (4).

Studies of the three species give us an opportunity to compare the diversification of two clades in V. campbellii, both of which are classified as members of the same species, with the speciation observed in V. hyugaensis and V. jasicida offering an opportunity to analyze speciation mechanisms in the genus Vibrio. In this study, we analyzed the genomic dissimilarity of V. jasicida and V. hyugaensis and compared the diversification of the two species with that in the two clades of V. campbellii. Comparative genomic analyses were used to test different hypotheses explaining the diversification of Harveyi clade bacteria, including the role of ecological diversification, gain and loss of gene families, and recombination. The results of these efforts were used to discuss speciation in the genus Vibrio.

RESULTS

Genome sequencing of four Harveyi clade strains.

Representative strains from the three Harveyi clade species were selected for resequencing and assembly into complete genomes. These strains were V. hyugaensis 090810a^T, V. jasicida 090810c, and V. campbellii 151112c. For the purpose of comparative analyses, the V. owensii strain 051011B was also sequenced and de novo assembled. These strains selected for the genomic analysis were all collected in the same geographical location over a period of 2 years and, in the case of V. hyugaensis 090810a^T and V. jasicida 090810c, on the same day (see Table S1 in the supplemental material). The four strains were isolated from seawater samples collected from shallow coastal waters of Miyazaki, Japan.

The genomes of the four strains were organized into two chromosomes, with chromosome I (ChrI) being 1.4 Mb larger on average than chromosome II (ChrII) (Table 1). This genomic organization into a large and a small chromosome is typical for Vibrionaceae (17, 18). The complete assembly of V. campbellii 151112c included two plasmids, and the V. owensii 051011B assembly included the sequence of a single plasmid. No replicons in the four strains had gaps.

TABLE 1.

Genome assembly statistics

Statistic	Vibrio hyugaensis 090810aT	V. jasicida 090810c	V. campbellii 151112c	V. owensii 051011B
Larger chromosome size (bp)	3,530,957	3,690,481	3,581,305	3,797,023
Smaller chromosome size (bp)	2,081,125	2,299,042	2,132,940	2,367,538
Plasmid size(s) (bp)	None	None	24,945, 82,427	96,819
Genome size (bp)	5,612,082	5,989,523	5,821,617	6,261,380
No. of CDSa	4,919	5,274	5,209	5,549
No. of tRNAs	130	132	135	136
No. of rRNAs	37	37	37	37

^aCDS, coding sequences.

Phylogeny and genome content for a highly diverse set of Vibrio strains.

The genome contents of nine Vibrio species from the Harveyi clade were analyzed. The analysis involved 69 strains (Table S1). The analysis focused on six species, V. campbellii, V. jasicida, V. hyugaensis, V. harveyi, V. owensii, and V. rotiferianus, with representative strains of V. parahaemolyticus, V. natriegens, and V. alginolyticus used as an outgroup during the analysis.

Phylogenetic analysis (Fig. 1) was based on sequences coding for 1,769 single-copy orthologues found in the 69 strains. Phylogenetic relationships revealed by the analysis were similar to those predicted for the Harveyi clade bacteria in previous studies (4, 5), including a close evolutionary relationship between V. jasicida and V. hyugaensis, as well as separation of V. campbellii into two clades that correspond to Vc-Gr-1 and Vc-Gr-2 from earlier studies (5). The analysis also revealed that strains CUB2 and MOR3, both previously identified as V. harveyi, share more recent common ancestors with the type strains of V. owensii and V. jasicida, respectively, than with the type strain of V. harveyi.

FIG 1.

(a) Phylogenic tree constructed based on concatenated 1,769-amino-acid sequences from single-copy orthologous genes. The numbers of gained (above the branch) and lost (below the branch) gene families along each branch of the phylogeny are indicated. (b) The Pfam domains with an enriched or reduced copy number in specific lineages are listed.

In addition to the phylogenetic analysis, ANIs were also calculated between the genome sequences of the 69 strains (19, 20). Based on recommendations for the Vibrio taxonomy (2), strains with an ANI of ≥95% were classified as members of the same species. The results of the ANI calculations (Table S2) were consistent with the results of the phylogenetic analysis, i.e., strains with ANI of ≥95% formed a monophyletic group. Strain CUB2 had an ANI of 96.3% with the type strain V. owensii DY05 and an ANI lower than 95% with strains from other species. Strain MOR3 had an ANI of 97.9% with the type strain V. jasicida LMG25398 (Table S2), while it had an ANI lower than 95% with strains from other species. Based on these results, in this study, CUB2 and MOR3 were classified as V. owensii and V. jasicida, respectively.

The mean ANI between strains of V. jasicida and V. hyugaensis (93.7%) is higher than that of the other pairwise species comparisons in this study (85.4% to 89.8%) (Table S2). However, the ANI between strains from the two species is below the 95 to 96% threshold recommended for members of the same Vibrio species. The ANIs between strains from both clades of V. campbellii (96.4%) are higher than that recommended for classification as members of the same Vibrio species.

The availability of complete genome sequences for eight strains in the Harveyi clade allowed us to calculate the ANIs between individual chromosomes of each species. The ANIs calculated for individual chromosomes were comparable to those calculated for whole genomes, with similar ANI values for the two chromosomes (Table S3). A difference was observed for ANIs calculated for individual chromosomes of V. hyugaensis 090810a^T and V. jasicida 090810c. The overall ANI between the two strains was 93.7%, while the ANIs calculated between ChrI and ChrII were 95.2% and 91.1%, respectively.

Genomic features specific to V. hyugaensis and V. jasicida.

Genomic comparison between five strains of V. hyugaensis and representative strains from five Harveyi clade species identified 88 Pfam domains enriched in V. hyugaensis, 30 of which were present only in V. hyugaensis. These 88 Pfam domains were enriched in functions related to DNA-mediated transposition (Table S4a). The following four toxin-related domains were also identified: the MqsR_toxin (motility quorum-sensing regulator, toxin of MqsA) domain, important for cell signaling and biofilm regulation; the GH-E domain (HNH/ENDO VII superfamily nuclease with conserved GHE residues), which was only identified in V. hyugaensis 100512A and can be found in several bacterial polymorphic toxin systems involving competition between related bacterial strains; the TNT (tuberculosis-necrotizing toxin) domain, which was annotated in the proteomes of three strains of V. hyugaensis, and the protein with this domain is homologous with type IV secretion protein Rhs (putative DNase RhsC); and the MORN_2 (MORN repeat variant) domains, which have been predicted to function in adhesion and active invasion (21).

A genome comparison of 18 strains of V. jasicida strains with representative strains from five Harveyi clade species also identified 60 enriched Pfam domains (4-fold copy number with Wilcoxon rank sum test [P ≤ 0.05]). Proteins within these 60 domains were enriched in V. jasicida 090810c that have GO functions related to a DNA restriction-modification system (Table S4b). For instance, proteins with the domains Eco57I (Escherichia coli Eco57I restriction-modification methylase) and BsuBI_PstI_RE (BsuBI-PstI family restriction endonuclease) are only present in V. jasicida but absent in all other five species.

Genomic features differentiating V. hyugaensis and V. jasicida.

We carried out a comparative genomic analysis between V. hyugaensis and V. jasicida. Compared to V. jasicida, we identified more gene loss in V. hyugaensis (Fig. 1). We quantified such differences by investigating the copy number difference in the Pfam domains found in both species and identified 122 Pfam domains which had 4-fold higher copy number in V. hyugaensis than those in V. jasicida (P ≤ 0.05; Wilcoxon rank sum test). Genes with these 122 Pfam domains were enriched in V. hyugaensis 090810a^T, with GO terms related to DNA metabolic processes, including transposition and DNA integration (Table S5a). The analysis also identified seven Pfam domains that were present in all V. hyugaensis strains but absent in V. jasicida (Fig. 1). These include DUF4177, leukocidin (related to bacterial invasion), AvrD (Pseudomonas avirulence D protein), NifW (nitrogen fixation protein), DDE_3 (DDE superfamily endonuclease), HTH_33 (winged helix-turn-helix), and RCC1_2 (regulator of chromosome condensation repeat).

Comparative genomic analysis also identified 131 Pfam domains which had a 4-fold higher copy number in V. jasicida than in V. hyugaensis. Genes with these domains were enriched in V. jasicida 090810c, with GO terms related to pathogenesis, protein secretion, and urea catabolic process (Table S5b). The analysis also identified 48 domains only present in V. jasicida, most of which are responsible for pathogenesis and protein secretion by the type III secretion system, including the domains of Type_III_YscX (type III secretion system YscX), TyeA (translocation of Yops into eukaryotic cells A), HrpJ (HrpJ-like domain), T3SS_needle_reg (YopR bacterial protein domain), and Yop-YscD_cpl (forkhead-associated domain) (Fig. 1).

We were interested in genes that may contribute to the divergence of these two species and reasoned that the phylogenetic relationship of these genes alone would correctly classify their role during divergence. After ranking the proteome according to the correlation of subtree branch length of V. jasicida and V. hyugaensis and orthologous sequence distances, the first is a hypothetical protein (e.g., Vja090810c_04160), and the second one is a protease 3 precursor (e.g., Vja090810c_03736). Phylogenetic trees constructed for both of these proteins agree with the classification of V. jasicida and V. hyugaensis in the species tree based on the concatenated alignment of 1,769 single-copy orthologues (Fig. S1a and b).

Syntenic differences between V. hyugaensis and V. jasicida and V. campbellii strains from two clades.

The syntenic differences between V. hyugaensis and V. jasicida were analyzed by comparing the chromosomes of V. hyugaensis 090810a^T and V. jasicida 090810c. The analysis identified syntenic coverages of 92.3% on ChrI and 92.1% on ChrII of strain 090810a^T and 89.7% on ChrI and 88.1% on ChrII of strain 090810c (Fig. 2 and Table S6). In a previous study, it was also observed that ChrI had more highly conserved regions than ChrII when comparing the synteny between strains from the two clades of V. campbellii (5). The analysis identified coverages of 85.4% on ChrI and 78.6% on ChrII of strain V. campbellii ATCC BAA-1116 and 87.4% on ChrI and 79.9% on ChrII of strain 1114GL. Analysis of GO enrichment of genes in the synteny break regions revealed abundant genetic mobile elements in ATCC BAA-1116 and 1114GL (Table S7). In addition, the analysis identified a large number of pathogenesis-related genes present in a synteny break region of V. jasicida 090810c (Table S7b).

FIG 2.

Synteny map between V. campbellii and V. hyugaensis genomes. Synteny blocks were identified by DAGchainer. The syntenic blocks aligned with the same strand (i.e., +/+) are linked in red, and the reversion (i.e., +/−) blocks are linked in blue.

Previous studies identified a 73-kb superintegron in V. campbellii 1114GL and no superintegron in V. campbellii ATCC BAA-1116 (5). The complete gapless assemblies allowed us to identify that the superintegron includes the integron-integrase gene (intI) (tyrosine recombinase XerD) and a contiguous recombination site (attI), similar to superintegrons found in Vibrio cholerae (22). Most genes in the superintegron are hypothetical proteins, and synteny was not established in this region. In our other complete V. campbellii genome, 151112c, this cassette array spans 81,643 bp. In addition, 124,037-bp and 59,942-bp superintegrons were found in V. hyugaensis 090810a^T and V. jasicida 090810c, respectively (Fig. S2). Interestingly, unlike the superintegrons of V. cholerae, which are located on ChrII, the superintegrons of V. campbellii 1114GL, V. campbellii 151112c, V. hyugaensis 090810a^T, and V. jasicida 090810c are located on ChrI. Genes adjacent to the superintegron exhibit a shared synteny, suggesting that the superintegron was either inserted at this specific region in each species or inserted in the most recent ancestor in this clade of Vibrio spp. and subsequently expanded or lost.

Effect of recombination.

The effect of recombination on the diversification of bacteria in the Harveyi clade was inferred based on the results of an analysis using ClonalFrameML (23). The analysis was done based on sequences coding for 1,769 single-copy orthologs found in the 69 strains, the same data set used for the phylogenetic analysis shown in Fig. 1. Phylogenetic reconstruction from the analysis (Fig. S3) showed a relationship between analyzed strains congruent with the relationship predicted using FastTree (Fig. 1). The results of the analysis were used to calculate the overall contribution of recombination and mutation (r/m) to clonal diversification. The results show that the effects of recombination are much more important than those of mutation in the node leading to the two V. campbellii clades (r/m = 5.47) compared to the node leading to V. hyugaensis and V. jasicida (r/m = 0.08).

Difference among V. jasicida strains isolated in Miyazaki and elsewhere.

We were interested to know if genetic differences exist between the four V. jasicida strains isolated in Miyazaki (V. jasicida 090810c, 200612G, 201212A, and 270813B) and the other 14 V. jasicida strains isolated elsewhere in the world. The comparisons between protein numbers in an orthologous group showed that the genes for putative protease YdeA and transcriptional regulator SlyA are the genes with the most biased distribution (P = 0.0049), being present in all strains isolated in Miyazaki but absent in the majority of the other strains (Table S8). The comparison of Pfam domain numbers showed that all Miyazaki strains carry at least one HOK_GEF domain (270813B has five domains), but only 4 out of 14 non-Miyazaki strains contain a single copy of this domain (Table S9). The HOK_GEF domain is related to the hok/sok system, which mediates plasmid stabilization by killing plasmid-free cells in bacteria.

DISCUSSION

This study reports the results of analyses aimed at understanding the speciation process in the Harveyi clade of Vibrio species. The study focused on the speciation process that led to the differentiation of two Vibrio species, V. jasicida and V. hyugaensis, as well as the process of diversification within V. campbellii. The results of the study allowed us to test hypotheses for speciation in the Harveyi clade.

In this study, we report the first complete genome sequences of representative strains from V. jasicida and V. hyugaensis. The complete genome sequences of representative strains of V. campbellii and V. owensii were also determined. Genomic analyses using these complete genome sequences revealed high synteny between the V. jasicida and V. hyugaensis chromosomes, which was higher than the synteny between the chromosomes of strains from different V. campbellii clades. The lower synteny between the V. campbellii strains was mainly due to several large synteny breaks at ChrII. The synteny between the ChrII of V. jasicida and ChrII of V. hyugaensis was higher than that between two V. campbellii clades; however, the ANI calculated between ChrII of V. jasicida and V. hyugaensis was lower than that calculated from the whole genome. These results indicate frequent gene gain and loss within V. campbellii and suggest that the V. campbellii ChrII underwent more changes during the diversification process. Considering that strains from both clades of V. campbellii are classified as members of the same species, these results suggest that frequent gene gain and loss are not requisites for the speciation process in the Harveyi clade Vibrio species. Small chromosomes of bacteria in the clade appear to have higher rates of gene turnover than do the large chromosome, but those changes are not necessary for speciation.

It has been hypothesized that differences in bacterial ecology contributed to the Vibrio speciation process. The V. jasicida genome was found to be enriched with hypothetical pathogenicity and type III secretion system genes which were absent in V. hyugaensis. This result would suggest that the adaptation to associate with marine animals played a role in how the two species diversified. However, only one V. jasicida strain was isolated from an association with animals, while the other 17 strains were isolated directly from seawater samples. The two species also have indistinguishable growth optima (15). Therefore, the genomic differences identified in this study should not be used to support the hypothesis that ecological diversification toward differential associations with marine animals played a role in the diversification of the two species. However, the results of this study also cannot be used to reject a hypothesis that ecological diversification contributed to the diversification of V. jasicida and V. hyugaensis. The Vibrio strains analyzed in this study were isolated using a wide range of methods from different geographical locations and over a long period of time, and this collection of bacteria is not well suited for an analysis of ecological diversification.

Unlike V. jasicida, strains of V. hyugaensis were found to be enriched in Pfam domains related to DNA metabolism and had a higher rate of gene loss. Interestingly, V. campbellii Vc-Gr2 was also found to be enriched in the DNA metabolism (5) and had more gene loss than with Vc-Gr1 (Fig. 1). The present study also found that recombination had a higher effect than mutation in the node leading to the two V. campbellii clades compared to the node leading to V. hyugaensis and V. jasicida. Similar results were found in previous studies, which showed differences in the inter- and intraspecies recombination frequency rates between the two species (4). These results imply that recombination is a major force driving diversification within V. campbellii and could be interpreted as evidence that the speciation process involved changes in the frequency of gene exchange between V. hyugaensis and V. jasicida. However, we cannot rule out a possibility that the change in the frequency of recombination occurred after the speciation process as a result of V. jasicida and V. hyugaensis diversification. It should be noted that the analysis of recombination was done using conserved regions of the Harveyi clade species' genome, and the analysis does not take into consideration auxiliary parts of the genome which might be affected by recombination at different rates.

Overall, the results of this study suggest that multiple processes contributed to the diversity of Vibrio spp. in the Harveyi clade. We observed significant gene loss and/or gain in some strains and identified potential ecological differences between V. campbellii strains and different recombination frequency rates. Some of the diversification processes can be observed when analyzing strains isolated from a limited geographical location. For example, V. jasicida strains isolated from the Miyazaki area are enriched with genes involved in plasmid stabilization and killing of plasmid-less cells, which could be interpreted as evidence for a local speciation process. However, when strains isolated from a wide range of geographic locations are used for comparative analyses, the observed diversification processes cannot be used to explain patterns of speciation observed in the clade. As a result, we can identify no apparent relationship between the observed processes increasing the diversity of the bacteria and the evolution of new Vibrio species. An alternative explanation for the observed results is that speciation occurred through a number of stochastic diversification and extinction events influenced by multiple evolutionary processes over a long period of time.

MATERIALS AND METHODS

DNA preparation, sequencing, and de novo assembly of the Vibrio genome.

Genome sequencing of 10 V. jasicida and V. hyugaensis strains (see Table S1 in the supplemental material) was done using the MiSeq platform (Illumina). Genomic DNA was isolated using the DNeasy blood and tissue kit (Qiagen), according to the manufacturer's instructions. Library preparation, sequencing, and assembly were done as described previously (9). We also resequenced V. hyugaensis 090810a^T, V. jasicida 090810c, V. campbellii 151112c, and V. owensii 051011B using the PacBio RSII platform (Pacific Biosciences, Menlo Park, CA). The four strains were cultured from a glycerol stock in LSW-70 broth at 30°C with shaking at 200 rpm for 16 h. The bacterial cells were pelleted by centrifugation at 5,000 × g for 10 min. Genomic DNA was extracted using the Qiagen Genomic-tip 100/G, according to the manufacturer's instructions. One to three single-molecule real-time (SMRT) cells for each strain were sequenced on the PacBio RSII platform. The filtered subreads were subjected to the Canu assembler (24). The consensus sequences were further polished by Quiver (25), as implemented by SMRT version 2.0.1. Each genome was circularized by Circlator (26).

Another 59 Vibrio genome assemblies were downloaded from NCBI (last retrieval date, 6 August 2016) (see Table S1). The selection of strains preferred those with less-fragmented genomes (<400 scaffolds). An exception was made for V. rotiferianus Oz08, because it was one of only three genome assemblies available for this species.

Predicted proteins and domains.

Our assemblies and the publicly available assemblies with >300 bp were annotated (Table S1) by the Prokaryotic Genome Annotation System (Prokka) pipeline. Functional annotation of the predicted proteins was obtained by Argot2. The protein domains of each gene were identified by pfam_scan.pl version 1.5 by comparing them against Pfam version 30.0.

Phylogenetic analysis.

Single-copy orthologous genes in 69 strains (Table S1) (5 strains of V. hyugaensis, 17 strains of V. jasicida, 13 strains of V. campbellii, 15 strains of V. harveyi, 10 strains of V. owensii, 3 strains of V. rotiferianus, 1 strain of V. parahaemolyticus, 4 strains of V. natriegens, and 1 strain of V. alginolyticus) were identified based on OrthoFinder (27). The sequences in each orthologous group were aligned by MAFFT (version 7.271) (28) with the local alignment option, and the orthologous group of alignments with more than 10% gaps was discarded. The resulting alignment had 1,722,426 characters. A phylogenetic tree was generated using FastTree with 1,000 bootstrap replicates from concatenating the remaining orthologous sequences alignments. The phylogeny was plotted using FigTree version 1.4.2 (http://tree.bio.ed.ac.uk/software/figtree/). The phylogenetic tree of proteins with HOC_GEF domains were aligned by MAFFT (version 7.271) (28) and trimmed by trimAl (with option -automated1) (29). The phylogeny was constructed by FastTree, with 1,000 bootstrap replicates. Orthologous proteins that represent the correct species classification and branch length of the V. jasicida and V. hyugaensis subtree were identified using coefficient correlations (Pearson correlation coefficient [PCC]) metric by Popmarker (30).

Repertoire, expansion, and reduction of protein domain family size.

Gene families' gain and loss along the species phylogeny were identified using dollop, as implemented in the PHYLIP package (version 3.69.650) (31). The enrichment of orthologues between 4 strains of V. jasicida collected in Miyazaki and 14 strains elsewhere was analyzed using Fisher's exact test. Enrichment of the Pfam domain copy number between two sets of interests was assessed by the Wilcoxon rank sum test (P ≤ 0.05). We compared the Pfam copy number between (I) 5 V. hyugaensis strains all collected in Miyazaki versus 18 V. jasicida strains, (II) 4 V. jasicida strains collected in Miyazaki versus 14 V. jasicida strains elsewhere in the world, and (III) 5 V. hyugaensis strains versus 4 V. jasicida strains inside Miyazaki. The expanded or reduced domains were assigned to the KEGG pathway by KAAS (32). GO enrichments were identified for significant domain gains/losses using TopGO (version 2.10.0) (33).

Synteny analysis.

The orthologous genes from the BLASTP search with an E value of <1 × 10⁻¹⁰ were used as input for DAGchainer (-Z 12 -D 3 -g 1 -A 3) (34) to define synteny blocks between two pairs of genomes, including (I) V. jasicida 090810c and V. hyugaensis 090810a^T, (II) V. hyugaensis 090810a^T and V. campbellii 1114GL, (III) V. campbellii 1114GL and V. campbellii 151112c, and (IV) V. campbellii 151112c and V. campbellii ATCC BAA-1116. The regions not covered by synteny blocks were defined as synteny breaks. Synteny between the genomes was plotted using R (35).

Recombination frequency.

The rates of recombination to mutation (r/m) were calculated based on R/theta, delta, and nu values predicted by an analysis using ClonalFrameML (23) using the (-embranch) option. The analysis was done based on a concatenated alignment for all single orthologs of length 1,722,426 characters.