EHEC strains used in this study were isolated from patients in Japan, Italy, or France (Table 1). The XbaI digestion patterns examined by pulsed field gel electrophoresis (PFGE) showed that the genomic DNA of EHEC strains is significantly divergent (Additional data file 1), while all possess stx1 and/or stx2 genes, and the eaeA gene encoding intimin (see 'Detection and subtyping of stx and eaeA genes' in Materials and methods). The results of the fluorescent actin staining (FAS) assay 22 indicated that all strains are potentially capable of inducing A/E lesions except for O111 strain 1. The efficiency, however, somewhat varied from strain-to-strain (data not shown).

The MLST analysis using seven housekeeping genes (aspC, clpX, fadD, icdA, lysP, mdh, and uidA) indicated that strains belonging to the O157, O26, O111, and O103 serotypes were clustered into three different phylogenic groups (O26 and O111 strains were clustered together; Additional data file 2). This result is basically consistent with those from previous MLST analyses using different genetic loci 2123. The type of intimin was classified as γ1, β1, γ2, and ε for O157, O26, O111, and O103, respectively.

The I-CeuI digestion of chromosomal DNA yielded seven fragments in 26 out of 29 EHEC strains (data not shown). Because I-CeuI specifically cleaves a 19 base-pair sequence in the 23S ribosomal RNA gene, it demonstrated that these strains have seven copies of the ribosomal operon (rrn), as in K-12 and O157. Estimated chromosome sizes of these strains were all much larger than that of K-12, with diverged sizes ranging from 5,102 to 5,945 kb (Table 2). O111 and O103 strains contained slightly smaller chromosomes than O157 strains. In contrast, most O26 strains contained relatively larger chromosomes. We could not estimate the chromosome sizes in two O157 strains (2 and 9) and one O103 strain (4), because all or the largest fragments repeatedly exhibited smear patterns.

Plasmid profiles indicated that all but one O157 strain contain one large plasmid of a similar size (Table 2; Additional data file 3). All of the non-O157 EHEC strains also contained at least one large plasmid except for O26 strain 1 (one small plasmid was present) and O103 strain 2 (no plasmid was detected). Several O26 and O111 strains possessed two or three large plasmids. The estimated total genome sizes of EHEC strains ranged from 5.27 Mb to 6.21 Mb.

We analyzed the gene contents of non-O157 EHEC strains by using the O157 oligoDNA microarray, and compared the results with those of O157 strains in our previous report 18 (Figures 1 and 2). More Sakai genes were absent from the non-O157 EHEC strains. In O157 strains, the absent genes were found mostly in Sp and SpLE regions, but in non-O157 EHEC strains, they were found not only in Sp and SpLE regions but also in various S-loops. The conservation tended to exhibit a serotype-specific pattern, but remarkable strain-to-strain diversity was also observed in each serotype.

To more precisely analyze the CGH data, we categorized the Sakai genes into three groups 18. Since most Sakai genes were represented by two oligonucleotide probes in our microarray, we first classified the probes into two groups by their homologies to the K-12 genome sequence; those with ≥90% identity into 'conserved in K-12' probes and others into 'Sakai-specific' probes. Each gene was then classified into 'conserved in K-12' genes, 'partly conserved in K-12' genes (genes represented by one 'conserved in K-12' probe and one 'Sakai-specific' probe), or 'Sakai-specific' genes. Repeated gene families that occurred in O157 Sakai more than once were analyzed separately from singleton genes (see Materials and methods for details on the classification and the presence or absence determination).

'Conserved in K-12' singleton genes were highly conserved in all serotypes: 3,596 (98.5%), 3,450 (94.5%), 3,331 (91.2%), and 3,542 (97.0%) out of 3,651 genes were fully conserved in O157, O26, O111 and O103, respectively, and 3,240 (88.7%) in all the test strains (Figure 3; Additional data file 4). 'Sakai-specific' singleton genes were relatively well conserved in O157 strains, but very poorly in non-O157 EHEC strains: 741 (64.3%), 221 (19.2%), 300 (26.0%), and 231 (20.0%) out of 1,153 genes were fully conserved in O157, O26, O111, and O103, respectively. Only 98 (8.5%) were conserved in all the test strains.

Among the 4,905 singleton genes, 101 were categorized as 'partly conserved in K-12' genes. They included 81 genes that are encoded on the backbone and 20 genes on S-loops or backbone/S-loop junctions. In O157, all but 5 (95.0%) of the 'partly conserved in K-12' genes were fully conserved. In non-O157 EHECs, however, many 'partly conserved in K-12' genes were categorized as 'uncertain' (7 to 42 genes in each non-O157 EHEC strain, 28 genes on average), because only one of the two probes yielded positive results. Therefore, only 44 (43.6%), 40 (39.6%), and 58 (57.4%) were fully conserved in O26, O111, and O103, respectively (Figure 3; Additional data file 4). This result suggests that most of the 'partly conserved in K-12' genes are present in the non-O157 EHEC strains but many have significantly divergent sequences from those of O157 Sakai.

O157 Sakai contains many repeated genes (542 out of 5,447 genes), such as transposase- and phage-related genes. They can be grouped into 151 families. Compared with the singleton genes, the repeated gene families were relatively well conserved in non-O157 EHECs. About half of the 'conserved in K-12' repeated gene families (11 out of the 23 families (47.8%)) were fully conserved in all the test strains, and 81 (63.3%), 74 (57.8%), 60 (46.9%), and 77 (60.2%) out of the 128 'Sakai-specific' repeated gene families were fully conserved in O157, O26, O111, and O103, respectively (Figure 3; Additional data file 4). Because most of the repeated genes were from lambda-like prophages and IS elements 818, this result indicates that non-O157 EHEC strains also contain multiple lambda-like prophages and IS elements very similar to those found in O157 Sakai.

Among the 3,651 'conserved in K-12' singleton genes, 224 (6.1%) were absent in at least one test strain. These genes were found to be absent more frequently in non-O157 EHEC strains than in O157 strains: 75 genes (2.1%) in O26 strains, 184 (5.0%) in O111, and 61 (1.7%) in O103, while only 37 (1.0%) in O157 (here we counted only the genes that were judged as 'absent' in at least one strain; therefore, these results do not include the genes that were 'uncertain' in some strains but 'absent' in no strain). These genes were dispersed on the chromosome and belonged to various functional categories (Additional data file 5); but as expected, none of them was listed as essential, either in the 'profiling of E. coli chromosome' (PEC) database 24 or in a systematic single-gene deletion study of E. coli K-12 25. We also identified 46, 83, and 30 'conserved in K-12' singleton genes that are fully absent in O26, O111, and O103, respectively. Among these, 22 genes, which are located in 12 different chromosomal loci, were absent in all non-O157 EHEC strains, and 10, 44, and 3 genes were specifically missing in O26, O111, and O103, respectively.

We categorized 'Sakai-specific' singleton genes according to the COG (clusters of orthologous groups of proteins) classification 26, and analyzed the gene conservation of each functional category (Figure 4). In O157, most genes were well conserved in all categories. Many genes for 'replication, recombination and repair' and for 'transcription' were variably present among O157 strains, but most of them were on Sps and SpLEs. In the non-O157 serotypes, however, the 'Sakai-specific' singleton genes belonging to almost every COG functional category exhibited poor conservation (many were classified as 'Fully absent'). The level of conservation was similar to that observed for the four sequenced pathogenic E. coli strains of different pathotypes 27282930 (Additional data file 4).

A relatively large number of genes for 'carbohydrate transport and metabolism' were fully conserved in non-O157 EHECs. Among these, genes for the sugar ABC transporter system (ECs0374-0378), and the N-acetylgalactosamine-specific PTS system (ECs4013-4014), and two genes for sugar utilization (ECs3242: fructokinase and ECs3243: sucrose-6 phosphate hydrolase) were conserved in all the tested strains. A relatively large number of genes for the 'cell wall/membrane biogenesis' category were also fully conserved. Most of them were the genes for lipopolysaccharide core biosynthesis (ECs2831 and ECs2836-2845). This is consistent with the fact that four serotypes examined here share the same core type (R3) 3132.

SpLE1 carries gene clusters for urease (ECs1321-1327) and tellurite resistance (ECs1343, 1351-1358). In an earlier report, the urease genes were found specifically associated with EHEC strains irrespective of their serotypes 33. Our present data, however, demonstrate that five EHEC strains (one O157, one O26, and three O103 strains) lack the urease genes. The tellurite resistance genes were also well conserved in non-O157 EHECs but absent in one O26 and two O103 strains.

In the COG classification, many of the virulence-related genes were classified into the 'not in COGs' category. We thus picked up all the known or suspected O157 virulence-related genes, and analyzed their conservation in non-O157 EHECs. Fimbria are important for virulence as an initial attachment factors to the host intestine. The O157 Sakai genome contained 14 fimbrial biosynthesis gene clusters (loci 1 to 14), all of which were completely conserved in every O157 strain except for strain 8, in which locus 11 was partially conserved (Table 3). Among the 14 clusters, four (loci 3, 5, 7, and 14) were completely conserved in K-12 and three (loci 1, 8, and 11) partially conserved. These seven loci were also completely or partially conserved in the non-O157 EHEC strains, suggesting that these gene clusters are widely conserved in various E. coli strains irrespective of their pathotypes. Genes on the remaining seven loci were almost completely absent in all non-O157 serotypes. Only loci 9 and 10 were partially conserved in several non-O157 EHEC strains. Thus, we may regard them as O157-specifc fimbrial gene clusters.

In addition to the fimbrial genes, 14 Sakai genes have been demonstrated or suspected to encode non-fimbrial adhesins (Table 4). They were relatively well conserved in the non-O157 EHEC strains. 'Regulators' and 'Toxins and their activators' showed similar levels of conservation as the genes related to adhesion (Table 4).