Comparative Genomic Analysis Using Microarray Demonstrates a Strong Correlation between the Presence of the 80-Kilobase Pathogenicity Island and Pathogenicity in Kanagawa Phenomenon-Positive Vibrio parahaemolyticus Strains

Abstract

Vibrio parahaemolyticus is a gram-negative marine bacterium. A limited population of the organisms causes acute gastroenteritis in humans. Almost all of the clinical V. parahaemolyticus isolates exhibit beta-type hemolysis on Wagatsuma agar, known as the Kanagawa phenomenon (KP). KP is induced by the thermostable direct hemolysin produced by the organism and has been considered a crucial marker to distinguish pathogenic strains from nonpathogenic ones. Since 1996, so-called “pandemic clones,” the majority of which belong to serotype O3:K6, have caused worldwide outbreaks of gastroenteritis. In this study, we used a DNA microarray constructed based on the genome sequence of a pandemic V. parahaemolyticus strain, RIMD2210633, to examine the genomic composition of 22 strains of V. parahaemolyticus, including both pathogenic (pandemic and nonpandemic) and nonpathogenic strains. More than 86% of the RIMD2210633 genes were conserved in all of the strains tested. Many variably present genes formed gene clusters on the genome of RIMD2210633 and were probably acquired through lateral gene transfer. At least 65 genes over 11 loci were specifically present in the pandemic strains compared with any of the nonpandemic strains, suggesting that the difference between pandemic and nonpandemic strains is not due to a simple genetic event. Only the genes in the 80-kb pathogenicity island (Vp-PAI) on chromosome II, including two tdh genes and a set of genes for the type III secretion system, were detected only in the KP-positive pathogenic strains. These results strongly suggest that acquisition of this Vp-PAI was crucial for the emergence of V. parahaemolyticus strains that are pathogenic for humans.

Vibrio parahaemolyticus is a halophilic bacterium that occurs naturally in marine or estuarine environments. A limited population of V. parahaemolyticus is capable of causing human diseases, such as acute gastroenteritis, which are often associated with consumption of raw or undercooked seafood (3). Almost all of the clinical V. parahaemolyticus isolates exhibit the Kanagawa phenomenon (KP), a beta-type hemolysis on a special blood agar (Wagatsuma agar) (32), and this phenomenon is caused by the thermostable direct hemolysin (TDH) (8). The KP has long been recognized as an effective marker for discriminating pathogenic from nonpathogenic strains and is used extensively as a clinical indicator for the virulence of V. parahaemolyticus (8, 23). A small number of clinical strains are KP negative, but these often carry the TDH-related hemolysin gene (trh) (8, 23). Even if they possess the tdh gene, trh-possessing strains rarely show hemolytic activity on Wagatsuma agar, probably due to low or no production of TDH from the tdh gene (34), so that strains having both tdh and trh are usually classified as KP negative. Thus, human-pathogenic V. parahaemolyticus strains can be divided into two groups, KP-positive and KP-negative (trh-possessing) ones.

Although infections caused by V. parahaemolyticus are usually sporadic cases caused by various serotypes of this organism, strains with O3:K6 and a few other serotypes, such as O4:K68, O1:K25, and O1:K untypeable (O1:KUT), have caused an increasing number of worldwide outbreaks of gastroenteritis since 1996 (1, 4, 5, 6, 7, 18, 19, 21, 26). These strains showed almost identical fragment patterns of pulsed-field gel electrophoresis (PFGE) and arbitrarily primed PCR and are referred to as “pandemic clones” (2, 19, 21, 26, 36).

In 2003, we determined the entire genomic sequence of a pandemic V. parahaemolyticus strain, RIMD2210633, with the O3:K6 serotype (16) and identified 4,832 protein-coding genes (ORFs) on the two circular chromosomes (chromosomes 1 and 2) of the strain. A large pathogenicity island (Vp-PAI) was identified on chromosome 2, where one set of type III secretion system (TTSS) genes and two copies of tdh genes encoding almost identical TDHs were located (16). Many genes encoding potential virulence factors were also identified at this chromosomal locus. While genome sequencing can disclose the presence of these potential virulence factors, genomic comparisons between pathogenic and nonpathogenic strains within a species should be useful for identifying virulence determinants crucial to the pathogenicity of this organism. DNA microarrays have successfully been used for comparative genomic analyses of many pathogenic bacteria to gain insights into genomic evolution or diversity in individual bacteria and to identify genes that correlate with diseases. In this study, we constructed a DNA microarray targeting all ORFs of V. parahaemolyticus RIMD2210633 and performed a comparative genomic hybridization (CGH) analysis of various V. parahaemolyticus strains to determine the difference in gene repertoire between pathogenic and environmental (nonpathogenic) strains and between pandemic and nonpandemic strains.

MATERIALS AND METHODS

Bacterial strains and culture conditions.

All V. parahaemolyticus strains used in this study are listed in Table 1. The strains were obtained from the Laboratory for Culture Collection, Research Institute for Microbial Diseases, Osaka University (Osaka, Japan). Bacterial strains were grown to stationary phase at 37°C in either marine broth (Difco) or LB broth containing 3% NaCl. There are two variants among the trh genes (trh1 and trh2). To investigate these trh-possessing strains in detail, we analyzed a disproportionately large number of KP-negative strains compared to their natural occurrence.

TABLE 1.

V. parahaemolyticus strains used in this study

ID no.a	Strain	Source	Yr of isolation	Serotype	KP	Presence or absence of geneb		Description
						tdh	trh
1	RIMD2212201(OM1107)	Seawater	1999	O3:K20	−	−	−	Nonpathogenic
2	RIMD2210384	Food	1976	O4:KUT	−	−	−	Nonpathogenic
3	RIMD2212470	Seawater	2001	O5:KUT	−	−	−	Nonpathogenic
4	RIMD2210001 (EB101)	Patient	1950	O1:K1	−	−	+	Nonpandemic
5	RIMD2210856 (AQ4673)	Patient	1991	O10:KUT	−	+	+	Nonpandemic
6	RIMD2210911 (AQ4729)	Patient	1992	O1:K1	−	−	+	Nonpandemic
7	RIMD2210536 (AQ4023)	Patient	1985	O3:K6	−	−	+	Nonpandemic
8	RIMD2210554 (AQ3996)	Patient	1983	O4:K11	−	−	+	Nonpandemic
9	RIMD2210883 (AQ4700)	Patient	1992	O4:K11	+	+	−	Nonpandemic
10	RIMD2210056 (HG-12-72)	Patient	1972	O4:K8	+	+	−	Nonpandemic
11	RIMD2210884 (AQ4701)	Patient	1992	O4:K8	+	+	−	Nonpandemic
12	RIMD2210086 (WP-1)	Patient	1968	O4:K12	+	+	−	Nonpandemic
13	RIMD2210706 (AQ4817)	Patient	1993	O4:K12	+	+	−	Nonpandemic
14	RIMD2211499 (AQ4953)	Patient	1994	O2:K3	+	+	−	Nonpandemic
15	RIMD2211932 (AQ4577)	Patient	1990	O8:K-	+	+	−	Nonpandemic
16	RIMD22121044 (KX-V1394-3)	Patient	2001	O4:K4	+	+	−	Nonpandemic
17	RIMD2210587	Patient	1996	O4:K12	+	+	−	Nonpandemic
18	RIMD2210612	Patient	1996	O4:K9	+	+	−	Nonpandemic
19	RIMD2210600 (KX-V204)	Patient	1996	O3:K5	+	+	−	Nonpandemic
20	RIMD2212686 (KX-V1192)	Patient	2001	O3:K6	+	+	−	Pandemic
21	RIMD2212924 (KX-V1433)	Patient	2001	O4:K68	+	+	−	Pandemic
22	RIMD2210633 (KX-V237)	Patient	1996	O3:K6	+	+	−	Pandemic

^aStrain identification numbers correspond to lanes in Fig. 1.

^b+, present; −, absent. The presence or absence of the tdh and trh genes was determined by PCR.

Cloning of V. parahaemolyticus ORFs.

Specific primers were principally designed to obtain PCR products with an average length of 520 bp. For ORFs smaller than 520 bp, the primers were designed to obtain PCR products covering, whenever possible, entire coding regions of each ORF. Primers were synthesized by SigmaGenosys Japan (Sapporo, Japan). All PCR amplifications of every ORF were performed by using a Takara ExTaq polymerase (Takara Bio Inc., Shiga, Japan) with RIMD2210633 genomic DNA as a template, with a protocol of an initial 5-min denaturation at 95°C followed by 30 amplification cycles of 30 s at 95°C, 30 s at 55°C, and 2 min at 72°C. All PCR products were analyzed by agarose gel electrophoresis to verify their product size. Each PCR product was cloned into the pCRII-TOPO vector using a TOPO TA cloning kit (both from Invitrogen) and introduced into Escherichia coli DH5α. All transformants were cultured and stored as frozen stocks in LB broth containing 20% (wt/vol) glycerol. Eight V. parahaemolyticus ORFs that were not present in RIMD2210633 were prepared by PCR using the following template DNA: genomic DNA of strain AQ3996 (trh1) and strain AQ4729 (trh2, ureR, nicD, and ureC) and pVP-RET1 DNA (ret, orf560, and orf218).

Construction of DNA microarray.

The microarray contains PCR products derived from all of the ORFs of V. parahaemolyticus strain RIMD2210633 (16) with the exception of VPA1386. Each V. parahaemolyticus ORF was amplified from the frozen stock using M13-Forward and M13-Reverse primers. After checking the quality and size of each of the PCR products by agarose gel electrophoresis, the PCR products were spotted on poly-l-lysine-coated glass slides (Matsunami Glass Ind., Ltd., Osaka, Japan) by using the SPBIO spotter (Hitachi Software Engineering, Tokyo, Japan).

Isolation and labeling of chromosomal DNA.

Genomic DNA was prepared from each strain using a DNeasy tissue kit (Qiagen) according to the manufacturer's instructions. Two micrograms of genomic DNA was suspended in 24 μl H₂O, combined with 20 μl of the random hexamer solution (supplied by the BioPrime Array CGH Genomic Labeling System [Invitrogen]), heated to 95°C for 5 min, and chilled on ice. Five microliters of 10× deoxynucleoside triphosphate-aminoallyl dUTP mix (5 mM dATP, dCTP, and dGTP, 2 mM dTTP, and 3 mM aminoallyl dUTP) and 1 μl of the Klenow fragment solution were added to the mixture and incubated for 1 h at 37°C. The aminoallyl-labeled DNA was purified by phenol-chloroform extraction and ethanol precipitation. Precipitated DNA was dried and resuspended in a 10-μl solution of 50 mM NaHCO₃ (pH 9.0), and 10 μl Cy3 or Cy5 monofunctional dye (GE Healthcare) dissolved in dimethyl sulfoxide was added to the solution. One-tenth of one reaction vial of Cy3 or Cy5 dye supplied by the manufacturer was used per reaction. After incubation for 1 h at room temperature in the dark, unincorporated dye was removed by using a CentriSep spin columns (Princeton Separations, Inc.). Cy3- or Cy5-labeled probes were recovered by means of ethanol precipitation and resuspended in 18 μl of H₂O. Equal amounts of probes labeled with Cy3 and Cy5 were mixed, and 3 μl of 10% sodium dodecyl sulfate (SDS), 6 μl of 10 mg/ml yeast tRNA, and 15 μl of 20× SSC (3 M NaCl, 0.3 M trisodium citrate·2H₂O, pH 7.0) were added to the mixture. The probe mixture was then heated for 5 min at 96°C, quickly chilled on ice, and centrifuged for 1 min at 14,000 rpm in a microcentrifuge. Each probe mixture was incubated for 5 min at 55°C just before being applied to a microarray slide.

Hybridization and detection of microarray signals.

The prewarmed probe mixture was applied to a microarray slide and covered with a MAUI AO lid (BioMicro Systems). After the microarray slide had been sealed, it was placed in the MAUI hybridization chamber and incubated for 16 h at 55°C. After hybridization, the microarray slides were washed twice with the 2× SSC-0.1% SDS solution at 55°C for 10 min, twice with the 0.2× SSC-0.1% SDS solution at room temperature for 10 min, and finally twice with the 0.2× SSC solution at room temperature for another 10 min. Washed microarray slides were dried by centrifugation and scanned with a microarray scanner, the ScanArray Express Lite (Perkin Elmer Life and Analytical Sciences), and the data were processed with ScanArray Express software. For each test strain, three experiments were carried out independently.

Data analysis.

The signal intensities of each spot were quantified with the ScanArray Express software, and further data analyses were performed with Microsoft Excel software and the microarray genomic analysis program GACK (http://falkow.stanford.edu/whatwedo/software) (13). After the data with low signal intensities or slide abnormalities were excluded, the data set was analyzed by GACK using cutoff lines, each of which was determined for every microarray experiment.

To determine the presence or absence of each of the spots, log₂ values of the test strain divided by those of the reference strain for each spot were analyzed by GACK; spots with 0% were considered “absent” and those with an estimated probability of presence between 0% and 50% “divergent.” Since we performed three independent experiments for each strain, the ORFs were categorized as “present,” “absent,” or “divergent” when they got the same judgments in more than one experiment. When the judgments of three experiments were all different, the ORFs were judged “uncertain.” Absence of some variably present genes and regions was verified by PCR and sequencing. When an ORF was absent or divergent in at least more than one test strain, the ORF was judged to be variably present. When absent or divergent ORFs continually appeared and occupied more than 80% of a locus, the locus was designated a variably present gene cluster.

Microarray data accession number.

The microarray data have been submitted to the Gene Expression Omnibus (series record number GSE10020). Final processed data are presented in the supplemental material.

RESULTS AND DISCUSSION

Overview of comparative genomic hybridization analysis.

The DNA microarray we constructed contained PCR products from almost all of the ORFs of V. parahaemolyticus strain RIMD2210633. In addition, eight additional ORFs selected from previously reported V. parahaemolyticus genes were spotted on the microarray (see Materials and Methods). These additional ORFs comprised two TDH-related hemolysin genes (trh1 and trh2) (14), three genes within the urease operon (ureR, nikD, and ureC) (29), and three genes within a retron element (ret, orf560, and orf218) (T. Shimamoto, personal communication).

With the aid of a CGH analysis using the DNA microarray, we compared the genomic composition of 22 strains of V. parahaemolyticus. These test strains were chosen from strains of various serotypes, years of isolation, and sources and consisted of three environmental strains and 19 pathogenic strains (14 KP-positive strains and 5 KP-negative strains) (Table 1). Of the 14 KP-positive strains, 3 were pandemic and the others nonpandemic strains. An overview of the results of CGH analyses is provided in Fig. 1. The final data sets from the CGH analyses are available in the supplemental material. Of 4,832 RIMD2210633 ORFs, a total of 675 (14.0%) were absent or highly divergent in at least more than 1 strain among the 22 strains (referred to as “variably present genes” in this paper; see Materials and Methods). The remaining 4,157 ORFs (86%) were conserved in all of the tested strains and represent the core gene set of V. parahaemolyticus. We analyzed the functions of the 675 variably present genes according to their clusters of orthologous groups (COG) classification (the database of the V. parahaemolyticus genome project [http://genome.gen-info.osaka-u.ac.jp/bacteria/vpara/genome.html]). A large proportion (51.1%) of the variably present genes were not assigned to any COG category, and this ratio was significantly higher than that for the whole genome (25.5%). The abundant functional group was that for “cell envelope biogenesis, outer membrane,” followed by that for “DNA replication, recombination, and repair.”

FIG. 1.

Genomic composition of V. parahaemolyticus strains. Each row corresponds to a spot on the microarray, and they are arranged according to the chromosomal positions of the RIMD2210633 genes. The numbers on the top correspond to strain numbers in Table 1. The colors indicate the status of the ORFs, as follows: blue, present; pale blue, divergent; yellow, absent; gray, uncertain. Representative variably present regions are indicated on the left of the columns. Asterisks indicate the positions of tRNA flanking the variably present regions. GC percentages of the genome are on the right.

More than half of all variably present genes were observed as clusters (Fig. 1). We identified 13 variably present gene clusters containing more than 10 ORFs and labeled them V. parahaemolyticus variably present gene cluster 01 (VPC-01) to VPC-13 (Table 2). PFGE analyses are often used to distinguish strains and phylogenetic analysis of V. parahaemolyticus (11, 35, 37). Genomic structures of V. parahaemolyticus strains are considered highly variable based on their PFGE profiles (26, 27, 35, 37). Our CGH study, however, revealed that the majority of the genes (more than 86%) are conserved among V. parahaemolyticus strains, which suggests that genomic diversities observed by the PFGE analyses may have been generated mainly as a result of gains or losses of large gene clusters.

TABLE 2.

Size, position, GC percentage, and content of the large variably present gene cluster identified in this study

Cluster IDa	Size (no. of ORFs)	Start site	End site	G+C content (%)	Gene product(s) or function	Other characteristicsb	GI IDa,c
VPC-01	52	VP0187	VP0238	40.6	O3:K6 antigen biosynthesis
VPC-02	28	VP0376	VP0403	40.9	Type I restriction modification system	T, P	VPaI-1
VPC-03	22	VP1071	VP1092	42.1	Phage family integrase, putative ATP-dependent DNA helicase	T, P	VPaI-3
VPC-04	38	VP1383	VP1420	43.3	Hypothetical protein, conserved hypothetical protein
VPC-05	39	VP1548	VP1586	47.5	Bacteriophage f237, Alteromonas phage PM2	P
VPC-06	94	VP1771	VP1864	41.8	Superintegron (hypothetical protein, conserved hypothetical protein)	P
VPC-07	14	VP2131	VP2144	38.9	Hypothetical protein, putative phage-related protein	T, P	VPaI-4
VPC-08	11	VP2900	VP2910	38.3	Hypothetical protein, putative phage protein	P	VPaI-5
VPC-09	15	VPA0434	VPA0448	44.2	Hypothetical protein, putative RNase HI, resolvase, recombinase	P
VPC-10	25	VPA0889	VPA0913	48.2	Hypothetical protein, phage-related protein, Alteromonas phage PM2	P
VPC-11	17	VPA1254	VPA1270	43.4	Putative IS1328 transposase, hypothetical protein	P	VPaI-6
VPC-12	87	VPA1310	VPA1396	39.2	Vp-PAI (pathogenicity island)	K	VPaI-7
VPC-13	10	VPA1700	VPA1709	49.1	Mannose metabolism

^aID, identifier.

^bT, flanked by tRNA gene; P, contains pandemic strain-specific genes; K, KP-positive clinical strain specific.

^cGI ID assigned by Hurley et al. (10).

Using a bioinformatic analysis of the genomic sequence of RIMD2210633, Hurley et al. identified seven genomic islands (GIs) (VPaI-1 to VPaI-7) larger than 10 kb which have aberrant base compositions compared to the core genome (10). Each of the seven islands possesses a phage-like integrase gene and is flanked by direct-repeat sequences. Many genes within those GIs were missing in many isolates in our microarray analysis (Table 2). In addition to these GIs, we identified seven gene clusters with normal or aberrant G+C content that were frequently missing in test strains.

Characteristics of the variably present gene clusters larger than 10 ORFs are listed in Table 2. Among them were five putative prophage regions (VPC-03, VPC-05, VPC-07, VPC-08, and VPC-10), with the genes in these prophage regions highly divergent or completely absent in many test strains (Fig. 1; Table 2). Such a variation of prophages has also been observed in Escherichia coli O157 strains (24, 25). Other notable characteristics of variably present gene clusters are the type I restriction-modification system (VPC-02), superintegron (VPC-06), insertion sequence (IS), and transposon (VPC-06, VPC-11, and VPC-12). IS and transposons are known as mobile elements which transfer horizontally. Three of the variably present gene clusters, including two prophage regions (VPC-03 and VPC-07), were flanked by the tRNA gene, which suggests that these clusters were integrated into the chromosome by using tRNA loci as target sites (Fig. 1; Table 2) (10). The presence of various mobile elements in the clusters may partly account for the relatively high percentage of genes for the “DNA replication, recombination and repair” COG category among variably present genes.

Most of the variably present genes that were classified in the “cell envelope biogenesis” category were found in VPC-01 (corresponding to VP0187 to VP0238). This region encodes the genes for the biosynthesis of lipopolysaccharides and capsular polysaccharides which serve as major antigens of V. parahaemolyticus. In fact, among the serotype O3:K6 strains with the same serotype as RIMD2210633 (strains 7, 20, and 22; lanes 1 to 3 in Fig. 2), genes in this region were fully conserved, whereas only the genes in the first half (VP0187 to VP0213) were conserved in the two test strains with serotypes O3:K20 and O3:K5 (strains 1 and 19; lanes 4 and 5 in Fig. 2). This suggests that the first and second halves of this region are involved in the biosynthesis of O and K antigens, respectively.

FIG. 2.

Comparison of locus VP0187 to VP0238 in various serotypes of V. parahaemolyticus strains. Serotype and number of strains are shown at the top. Blue, yellow, pale blue, and gray show, respectively, present, absent, divergent, and uncertain genes.

Genomic comparison between pandemic and nonpandemic strains.

Although a variety of serovars of V. parahaemolyticus can cause human diseases, a predominant occurrence of O3:K6 and a few other serotypes (O4:K68, O1:K25, and O1:KUT) has been reported worldwide since 1996, and these strains are known as “pandemic clones” (4, 19, 21, 26). To gain insight into the mechanism of the worldwide outbreaks of limited serotypes of strains, the genomic composition of pandemic strains with serotypes of O3:K6 and O4:K68 was compared to that of nonpandemic strains.

We first compared two O3:K6 strains isolated in 1996 and 2001 (strains 20 and 22, respectively) and one O4:K68 strain (strain 21) isolated in 2001. Comparison of the genomic content among these three pandemic strains showed that their genomic organizations closely resemble each other. A clear difference between these strains was seen only in the region specifying O and K antigens. In addition, the region encoding VP1884 to VP1891 was absent in the two pandemic strains isolated in 2001 (strains 20 and 21). These results agree with the hypothesis based on the findings of PFGE (2, 5) and arbitrarily primed PCR analysis (19, 26) that the O4:K68 strain originated from the pandemic O3:K6 clone.

We next examined the pandemic and nonpandemic strains for genetic differences and identified 65 pandemic-strain-specific genes that are located in 11 chromosomal regions (see the supplemental material). Many of these genes were hypothetical protein genes but were located in variably present gene clusters, such as type I restriction-modification system, phage, super-integron, or IS element. This suggests that the pandemic strains emerged not via a single genetic event or mutation from the nonpandemic strains but via multiple genetic events, including insertion of several large gene clusters.

In some previous studies, pandemic and nonpandemic strains of the same serotype (O3:K6) were compared by means of the DNA subtraction method, resulting in the identification of several pandemic-strain-specific regions (28; M. Nishibuchi et al., personal communication). Our microarray analysis confirmed their results but also indicated that more pandemic-strain-specific regions are present in pandemic strains.

Hurley et al. found by using PCR analysis that four of the GIs (VPaI-1, -4, -5, and -6) were present specifically in the pandemic group of V. parahaemolyticus isolates (10). Genes on the four pandemic-group-specific GIs (corresponding to VPC-02, -07, -08, and -11) showed a rather similar distribution to that among the isolates used in our analysis (Fig. 1) (see the supplemental material) because most genes within these regions were present only in pandemic group isolates. In addition, VP1071 to VP1085 on VPC-03 (corresponding to VPaI-3 in the Hurley study) were also pandemic group specific in our analysis.

In many pathogenic bacteria, proteins encoded on prophage regions are known to affect cellular processes of the host cells (17). It is thus possible that some of the pandemic-strain-specific ORFs in our study may have added to the organism some novel phenotype which is related to pandemic phenotypes. BlastP analysis of the pandemic-strain-specific hypothetical genes showed that several ORFs were homologous to nuclease, acetyltransferase, and RTX toxin but most showed no significant homology to proteins with known functions. A systematic functional analysis of the pandemic-strain-specific ORFs may be required to identify the genes that are crucial for the emergence of pandemic clones.

Genomic comparison between pathogenic and nonpathogenic strains of V. parahaemolyticus.

To determine what genes are responsible for the virulence of V. parahaemolyticus for humans, we compared the genomic content of three environmental (nonpathogenic) isolates (strains 1 to 3) and 19 clinical isolates (strains 4 to 22) to identify the genes or DNA regions present specifically in pathogenic strains. We could not find any genes that were common to all the clinical isolates but not to the environmental ones. However, by comparing the KP-positive strains (strains 9 to 22) with KP-negative clinical (strains 4 to 8) and environmental (strains 1 to 3) strains, we found that only VPC-12 (from VPA1310 to VPA1396) was exclusively conserved in KP-positive pathogenic strains and not in KP-negative strains (Fig. 1 and 3) (see the supplemental material). This cluster was previously described as the “pathogenicity island (Vp-PAI)” (16). We could not find any other genes that were unique and common to all of the KP-positive clinical strains, thus suggesting that VPC-12 has a strong correlation with pathogenicity in KP-positive V. parahaemolyticus.

FIG. 3.

Comparison of VPC-12 genes (VPA1310 to VPA1396) among pathogenic (both KP-negative and -positive) and nonpathogenic V. parahaemolyticus strains. Features of each strain are shown at the top. Blue, yellow, pale blue, and gray show, respectively, present, absent, divergent, and uncertain genes. Representative gene functions are indicated on the right.

KP has long been considered a phenotypic marker of the pathogenic V. parahaemolyticus strains, and TDH is known to be responsible for KP (8, 22, 30). The tdh genes are embedded in VPC-12. In addition, it has been reported that other putative virulence genes, including TTSS2, are encoded in VPC-12 (20). The involvement of the TTSS2 genes in the cytotoxicity and enterotoxicity of V. parahaemolyticus was demonstrated previously (15, 31). Our present findings strongly suggest that not only the tdh gene but the whole region of VPC-12 is required for the pathogenicity of KP-positive clinical V. parahaemolyticus strains.

Although this Vp-PAI of ca. 80 kb had previously been predicted on chromosome 2 of strain RIMD2210633 because of its low G+C content (39.8%) compared with the average G+C content of the genome (45.4%) (10, 16), the genome sequence data were not sufficient to define the boundary of this region. Our genomic comparison between KP-positive clinical strains and environmental ones more precisely defined the boundaries of the Vp-PAI (VPA1310 to VPA1396) (see the supplemental material). Investigation of the sequences around the boundaries may provide clues as to how the KP-positive clinical isolates of V. parahaemolyticus acquired the Vp-PAI.

The Vp-PAI was not identified in the KP-negative clinical strains (strains 4 to 8) (Fig. 1) (see the supplemental material). Many of the KP-negative clinical isolates are known to possess the TDH-related hemolysin (trh) and urease operon (ure) genes in close proximity to each other (11, 12, 29), and the KP-negative clinical isolates used in this study also possessed the trh and ure genes (see the supplemental material). Since the presence of a pathogenicity island-like structure has been predicted on the genome of KP-negative strains (12), investigation of the regions around the trh and ure genes may lead to the identification of additional virulence factors or genomic island(s) unique to the KP-negative clinical isolates of V. parahaemolyticus.

Comparative genomic analysis by microarray is a powerful tool to identify the characteristic genes or chromosomal regions within a species. In the study presented here, we used this type of analysis for V. parahaemolyticus and were able to identify the genes and genomic regions that are specifically present in pathogenic or pandemic strains. In particular, it was demonstrated that VPC-12 on chromosome 2 has a strong correlation with the pathogenicity of KP-positive V. parahaemolyticus. However, how the genes within VPC-12 contribute to the pathogenesis of KP-positive V. parahaemolyticus is not yet fully understood. Expression profiling analysis using microarray and functional analysis of putative virulence genes and hypothetical genes within VPC-12 can be expected to lead to a more detailed understanding of the mechanism of pathogenicity by V. parahaemolyticus.