Molecular characterization of cytolethal distending toxin gene-positive Escherichia coli from healthy cattle and swine in Nara, Japan

Background Cytolethal distending toxin (CDT)-producing Escherichia coli (CTEC) has been isolated from patients with gastrointestinal or urinary tract infection, and sepsis. However, the source of human infection remains unknown. In this study, we attempted to detect and isolate CTEC strains from fecal specimens of healthy farm animals and characterized them phenotypically and genotypically. Results By PCR analysis, the cdtB gene was detected in 90 and 14 out of 102 and 45 stool specimens of healthy cattle and swine, respectively, and none from 45 chicken samples. Subtypes of the cdtB genes (I to V) were further examined by restriction fragment length polymorphism analysis of the amplicons and by type-specific PCRs for the cdt-III and cdt-V genes. Of the 90 cdtB gene-positive cattle samples, 2 cdt-I, 25 cdt-III, 1 cdt-IV, 52 cdt-V and 1 both cdt-III and cdt-V gene-positive strains were isolated while 1 cdt-II and 6 cdt-V gene-positive were isolated from 14 cdtB positive swine samples. Serotypes of some isolates were identical to those of human isolates. Interestingly, a cdt-II gene-positive strain isolated from swine was for the first time identified as Escherichia albertii. Phylogenetic analysis grouped 87 E. coli strains into 77 phylogroup B1, 6 B2, and 4 D, respectively. Most of the B1 strains harbored both lpfA O113 and ehaA. Three and twenty-two cdt-V gene-positive strains harbored eaeA and stx genes, respectively, and seven possessed cdt-V, stx and subAB genes. The cnf2 gene, normally present in cdt-III gene-positive strains, was also detected in cdt-V gene-positive strains. Conclusions Our results suggest that healthy cattle and swine could be the reservoir of CTEC, and they could be a potential source of human infections.

The cdt operon contains three adjacent genes, cdtA, cdtB and cdtC, and expression of all the genes is necessary for maximum toxin activity. While CdtB acts as an active subunit with DNase I activity, CdtA and CdtC facilitate binding of CDT to a yet-to-be-identified receptor molecule(s) on susceptible cells and entry of CdtB into the cytoplasm. As a result, CDT induces distention and eventual death of certain cultured eukaryotic cell lines by causing an irreversible arrest of the cell cycle at the G 1 or G 2 phase [4].
In CDT-producing E. coli (CTEC), five subtypes of CDT (I through V) have been reported based on the amino acid sequences and the genomic location of their genes [4]. Although CTEC strains have been isolated from children with diarrhea [4], case control studies conducted in children up to 5 years of age in Brazil (used DNA probes for CDT-I) [5], Bangladesh (for CDT-I) [6] and Nigeria (for CDT-I and CDT-II) [7] failed to demonstrate significant association of CTEC with acute diarrhea. However, animal experiments with recombinant CDT of Shigella dysenteriae and Campylobacter jejuni CDT knockout mutants indicated that CDT is involved in diarrhea and inflammatory response [2]. Moreover, Pandey et al. [8] reported that high titer CDT-I-producing enteropathogenic E. coli (EPEC) were isolated from patients with bloody diarrhea in India while low titer producers were isolated from patients with acute watery diarrhea. We also demonstrated that an E. coli strain isolated from a child with bloody diarrhea in Japan, which was initially suspected to be Shiga toxin-producing E. coli (STEC), did not possess the stx genes rather it produced CDT-I by a retrospective analysis [9]. Furthermore, we have recently reported presence of various subtypes of the cdtB (cdt-I to cdt-V) genes in diarrheal stool specimens of children at a high rate (~9.7%). Moreover, out of 30 CTEC isolates, which produced any of the 5 subtypes of CDT (CDT-I to CDT-V), 23 were isolated as a sole pathogen [10] suggesting possible association of CTEC with diarrhea in children.
E. coli normally resides in the intestine of warm-blooded animals which are suspected to be the reservoir and possible source of human infection of pathogenic E. coli. For example, major natural reservoirs for STEC, one of the most important groups of food-borne pathogens, have been established to be domestic ruminants, such as cattle, sheep, and goats [11]. During the processing of carcasses, fecal contamination or transfer of bacteria from animal's skin to the carcass can facilitate transmission of STEC to the meat [12]. Indeed, on a number of occasions, CTEC also have been isolated from various farm animals [13][14][15][16], and these were associated with diseased animal.
In this study, we attempted to detect cdtB gene in stool specimens of apparently healthy domestic animals including cattle, swine and chickens from Nara prefecture in Japan. We further isolated and characterized CTEC strains from these farm animals by serotyping, phylogenetic grouping and virulence gene profiling and compared with the strains of human origin.

Detection and isolation of cdtB gene-positive bacteria
For analyzing the presence of CTEC in healthy farm animals, 102 stool specimens collected from cattle in a farm and 45 rectal swabs collected from swine and chickens in another farm were subjected to PCR-RFLP analysis which can specifically amplify so far known E. coli cdtB genes followed by subtyping them as cdt-I to cdt-V based on restriction site polymorphism. As shown in Table 1, 90 and 14 samples from cattle and swine, respectively, produced a 588-bp long PCR fragment containing the cdtB gene, while no PCR product was obtained using samples of chicken origin. The 90 cdtB gene-positive amplicons obtained from cattle stools were found to be comprised of 2 cdt-I, 87 cdt-III/V and 1 cdt-IV. Although same number of bacterial strains carrying the cdt-I and cdt-IV genes was successfully recovered, in the case of cdt-III/V, 78 bacterial isolates were obtained out of 87 PCR-positive cases. Similarly, the 14 amplicons derived from swine samples were identified as 1 cdt-II and 13 cdt-III/V. Analysis of bacterial cells allowed us to recover 1 and 6, as cdt-II and cdt-III/V, respectively ( Table 1). The cdtB-positive isolates were confirmed to carry cdtA, cdtB and cdtC genes by colony hybridization using corresponding gene probes (data not shown).

Discrimination of cdt-III/V-positive bacteria
We attempted to further discriminate cdt-III/V-positive bacteria by type-specific PCR assays as reported previously [10,17]. However, the type-specific PCR failed to differentiate cdt-III and cdt-V genes in 2 cdt-V genepositive E. coli (CTEC-V) OUT:H48, 1 both cdt-III and cdt-V gene-positive E. coli (CTEC-III and V) of cattle, and 5 CTEC-V O98:H10 and 1 OUT:HUT of swine as indicated by asterisk in Table 2. Therefore we developed new type-specific PCR primers for cdt-III and cdt-V genes in this study as shown in Figure 1. Using these primers all cdt-III/V positive isolates were clearly differentiated according to the subtypes of cdt, except for one isolate in which both cdt-III and cdt-V genes were detected as given in Table 2. Finally, among 81 cdtB genepositive isolates of cattle origin, 2 were found to harbor cdt-I, 25 cdt-III, 1 cdt-IV, 52 cdt-V and 1 both cdt-III and cdt-V, whereas 1 and 6 out of 7 cdtB gene-positive isolates from swine contained cdt-II and cdt-V, respectively.

Identification of CTEC
All cdtB gene-positive isolates from cattle and swine were confirmed as E. coli by biochemical tests except for a cdt-II gene-positive strain from swine (strain Sw-9). By API 20E testing, the strain Sw-9 was identified as E. coli (74.6%) with a doubtful api profile of 51445021 (https://apiweb. biomerieux.com/jsp). However, unlike typical E. coli, strain Sw-9 was nonmotile at 37°C and indole-negative, did not  Cattle CDT-I O112ac:H20 B1 1    . Sugar utilization tests of dulcitol, D-mannitol, D-melibiose, L-rhamnose and D-xylose also suggested that strain Sw-9 was E. albertii and not as E. coli [18,19]. Multilocus sequence (MLS) analysis based on the nucleotide sequence variation at 7 housekeeping loci (a total of 3,423 bp) in the genome revealed that strain Sw-9 belongs to the E. albertii lineage (Figure 2), consistent with the data of biochemical tests and 16S rRNA gene sequencing. Considering these findings together, the strain Sw-9 was identified as E. albertii.

Serotyping and phylogenetic grouping
To characterize the CTEC strains further, their serotype and phylogenetic groups were determined ( Table 2) Table 2. One isolate which was positive for both cdt-III and cdt-V genes was identified as O2: HUT (B2). Five and one CTEC-V isolates from swine were identified as O98:H10 (B1) and OUT:HUT (B1), respectively. Interestingly, the E. albertii strain Sw-9 showed cross reaction with the E. coli O84 antiserum.

Virulence gene profile
To analyze the virulence gene profile of the CTEC and E. albertii strains isolated in this study, genes for DEC, NTEC and putative adhesins reported in STEC (see details in Material and Methods section) were investigated by colony hybridization assays (Table 2). In agreement with the previous report [20], all the CTEC-III strains possessed the cnf2 gene, indicating that cdt-III of these strains could be located on pVir-like plasmid. Surprisingly, 7 of the CTEC-V strains also possessed cnf2. The eaeA gene that encodes an outer membrane protein called intimin, which is necessary for intimate attachment of EPEC and EHEC strains to epithelial cells, was detected in the E. albertii strain Sw-9 from swine and all of the 3 CTEC-V O156:HUT (B1) strains from cattle ( Table 2). The intimin subtype of three CTEC-V O156 strains was determined as θ/γ2 by PCR-RFLP, but the amplicon was not obtained in E. albertii strain Sw-9. Sixteen CTEC-V isolates (6 O22, 10 OUT) were positive for the stx1 and stx2 genes, while 6 CTEC-V strains (5 O113, 1 OUT) were positive for only stx2. Cytotoxicity assay using Vero and CHO cells, which are susceptible and unsusceptible to Stx intoxication, respectively, indicated that all the stx gene-positive CTEC strains produced functional Stx (titer ranging from 16 to 128<) and CDT (1 to 64) ( Figure 3). However, 7 strains caused unexpected Figure 2 Neighbor-joining tree based on nucleotide variation at 7 conserved housekeeping loci. The partial gene sequences for 7 housekeeping loci of CTEC-II strains Sw-9 and AH-5, E. coli, E. fergusonii and E. albertii were concatenated in the order adk, fumC, gyrB, icd, mdh, purA and recA and aligned. Based on 3,423 bp of the concatenated sequences, a neighbor-joining tree was constructed by using MEGA 4 software. Figure 1 Schematic representation of PCR primer binding region of type specific PCR for cdt-III and cdt-V. White (Cdt-IIIAf, Cdt-IIICr and CdtIIIC-R), black (CdtVC-R, P2-A2, cdtA-F, cdtC-F and P2-C3) and gray (CdtIII/VB-F) arrows indicate PCR primers which specifically bind to cdt-III, cdt-V and both cdt-III and cdt-V genes, respectively. morphological change to CHO cells, indicating that these strains might produce a third toxin. Since the observed morphological change resembled to that induced by SubAB, an AB 5 toxin discovered in LEE-negative STEC [21], the 7 strains were subjected to PCR analysis specific to the subA and subB genes and all the strains were positive for both the genes. Collectively, these data indicate that the 7 E. coli strains produced CDT-V, Stx and SubAB toxins.

Discussion
Clinical importance of CTEC in humans including intestinal and extra-intestinal infections is not yet fully understood. Several studies, however, showed that on several occasions CTEC strains were isolated from patients with diarrhea, septicemia, or urinary tract infection [4], suggesting that CTEC might be associated with human diseases. To understand the possible reservoir and potential source of CTEC infection, we have screened feces of healthy farm animals (cattle, swine and chicken) for the presence of E. coli cdtB gene by a PCR-RFLP assay, which can detect and differentiate 5 subtypes of the E. coli cdtB gene [10]. In addition, we isolated CTEC strains from the cdtB gene-positive samples and characterized them for serotypes, virulence gene profiles and phylogenetic groups to compare with those of CTEC strains from diarrheal patients. There is a report regarding the isolation of CDT-V-producing E. coli O157 from healthy cattle by Tóth et al. [23]. In most of the previous studies, however, CTEC strains were isolated from diseased animals with various symptoms [13][14][15][16]. In this study, to avoid any bias, we have isolated CTEC strains from cdtB-positive fecal sample of apparently healthy cattle and swine.
A total of 81 and 7 CTEC strains have been isolated from 90 and 14 cdtB gene-positive fecal samples of cattle and swine, respectively ( Table 1). The 81 strains from cattle samples were grouped into 12 O serogroups and 31 O:H serotypes ( Table 2). In our previous work, we showed that CTEC-I belonging to the O2 serogroup and B2 phylogenetic group was most predominant among the CTEC strains isolated from children with diarrhea in Japan [10]. Although 6 CTEC strains belonged to the O2 serogroup and B2 phylogenetic group were isolated in this study, none of them were CDT-I producers (4 CTEC-III, 1 CTEC-V, and 1 CTEC-III and V). This may be because of different geographical background between clinical and animal samples collected. Alternatively although cattle and swine carry a variety of CTEC strains, all the CTEC strains in cattle and swine may not be associated with human diseases. Since all types of CTEC have been isolated from patients with diarrhea, CTEC strains found in cattle and swine in this study might be associated with  human diseases in future. Results obtained in this study indicate that further studies on prevalence of CTEC in food animals in several farms and meats are needed. Tóth et al. [23] reported the isolation of CDT-Vproducing E. coli O157 from healthy cattle in Hungary. However, all the CTEC strains isolated in the present study did not belong to O157 serogroup. It might be due to difference of the strategies. In their study, they tried to isolate only E. coli O157 from healthy cattle samples by using cefixime-tellurite-sorbitol-MacConkey agar and also by following the International Organization for Standardization reference method (ISO 16654) using an O157-specific immunomagnetic beads. On the other hand, we targeted CTEC by using PCR-RFLP for detection of all five subtypes of the E. coli cdtB gene. We further characterized only one strain from each cdtB gene-positive sample. Thus, we cannot exclude the possibility that CTEC O157 was present in our samples, but we could not isolate CTEC O157.
Presence of the cdt-I and cdt-IV, and cdt-III genes were reported to be strongly associated with that of the cnf1 and cnf2 genes, respectively [13,24]. It has also been reported that the cdt-III genes were located on a plasmid harboring the cnf2 gene [20], whereas cdt-V was chromosomal and carried by bacteriophage [25], suggesting that detection of the cnf2 gene could be one of the genetic markers to differentiate cdt-III and cdt-V gene-positive strains. Indeed, all the 25 strains with cdt-III were also positive for cnf2. However, 7 out of the 52 cdt-V genepositive strains from cattle also contained cnf2 and this gene arrangement has not yet been reported. Since homology between cdt-III and cdt-V genes is very high (cdtA, 97.3%; cdtB, 99.7%; cdtC, 96.5%) [4], it is difficult to differentiate the cdt-III and cdt-V genes by PCR, suggesting that some of the cdt-III and cdt-V genes might have been misidentified. In the present study, three PCR primer sets, cdt-IIIABC, cdt-Vup, cdt-Vdown, each targeting the internal region of cdt-III [10], the 5′ and 3′ flanking regions of cdt-V [17], failed in producing specific amplicon in 1, 9 and 3 strains, respectively, out of the 58 CTEC-V and 1 CTEC-III and V (Table 2). However, the typespecific PCR developed in this study using two primer sets each targeting cdt-III or cdt-V (Figure 1) could produce specific amplicon either for cdt-III or cdt-V. The cdt-IIIand cdt-V-specific PCR designed in this study is more reliable to differentiate these genes and to generate more precise epidemiological data. In fact, using the typespecific PCR, we identified a both cdt-III and cdt-V genepositive E. coli strain. To our knowledge, this is the first report to describe the isolation of CTEC-III and V strain.
Since reservoir for STEC has been identified to be ruminant such as cattle and this study also indicates that reservoir for CTEC could be the same, similar genes for adhesion might be associated with colonization of both STEC and CTEC. In addition to the eaeA gene, saa, iha, lpfA O113 and ehaA genes have also been reported to encode putative adhesins in STEC O157 and non-O157 [26][27][28][29]. Recently Wu et al. [22] described a probable association of these 4 genes, in particular lpfA O113 and ehaA genes, with the long-term STEC shedding from cattle. When virulence gene profiling, in particular, for adhesin were analyzed in this study, 86 and 83% strains from cattle and swine, respectively, were found to be positive for lpfA O113 and ehaA genes, while 100% stx gene-positive CTEC isolates were all positive for saa, lpfA O113 , ehaA and iha genes. Furthermore, almost all of them were positive for cdt-III or cdt-V whereas 2 strains were positive for cdt-I genes. In this study, 97% of cdt genes detected in the feces of cattle was cdt-III or cdt-V whereas only 2 and 1% of cdt genes were cdt-I and cdt-IV, respectively. Clark et al. [13] also reported that the cdt-III genotype was more prevalent in animal strains although the majority of cdt genotypes isolated from humans was cdt-I and cdt-IV [10]. Taken together, these data indicate that LpfA O113 and EhaA could be associated with adhesion of CTEC in cattle intestine, especially CTEC-III and CTEC-V.
Strain Sw-9 initially identified as CTEC-II O84:NM by biochemical test was re-identified as E. albertii, a newly emerging diarrheagenic pathogen [19], by a MLS analysis and sugar utilization tests. This may be the first report showing isolation of E. albertii from swine in Japan. Furthermore, this finding prompted us to reinvestigate if previously identified CTEC-II strains were of E. albertii or not. Indeed the CTEC-II strain AH-5, previously identified as OUT:NM [10], was found to be E. albertii (Figure 2). Ooka et al. [19] recently reported that 26 out of 179 eaeA gene-positive E. coli strains, isolated from humans, birds and the environment in Japan, were identified as E. albertii by MLS analysis and cdtB gene of CDT-II/III/V subtypes group was detected by PCR in all the E. albertii strains except 1 strain. EPEC isolates, previously identified as E. coli O86:K61 and contained the cdtB gene, were also identified as E. albertii [30]. The cdt genes of E. albertii strain 19982 (GenBank: AY696755) are highly homologous to the cdt-II genes present in E. coli strains. These data suggest that E. albertii might have been misidentified as not only EPEC but also CTEC-II. Since there is no reliable method to identify E. albertii other than MLS analysis to date, the development of simple and reliable identification method of E. albertii is required. The cdt-II genes could be one of useful genetic markers for this purpose although discrimination of E. albertii from true CTEC-II is still necessary.

Conclusions
We could isolate a number of CTEC strains from cattle and swine, which had diverse variations in serotype and genotype. Some of the CTEC strains possessed virulence genes associated with human diseases and serotype that are frequently detected among human clinical strains. Thus, cattle and swine could be possible reservoirs of CTEC and serve as potential sources of infection to human. To the best of our knowledge, this might be the first report regarding comprehensive surveillance and characterization of CTEC strains isolated from healthy food animals. Because of the limited number of animals and farms examined, further studies are of course needed to verify the probability that these animals are indeed the source of CTEC infection to humans.

Sample collection
In August 2004 in Japan, stool specimens from the rectum of 102 cattle (around 1 year of age), including 95 cross breeding cattle (from Bv-1 to Bv-95) and 7 Holstein cow (Bv-96 to Bv-102), and rectal swabs from 45 cross breeding swine (<6 month-old) and 45 broiler chickens (<1 year-old) were collected in Nara, Japan. The cattle were kept in several barns in a farm, the swine in several pens in a barn, and the chickens in a windowless broiler house. All the animals were healthy and asymptomatic. The samples were transported to the laboratory at ambient temperature and processed within 6 h of collection. Fecal sampling in the present study was approved by Laboratory of Animal Research of Nara Prefectural Livestock Experiment Station and performed according to the Guidelines for Animal Experimentation of Nara Prefectural Livestock Experiment Station.

Detection of cdtB gene by PCR
Aliquot of stool specimens and rectal swabs were inoculated into 3 mL of tryptic soy broth (Nissui Pharmaceutical Co., Tokyo, Japan) for enrichment and incubated overnight at 37°C with shaking. Fifty microliter of the culture was added into 450 μL of TE buffer (10 mM Tris-HCl, 1 mM EDTA [pH 8.0]) and boiled for 10 min. After centrifugation at 12,000 × g for 3 min, the supernatant was used as a template for PCR in a thermal cycler (GeneAmp PCR System 9700; Life Technologies, Carlsbad, CA, USA). In the PCR assay, the cdtB gene was detected by using the cdtB common primer set which can detect all five subtypes of the E. coli cdtB gene [10] ( Table 4). E. coli strain C600 and E. coli O86 strain GB1371 (harboring the cdt-I genes) were used as negative and positive controls, respectively. To examine the CDT subtypes, a PCR-RFLP assay was employed as reported previously [10]. Briefly, PCR products were digested by either EcoRI/EcoRV or MspI (Takara Bio Inc., Shiga, Japan) and the digests were analyzed by electrophoresis in 3.0% agarose gels (NuSieve 3:1 Agarose; Takara Bio Inc.). Since differentiation of cdt-III from cdt-V by PCR-RFLP assay was not successful, type-specific PCRs were performed to further discriminate cdt-III from cdt-V by using specific primers such as Cdt-IIIAf and Cdt-IIICr [10], P2-A2/ cdtAF and cdtC-F/P2-C3 [17] and newly designed primers such as CdtIII/VB-F/CdtIIIC-R and CdtIII/VB-F/CdtVC-R in this study (Figure 1). The primer sequences and PCR conditions are presented in Table 4.

Identification of cdt-harboring organisms
Enrichment culture in which cdtB gene was detected by the PCR was serially diluted in sterile 10 mM phosphate buffered saline (pH 7.4) and 100 μL of each dilution was spread on MacConkey agar (Difco Laboratories, Detroit, MI, USA). Colonies were transferred to nitrocellulose (Schleicher & Schuell, Dassel, Germany) or Hybond-N + membrane (GE Healthcare, Buckinghamshire, UK) by a replica blotting method and a colony hybridization assay was carried out by using specific DNA probes under high stringent condition. For preparation of probes, the cdt-IB, cdt-IIB, cdt-IIIB and cdt-IVB genes were PCR amplified using template DNAs isolated from E. coli strains NT3363 [8], AH-5, AH-6 and AH-8 [10], respectively, and common primer sets (Table 4) followed by labeling of each PCR product by random priming method using the MultiPrime DNA Labeling System (GE Healthcare) and (α-32 P)-dCTP (111 TBq/mmol) (Perkin Elmer, Wellesley, MA, USA). Hybridization positive colonies were detected from the corresponding master plate and reconfirmed by cdtB-specific PCR using the common primers (Table 4). To identify cdtB-positive colonies as E. coli, bacterial cells were further analyzed by the API 20E System (bioMérieux, Marcy-l'Etoile, France) and by conventional biochemical tests [31]. When the results of biochemical tests were ambiguous, further confirmation was done by 16S rRNA gene sequencing (approximately 500 bp in size) by using the MicroSeq 500 16S rDNA Sequencing Kit and an ABI PRISM 3100 Genetic Analyzer (Life Technologies). Serotyping was carried out by tube agglutination method using somatic (O1-O173) and flagellar (H1-H56) antisera [31], which were prepared at the Osaka Prefectural Institute of Public Health, Osaka, Japan.

Multilocus sequence analysis
Multilocus sequence (MLS) analysis was applied to the cdt-II-positive strain according to the protocol by University of Warwick (http://mlst.warwick.ac.uk) with minor modifications. Briefly, partial gene sequences for 7 housekeeping loci (adk, fumC, gyrB, icd, mdh, purA, recA) were determined by sequencing their PCR products using the BigDye Terminator Sequencing Kit (Life Technologies). Obtained sequences were aligned and trimmed to a uniform size by using Seqman (DNASTAR, Madison, WI, USA) and concatenated. Based on the concatenated sequences, a neighbor-joining tree was constructed using  [32].

Phylogenetic grouping of CTEC
Phylogenetic groups of each CTEC isolates were determined by PCR developed by Clermont et al. [33].

Detection of virulence genes
Presence of virulence genes including cdt in diarrheagenic E. coli (DEC) and necrotoxigenic E. coli (NTEC) and putative adhesin genes of STEC were analyzed by colony hybridization assays using appropriate DNA probes ( , CTEC strains AH-1 (cnf1), AH-5 (cdt-IIA, cdt-IIC), AH-6 (cdt-IIIA, cdt-IIIC, cnf2), AH-8 (cdt-IVA, cdt-IVC) and AH-10 (cdt-VA and cdt-VC) were used as positive controls. The DNAs of these control strains were also used as template to PCR amplify each of the virulence gene followed by preparation of DNA probes. The E. coli eaeA gene was PCR amplified using the eae-F and eae-R primer set and subtyped by PCR-RFLP with MspI as described previously [34].

Cytotoxicity assay
Cytotoxicity assay was performed as described earlier [10]. Briefly, test strains were grown overnight in 3 mL of tryptic soy broth at 37°C overnight with shaking. Bacterial cells were lysed by sonication using an Astrason ultrasonic processor (Heat-System 7 Ultrasonics, Farmingdale, NY, USA) and each sonic lysate was passed through sterile disposable filter with 0.22-μm pore size and each filtrate was used for cytotoxicity assay. Vero and CHO cells were seeded at density of 1 × 10 4 cells in a 96 well plate (Asahi glass Co., Ltd., Tokyo, Japan) respectively, and 20 μL of 2-fold serially diluted each toxin solution was added to assay their cytotoxic effects. After 9 h of incubation, 100 μL of fresh medium was added per well and cytotoxic effect of each test sample, if any, was examined microscopically after 72 h of incubation. The toxin titer was expressed as the reciprocal of the highest dilution that caused 50% of the Vero and CHO cells in a well to be killed and distended, respectively. E. coli strains Sakai and GB1371 were always used as positive controls and as a negative control we used E. coli strain C600. Vero and CHO cells were cultured in Minimum Essential Medium (MEM) and MEM-α (Life technologies), respectively, containing 10% fetal bovine serum (EuroClone S.p.A., Pero, Italy), and 1% antibiotic-antimycotic (100x) (Penicillin G sodium [10,000 U/mL], streptomycin sulfate [10,000 μg/mL], and 25 μg/mL amphotericin B in 0.85% saline [Life technologies]). Cells were cultured at 37°C under 5% CO 2 in air.

Sequence analysis of cdt-III and cdt-V
To determine the entire sequence of the cdt genes, the cdt gene-cluster including their flanking regions were PCR amplified followed by sequencing as previously described [10]. For the cdt-III genes, PCR product obtained by the pVir-u and pVir-d primers specific to the flanking eae-R ATATYTATTTGCWGSVCCCCAT region of cdt-III on the pVir plasmid was sequenced. For the cdt-V genes, PCR products obtained by the P2-A2 and CdtVC-D2 primers and the CdtIII/VB-F2 and P2-C3 primers were sequenced (Figure 1). Each PCR product was purified by the QIAquick PCR Purification Kit (QIAGEN, Hilden, Germany) and the nucleotide sequence of the PCR product was determined as described above. Nucleotide and amino acid sequences were analyzed and compared with each subtype using the BLAST program through the DDBJ (DNA Data Bank of Japan), and the DNA Lasergene software package (DNASTAR).

Nucleotide sequence accession numbers
All nucleotide sequences obtained in this study have been registered in the DDBJ database. The accession numbers are AB839651-AB839676 (for the cdt genes) and AB839677-AB839690 (for 7 housekeeping genes used for MLS analysis).