We have previously reported that aEPEC isolates obtained during a water quality study were heterogeneous in terms of serotype, intimin type and patterns of adherence to HEp-2 cells 20. This overall heterogeneity was confirmed by MLST analysis, which showed that 56 of the 79 aEPEC strains of human origin investigated in the study belonged to one of 11 different clades and that 23 strains could not allocated to a clade (Figure 1). As observed with phylogenetic analyses of A/E strains of E. coli in general, there was a tendency for each clade to contain strains with the same intimin and flagellar type. Five of 11 clades which contained aEPEC strains in this study were clades that include either tEPEC or STEC strains, whereas six clades were apparently distinct for aEPEC. These six clades comprised one which contained three strains with intimin-β and H7 (and O-antigens, O25 or O153); one clade with seven strains with intimin-ν and H19 (all O-nontypable [nt]); one clade with six strains with intimin-θ and H21 (O119 and Ont); a clade with five strains with intimin-ι and H8 or H- (O98, O107 or O177); one with four strains with intimin-κ and H10 or H- (O49, O88, and O153), and one with 13 strains with intimin-α and H6 or H34 (O71, O125, O126, and Ont). The last-mentioned clade was closely related to a tEPEC clade (EPEC-1), which also comprises strains with intimin-α and flagellar antigen, H6.

Of the strains that clustered with known clades of tEPEC or EHEC, three (all intimin-γ and O55:H7) belonged to the EHEC-1 clade, which also includes the pandemic, prototypical O157:H7 EHEC clone. Five aEPEC isolates grouped within the EPEC-2 clade which includes pEAF/BFP-positive strains with intimin-β and H2. The aEPEC serotypes in this clade included O15:Hnt, O114:H2, O117:H2, O128:H2, and Ont:H2. This clade also contained the prototypical aEPEC strain, E128012 12, two rabbit-specific EPEC (REPEC) strains, 84/110/1 and E22 (both of which carry intimin-β and are serotype O103:H2), and a calf isolate, also O103:H2, but with intimin-ε. Four aEPEC strains were assigned to the EPEC-4 clade (intimin-β and H6). All of these strains were intimin-β; two were O33:H6, and the others were O139:H14 and Ont:H6. Two aEPEC strains, both O2:H4, intimin-κ, belonged to the EPEC-3 clade, which also included tEPEC strain O86:H34, intimin-κ. Overall, three of 75 Australian and New Zealand isolates from humans reported here belonged to EHEC clades, 11 fell within EPEC clades, 38 belonged to clades distinct for aEPEC, and 23 could not be classified.

Of the eight Australian calf isolates examined, three were not assigned to a particular clade, and four belonged to the EHEC-2 clade (intimin-β and H11), but were of different serotypes from each other (O111:H-, O111:H11, O177:H11 and Ont:H-). This clade also included archetypal aEPEC strains E65/56 (O26:H-) and F41 (Ont:H-), which were isolated in Europe more than 50 years ago, and a well-studied, REPEC strain, 83/39 (O15:H-), in which Ral, a K88-like adhesin, was first identified 21. None of the Australian or New Zealand strains of human origin investigated in this study belonged to this clade, suggesting that there was no major exchange of aEPEC strains between the cattle and humans from whom these bacteria were obtained. One calf isolate (O103:H2) was assigned to the EPEC-2 clade, which also contained REPEC strains E22 and 84/110-1 (also O103:H2), tEPEC strain, Stoke W (O111:H2), and the prototypical aEPEC strain, E128012 (O114:H2). No aEPEC strains carrying intimin-β2, -ξ, -o, -ρ or -σ were identified in the entire collection of 87 test and 8 reference strains reported here.

None of the 67 Australian aEPEC strains of human origin investigated in this study was positive in the PCR for genes encoding the following adhesins of pathogenic E. coli: BFP, Lda, Pap, Saa, Afa, Sfa/Foc, K88, K99, Af/R2 or RalG (Table 1). On the other hand, all strains were positive in the PCR for FimH of Type 1 pili. Moreover, all isolates exhibited mannose-sensitive haemagglutination, indicating that they produced functional Type 1 pili.

One strain, NT731 (O45:H19, intimin-γ) was the only one of the 67 aEPEC isolates investigated that was positive in the PCR for Afa, a virulence determinant of uropathogenic E. coli that also serves as an adhesin in some EPEC strains 2223.

Two strains, R89 and R104, which belonged to a distinct aEPEC clade, characterised by intimin-β and H7 flagella, were positive in the PCR for AF/R1, a fimbrial adhesin and essential virulence determinant of the prototypical REPEC strain, RDEC-1 24. One strain, R219 (O51:H49, intimin-α), was the only one that tested positive for AggA, the pilin subunit of enteroaggregative E. coli 25. Three strains, NT286, NT731, and NT786, were positive in the PCR for Iha, an adherence-conferring protein that is highly prevalent in STEC 26. The Iha-positive aEPEC strains identified here were unrelated to each other and to STEC in terms of MLST type, intimin type and serotype (Table 1, Figure 1).

The PCR for long polar fimbriae (Lpf) was positive in 13 (19%) of 67 strains, all of which were from patients with diarrhoea. Lpf-positive strains occurred within the EPEC-2 clade (3 strains), the aEPEC clade with intimin-ν and H19 (7 strains), and the aEPEC clade with intimin-β and H7 (3 strains). Overall, seven of the LPF-positive isolates carried intimin-ν, and six had intimin-β.

We and others have previously reported that certain determinants of EPEC, which are not encoded by LEE or the EAF plasmid, such as Efa1, NleB and the cytolethal distending toxin (Cdt) are associated with virulence in attaching-effacing E. coli 1527. NleB was detected in 20 (30%) of the 67 strains tested, whereas Efa1 was detected in 8 strains (12%), all of which were also positive for NleB. NleB-positive strains were distributed amongst the following clades: EHEC-1 (3 strains), EPEC-2 (2 strains), aEPEC-1 (intimin-β, H7; 3 strains), aEPEC-3 (intimin-θ, H21 [or H6]; 6 strains). Six NleB-positive isolates could not be assigned to a clade, although all carried intimin-θ (Table 1). The Efa1-positive strains occurred within the EHEC-1 and EPEC-2 clades, as well as within the aEPEC-1 clade that was characterised by strains with intimin-β and H7. Seven (10%) strains were positive in the PCR for Cdt. Three of these strains belonged to the aEPEC-6 clade (intimin-α and H34), one belonged to EPEC-2 (intimin-β, H2), and three were unassigned (Table 1).

To determine if aEPEC carry DNA sequences related to those that code for the production of BFP, but were not amplified by the PCR for BfpA, we investigated the aEPEC strains by DNA hybridisation using probes derived from the bfpA and bfpB genes of EPEC strain E2348/69. Unexpectedly, six isolates (ESA212, R176, R182, R228-1, R281, and W114) hybridised with the BfpA probe at high stringency. Three of these strains belonged to the aEPEC clade with intimin-ι and H8, but they belonged to different O-serogroups. The other three probe-positive strains also differed from each other. Six strains hybridised with the BfpB probe. Four of these were positive for intimin-α, three carried H34, two carried H6, but all were of different serotypes. No strain hybridised with both Bfp probes.

Some aEPEC strains from humans and animals express adhesins that are homologous to the K88 fimbriae of enterotoxigenic E. coli 2128. To determine if the aEPEC strains in our collection carried similar sequences, we probed these strains for the fae gene of K88, but none of the aEPEC hybridised with this probe, even when tested at low stringency.

After incubation for three hours with HEp-2 cells, 54 (81%) of 67 aEPEC strains were adherent: 24 strains adhered in an aggregative pattern, and two in the pattern termed "localised-like adherence", because it resembles BFP-mediated localised adherence, but the bacteria are more loosely associated with each other than BFP-bearing strains. Twenty-eight strains showed an indeterminate pattern of adherence described previously 20, in which bacteria adhere in a mixed pattern of diffuse and localised-like adherence. Thirteen strains did not adhere to HEp-2 cells after 3 hours. Extending the contact period of the bacteria with the HEp-2 cells from 3 to 6 hours allowed some strains that had not adhered after 3 hours to adhere in an indeterminate pattern, but the pattern of strains that were adherent after 3 hours did not change.

Five of the six strains that hybridised with the BfpA probe were non-adherent after three hours, whereas five of the six BfpB-positive strains showed aggregative adherence and one showed localised-like adherence. Adherence to HEp-2 cells was not associated with a positive PCR for either Lpf or Efa.

The 67 aEPEC strains we investigated by PCR, DNA hybridisation and for adherence to HEp-2 cells originated from individuals with different clinical presentations (Table 2). Fifty-seven isolates were obtained from patients with diarrhoea, and ten were from asymptomatic individuals. Eleven isolates were from children with persistent diarrhoea (i.e., diarrhoea lasting more than 14 days), and 12 were from children with diarrhoea less than 14 days in duration. Thirty-four strains were from patients in whom the duration of diarrhoea was not known. To determine if any of the putative accessory virulence determinants of aEPEC that were sought in this study were associated with a particular clinical presentation, we compared the frequency of these determinants in isolates from patients with and without diarrhoea, and those known to have acute or persistent diarrhoea. The results showed that the frequency of the factors investigated did not differ significantly between the groups under comparison (P > 0.1, Fisher's exact test, two-tailed).

For the purposes of this study, aEPEC were defined as strains of E. coli that were positive by PCR for the eae gene, but negative by PCR for the genes for BfpA and Shiga toxins 1 and 2, using the PCR primers and conditions described previously 14. Sixty-seven of the aEPEC strains investigated in this study were isolated in our laboratory during the course of four separate studies: a Melbourne water-quality study (21 strains with a prefix of W 20, a study examining the clinical features of children infected with aEPEC at the Royal Children's Hospital, Melbourne, Australia (28 strains with a prefix of R 14), a study of the prevalence of STEC in Melbourne (3 strains, prefixed ESA), and a study of acute diarrhoea in children in the Northern Territory, Australia (15 strains, prefixed NT 33). The remaining clinical aEPEC isolates were E128012, from a case of sporadic infant diarrhoea in Bangladesh 12, F41 (Denmark 45), E65/56 and D5301 (England 464748), all of which are archetypal aEPEC strains 49. We also tested 8 clinical aEPEC strains from New Zealand (kindly supplied by Jenny Bennett, ESR Ltd., Porirua, New Zealand) and eight aEPEC strains isolated from symptomatic cattle in Australia 18 (kindly supplied by Dr Steven Djordjevic, Elizabeth Macarthur Agricultural Institute, Camden, NSW, Australia).

Reference strains of E. coli included in the phylogenetic analysis of the aEPEC strains were: tEPEC (eae+ bfpA+) strains, E2348/69, E990, Stoke W and C771 1249; REPEC strains, E22 50, 83/39, 84/110-1 51, and an STEC O157:H7 strain, EDL933, which is LEE-positive and classified as enterohemorrhagic E. coli (EHEC) 52. E. coli strains used as controls for PCR included enteroaggregative E. coli strain 17-2 53; STEC strains, EH41 54, and EH52 (this study); enterotoxigenic E. coli strain K88 and E. coli K12-K99+ (courtesy of Professor Peter Reeves, University of Sydney, Sydney, NSW, Australia); REPEC strains, B10 55, 83/39 and RDEC-1 56, and uropathogenic E. coli strain J96 57. Adherent-invasive E. coli strain LF82, which was isolated from a chronic ileal lesion of a patient with Crohn's disease, and 52D11 (an isogenic fimA mutant of LF82) 43 were kindly supplied by Dr Arlette Darfeuille-Michaud, Université d'Auvergne, Clermont-Ferrand, France, and used as controls to test for mannose-sensitive haemagglutination. Unless otherwise specified, bacteria were routinely subcultured on horse blood agar or Luria-Bertani agar (BD Difco, Franklin Lakes, NJ) at 37°C.

Genomic DNA was isolated from E. coli using hexadecyltrimethylammonium bromide (CTAB) as described in Ausubel et al. 58, and was used as the template for all experiments requiring DNA.

Eighty-three test strains isolated from humans or cattle in Australia and New Zealand, together with four archetypal aEPEC and eight A/E E. coli control strains were subjected to MLST analysis using the methods described on the EcMLST website http://www.shigatox.net/mlst. Briefly, seven housekeeping genes (aspC, clpX, fadD, icdA, lysP, mdh and uidA) were amplified with AmpliTaq Gold in 50 μl reaction volumes. PCR products (5 μl) were electrophoresed on 1% agarose gels to check the size and yield. The remaining 45 μl was purified using the QIAquick PCR Purification Kit (Qiagen, Valencia, CA) and eluted in 20 μl elution buffer. Both strands of each gene were sequenced using ABI PRISM BigDye Terminator (Applied Biosystems, Foster City, CA) according to the manufacturer's instructions. Sequences were checked and cropped to the required length using Sequencher 4.0 (Gene Codes, Ann Arbor, MI). Sequences were concatenated for phylogenetic analyses, and aligned with the ClustalW algorithm using the MegAlign module of the Lasergene software (DNASTAR Inc., Madison, WI). Neighbour-joining trees were constructed using the Kimura two-parameter model of nucleotide substitution with the MEGA3 software (Center for Evolutionary Functional Genomics, Tempe, AZ) 59. The inferred phylogenies were each tested with 500 bootstrap replications.

The sequences of the aspC, clpX, fadD, icdA, lysP, mdh and uidA genes used for the MLST analysis have been deposited in the GenBank data base under accession numbers GQ130379 to GQ131022.

The eae gene was subtyped by using the restriction fragment length polymorphism assay described by Ramachandran et al. 60. This method permits detection of the following intimin types: α (alpha), β (beta), β2, γ (gamma), ε (epsilon), ζ (zeta), θ (theta), ι (iota), κ (kappa), λ (lambda), ν (nu), ξ (xi), o (omicron), ρ (rho), and σ (sigma).

PCR amplifications were performed in a GeneAmp PCR System 9700 thermal cycler (Applied Biosystems) or an iCycler (Bio-Rad Laboratories, Hercules, CA) with AmpliTaq Gold polymerase (Applied Biosystems) in a reaction volume of 20 μl. The genes, primers, amplicon size and PCR conditions used for these studies are listed in the additional file (see Additional file 1). The test strains for these analyses, and those described below, were the 67 aEPEC strains obtained from humans in Australia. The following E. coli strains were used as positive controls: E2348/69 (bfpA), 83/39 (efa1, ralG), EDL933 (iha, nleB1), EH41 (saa, lpfD_O113); K88 (fae operon), K12-K99+ (fan operon), 17-2 (aggA); J96 (fimH, papA, sfa/focDE, focG), EH52 (afaC), RDEC-1 (afr1), B10 (afr2), and E990 (cdt). PCR products were electrophoresed on 1–1.5% Tris-acetate-EDTA agarose gels and stained with ethidium bromide before visualisation on a UV transilluminator.

Genomic DNA was spotted onto Magna Nylon Transfer Membranes (GE Osmonics, Trevose, PA) and denatured and neutralised according to the "DIG System User's Guide for Filter Hybridisation" (Roche, Mannheim, Germany). Transferred DNA was UV-crosslinked using a Spectrolinker XL-1000 UV crosslinker (Spectronics Corp., Westbury, NY). Digoxigenin-labelled DNA probes were prepared by PCR (Roche) using primers to detect bfpA (Table 1); primers MP-bfpB-F (GATAAAACTGATACTGGGCAGC) and MP-bfpB-R (AGTGACTGTTCGGGAAGCAC) to detect bfpB 61; and primers faeEF (ATGCGCCGGGTGATATCA) and faeER (TTATTTCTGCTCTGCGGT) to detect faeE. EPEC E2348/69 was used as template for the bfpA and bfpB probes and enterotoxigenic E. coli strain K88 was used as template for the faeE probe. These strains were also included as positive controls on the appropriate membranes. Before use, probes were sequenced using ABI PRISM Big Dye Terminator as described above. Sequencing reactions were purified using MgSO₄ and submitted to the Australian Genome Research Facility (Parkville, Vic, Australia). Hybridisation was performed at low stringency (hybridisation at 50°C followed by washing in 2× SSC at room temperature), medium stringency (hybridisation at 59°C, washing in 0.1× SSC at 59°C), and high stringency (hybridisation at 68°C, washing in 0.1× SSC at 68°C), and detected by using chemiluminescence as recommended by the manufacturer. Bacteria were considered probe-positive if the intensity of the spot was similar to that of the positive control.

The Center for Vaccine Development method was used to determine the pattern of bacterial adherence to HEp-2 epithelial cells 62. The criteria used to assign bacterial adherence to a particular pattern have been described previously 20.

The expression of Type I pili was determined by investigating bacteria for mannose-sensitive haemagglutination of guinea pig erythrocytes. For these assays, bacteria were grown in Brain Heart Infusion broth (Oxoid Ltd., Basingstoke, England) or Antibiotic Medium No. 3 (Penassay broth, PAB; Oxoid Ltd.) at 37°C without shaking, and tested for haemagglutination using the method described by Iida et al. 63. E. coli strains, LF82 and 52D11, were used as positive and negative controls, respectively. Strains that were haemagglutination-negative were retested after passage through Brain Heart Infusion broth to enhance the expression of Type I pili.