INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Enteropathogenic Escherichia coli (EPEC) is a leading cause of infantile diarrhea, particularly in developing countries (36). Although EPEC was one of the first classes of enterovirulent E. coli to be identified, the precise mechanism by which EPEC strains cause diarrhea is still under investigation. Two definitive features that are observed when EPEC infects epithelial cells are the attaching-and-effacing (A/E) lesion and localized adherence (LA) (36). The A/E histopathology is characterized by bacteria attached intimately to the host cell (33). Following signal transduction, host cytoskeletal proteins, including actin, accumulate at the site of attachment. Eventually, the host cell undergoes structural modifications resulting in the elevation and cupping of the bacterial cell on a pedestal-like protrusion supported by host cell-polymerized actin (33). In the intestine, this leads to effacement of microvilli. EPEC strains possess a chromosomal pathogenicity island, called the locus of enterocyte effacement (LEE), which contains the genes required to produce A/E lesions (28). The LEE encodes a 94-kDa outer membrane protein required for intimate attachment called intimin and a type III secretion system as well as proteins that are translocated via this system (14).

Typical EPEC strains adhere to host epithelial cells in tight, three-dimensional microcolonies. Expression of a type IV pilus, the bundle-forming pilus (BFP), is required for this LA phenotype (17). The genes required for biogenesis of the BFP are carried on a large 50- to 70-MDa virulence plasmid, the EPEC adherence factor (EAF) plasmid (51), which varies in sequence among different EPEC strains but is somewhat conserved (37). The presence of this plasmid has been shown to greatly enhance the virulence of EPEC, and hence strains carrying the EAF plasmid are strongly associated with diarrhea in epidemiological studies and produce diarrhea in volunteers (5, 24, 36). In addition to the bfp genes, the plasmid also carries genes encoding a regulator (Per) (18). The per operon consists of three genes, the first of which, perA, encodes a protein that belongs to the AraC family of transcriptional activators (18). The other two genes, perB and perC, encode proteins that enhance the activity of PerA by an unknown mechanism (18) (53). Per has been shown, to date, to activate genes within and outside the LEE via the LEE-encoded regulator (Ler) (13, 30) and the bfp operon (53) as well as a chromosomal gene involved in microcolony formation, trcA (54). Per also activates its own promoter (26, 34). Volunteer studies conducted by Bieber et al. demonstrated that a perA mutant (referred to as a bfpT mutant in that study) is less virulent than its isogenic wild-type strain (5). Per appears, therefore, to play a pivotal role in the regulation of essential virulence genes in EPEC.

Most EPEC strains are genetically related and belong to a limited number of O:H serotypes. Classic EPEC O serogroups include O55, O86, O111, O114, O119, O128, and O142. The most common H antigens associated with EPEC are the H6 and H2 antigens. EPEC strains have been subtyped into two major genetic lineages by multilocus enzyme electrophoresis. The EPEC1 lineage includes predominantly H6 strains, and the EPEC2 lineage includes predominantly H2 strains (40, 56). A less common EPEC H type is H34, and a number of EPEC strains are nonmotile in conventional tests and classified as H. EPEC strains belonging to nonclassic serotypes have been reported, but these rarely possess the EAF plasmid. EPEC strains from classic serogroups appear to have arisen from a few clones that acquired virulence loci, such as the LEE and the EAF plasmid, in a stepwise manner (43). Studies with other EPEC virulence genes, including eae (which encodes intimin), have shown that horizontally acquired genes often, but not always, show allelic variation that correlates with clonality (1, 6, 43, 56). The per operon is part of a plasmid and itself has a lower G+C content than most other regions of the plasmid. In strains E2348/69 and B171-8, the per operon is known to be flanked by insertion sequences (18, 52). We sequenced the perA, -B, and -C genes from multiple EPEC strains belonging to different serotypes and investigated potential effects on virulence gene activation resulting from sequence differences. We also sequenced the region immediately upstream of the perABC operon to study the relationship between the per sequence, EAF structure, and genomic background with a view towards better understanding the evolution of EPEC.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Strains. Thirty-one EPEC strains that carry an EAF plasmid and were isolated from patients with diarrhea in different countries were employed in the study (Table 1). Thirty strains, representing nine EPEC O groups and 13 serotypes, were from human with diarrhea and had been characterized to various degrees in previous studies. One EAF-positive canine isolate was also studied. PCR-restriction fragment length polymorphism (PCR-RFLP) typing based on the fliC gene was performed on all strains in order to verify H type and identify H types of nonmotile strains as an EAF-independent subclassification (15). E. coli K-12 strain MG1655 (7), enterohemorrhagic E. coli strain EDL933 (44), and JPN15 (20), the plasmid-cured derivative of EPEC reference strain E2348/69, were employed in PCR and hybridization experiments as controls.

Screening for EAF plasmid genes. The EAF probe (an empirically derived probe from the EAF plasmid of EPEC strain E2348/69) (35) was used to screen all the strains by colony hybridization. The presence of other, previously described, EAF plasmid-encoded genes was determined by colony hybridization and PCR. PCR was carried out under high-stringency conditions, employing primers designed to anneal to the sequence of the target gene from plasmid B171-8 (accession number AB024946). Colony hybridizations were conducted under high-stringency conditions at 65°C employing probes that were labeled by random priming with [-³²P]dCTP as described previously (39). Fragment probes for eae, bfpA, the EAF, and perABC were prepared from plasmid clones described in Table 2. Probes for bfpG, trcP, orf35-36 (truncated EHEC toxB homolog), orf61-62 (truncated gadB and gadC homologs), and orf67-68 (stbAB homolog) were prepared by PCR with primers listed in Table 3, using strain B171-8 as the template, and verified by restriction analysis.

LA and A/E tests. Fluorescent actin staining of HEp-2 monolayers infected with each of the E. coli isolates was carried out as described by Knutton et al. (22), with modifications. HEp-2 cells were grown overnight to 50% confluence in Dulbecco's minimum Eagle medium (DMEM) (Gibco BRL, Gaithersburg, Md.) containing penicillin, streptomycin, and 10% fetal bovine serum on coverslips in 24-well plates. Bacteria were grown for 16 h in Luria broth (LB) without shaking. The HEp-2 cells were washed three times with phosphate-buffered saline (PBS), and the medium was replaced with DMEM containing 10% fetal bovine serum and 1% mannose. Ten microliters of bacterial suspension was added per well, and the slides were incubated at 37°C and 5% CO₂ for 3 h. After a 3-h incubation period, the cells were washed three times with PBS, fixed with formalin, permeabilized with 0.1% Triton X-100 in PBS, and stained with fluorescein isothiocyanate-labeled phalloidin (Sigma). The monolayers were then washed with 2× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate) and treated with RNase A (10 mg/ml) before being stained with propidium iodide (Molecular Probes, Eugene, Oreg.). The HEp-2 cells were observed for the adherence pattern of the red-stained bacteria as well as actin accumulation (green) at the site of bacterial attachment under incident fluorescent light with a Zeiss Axioskop routine microscope. Each strain was tested in duplicate. EPEC strain E2348/69 and E. coli K-12 strain MG1655 were used as positive and negative controls, respectively.

Western immunoblotting. Overnight bacterial cultures in LB were subcultured into DMEM and grown, with shaking, for 3 to 4 h to an optical density (A₆₀₀) of 1.0. Bacterial suspensions were pelleted and resuspended in PBS. Whole-cell lysates were prepared by boiling samples in sodium dodecyl sulfate sample buffer. Ten micrograms of total protein was loaded per lane, subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis, and transferred to 0.45-µm-pore-size polyvinylidene difluoride membranes (Millipore, Bedford, Mass.). Membranes were incubated with polyclonal rabbit antisera prepared against intimin, bundlin, and the LEE-encoded chaperone CesD, which had been prepared from proteins purified from EPEC strain E2348/69. The membranes were washed with PBS-0.05% Tween 20, immunostained with a 1:35,000 dilution of goat anti-rabbit horseradish peroxidase antibody conjugate (Kirkegaard and Perry Laboratories, Gaithersburg, Md.), and developed with enhanced-chemilumniscence reagents according to the instructions of the manufacturer (Amersham, Pharmacia Biotech, Piscataway, N.J.). Bands were detected by exposure to X-ray film (Kodak X-Omat blue).

Cloning and sequencing of perABC and the per upstream regions. Overlapping clones for the per region were prepared by amplifying with the K1547-K1693 and K182-K696 or K1694-K696 primer pairs (Table 3) with the proofreading enzyme Pwo polymerase (Boehringer Mannheim GmbH, Mannheim, Germany) and cloning the resulting fragments into the SmaI site of pBluescript (Stratagene) or into pCR-Blunt II TOPO (Invitrogen, Carlsbad, Calif.). Boiled colonies of the test strains were used as templates. Because of the tendency to select PCR-derived per deletion mutants when cloning per genes into high-copy-number vectors, clones with inserts in the inverse orientation were selected for sequencing. Clones that contained mutations were verified by carrying out another PCR with a fresh template, recloning, and resequencing the region or in a few cases by restriction analysis of the new PCR product with the enzymes SspI and DraI (these enzymes gave altered restriction profiles for the perA alleles of O142#15 [SspI only], MB80, and CA 4225). The upstream region of per from strain MB80 was cloned by partially digesting genomic DNA with the frequently cutting enzyme TaqI (Gibco BRL) to reduce the number of nonspecific products and subjecting the resulting digest to Taq polymerase PCR and annealing at 50°C with primer pair K1547-K1549. The resulting multiple fragments were gel purified, and fragments that yielded an expected 350-bp product after BamHI (Gibco BRL) digestion were cloned and sequenced. On the basis of this sequence, primer K1978 was designed and used with K1549 to amplify the upstream regions of strains MB80, O142#15, and CA 4225 under high-stringency conditions (annealing at 52°C with Pwo polymerase) and cloned for sequencing. Plasmid DNA was purified with a Wizard miniprep kit (Promega, Madison, Wis.) and subjected to automated sequencing at the University of Maryland Biopolymer Facility on an Applied Biosystems, Inc., sequencer. Primers K182, K1691, K696, K1547, and K1978 were used with vector-priming M13 oligonucleotides to obtain double-stranded sequence. Clones of the per operon used in functional experiments were prepared by amplifying the entire operon with the primers K1547 or K1978 and K696, which carries a BamHI site in its tail, digesting with BamHI, and cloning the resulting product into the BamHI site of pBR322. This produced 1.8-kb clones of the per operon that were analogous to pJLM171 (30) and included the 182 bp of the region upstream of the perA start codon.

Sequence analysis. Sequences were assembled by the Sequencher 3.1 program (Gene Codes, Inc.) and annotated using MacDNASIS version 3.5 (Hitachi Software Engineering America, Ltd.). The sequences of per operons from strains E2348/69 and B171-8 were retrieved from GenBank (accession numbers Z48561 and L42638, respectively) and included in the analysis. Multiple alignments were performed with the CLUSTAL W program. Phylogenetic trees were derived with the program MEGA2 (http://www.megasoftware.net; Sudhir Kumar, Koichiro Tamura, Ingrid Jakobsen, and Masatoshi Nei, 2000, Institute of Molecular Evolutionary Genetics, The Pennsylvania State University). The significance of synonymous and nonsynonymous substitutions was determined by the method of Nei and Gojobori (38) with the MEGA2 program.

-Galactosidase assays. Clones of the perABC operon from representative strains in pBR322 were assayed for their ability to activate single-copy chromosomal fusions of the upstream region of the LEE1 gene operon (nucleotides 211 to +221 relative to the transcriptional start site) and bfpA (nucleotides 688 to +124 relative to the transcriptional start site) to lacZ in an E. coli K-12 (MC4100) background. Fusions were constructed by the method of Simons et al. (49) as described previously (30). Overnight bacterial cultures were subcultured into LB and grown to an optical density at 600 nm of 0.4 to 0.5. The -galactosidase activities of culture lysates on the substrate ortho-nitrophenyl--D-galactopyranoside (Sigma) was measured in Miller units as described by Miller (32). Statistical analysis of the results was performed by Student's t test.

SspI-based PCR-RFLP. The region encompassing nucleotide 54 to the end of the perA open reading frame was amplified by Taq PCR with primers K1694 and K1693. The resulting 770-bp products were digested at 37°C for 90 min with SspI (Gibco BRL). Fragment length polymorphisms were distinguished by electrophoresis on a 2% agarose gel.

Nucleotide sequence accession numbers. The nucleotide sequences of the perA, -B, and -C genes obtained in this study were deposited in the GenBank database under the accession numbers listed in Table 4. Sequences of per genes from strains E2348/69 (Z48561 [18]) and B171-8 (L42638 [53]) were retrieved from previous depositions.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

EPEC strains vary in virulence factor expression. Expression levels of two established virulence proteins of EPEC that are regulated by Perbundlin (the BFP structural subunit) and intiminwere examined in different EPEC strains. CesD, a LEE-encoded (and therefore Per-regulated) chaperone, was also evaluated because, unlike the other two proteins, it is not surface exposed and is less likely to vary antigenically among different strains. As shown in Fig. 1, there was considerable variation in the expression levels for bundlin, intimin, and CesD as well as in the ability of different isolates to demonstrate the A/E and LA phenotypes. Although the expression of different initimin alleles may account for some of the differences in levels of intimin expression (1), similar antigenic differences have not been demonstrated for bundlin (6) and are not expected to occur with CesD, a protein that is highly conserved between widely divergent A/E pathogens (41, 57). We therefore hypothesized that differences in the levels of expression of these proteins may be related to defects in specific genes or in genes encoding key regulators. Per is known to play a central role in the regulation of these genes and phenotypes and was therefore selected for comparative sequence analysis.

View larger version (28K): [in this window] [in a new window]   FIG. 1.   Expression of intimin, bundlin, and CesD, virulence proteins known to be directly or indirectly regulated by per, in different EPEC isolates. , antibody.

EAF plasmids are partially conserved, and the per operon is present in almost all EAF-positive EPEC strains. In concordance with the results of previous studies (16, 18, 37), we found that, although there is variation in the gene contents of EAF plasmids of different EPEC isolates, the bfpA gene and the stbAB region are universally present in all strains examined. The perA, -B, and -C genes were shown, by hybridization and PCR, to be present in all but one O119:H6 isolate (O119#5). The empirically derived EAF probe was shown in previous studies to hybridize with most but not all strains that have the virulence plasmid (35) and hybridized with all but one of the strains employed in this study. As shown in Table 4, other regions of the EAF plasmid are less conserved. The bfpG gene was present in 28 (90.3%) strains. The three strains that were negative for bfpG were the O119:H2 (MB80) and O128:H2 (MB21) strains, which have previously been reported to lack this gene and the rest of the bfp operon (8), and the canine isolate CA 4225. Although all three isolates hybridized with the bfpA probe, bfpA could not be amplified by PCR, suggesting that the 3' end of the gene is deleted as described by Bortolini et al. (8). As expected, all three isolates fail to produce bundlin detectable by Western blotting (Fig. 1). The region orf35-36 in plasmid pB171 (52) (truncated homolog of the EHEC toxB gene) was found in 24 (77.4%) of the EPEC strains, while the region orf61-62 (gadB [truncated] and gadC homologs) was present in only 13 (41.9%) of the strains examined. The trcP gene, which was originally found during the sequencing of the EAF plasmid of O111:NM strain B171 (52), was in this study detected in strains E990 (O86:NM), 065542, 065900 (O114:H2), 023-220982 (O126), 012-050982 (O142:H6), and all the O111 strains but was absent in other strains. In summary, the O55, O86, O111, O114, O126, and O127 strains and a subset of the O142 strains were positive for most of the pB171-8-derived probes, while the O119 and O128 strains and the canine isolate were more likely not to hybridize with one or more of the probes.

Nucleotide sequence variation in the per operon is minimal. per sequences for at least one strain belonging to each serotype were subjected to comparative analysis. SspI restriction fragment length polymorphisms revealed that two O142:H6 strains showed a unique profile, and therefore per from strain O142#15 was sequenced in addition to that from O142:H6 strain E851/71. Two previously determined per sequences (from strains E2348/69 and B171-8) were also included in the analysis, and the accession numbers of sequences determined in this and other studies are listed in Table 1. The sequences of the per genes from different strains show very little variation. We found 94 to 99% identity among the sequences of perABC and the upstream 100 bp at the nucleic acid level. Pairwise analysis by the Nei-Gojobori method revealed that most of the sequences were not significantly different from one another. The major exceptions were the O119:H2 (MB80), O142:H6 (O142#15), and canine (CA 4225) perABC operons, which showed significant differences from the others (P < 0.05). In addition to other substitutions, these strains have mutating frameshifts in the perA sequence. The perA nucleotide G114 is deleted in strains MB80 (O119:H2) and CA 4225 (canine), and TA is inserted after nucleotide 313 in strain O142#15 (O142:H6), which results in truncation of the gene (Fig. 2).

View larger version (66K): [in this window] [in a new window]   FIG. 2.   Alignment of the nucleotide sequences of perA alleles from prototype strain E2348/69 (O127:H6) and strains MB80 (O119:H2) and O142#15 (O142:H6), which contain mutating frameshifts. Mutating frameshifts are shaded, and stop codons are boxed. SspI sites are indicated by bold underlined text.

View larger version (94K): [in this window] [in a new window]   FIG. 3.   Alignment of the PerA predicted amino-acid sequences of strains that have a full-length gene showing the locations of conserved (gray background) and nonconserved (white background) substitutions seen in different strains. No substitutions occurred within the DNA-binding helix-turn-helix (HTH) motif region annotated with arrows and dashed lines.

The sequence upstream of the perA promoter region varies in different strains. The sequence of the EAF plasmid from EPEC strain B171-8 reveals that a partial IS1() homolog lies immediately upstream of the per operon (52). The region upstream of the per promoter (500 bp relative to the transcriptional start site of perA) was shown by sequencing to be almost identical to the B171-8 sequence in O55, O86, O111, O114, O119:H6, O126, and O127 strains and subsets of O128:H2 and O142 strains. In O119:H2, O142:H6, and the canine isolate, this sequence is replaced by a partial IS1294 element inserted 180 bp upstream of the perA start codon (Fig. 4). We carried out multiplex PCR with primers K1547, K1549, and K1978, which gives a 500-bp product in strains with the IS1294 element upstream of perA, compared with the 650-bp product seen in strains that have an upstream IS1(). We were able to determine that the IS1294 element was restricted to the strains that have frameshifts in perA, namely, MB80 (O119:H2), CA 4225 (canine), MB21 (O128:H2), O142#15, and O142#19 (O142:H6).

View larger version (15K): [in this window] [in a new window]   FIG. 4.   The bfp-per region of EAF plasmids from different EPEC strains. Diagram 1, EAF plasmid of B171-8 representative of typical EPEC2 strains belonging to O111, O86, and O126 serogroups; diagram 2, EAF plasmid of E2348/69, representative of EPEC1 strains O55, O119:H6, and O127 and some O128 and O142 strains; diagram 3, EAF plasmid of O142#15, representative also of O142#19; diagram 4, EAF plasmid of MB80, representative of the O119:H2, O128:H2 (MB21), and canine strains. Asterisks indicate truncated genes, and sequences for which no data are available are represented by dashed lines.

Genetic relationships among per sequences and correlation with multilocus enzyme electrophoresis-derived EPEC clonal lineages. A phylogenetic tree generated from alignments of the perABC sequence and upstream (to 106 bp) nucleic acid sequence shows the close homology of the sequences (Fig. 5). Although all the sequences are closely related, there are two major clusters within the tree. The first comprises per from strains MB80 and O119#6, O142#15, and CA 4225, which cluster away from all the other strains. A subcluster of the second group comprises O55, O127:H6, O126, O86, and O111 per genes. Strains within this subcluster include conventional EPEC1 (O55, O127) and EPEC2 (O86, O111, and O126) strains (56) and possess -bundlin alleles (according to the classification of Blank et al. [6]). The per sequences from strains 0065900 (O114:H2), 1092-80 (O127:NM), and E851/71 (O142:H6) did not cluster with either of the two major groups but are closer to the EPEC1 and EPEC2 strains than to those with frameshifts. Interestingly, strains with IS1294 elements upstream of perAO142#15 (O142:H6), MB80 (O119:H2), and CA 4225 (canine)and the majority of the trcP-positive strains (those belonging to the O126, O86, and O111 serogroups but not the O114:H2 strain 0065900) form distinct clusters even though these sequences were not included in the region from which the tree was generated.

View larger version (22K): [in this window] [in a new window]   FIG. 5.   Unrooted tree for the nucleotide sequence of perABC and the upstream 106 nucleotides derived from the neighbor-joining alogarithm based on the gamma distance with  equal to 2. Bootstrap values from 1,000 tree replicates generated using MEGA2 are given at the branch points. The scale on the x axis indicates the number of substitutions per 100 nucleotides. Aligned sequences used to construct the tree were obtained from strains C54-58 (O55:H6), 5513-56 (O55:NM), E990 (O86:NM), 2966-56 (O111:H2), B171-8 (O111:NM), 065900 (O114:H2), MB80 (O119:H2), O119#6 (O119:H6), 023-220982 (O126), E2348/69 (O127:H6), 1092-80 (O127:NM), E56/54 (O128:H2), 15 (O142:H6)(15), E851/71 (O142:H6), and CA 4225 (canine).

Cloned frameshifted perA, -B, and -C alleles are unable to activate virulence gene promoters. Figure 6 shows that cloned per genes from O55:H6 (C54-58), O86:NM (E990), O111:NM (B171-8), O127:H6 (E2348/69) and O142:H6 (E851/71) isolates were able to mediate 4- to 5-fold increases in the activation of the LEE1 gene (P < 0.0005) and 9- to 11-fold increases in the activation of bfpA (P < 0.0001) -galactosidase promoter fusions. Conversely, clones from O119:H2 (MB80) and O142:H6 (O142#15) strains with frameshifts in perA were unable to activate these promoters above background level (P > 0.05). The cloned perA, -B, and -C genes from the O119:H6 strain O119#6 were able to activate both promoters but to lower levels: two- and threefold for the LEE1 gene and bfpA promoter fusions, respectively (P < 0.04). Although this level of activation was statistically significant, it was significantly lower than that seen with clones from E2348/69 and other strains with fully active per genes (P < 0.01).

View larger version (34K): [in this window] [in a new window]   FIG. 6.   -Galactosidase activities on LEE1 (A) and bfpA (B) promoter fusions of cloned perABC genes from different strains. (A) Activation of fusion JLM183 (211 to +221 relative to the LEE1 transcriptional start site fused to lacZ); (B) activation of fusion SS630 (688 to +124 relative to the bfpA transcriptional start site fused to lacZ). Clones from strains C54-58 (O55:H6), E990 (O86:NM), B171 (O111:NM), and E851/71(O142:H6) were able to activate these fusions at levels comparable to those of reference strain E2348/69 (O127:H6). Clones from strains MB80 (O119:H2) and O142#15 (O142:H6), with frameshifts in perA, were unable to activate the fusions above background while strain O119#6 (O119:H6) showed activation that was significantly above background (P < 0.04) but significantly below what was seen with clones that contained untruncated perA genes (P < 0.01).

Strains bearing inactive per genes can be identified by PCR-RFLP typing. We observed two varieties of frameshifts. The first was seen in the MB80 (O119:H2) and CA 4225 (canine) isolates. The second was seen in O142:H6 strain O142#15. In both alleles, there is an alteration in SspI restriction profile of the gene due to point mutations (Fig. 2). We therefore devised a PCR-RFLP analysis in which the perA gene is amplified with primers K1694 and K1693 and digested with SspI. We were able to distinguish three different patterns by agarose gel electrophoresis. Using this PCR-RFLP test, we were able to identify other strains with these alleles without sequencing (Fig. 7). O128:H2 strain MB21 produced a pattern consistent with that of the MB80 perA allele, and O142#19 (H6) showed a perA RFLP pattern similar to that of O142#15 (H6).

View larger version (63K): [in this window] [in a new window]   FIG. 7.   SspI-based PCR-RFLP typing distinguishes normal perA alleles (lanes 2 to 8 and 11 to 12) from those with O119:H2, O128:H2, or canine-type frameshifts (lanes 9 and 13) and O142:H6-type frameshifts (lanes 10). Test strains shown are E2348/69 (O127:H6) (lane 2), C54-58 (O55:H6) (lane 3), 5513-56 (O55:NM) (lane 4), E990 (O86:NM) (lane 5), 2966-56 (O111:H2) (lane 6), B171-8 (O111:NM) (lane 7), 065900 (O114:H2) (lane 8), MB21 (O128:H2) (lane 9), O142#15 (O142:H6) (14) (lane 10), C771 (O142:H6) (lane 11), E851/71 (O142:H6) (lane 12), and CA 4225 (canine) (lane 13). Lane 1, 1-kb ladder plus (Gibco, BRL).

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

EPEC strains were first described as diarrheagenic E. coli more than 50 years ago. However, insights into their pathogenic mechanisms have been uncovered only in the last 2 decades (36). The first molecular correlate of EPEC pathogenicity to be described was the EAF plasmid, which has subsequently been shown to carry genes encoding the BFP, Per, and other putative virulence factors (12, 18, 52). All the currently known virulence genes in prototype EPEC strain E2348/69 are activated by Per, either directly or via the LEE gene-encoded regulator, Ler (13, 18, 26, 30, 53, 54). Per is known to be required for the full virulence of EPEC (5). We therefore sequenced the perABC operons from representative EPEC strains and their upstream regions to determine if sequence variation could account for differences in levels of virulence gene expression and the epidemiological prominence of some EPEC serotypes.

A study by Nataro et al. (37) showed that EAF plasmids are at least partially conserved, and Gomez-Duarte and Kaper (18) noted that most EPEC strains harbor per genes. By examining the prevalence of perABC and other EPEC sequences, we were able to corroborate both reports and infer that, although there is some heterogeneity between EAF plasmids, the presence of the bfpA or per gene may be considered definitive for EAF plasmids. This is because all EAF-positive EPEC strains possess as least vestiges of the bfpA gene and all but one strain possess functional or mutated perA, -B, and -C genes, even though the intervening sequences differ among strains. The only other ubiquitous locus was the stbAB region, which is found in other plasmids (52). The product of the bfpA gene, bundlin, is surface expressed during infection and is therefore subject to evolutionary pressure from the host cell response, in a manner similar to that of intimin, evident from the mosaic nature of the surface-exposed region of BfpA (6). In addition, a minority of EPEC strains have lost the 3' end of their bfpA gene along with the rest of the operon (8). Per proteins are expressed intracellularly, and with the exception of strain O119#5, the perABC region was present in all the test strains. The per genes therefore represent a good choice for examining the phylogeny of EPEC plasmids in addition to evaluating structure-function relationships of the operon.

Most molecular and volunteer studies with EPEC have been conducted with prototype EPEC1 and EPEC2 strains, E2348/69 (O127:H7) and B171-8 (O111:NM), respectively. The perA, -B, and -C genes were cloned and sequenced by Gomez-Duarte and Kaper from strain E2348/69 (18). Subsequently, almost identical per genes were reported from strain B171-8 and designated bfpTVW (53). Comparison of these sequences showed that perBC and bfpVW are identical and that bfpT and perA differ very slightly (21). By sequencing the per operon from other strains, we demonstrate that, although these genes are highly conserved, point mutations and frameshifts in a few strains lead to reduction or abolishment of activity. We found that strains in which the perA gene is truncated before the C-terminal helix-turn-helix motif of the protein do not express a functional plasmid-encoded regulator. In all other strains, this helix-turn-helix region is 100% conserved. In all but the O119:H6 strain O119#6, which has multiple substitutions in per, all the per operons that were cloned demonstrated full, or in the case of those with frameshifted perA genes, no detectable activity. In O119#6, activity is significantly reduced but none of the many substitutions occur in the C-terminal DNA-binding domain of perA, suggesting that other regions are required for full activity.

Unlike the results seen with some virulence genes, notably eae (encoding intimin) (29), we found very little variation in the sequences of the perABC operons from different EPEC strains. Sequence comparison between the EPEC E2348/69 and enterohemorrhagic E. coli (EHEC) EDL933 LEE pathogenicity islands suggests that, while genes encoding proteins required for the regulation and delivery of virulence proteins (such as Ler and the type III secretion apparatus) are highly conserved, those encoding proteins that are exposed to the host cell (intimin and secreted proteins) are more divergent (41).

Although the sequence immediately upstream of the per genes is highly conserved, the sequence further upstream of this 180 bp is strikingly different. The presence of both the bfp and per operons in different strains and the homology between the bfpA and perA promotor regions (26) appears to suggest that the regions were acquired together and that the intervening DNA was acquired later. That different IS elements are found in this region may suggest that it is a hot spot for recombination or transposition. Because the bfp and perABC operons and the trcP gene of strain B171-8 have G+C content percentages significantly below that of most of the EAF plasmid (30 to 38% relative to 46%) and the G+C content percentage of intervening DNA is much higher (42 to 60%) (52), we cannot rule out the possibility that all three loci were acquired independently. Other IS elements have been found flanking virulence genes on large plasmids in pathogenic E. coli, suggesting that they play key roles in the evolution of virulence plasmids (9, 10, 25). Tobe et al. found that 29.5% of the sequence of plasmid pB171-8 was comprised of IS elements, many of which were truncated and most of which were located at truncation sites within other genes (52). This study and that of Bortoloni et al. (8) have found IS elements associated with the plasmids of other strains that are not present in the B171-8 plasmid. The variability of gene content among strains and the abundance of IS elements suggest that these elements may have played an important role in generating the variability in EAF plasmids among strains.

Cloned perABC operons from MB80 and O142#15 failed to activate bfpA and LEE1 gene fusions, and we therefore conclude that Per in these strains is inactive. We were able to detect expressed intimin but not bundlin in MB80 (O119:H2) and MB21 (O128:H2) strains with inactivating frameshifts in perA. The lack of expression of bundlin is, however, a function of a conserved deletion in the bfp operon at the 3' end of bfpA due to an IS66 insertion, previously identified by Bortolini et al. (8); hence, their inability to express bundlin is expected. The observation that the canine strain is similarly truncated in the bfp operon and perA genes provides an explanation for the findings of Beaudry et al., who observed that although bfpA- and EAF-homologous sequences are present in canine EPEC strains, most of these strains (including CA 4225) were unable to demonstrate LA (3). The ability of these strains to express intimin and CesD and of strain O142#15 (O142:H6) to be unimpaired in bundlin formation can be explained by analogy to the related pathogen EHEC, which possesses the LEE but lacks the EAF plasmid and its per genes. It is reasonable to believe that, like EHEC, EPEC strains that lack functional perABC operons regulate their virulence genes by other mechanisms. The absence of activity in perABC clones that are truncated before the helix-turn-helix motif in PerA provides evidence that this region is essential for activity and that the reduced activity in the clone from O119#6 (an H6 strain) suggests that other residues within the PerA protein are required for full activity. The perABC operon is the only EAF operon not involved in plasmid maintenance that has not been found to be truncated or interrupted by IS elements in at least some EPEC strains. This may suggest that there is some pressure to maintain intact per genes. Interestingly, although O142#15 (H6 strain) was found to have an inactivating frameshift mutation in perA, it showed no deficiency in virulence gene expression or phenotype in this study. Conversely, strain 2966-56 (O111:H2) possessed an intact per operon but failed to elaborate detectable intimin or to elicit the LA and A/E phenotypes. These strains warrant further study.

The identification of EPEC isolates from cases of diarrhea that lack a functional plasmid-encoded regulator poses questions about the pathogenic potential of these strains. We found that most of these strains were deficient in the expression of key virulence genes and in elaborating LA and A/E phenotypes in in vitro assays. Although there is currently no animal model for EPEC infection, it is known that EPEC strains that lack the EAF plasmid are rarely associated with diarrheal disease in epidemiological studies and, although not completely attenuated, are less likely to elicit diarrhea in human volunteers (5, 24, 36, 55). Data from this study and that of Bortolini et al. (8) imply that some strains lack the two major virulence factors present on the EAF plasmid. An antecedent from this is the fact that major serotypes associated with sporadic diarrhea and outbreaks caused by EPEC are O55:H6, O111:H2/NM, O114:H2, O119:H6, and O127:H6 (2, 55), all of which were found to possess intact perABC operons in this study. Also, the majority of O119 strains recovered from patients with diarrhea express the H6 flagellar antigen while O119:H2 isolates are relatively rare (19). The relative success of the H6 strains may be related to their possession of intact bfp and per operons. It is desirable to determine the epidemiological significance of per by determining the relative frequency of strains with nonfunctional operons among the heterogeneous O128:H2 and O142:H6 serotypes. Using tools generated in this study, it should be possible to examine a large number of EPEC by PCR-RFLP typing and multiplex PCR with a view to determining which strains contain frameshifts in perA and possess an IS1294 operon upstream of the per operon.

The results of this study have provided further evidence for the essential role of per in the pathogenicity of disease due to EPEC strains belonging to major enteropathogenic clones. Furthermore, in spite of the high level of conservation associated with per, EAF plasmids appear to be show some variation in gene arrangement and content within the bfp-per regions and at other loci.