We used PCR amplification and Southern blot hybridization to detect 9 virulence factors among 40 EHEC type-strains from the STEC Center, National Food Safety and Toxicology Center, Michigan State University (East Lansing, MI, USA). The virulence factor targets were the following: 1 chromosomal (eae [8]), 2 phage (stx1 and 2), and 6 plasmid (open reading frame [ORF] 1, ORF2 of pOSAK1 [1,2]; espP [9], hlyA [4,10], katP [11], and tox B [12] of pO157) (Table). DNA-DNA hybridization probes were made from virulence factors amplified from O157:H7 EDL933 genomic DNA.

PCR amplification was carried out with PCR primers (20 pmol/L each per 50 µL reaction) (Integrated DNA Technologies, Coralville, IA, USA) (Table) and 1 µL genomic DNA (extracted from overnight Luria-Bertani broth cultures according to PureGene DNA isolation kit instructions [Gentra Systems, Minneapolis, MN, USA] and dissolved in 50 µL 10 mmol/L Tris, pH 8.3) in a PCR cocktail containing 1× PCR buffer, 1.5 mmol/L MgCl₂, 1 U Vent exo(–) polymerase from New England BioLabs (Beverly, MA, USA), and 200 µmol/L each dATP, dGTP, dTTP, and dCTP. The mix was incubated for 30 cycles of 94°C, 40 s; annealing (for temperatures, see Table), 45 s; 72°C, 60 s, and a final 10-min extension at 72°C. Amplification products were confirmed by DNA sequencing.

³²P-labeled DNA probes were made from 2 µg PCR amplicons (purified by Montage PCR Cleanup Spin Column (Milipore Corp., Burlington, MA, USA). The DNA was denatured at 94°C, 40 sec; annealed (temperatures in Table) with 50 pmol/L of the appropriate PCR primers, 45 s extended for 2 h at 72°C. The 1× buffer contained the following: 1.5 mmol/L MgCl_2; 0.4 mmol/L each dATP, dGTP, dTTP; 2.0 µL 3,000 Ci/mmol α-³²P-dCTP (MP Bioscience, Buxton, UK); and 1.25 U Taq polymerase in a 50 µL final volume. Unincorporated ³²P-nucleotide was removed by Sephadex G-50 in Tris-EDTA, 1% sodium dodecyl sulfate (SDS).

Bacteria (800 µL overnight cultures) were transferred to Hybond-N+ nitrocellulose membrane (Amersham Biosciences UK Ltd, Buckinghamshire, UK) by dot-blot vacuum filtration apparatus (Schleicher and Schuell, Keene, NH, USA). Lysis and binding of genomic DNA fixation were carried out by exposure to lysis solution (1.5 mol/L NaCl, 0.5 mol/L NaOH) twice for 5 min each, and twice with neutralization solution (1 mol/L Tris-Cl, pH 7.4; 1.5 mol/L NaCl) for 5 min each. The filter was then submerged in 2× SSC with gentle agitation, air dried, and the DNA UV (254 nm) cross-linked at 120,000J/cm² (CL-1000 cross-linker, Fisher Biotech, Pittsburgh, PA, USA).

Probe hybridization was carried out in rotating hybridization bottles (Fisher Scientific Isotemp hybridization oven, Fisher Biotech) in 20 mL 6× SSC, 1% SDS at 68°C. Membranes were washed twice, for 1 min, in room temperature 2× SSC, 0.1% SDS, and twice at 45°C for 1 h in 1× SSC, 1% SDS. Hybridized membranes were exposed overnight with a phospho-imaging screen (Bio-Rad, Hercules, CA, USA) and visualized with a Personal Molecular Imager FX (Bio-Rad). The 3 chromosomal targets (stx1, stx2, and eae) were detected with 100% efficiency by both PCR and hybridization (no. positive by PCR/no. positive by hybridization): 21/21 stxI, 19/19 stxII, and 37/37 eae. Plasmid-associated genes, however, were detected with less efficiency relative to hybridization: katP: 15/17 (88%), hlyA: 26/27 (96%), espP: 19/23 (83%), toxB: 13/16 (81%), and both ORF1 and ORF2: 4/5 (80%) (Appendix Table).

Seventy-five percent (30/40) of the pathogenic E. coli strains tested contained at least 1 stx gene, 23% (9/40) were positive for both stx1 and stx2. The most common gene detected was intimin (eae), which was positive by both PCR and hybridization in 37/40 (93%) of the strains. While eae is strongly correlated with Shiga toxin, the adherence phenotype conveyed may be sufficient to cause a pathogenic state because 4 of the clinical isolates investigated contained only the eae gene.

Six plasmid virulence factor genes of pO157 and pOSAK1were targeted. Thirty-one (78%) of the 40 pathogenic strains tested were positive for at least 1 (by hybridization, PCR, or both) of the 4 genes, toxB, katP, hlyA, espP, which are usually carried on the archetypal pO157 plasmid: 11/31 (35%) retained all 4, 4/31(13%) carried three, 9/31 (29%) carried 2, and 5/31 (16%) carried only 1 (the sole PCR-positive/hybridization-negative isolate [espP in ED-31] was presumed to result from a nonspecific amplification).

Five EHEC strains (13%) hybridized to both ORF1 and ORF2 (from pOSAK1) (1,4), but only 4 were amplifiable. These same 4 also contained eae and at least 1 Shiga toxin gene (the 1 that failed to amplify [E851/71] lacked both stx1 and stx2).

The current accepted standard for EHEC identification is amplification of stx1, stx2, eae, and hlyA by PCR. However, this technology is generally only available at large hospital or state health laboratories. Hybridization is superior to culture screening methods and largely complimentary to PCR, but has a potentially broader epidemiologic application since it is unaffected by minor sequence variations that can completely inhibit PCR.

Only 3% of the 360 virulence factor hybridizations made in this study did not amplify. PCR failure is expected with its relatively higher sensitivity to single base primer-hybrid mismatch compared to whole amplicon hybridization. Notably, however, all 12 variations detected were among plasmid-associated virulence factors: 95% (228/240) of the plasmid hybridizable targets were amplified, compared to 100% (120/120) of the hybridizable chromosomal targets.

Although we detected the variable presence of genes ostensibly associated with 2 plasmids (pO157 and pOSAK1) and the bacterial chromosome, we did not attempt to verify either plasmid or chromosomal locations for any of the amplicons or DNA:DNA hybrids. While all virulence factor targets summarized in this study are subject to change there have been reports of any of the putative chromosomal or plasmid virulence factor targets in this study being found elsewhere.

Prager et al. (7) recently reported, using PCR alone, a wide variety of 25 virulence factor combinations among 266 pathogenic E. coli isolates representing 81 serotypes. Such diversity speaks directly to the need to accurately assess virulence factor presence to evaluate epidemiologic and clinical correlations. A similar 5% failure of the plasmid-associated virulence factor amplifications could have implications in such virulence factor correlations. Overall, however, these results are very similar to those of this study of prospective control strains. The use of a single control, such as EDL 933, will inherently bias PCR detection schemes since a failure of amplification in a test will be read as the absence of virulence factor element because it was amplifiable in the control.

If amplification failure is a measure of template variation, we find a much greater variability among plasmid-associated virulence factors. Although pO157 has been reported in most O157 H7 strains (13), our study demonstrates a high variability in the putative virulence factor content of pO157 as well as a highly variable content of pO157-associated virulence factors among the O157 isolates screened. Finally, pO157-associated virulence factors were detected among all but 4 of the 20 E. coli serotypes examined.