Evolution of Enterohemorrhagic Escherichia coli Hemolysin Plasmids and the Locus for Enterocyte Effacement in Shiga Toxin-Producing E. coli

Abstract

This study assessed the diversity of the enterohemorrhagic Escherichia coli (EHEC) hemolysin gene (ehxA) in a variety of Shiga toxin-producing E. coli (STEC) serotypes and the relationship between ehxA types and virulence markers on the locus for enterocyte effacement (LEE). Restriction fragment length polymorphism of the ehxA gene and flanking sequences and of the E. coli attaching and effacing (eae) gene was determined for 79 EHEC hemolysin-positive STEC isolates of 37 serotypes. Two main groups of EHEC hemolysin sequences and associated plasmids, which corresponded to the eae-positive and the eae-negative isolates, were delineated. Comparisons of the ehxA gene sequences of representative isolates of each group showed that this gene and the rest of the EHEC hemolysin operon are highly conserved. Digestion of an ehxA PCR product with the restriction endonuclease TaqI showed a unique restriction pattern for eae-negative isolates and another one for isolates of serotypes O157:H7 and O157:NM. A conserved fragment of 5.6 kb with four potential open reading frames was identified on the EHEC hemolysin plasmid of eae-positive STEC. Phylogenetic analysis of a subset of 27 STEC isolates, one enteropathogenic E. coli isolate, and a K-12 reference isolate showed that eae-positive STEC isolates all belong to a single evolutionary lineage and that the EHEC hemolysin plasmid and the ehxA gene evolved within this lineage without recent horizontal transfer. However, the eae gene and the LEE appear to have been transferred horizontally within this STEC lineage on several occasions. The reasons for the lack of transfer or maintenance of the LEE in other STEC lineages are not clear and require further study.

Shiga toxin-producing Escherichia coli (STEC) isolates have emerged as an important group of enteric pathogens with clinical manifestations that include asymptomatic carriage, watery diarrhea, hemorrhagic colitis, and severe and life-threatening hemolytic-uremic syndrome (2, 21, 28). Both sporadic and epidemic infections with STEC have been increasingly reported in North America and more recently in other parts of the world (2, 21, 28). The majority of clinical STEC infections and outbreaks have been associated with E. coli serotype O157:H7 (21). However, the frequency of this unique serotype may be, at least partially, the result of methodological bias (19), and the diversity of STEC strains and serotypes associated with clinical disease in humans is relatively high (28).

In addition to producing Shiga toxins (30, 35), most STEC isolates involved in human disease are able to attach tightly to enterocytes and to induce a reorganization of the underlying cell structure (14). The related cascade of events is referred to as enterocyte attachment and effacement and presents similarities to an equivalent mechanism in enteropathogenic E. coli (EPEC) (13). The genes responsible for this phenomenon, including the E. coli attaching and effacing gene (eae) and EPEC-secreted protein B (espB), are clustered within the locus of enterocyte effacement (LEE) located on the chromosome of EPEC and STEC (32). In EPEC, expression of the eae gene and several other chromosomal genes is regulated by products of an operon, the per locus, located on the large EPEC adherence factor (EAF) virulence plasmid (20). Despite the presence of large plasmids in disease-associated STEC isolates (enterohemorrhagic E. coli [EHEC]), no equivalent of the per locus has been found in these STEC strains (20). However, the large plasmid found in STEC encodes a toxin of the RTX family (3, 50) called enterohemorrhagic E. coli hemolysin (EHEC hemolysin) (42). This toxin is present in most isolates of EHEC serotypes (5, 6, 40, 42), and has been proposed as a marker for detection of EHEC (7).

The EHEC hemolysin gene (ehxA) of serotype O157:H7 STEC and its associated sequences (ehxC, ehxB, and ehxD) (42) have been studied extensively (3, 42), but little information on the ehx genes of other serotypes is available (43). The aim of the present work was to investigate the diversity of the EHEC hemolysin gene in a large variety of STEC serotypes of diverse origins and to examine the relationship of ehxA with virulence markers on the LEE and with the total genomic background of STEC. The results of this study will aid in understanding the evolution of EHEC hemolysin genes, as well as the plasmid on which they are carried, and of the LEE found in STEC populations.

MATERIALS AND METHODS

Bacterial isolates.

A total of 79 STEC isolates which produced EHEC hemolysin and belonged to 37 serotypes were studied (see Fig. 1). Forty-nine STEC isolates originated from the feces of healthy cattle in Ontario, Canada (40), and 30 were of human origin. Among the latter, 3 isolates were from Canada, 12 were from the United States, 9 were from Germany, 5 were from Australia, and 1 was from Swaziland. The EPEC strain 2348/69 (serotype O127:H6) was used for comparative analysis of the eae gene. Three ehxA- and eae-negative bovine STEC isolates of serotypes O113:H4, O153:H31, and O156:NM and the E. coli K-12 strain JM109 (55) were used for the genomic analysis.

FIG. 1.

Estimated degrees of nucleotide substitutions between variants of the ehxA gene of 79 STEC isolates. The dendrogram was generated by the UPGMA method of clustering by using a matrix of the proportion of nucleotide differences between pairs of variants based on restriction maps of a 2,862-bp PCR product of the ehxA gene. The numbers in the columns labeled EcoRI and PstI represent subtypes obtained by restriction analysis of the sequences upstream of the ehxA gene. The numbers and letters in the column labeled eae represent subtypes obtained by RFLP analysis of the eae gene, where numbers 1 and 2 indicate the two major clusters; a minus represents an eae-negative isolate. The numbers in parentheses represent the number of isolates for each subtype. NT, nontypeable because of resistance to endonuclease digestion due to methylation or because of degradation by high levels of nucleases in the plasmid preparations; B, bovine; H, human; AU, Australia; CA, Canada; D, Germany; SW, Swaziland; US, United States.

Serotyping.

All isolates were serotyped at the Health of Animals Laboratory, Health Canada, Guelph, Ontario.

PCR amplification of ehxA, eae, and espB.

Cultures of the isolates were grown overnight on Trypticase soy agar (Difco Laboratories, Detroit, Mich.). A loopful of culture was scraped from the plate and then washed once in phosphate-buffered saline solution (pH 7.4). The bacteria were resuspended in 1 ml of water, boiled for 10 min, and then pelleted by centrifugation; the supernatant was used as the DNA extract for PCR assays. A 2,862-bp fragment of the ehxA gene (bases 71 to 2932) was amplified by using the primers 5′-CATCATCAAGCGTACGTTCC-3′ (hly3) and 5′-ATGCTAATCGTTCATCACCT-3′ (hly6). The amplification reactions were performed in a final volume of 100 μl containing 1× PCR buffer (Gibco-BRL, Gaithersburg, Md.), 0.2 mM deoxynucleoside triphosphates, 1.5 mM MgCl₂, 0.5 μM each primer, 2.5 U of Taq polymerase (Gibco-BRL), and 5 μl of the DNA extract. Thirty-five amplification cycles were performed; each cycle consisted of 40 s of denaturation at 96°C, 60 s of annealing at 61°C, and 90 s of extension at 72°C, followed by a final extension of 10 min at 72°C. A 1,110-bp fragment of the eae gene (bases 803 to 1912) was amplified by the method of Sandhu et al. (41). PCR conditions described by Wieler et al. (53) were used for the espB gene.

Restriction fragment length polymorphism (RFLP) analysis of PCR products.

A 5-μl volume of the amplification products was digested with 5 to 10 U of restriction endonuclease for 3 to 4 h in a final volume of 20 μl. The enzymes HindIII, HinfI, Sau3A, and TaqI (Pharmacia Biotech, Uppsala, Sweden) and MspI and RsaI (Boehringer, Mannheim, Germany) were used to digest the ehxA amplification products, and the enzymes HaeIII (Pharmacia Biotech), HinfI, TaqI, MspI, and RsaI were used to digest the eae amplification products. The resulting DNA fragments were analyzed by electrophoresis in 2 to 4% agarose gels in TAE (40 mM Tris-acetate, 1 mM EDTA [pH 8.5]) and stained for 30 min in a 1-μg/ml ethidium bromide solution. Restriction maps were constructed for each restriction pattern by using the known sequences of the ehxA (42, 43) and eae genes (56). The magnitude of nucleotide differences between pairs of isolates for the sequences under investigation was estimated by using the proportion of shared restriction sites (31).

A distance-based (phenetic) analysis of the resulting pairwise divergence matrix was carried out by using the unweighted pair-group method with arithmetic means (UPGMA) (44). The relationships between variants of the ehxA genes and between the eae genes were additionally confirmed by the Dollo parsimony method (17) using the PAUP program v3.1.1 (47), a character-based (cladistic) method. These two phylogenetic procedures differ in their use of the data set. In contrast to the phenetic approach, which transforms restriction site variation to a single divergence value for all pairwise comparisons of the genes of the isolates, the cladistic method uses only those restriction sites that are derived and shared by two or more variants to infer patterns of ancestry. Confidence is increased when these two methods converge on a single topology of phylogenetic relationships.

Plasmid isolation.

Small-scale plasmid preparations were obtained by the alkaline lysis method (39), with one phenol-chloroform extraction. Treatment with proteinase K (Boehringer; 50 μg/ml for 60 min at 55°C) followed by two phenol-chloroform extractions was necessary for isolates with high nuclease activity (mainly eae-negative isolates) to avoid subsequent degradation of the plasmid DNA during digestion with restriction enzymes.

RFLP of plasmids, transfer, and hybridizations.

Each plasmid preparation (10 μl) was digested with 20 U of either EcoRI or PstI (Pharmacia Biotech) for 4 h in a final volume of 20 μl, as described by the manufacturer. The resulting fragments were separated in 0.7% agarose gels in TAE. The DNA was transferred by vacuum blotting on Hybond N membranes (Amersham Life Science, Little Chalfont, England) after depurination and denaturation and fixed by UV cross-linking. A probe specific for ehxA was prepared by random labeling of the hly3-hly6 PCR product of the plasmid pEO40 (42) by using the digoxigenin (DIG) DNA labeling kit (Boehringer) as described in the instructions of the manufacturer. Hybridization reactions were done overnight at 37°C in hybridizing solution containing 50% formamide, 2× SSC (20× SSC is 3 M NaCl plus 0.3 M sodium citrate [pH 7.0]), and 2% blocking reagent (Boehringer). Washing was done under high-stringency conditions at 65°C in 0.2× SSC. The hybridization patterns were revealed by chemiluminescence using the DIG luminescent kit (Boehringer) as described in instructions of the manufacturer. For the comparison of EHEC plasmids from strain 4304 (serotype O157:H7) and strain EC930073 (serotype O22:H8), plasmid DNA from the two strains and from the EPEC strain 2348/69 was extracted from 100-ml overnight cultures in LB broth by using the Qiagen (Santa Clarita, Calif.) plasmid extraction kit. Probes were prepared by random priming with the DIG DNA labeling kit using the whole plasmids after digestion with the restriction endonuclease HinfI and subsequent purification with the Gene Clean kit (Bio 101 Inc., La Jolla, Calif.). Hybridizations were performed as described above.

Cloning and sequencing.

Plasmids were prepared from 100-ml overnight cultures in LB broth by using the Qiagen plasmid extraction kit. After restriction digestion and electrophoresis, the desired plasmid fragments were purified with the Gene Clean kit (Bio 101) and ligated into the pBluescript II KS(+/−) vector (Stratagene, La Jolla, Calif.) by using standard protocols (39). The resulting constructs were transformed by either calcium chloride treatment and heat shock (39) or electroporation with an E. coli pulser (Bio-Rad, Hercules, Calif.) into E. coli JM109, as described in the instructions of the manufacturer. The cloned fragments were sequenced by the chain termination method with the ABI PRISM dye terminator sequencing ready reaction kit (PE Applied Biosystems, Foster City, Calif.) and the ABI PRISM 377 DNA sequencer (PE Applied Biosystems), as described in the instructions of the manufacturer. Both strands were sequenced for each fragment. The resulting sequences were aligned with known sequences by use of the ALIGN program (Scientific & Educational Software, State Line, Pa.). Potential open reading frames (ORFs) were identified with the Gene Runner program 3.4 (Hastings Software Inc.). A search for homologous sequences in the GenBank, EMBL, DDBJ, PDB, and SwissProt databases was done with the BLAST program (1). A search for prokaryotic promoter sequences was done with the Promoter Prediction by Neural Network program available at the Baylor College of Medicine internet site (http://gc.bcm.tmc.edu:8088/search-launcher/launcher.html), with a score cutoff of 0.80. Potential ribosome binding sites were searched by inspection of the region preceding each ORF.

AFLP.

Amplified fragment length polymorphism (AFLP) analysis is a powerful method for gathering phylogenetically informative DNA fingerprints because it combines the principle of RFLP with the capacity to sample the whole genome by using selective PCR. Moreover, a large number of markers on a single automated gel apparatus can be assayed. For the present study, cells were grown overnight at 37°C in brain heart infusion broth (Becton Dickinson, Cockeysville, Md.). Five hundred microliters of each culture was centrifuged and used for DNA isolation by the protocol recommended for use with the Nucleon I kit (Scotlab, Lanarkshire, Scotland). The cells were resuspended in 340 μl of reagent B (400 mM Tris-HCl, 60 mM EDTA, 150 mM NaCl, 0.1% sodium dodecyl sulfate [SDS; pH 8.0]) and incubated at 37°C for 30 min. The lysates were mixed with 100 μl of 5 M sodium perchlorate and incubated at 37°C for 20 min followed by 20 min at 65°C. The mixtures were then extracted with 580 μl of chloroform in a Nucleon microtube with a silica plug. The DNA was then precipitated with 880 μl of 100% ethanol, and the pellets were washed twice with 1 ml of 70% ethanol and resuspended in 100 μl of sterile distilled water. Final DNA preparations were quantified spectrophotometrically, diluted to a concentration of 10 ng/μl, and stored at −20°C. DNA from E. coli ATCC 25922 was used as a control for DNA extraction. Five-microliter aliquots of the DNA preparations were cut with the restriction endonucleases EcoRI and MseI (New England Biolabs, Beverly, Mass.) and then ligated to EcoRI and MseI oligonucleotide adaptors in accordance with the AFLP microbial fingerprinting kit protocol (PE Applied Biosystems). The EcoRI-MseI restriction fragments tagged with specific adaptors were then selectively amplified with three pairs of primers, EcoRI-A(FAM)–MseI-CA, EcoRI-A(FAM)–MseI-G, and EcoRI-C(TAMRA)–MseI-G, by using the GeneAmp PCR system 9600 (PE Applied Biosystems). Primer notations and amplification procedures were as described in the protocol of the AFLP kit. The PCR products were separated on a Long Ranger gel (J. T. Baker, Toronto, Ontario, Canada) with 0.6× Tris-borate-EDTA (TBE) buffer (Bio-Rad, Mississauga, Ontario, Canada) for 3.5 h by using the ABI PRISM 377 DNA sequencer (PE Applied Biosystems). The GSROX-500 size standard (PE Applied Biosystems) was used as an internal standard for all of the AFLP experiments. The data were analyzed with the ABI PRISM GeneScan 2.02 software and then with the ABI PRISM Genotyper 2.0 (PE Applied Biosystems) to tabulate fragments by size and fluorescence intensity. Binary files were then generated manually from the tabulated fragments on the basis of their presence or absence by using Microsoft Excel 97. As with the previously described RFLP data sets, a cladistic analysis of the AFLP characters was performed by using maximum parsimony in PAUP v3.1.1 (47). Parsimony procedures search for phylogenetic trees that minimize the number of evolutionary steps (i.e., tree length measured as gains and losses of restriction sites) required to interpret the data set (24, 54). The ancestral condition of the binary characters was identified by including the outgroup JM109 in the analysis. Because the loss of an existing restriction site is more probable than independent gains of the same site at a specific location (11), the Dollo parsimony criterion requiring that every derived restriction site be uniquely derived within the phylogeny was applied. Noise obscuring the phylogenetic signal (homoplasy) was reported by the consistency and retention indices (54). Confidence in each clade was assessed by bootstrapping the data by using 1,000 pseudoreplicates (18) and by evaluating the decay index (9) at monophyletic groupings by using the program AutoDecay v2.9.6 (15). Bootstrapping results were reported as the proportion of bootstrap trees that validate each grouping, while the decay index showed support for a monophyletic group by calculating the difference in tree length between the shortest trees with and without that group. Hypotheses relating to the monophyly of eae-positive STEC isolates and to the evolution of intimin variants were evaluated by using a topology-dependent permutation tail probability test (T-PTP) (16). The result of this test is analogous to the decay index in that the length difference between the shortest trees with and without the monophyletic grouping is evaluated against length differences of trees calculated from permuted data sets. A monophyletic grouping was declared significantly robust when T-PTP was <0.05.

RESULTS

Restriction analysis of the ehxA gene.

Eleven distinct restriction types were identified on the basis of restriction enzyme digestion of the amplification product of the ehxA gene for the 79 STEC isolates examined (Fig. 1). Only the O22:H8 and O111:NM serotypes comprised isolates of more than one restriction type (i.e., two types each). All 18 O157:H7 and O157-NM isolates (from Canada, Germany, the United States, Australia, and Swaziland) belonged to a unique type not found in any other serotype. The ehxA gene of these two serotypes possessed a restriction site for the enzyme TaqI at position 548, which resulted in a specific restriction pattern with this enzyme (Fig. 2). Analysis of the relationships between the ehxA sequences studied by the UPGMA method of clustering (Fig. 1) shows a clear separation into two major groups (I and II). Although there were only eight cladistically informative characters, the same division was observed within the single tree obtained by maximum parsimony (length = 8 steps; tree not shown). All characters were in agreement over the branching structure. All the isolates with a group I ehxA gene were eae positive by PCR, whereas all the isolates with a group II ehxA gene were eae negative (Fig. 1). All the isolates of group II showed the same unique TaqI restriction pattern (Fig. 2).

FIG. 2.

Restriction patterns of a 2,862-bp PCR product of the ehxA gene obtained by digestion with the enzyme TaqI. Electrophoresis was done in a 2% agarose gel in TAE. Lanes: 1 to 3, patterns specific for the eae-positive isolates; 4, pattern specific for the eae-negative isolates; 1, profile specific for serotypes O157:H7 and O157:NM; MW, 123-bp ladder.

Sequence of a group II ehxA gene and comparison with published sequences from group I ehxA genes.

The DNA sequence of the cloned ehxA gene of isolate EC920006 (serotype O8:H19, eae negative, dairy cattle origin, Canada) has been deposited under accession no. AF043471 in the GenBank database. The ehxA gene of this isolate has 98.0 and 97.3% identity with the published sequence for serotype O157:H7 (42) at the nucleic acid and amino acid levels, respectively. Alignment of the amino acid sequences of the EHEC hemolysin of isolate EC920006 with the previously published sequences of isolates of serotypes O157:H7 (42) and O111:NM (43) showed a nonhomogeneous distribution of substitutions (Fig. 3). Two regions encompassing amino acids 661 to 834 and amino acids 949 to 998 are highly conserved. The first 300 N-terminal amino acids of EhxA are also relatively conserved (1.3% substitutions). The most variable regions of the toxin are located between amino acids 312 and 660 (5.7% substitutions) and between amino acids 835 and 948 (7% substitutions). The lysine residues at positions 550 and 675 are conserved among all three serotypes.

FIG. 3.

Distribution of substitutions in the EHEC hemolysin of serotypes O157:H7 (42), O111:NM (43), and 08:H19 (present study). Numbers and arrows below the horizontal bar represent the positions of numbered amino acids in EhxA. Each vertical bar represents a substitution. N, N-terminal end of the hemolysin; C, C-terminal end of the hemolysin; REPEAT, tandem repeats involved in calcium binding.

RFLP and Southern blotting of EHEC hemolysin and flanking sequences.

A set of 20 isolates of 11 different serotypes was initially tested with the restriction enzymes BamHI, EcoRI, HindIII, PstI, and SalI. The two most discriminatory enzymes, EcoRI and PstI, were chosen for further typing of the whole collection of isolates. Plasmids from two isolates (serotypes O91:H7 and O98:H25) remained nontypeable because of degradation during restriction enzyme analysis, despite the precautions that were taken. The plasmids of 11 other isolates could not be digested with one of the two restriction enzymes used (two with EcoRI and nine with PstI). From the 10 serotypes with multiple typeable isolates, 5 showed more than one restriction pattern per serotype (Fig. 1). By using the two endonucleases EcoRI and PstI, 4 of the 8 restriction sites located within the ehx operon were conserved among all typeable isolates, whereas all 28 restriction sites located upstream of this operon were variable. These data suggest a higher level of diversity upstream than within the ehx operon.

RFLP of the eae PCR product and espB PCR.

Six restriction types were distinguished on the basis of RFLP analysis of the amplified 1,110-bp eae fragment. Except for serotype O26:H11, all the serotypes with multiple isolates were homogeneous in terms of the RFLP pattern of this fragment. Analyses of the phylogenetic relationships among RFLP patterns by UPGMA (Fig. 4) and maximum parsimony (length = 6 steps; tree not shown) show a clear division into two major groups without ambiguities in the branching of the trees. The eae restriction profile of the O127:H6 EPEC strain did not differ significantly from the profiles of the STEC strains (Fig. 4). All the isolates with espB-positive PCR results, and consequently presenting related subtypes of this gene, also fall into the same eae RFLP major group 1 and the same two eae RFLP types (Fig. 4), thus confirming that our results obtained by RFLP of the eae gene are likely valid for the whole LEE.

FIG. 4.

Estimated degrees of nucleotide substitutions between variants of the eae gene of 18 STEC serotypes and one EPEC strain (2348/69) of serotype O127:H6. The dendrogram was generated by the UPGMA method of clustering by using a matrix of proportion of nucleotide differences between pairs of variants based on restriction maps of a 1,110-bp PCR fragment of the eae gene. The espB column shows PCR results identifying espB subtypes within the eae-positive isolates (53). −, no PCR product; +, PCR product of the expected size.

eae-associated fragments on EHEC hemolysin plasmid.

Two EcoRI fragments (2,213 and 3,399 bp) were isolated following the comparison of the EHEC hemolysin plasmids of an eae-positive (strain 4304, serotype O157:H7) and of an eae-negative isolate (strain EC930076, serotype O22:H8) by restriction analysis with the enzyme EcoRI and Southern blotting. These two fragments hybridized specifically with EHEC hemolysin plasmids of nine eae-positive STEC isolates but not with plasmids of eight eae-negative STEC isolates (Fig. 5) nor with a representative EPEC isolate of serotype O127:H6 (data not shown). Further analysis using PCR confirmed that these two fragments are contiguous and are part of a single continuous stretch of DNA. Using the same approach, an EcoRI fragment of approximately 4 kb on the EHEC hemolysin plasmid of the O22:H8 isolate was shown to hybridize only very weakly with the EHEC hemolysin plasmids of the nine eae-positive isolates examined. However, this fragment hybridized strongly with the EHEC hemolysin plasmids of the eight eae-negative isolates, with the EAF plasmid of the EPEC strain 2348/69, and with other large cryptic plasmids of eae-positive isolates. Based on these consistent differences, EHEC hemolysin plasmids of eae-positive and -negative isolates were named type I and type II plasmids, respectively.

FIG. 5.

Conserved EcoRI fragments of the EHEC hemolysin plasmid of eae-positive isolates. Southern blot hybridization of purified plasmid preparations of STEC probed with ehxA probe (A), the labeled 2,213-bp EcoRI fragment of the EHEC hemolysin plasmid of strain 4304 (B), and the 3,399-bp EcoRI fragment of the EHEC hemolysin plasmid of strain 4304 (C) is shown. Lanes: 1 to 9, EHEC hemolysin plasmids of eae-positive isolates of serotypes O157:H7, O145:NM, O111:H8, O111:NM, O26:H11, O103:H2, O5:NM, O76:H25, and O118:H16, respectively; 10 to 17, EHEC hemolysin plasmids of eae-negative isolates of serotypes O153:H25, O113:H21, O22:H8, O8:H19, O115:H8, O46:H38, O39:H49, and O88:H25, respectively.

The sequences of the two fragments of strain 4304 conserved among type I plasmids (deposited under accession no. AF043470 in the GenBank database) showed significant homologies (Fig. 6) with several regions of the large virulence plasmid of Shigella flexneri (msbB [37], virK [34], shf1 and shf2 [38], sepA [4]), with chromosomal and plasmid sequences of E. coli (msbB [29], a hypothetical 66.6-kDa protein, [8], perA to perD [20], bfpT to bfpW [48]), and with Salmonella enterica (sinR [22]). The sequence region between bp 4578 to 5612 also was almost 100% identical to the 5′ part of an O157:H7 EHEC hemolysin plasmid region recently described by Brunder et al. (10). Four major ORFs (designated ecf1 to ecf4 for eae-positive conserved fragments 1 to 4) found on the cloned fragment were all preceded by potential promoter sequences (Fig. 6). Three of these ORFs were highly similar to previously described proteins. Ecf4 probably extends beyond the cloned region of the EHEC hemolysin plasmid and has 66% identity and 79% similarity with the N-terminal part of the chromosomally encoded MsbB protein of E. coli (29). It also has 70% identity and 83% similarity with the 131-amino-acid N-terminal sequence of the hypothetical plasmid-encoded MsbB protein of S. flexneri (37). However, no obvious ribosome binding site was found upstream of the ecf4 start codon. Ecf3 corresponds to a hypothetical protein of 582 amino acids with a molecular size of 67.3 kDa. It has an overall 80% identity and an 86% similarity with the hypothetical chromosomally encoded 66.6-kDa YIJP protein of E. coli (8). Ecf2 corresponds to a hypothetical protein of 368 amino acids with an expected size of 40.8 kDa. This hypothetical protein did not have significant sequence similarity with any other known protein. Ecf1 corresponds to a 273-amino-acid protein with an expected size of 32.2 kDa. It has 70% identity and 80 to 83% similarity with the sequences of the plasmid-encoded Shf1 and Shf2 hypothetical proteins of S. flexneri (38). The coding regions for these two proteins on the STEC plasmid examined is not separated by a stop codon, and they may be translated as a single peptide. All three ORFs, ecf1 to ecf3, are preceded by potential ribosome binding sites.

FIG. 6.

Regions of homologies with known sequences and potential ORFs on the 5,612-bp conserved fragment of the type I EHEC hemolysin plasmid. The rectangular box represents the 5,612-bp fragment; vertical ticks mark 500-bp portions of the fragment. The shaded areas above and below the box and the corresponding labels represent zones of homology with known sequences at the nucleic acid level (see text for details). The arrows represent the four potential ORFs ecf1 to ecf4 (from right to left). The stars represent the potential promoter regions for the corresponding ORFs. The shaded region in the box represents the zone of homology with the recently published sequence of an adjacent EHEC hemolysin plasmid fragment from an O157:H7 isolate (10), and the EcoRI-labeled arrows indicate the restriction sites used for cloning.

AFLP.

Of the 313 variable AFLP characters, 194 were cladistically informative. Maximum parsimony analysis resulted in a single minimal-length tree with 920 steps (Fig. 7). Although our analysis was heavily impacted by homoplasy (noise obscuring phylogenetic signal), the most parsimonious tree grouped all of the eae-positive STEC isolates (also containing a type I EHEC hemolysin plasmid) into one single lineage (T-PTP = 0.015). All but one of the eae-negative isolates (strain EC920053) were distributed into other dispersed lineages. However, two of the four equally parsimonious trees that were only one step longer grouped the strongest supported clade, EC920016-EC920370, with EC920158-EC920185 while forcing EC920053-2348/69 outside of the monophyletic eae-positive STEC clade. Even though these trees with one extra step do not meet the criterion of maximum parsimony, they account for the low bootstrap values at the base of the eae-positive STEC clade (the bootstrap was 78% when EC920053 and 2348/69 were excluded) and suggest a possible alternative relationship among isolates. Nonetheless, comparison of the clustering obtained by RFLP analysis of the EHEC hemolysin gene (Fig. 1) and by AFLP (Fig. 7) showed a high degree of support for a single evolutionary origin of the eae-positive STEC isolates. No such agreement was observed for the eae-negative isolates, and bootstrapping tests showed that the topology of the genomic tree was also much less reliable for these latter isolates.

FIG. 7.

The single most parsimonious cladogram showing the phylogenetic relationships among 27 STEC isolates and one EPEC isolate based on AFLP data from three sets of PCR primers (length = 920 steps; consistency index = 0.21; retention index = 0.81 with 194 characters). The tree was resolved from a heuristic search using PAUP v3.1.1; isolates were added randomly during 25 replications with MULPARS and steepest-descent options invoked and with branch swapping by the tree bi-section-reconstruction algorithm. Characters were polarized in accordance with Dollo’s rules, and the laboratory E. coli K-12 strain JM109 was used as an outgroup. Bootstrap percentages from 1,000 pseudoreplicates are shown at each node, followed by the decay index. The data set is available upon request at cycles@ovcnet.uoguelph.ca. The numbers and letters in the column labeled eae represent subtypes obtained by RFLP analysis of the eae gene, where numbers 1 and 2 indicate the two major clusters; a minus represents eae-negative isolates. The data in the column labeled espB are PCR results identifying espB subtypes within the eae-positive isolates. nd, not determined. The numbers and letters in the column labeled ehxA represent subtypes obtained by RFLP analysis of the ehxA gene, where I and II indicate the two major clusters; a minus represents an ehxA-negative isolate.

The RFLP phylogeny of the eae gene revealed two distinct lineages (1 and 2 in Fig. 4). Although all eae-positive isolates are members of the same clade, and although isolates with distinct eae lineages do form small subclades (Fig. 7), no cladistic support was obtained for the monophyletic groupings of either eae-positive group 1 (T-PTP = 0.157) or group 2 (T-PTP = 0.198), which did not represent a significant departure from a random assortment of isolates. This lack of correlation between major eae RFLP groups and genomic lineages is further supported by the results of the espB PCR (Fig. 7).

DISCUSSION

Despite their increasing importance for human health and food safety and a growing body of knowledge on the pathogenic mechanisms of STEC-related diseases, little is known about the evolution of STEC isolates and their virulence factors. The Shiga toxins are encoded by bacteriophages (46), whereas eae (56) and the other genes of the LEE (32) are located on a chromosomal pathogenicity island (33). The genes encoding the EHEC hemolysin (ehxA), its activation (ehxC), and part of its secretion machinery (ehxB and ehxD) are located on large plasmids having some similarities with the EAF plasmid of EPEC (23). All of these factors are located on potentially mobile elements, and their recent acquisition has been suggested as being responsible for the emergence of STEC as a new major human pathogen (51). The analysis of the diversity of the ehxA and eae genes and associated sequences performed in the present work will help to obtain a better understanding of the evolution of virulence-associated factors in STEC populations and to confront some of these hypotheses.

In the present study, we used RFLP data to estimate the levels of substitutions in the ehxA genes of isolates of 37 STEC serotypes. The results obtained by this method are in full agreement with those obtained by sequencing the complete ehxA gene and show that, with an estimated maximum of less than 4% of nucleic acid substitutions between isolates, this gene is highly conserved among a large variety of STEC serotypes. The analysis of a limited number of restriction sites in the region flanking the ehxA gene by Southern blotting suggests that, unlike sequences located further upstream on the EHEC hemolysin plasmid, the ehxC, ehxB, and ehxD sequences are highly conserved. The alignment of the amino acid sequences of three EHEC hemolysin isolates shows that regions of the EHEC hemolysin that correspond to active sites of the toxin are particularly well conserved. These include the region between amino acids 661 and 834, which contains the tandem repeats (42) typical of all RTX toxins and which are involved in calcium binding (50); the lysine residues at positions 550 and 675, which are involved in the activation of the toxin by fatty acylation (36); and the 50 C-terminal amino acids, which are involved in secretion (3, 50). The high level of conservation in the ehx operon and particularly of some active sites of the toxin in an apparently more variable region of the EHEC hemolysin plasmid suggests that the toxin is under strong selective constraints and plays an important role in the survival of STEC. However, the mechanisms by which STEC may take advantage of the EHEC hemolysin remain to be elucidated.

The RFLP analysis of the EHEC hemolysin shows that the serotypes O157:H7 and O157:NM possess a specific TaqI restriction site at position 548. All of the eae-negative serotypes also possess a unique TaqI restriction pattern. The serotypes O157:H7 and O157:NM are of particular epidemiological importance because they represent the most frequently isolated serotypes in clinical disease and in food-borne outbreaks of STEC infections (2). In addition, most O157:H7 isolates produce EHEC hemolysin (5, 40, 42). Finally, eae-negative STEC serotypes are less frequently associated with disease in humans than are eae-positive ones (41). Thus, RFLP analysis of the ehxA PCR product (Fig. 2) can be used to detect EHEC hemolysin-positive E. coli isolates, to identify the epidemiologically highly relevant O157:H7 and O157:NM serotypes, and to distinguish eae-negative from eae-positive serotypes, all in a single step.

The clear division into two main groups observed for the ehxA gene (Fig. 1) is probably only a marker for the evolution of two lineages in the EHEC hemolysin plasmid of STEC (type I and type II plasmids). The presence of a conserved 5.6-kb fragment on the type I EHEC hemolysin plasmid, lacking or significantly different in type II plasmids, strongly supports this hypothesis. This is also confirmed by the presence of a conserved sequence within a 4-kb fragment on the type II EHEC hemolysin plasmid, which has only low homology with the type I EHEC hemolysin plasmid. In addition, type II plasmids also seem to be consistently slightly larger than type I plasmids (Fig. 5).

Several stretches of DNA on the conserved 5.6-kb fragment of a type I plasmid showed strong homologies with the sequences of the large virulence plasmid of S. flexneri. However, most of the regions homologous to the S. flexneri plasmid found on the DNA fragment seem to be incomplete or disrupted (Fig. 6). Some sequences with homologies to the E. coli chromosome (8) and others previously thought to be specific for Salmonella spp. (22) suggest that the conserved 5,612-bp fragment is a composite entity resulting from multiple recombination events. It may have a common origin with the large Shigella virulence plasmid but has been subjected to additional recombinations not observed in the latter. Four potential ORFs related to the four major zones of homology detected in the DNA sequence (Fig. 6) were found on the 5,612-bp fragment. At least three of these four ORFs (ecf1 to ecf3) were preceded by both potential promoter sequences and ribosome binding sites and probably represent functional genes. The functions of the hypothetical proteins described here and of their homologs are not known in detail, and further studies to demonstrate their expression and to clarify their possible role in STEC isolates are needed. However, preliminary results obtained in our laboratory suggest that, like the chromosomal msbB gene (45), these ORFs may be involved in lipopolysaccharide biosynthesis and their regulation may be associated with attachment of STEC on eukaryotic cells.

AFLP has recently been used as a valuable tool for systematic and phylogenetic studies (25, 26). Our analysis using this technique shows that eae-positive STEC isolates with a type I EHEC hemolysin plasmid form a monophyletic lineage within STEC populations. The correlation between the groups defined by RFLP analysis of ehxA and by AFLP analysis of the whole STEC genome for eae-positive isolates clearly suggests a clonal evolution of the type I plasmid within a single evolutionary lineage. This conclusion is in agreement with a previous study showing that the EHEC hemolysin plasmid may have been transferable in the past but is now transfer defective (23). Type I plasmids and the EHEC hemolysin are therefore not a new entity in STEC and were present in STEC before the serotypes composing the eae-positive phylogenetic lineage diverged. The lack of such a correlation for the ehxA gene of eae-negative isolates can be explained in two ways. The low reliability of the genomic tree topology and the limited number of informative characters for the ehxA gene alone could hide the clonal evolution of type II plasmids in STEC. Alternatively, transfers of type II plasmids may also be at the origin of a true lack of clonality for this ehxA gene in eae-negative isolates. The previous demonstration of a lack of transferability of EHEC hemolysin plasmids was based mainly on plasmids of eae-positive isolates (23), and comparative studies between type I and type II plasmids may show differences in transfer-associated factors between these two categories of plasmids. The generally larger size of type II plasmids (Fig. 5) and the presence of conserved sequences on EHEC hemolysin type II plasmids and other large cryptic plasmids of STEC that are apparently missing or much less conserved in type I plasmids are in agreement with this hypothesis.

The lack of correlation between RFLP results for the eae gene and AFLP data show that the LEE has been transferred horizontally between STEC strains or between STEC and EPEC strains on several occasions in the past. PCR results for the espB gene also support this conclusion. These results are consistent with those of others (52), showing that the LEE has different locations in different STEC phylogenetic lineages.

With regards to the reasons for the observed confinement of the LEE to a single major STEC lineage, the 2348/69-EC920053 clade represents an interesting case. Cladistic permutation tests show that on the basis of our AFLP data, this clade cannot be confidently attributed to either the eae-positive STEC lineage or to an eae-negative lineage. If this pair of strains does not belong to the eae-positive STEC lineage, then the strain 2348/69 is a good illustration that the mobile LEE can be stably inherited in lineages other than the eae-positive STEC lineage. This stable inheritance is probably related to the regulatory and adherence-associated factors encoded by the EAF plasmid in EPEC (12, 20). Similar mechanisms in STEC would not be surprising, and a lack of these factors or of some factors necessary for the transfer of the LEE in some STEC lineages would explain the confinement of the LEE to a broad but unique phylogenetic lineage. If, as shown in Fig. 7, the 2348/69-EC920053 clade belongs to the eae-positive STEC lineage, then the strain EC920053 strongly suggests that the LEE in STEC is associated with the presence of the EHEC hemolysin type I plasmid factors and not with a specific genomic lineage. Indeed, the EHEC hemolysin plasmid of eae-positive STEC has been implicated in the attachment of STEC to epithelial cells (27, 49). The two categories of EHEC hemolysin plasmids may have evolved differently with regards to these and other adherence-associated factors. Finally, the high nuclease activity observed in eae-negative STEC during the plasmid isolation procedure could also be a marker for a mechanism limiting the spread of foreign DNA like the LEE among STEC populations. Further studies to test these hypotheses and to find the reasons for the confinement of the LEE to a single phylogenetic lineage containing type I plasmids in STEC are clearly needed.

In conclusion, our results show that the EHEC hemolysin operon and the ehxA gene in particular are highly conserved among a large variety of STEC serotypes. The ehxA gene may serve not only as a target for detection of clinically important STEC isolates (7) but also as a tool for differentiation of STEC isolates with different levels of clinical relevance. Two major groups of EHEC hemolysin plasmids that may differ significantly in terms of virulence-associated factors were delineated. Horizontal transfer of the LEE has taken place on several occasions in the past, but the presence of the LEE in STEC seems to be limited to a single phylogenetic lineage harboring type I EHEC hemolysin plasmids.