Genotypic Characterization and Prevalence of Virulence Factors among Canadian Escherichia coli O157:H7 Strains

Bacterial strains.Initially, 69 E. coli O157:H7 strains (set 1), comprising 37 strains from humans and 32 strains from cattle, were selected from the culture collection of the Laboratory for Food-Borne Zoonoses based on their PT (Table 1) . These strains had been isolated from the provinces of Alberta, Ontario, Saskatchewan, and Quebec. A total of 43 additional human strains and 75 additional bovine strains were subsequently included to examine certain trends that became evident after initial analysis of set 1. Altogether, 187 test strains comprising 80 human and 107 bovine sources were selected (Table 1). In addition, 10 E. coli O157:H7 reference strains of known OBGS and LSPA6 type (28) were used in the study as positive controls in the LSPA6 and OBGS studies (Table 2). All strains were stored at −70°C in brain heart infusion broth (Difco) with 15% glycerol. When required, the strains were grown in 1 ml of brain heart infusion broth overnight at 37°C.

Phage typing and virulence gene profiling.PTs were determined for all of the strains as described previously (1, 26). All of the strains were tested by PCR to determine their Shiga toxin (stx) type, the presence of the EHEC hemolysin (ehxA) and the attaching and effacing genes (eaeA) (40, 57). Set 1 strains were stx₂ subtyped (4). Since the method used for stx₂ subtyping does not differentiate between stx_2c, stx_2da, and stx_2v(521) (57), all strains were tested by an stx_2c-specific PCR developed by Rolgaard and coworkers that successfully detected 97% of Stx_2c-producing O157:H7 strains (B. Rolgaard, unpublished data). The primer sequences were designed to specifically target genes upstream of the stx gene, namely, the q and ileZ genes encoded by the Stx phages, 933w (stx₂) and Nil2 (stx_2c). The primer sequences were as follows: the q_Nil2 forward primer, 5′-TGC CGA CGG GAA GTT GAC-3′; the q_933w forward primer, 5′-GGC GGC AAA TAA CTA TGA GG-3′; and the ileZ reverse primer, 5′-AGC GAC CTG GCG ATT ATG AG-3′. Briefly, each 25-μl reaction mixture contained 5 μl of cell template, 1× PCR buffer (50 mM KCl, 10 mM Tris-HCl [pH 8.3], 1.5 mM MgCl₂, and 0.01% [wt/vol] gelatin; Applied Biosystems, Foster City, CA), 200 μM concentrations of each deoxynucleoside triphosphate (Applied Biosystems), each primer at a concentration of 0.5 μM, and 0.1 U of AmpliTaq (Applied Biosystems)/μl. Amplification was conducted at 94°C for 4 min, followed by 30 cycles of 94°C for 30 s, 47.0°C for 60 s, and 72°C for 90 s. Products of 478 and 886 bp were produced for the O157:H7 strains containing stx_2c and stx₂, respectively.

The presence of the colicin genes K, V, B, and M was determined on set 1 strains by using previously described PCR assays (21, 50). Colicins Ia/Ib and E2/E3/E6/E7 genes were detected in set 1 strains by PCR using previously described primers (50) with the following conditions: each 25-μl PCR mixture contained 1 μl of cell template, 1× Gold buffer (Applied Biosystems) 1 μM concentrations of each primer, 0.2 mM concentration of each deoxynucleotide triphosphate (Applied Biosystems), 1 mM MgCl₂, and 0.036 U of AmpliTaq Gold (Applied Biosystems). Thermal cycling conditions were the same as in the published report (50) except that the initial denaturation step was at 94°C for 10 min. The presence of the colicin D gene, cda, was determined on all test strains by PCR, as previously described (21).

LSPA6 and OBGS typing.All strains were typed by the LSPA6 method as two reactions rather than the one multiplex reaction as originally described (54). The LSPA6 primer sequences used in the present study are listed in Table 3. One multiplex reaction contained folD, Z5935, rbsB, and arp primers and the second multiplex reaction contained yhc and rtcB primers. Lysates for PCRs were prepared by boiling washed cells (44). Amplification reactions contained 4 μl of the cell lysate combined with 16 μl of the mastermix that consisted of 1× PCR buffer (50 mM KCl, 10 mM Tris-HCl [pH 8.3], 1.5 mM MgCl₂ and 0.01% [wt/vol] gelatin), 250 μM concentrations of each deoxynucleoside triphosphate, each primer at a concentration of 0.5 μM, and 0.05 U of AmpliTaq (Applied Biosystems)/μl. Amplification was conducted at 94°C for 5 min, followed by 30 cycles of 94°C for 30 s, 52.2°C for 30 s, and 72°C for 45 s. Aliquots (5 μl) of the amplified products were analyzed by standard submarine gel electrophoresis using a 3% Metaphor agarose gel (Cambrex, Rockland, ME) stained with ethidium bromide, and the bands visualized with UV transillumination. Images of the amplicon gel profiles were captured by using a Gel Doc 100 (Bio-Rad, Mississauga, Ontario, Canada) and analyzed in Bionumerics (version 3.5; Applied Maths, Austin, TX).

LSPA6 alleles and genotypes were defined by using a binary character table that was generated in Bionumerics and exported to Microsoft Excel. Alleles shared with the lineage I control strain were designated 1, and those shared with the lineage II control strain were designated 2. Unique alleles were designated 3, and a zero character was assigned if no band was present. To generate the genotype classification, the alleles were placed in the following order: folD, Z5935 gene, yhcG, rtcB, rbsB, and arp-iclR.

Thirty-five strains from set 1 were also tested by traditional OBGS, using the primer pairs KOCT22A/KOCT6C, KOCT21/KOCT13C, KOCT19/6C, KOCT3B/KOCT4C, and KOCT14/KOCT1C. The OBGS reference strains were included in each gel. Binary files were created in Microsoft Excel from printed copies of the images produced by an Alden 9315CTP photographic quality thermal printer (Alden Electronics, Westborough, MA). The files were generated from the presence or absence of bands between 200 and 1,500 bp. The lineage designation was obtained by comparison of the binary strings of each sample to the binary string of the OBGS reference strains.

All strains that were LSPA6 genotype 111111 were classified as lineage I. Most other allele combinations corresponded to OBGS lineage II. LSPA6 genotype 211111 was originally considered a lineage II strain (54), but recent reports indicate that LSPA6 genotype 211111 strains have characteristics of both lineage I (LSPA6 111111) and lineage II strains (52, 55). Accordingly, LSPA6 genotype 211111 was designated as lineage I/II and was not considered a lineage II LSPA6 genotype in the strain comparisons for the present study.

Shiga toxin production.The levels of Stx₂ produced by a subset of strains were quantified by using a Stx-enzyme-linked immunosorbent assay (Stx-ELISA) with monoclonal antibodies that recognize Stx₂, Stx_2c, Stx_2d, and Stx_2e variants (2). Quantification was achieved by generation of a standard curve with known concentrations of purified Stx₂. The results were recorded as picograms of toxin per 10⁸ CFU. Cultures were prepared with or without induction by mitomycin C as previously described (46) except that 1 ml of the cells were lysed by incubation at 37°C for 5 min after the addition of a warm solution of polymyxin B (Sigma) to a concentration of 1.5 mg/ml rather than by sonication.

Antimicrobial resistance testing.Antimicrobial resistance was determined as previously described (42) using the following antibiotics and their respective concentrations: amikacin (16 μg/ml), ampicillin (32 μg/ml), carbadox (30 μg/ml), cefotaxime (64 μg/ml), cefoxitin (32 μg/ml), ceftiofur (8 μg/ml), ceftriaxone (8 μg/ml), cephalothin (32 μg/ml), chloramphenicol (32 μg/ml), ciprofloxacin (0.125, μg/ml), cotrimoxazole (80 μg/ml: sulfisoxazole and trimethoprim at 76 and 4 μg/ml, respectively), florfenicol (16 μg/ml), gentamicin (16 μg/ml), kanamycin (64 μg/ml), nalidixic acid (32 μg/ml), nitrofurantoin (64 μg/ml), spectinomycin (64 μg/ml), streptomycin (64 μg/ml), sulfisoxazole (512 μg/ml), tetracycline (16 μg/ml), tobramycin (8 μg/ml), and trimethoprim (16 μg/ml). These antibiotics were chosen since they are or have been used therapeutically in veterinary or human medicine or as growth promoters in animal agriculture and may provide a selective advantage to resistant strains. Quality control strains included American Type Culture Collection (ATCC) strains Pseudomonas aeruginosa ATCC 27853, E. coli ATCC 25922, and E. coli ATCC 35218 (36, 37), as well as E. coli strain R1022, a multiresistant bovine isolate.

Plasmid profiling.Plasmid DNA was isolated by the method of Crosa and Falkow (10), except that the strains were grown overnight on veal infusion yeast extract agar, scraped off the surface of the agar with a sterile toothpick, and suspended in lysis buffer (pH 12.4). The plasmid DNA was subjected to electrophoresis in a horizontal 0.7% agarose gel in Tris-acetate buffer and stained with ethidium bromide, and the bands were visualized by UV transillumination. Images were captured by using a Gel Doc 100 (Bio-Rad). Plasmids from PT 31 and PT 33 strains of E. coli O157:H7 appeared to be degraded by nucleases. For these strains, the addition of 50 μM thiourea (Sigma) to the running buffer was used to protect against degradation (43).

Statistical analysis.All pairwise associations between LSPA6 genotypes, PTs, the presence of stx_2c and cda genes, and AMR were tested by using the Fisher exact test (http://www.matforsk.no/ola/fisher.htm ). (For a comprehensive listing of P values, see Tables S1, S2, and S3 in the supplemental material.)

Distribution of LSPA6 and OBGS types.In order to examine the frequency of specific LSPA6 types in the collection of E. coli O157:H7 strains, an initial subset of 69 strains (set 1, Table 1) was selected based on the approximate proportions of each PT in our collection (Table 1). PT 14 was the most common PT in the collection, which is consistent with what has previously been reported for Canada (11). The other 16 PTs included: 1, 2, 4, 8, 21, 23, 31, 32, 33, 34, 45, 54, 67, 74, 82, and 87. Among the 69 strains, five LSPA6 genotypes were identified: 57 (82.6%) were of LSPA6 genotype 111111 (LSPA6-LI), while the remaining 12 (17.4%) strains were of other LSPA6 genotypes. These included four (5.8%) LSPA6 genotype 211111 (LSPA6-LI/II) strains, five (7.2%) LSPA6 genotype 222222 strains, two (2.9%) genotype 222212 strains, and one (1.4%) genotype 222213 strain. LSPA6 and OBGS typing results from the 35 strains tested by both methods agreed with respect to lineage designation, with the exception of three strains that could not be typed by the OBGS method but which were resolved by LSPA6 typing (data not shown).

Association between LSPA6 type, PT, and source of isolation.A strong association between PT and LSPA6 genotype was observed in the 69 strains from set 1 (Table 1). However, certain PTs had very few representatives, and additional strains of the same PTs were selected at random from the larger collection of E. coli O157:H7 strains to further examine this association. In this expanded set of strains, PTs 4, 14, 21, 31, 33, and 87 were all genotype LSPA6 111111 (LSPA6-LI), while three PTs (45, 67, and 74) had minor differences in LSPA6 genotypes but remained within lineage II (e.g., PT 67 had six strains of LSPA6 genotype 222222 and two of LSPA6 genotype 222223) (Fig. 1). Two PTs (PT 8 and PT 32) had strains of LSPA6 111111 and 211111. Only 2 of 17 PTs (PT 1 and PT 34) had strains from both LSPA6-LI and LSPA6-LII genotypes. To the best of our knowledge, this is the first time that an association has been demonstrated between PT and LSPA6 genotype.

• Open in new tab
• Download powerpoint

FIG. 1.

Relationship between PT and LSPA6 genotype for the 187 E. coli O157:H7 strains in Table 1.

Within this expanded set of strains, the LSPA6-LI genotype made up 73.8% (59 of 80) and 49.5% (53 of 107) of E. coli O157:H7 strains isolated from human and bovine sources, respectively (Fig. 2). The relative overrepresentation of LSPA6-LI among human isolates was also observed by Yang et al. (54) in their study of a large collection of bovine and human isolates. Nine LSPA6-LI/II (211111) genotype strains were of human origin and five were of bovine origin. LSPA6-LII genotypes represented 15.0% (12 of 80) of the isolates from humans (222222, 222212, 222213, 232212, 212222, and 222121) and 45.8% (49 of 107) of the isolates from cattle (222222, 222223, 222212, 222213, 232212, 212222, 212232, 222121, 232222, and 232212). The bovine source bias among LSPA6-LII genotypes was highly significant (Fisher exact test; P = 8.047 × 10⁻⁶). Certain lineage II-related PTs such as PT 23 also had a significant (Fisher exact test; P = 1.681 × 10⁻²) bovine host isolation bias. The strong association between PT and LSPA6 genotype suggests that some PT/LSPA6 genotype combinations are stable, although further investigation will be required to explore whether this association is widespread.

• Open in new tab
• Download powerpoint

FIG. 2.

Relationship between source and LSPA6 genotype for the 187 E. coli O157:H7 strains in Table 1.

Association between LSPA6 genotype, PT, Stx₂ genotype, and toxin production.Considerable variation was found among LSPA6 genotypes in stx gene content and stx₂ genotype. stx₁ and stx₂ were identified in 79% of strains tested, stx₂ alone was present in 17% of the strains, stx₁ alone was present in 3% of the strains, and one strain did not contain either toxin-producing gene. The stx_2c variant was detected by the stx_2c specific PCR in 59 of 61 LSPA6-LII strains (96.7%). An additional band of approximately 660 bp was noted for all strains regardless of the stx₂ content; however, the nature and significance of this band is not known (data not shown). The remaining two strains contained only the stx₁ gene. In contrast, the stx_2c variant was only detected in two of the 112 LSPA6-LI strains (1.8%) and seven of 14 LSPA6-LI/II strains (50%) (Table 4). The two LSPA6-LI strains contained both stx_2c and stx₂. The difference in the carriage rates of stx_2c between LSPA6-LII strains and LSPA6-LI genotype strains was found to be statistically significant (Fisher exact test; P ≥ 3.089 × 10⁻⁴¹). Although Stx_2c has been reported to have the same level of toxicity as Stx₂ in mice (32), the risk of developing HUS after infection with stx_2c-containing STEC has been reported to be significantly lower than with STEC possessing the stx₂ toxin genotype (14). This is also in agreement with other studies that report that most HUS cases were associated with E. coli strains carrying stx₂ rather than other stx₂ variant genes (6, 49).

Although it is possible that these Stx₂ variants differ in toxicity for humans, it has been suggested that Stx_2c-producing strains simply produce less Stx (7, 12, 38). To determine whether this was the case, Stx₂ production by eight Stx₂-producing strains (seven LSPA6-LI strains and one LSAP6-LI/II strain) and nine Stx_2c-producing strains (all LSPA6-LII) was quantified by an Stx-ELISA that reacts with all Stx₂ variants. Figure 3 shows the amount of toxin produced by selected strains from LSPA6-LI, -LII, and -LI/II before and after induction with mitomycin C. For five of the nine LSPA6-LII Stx_2c-producing strains, the amounts of toxin produced without induction with mitomycin C were very low compared to other LSAP6-LII strains and the Stx₂-producing LSAP6-LI and LI/II strains. Since the low-toxin-producing strains were from PTs 23 and 67 strains, an additional five PT 23 and five PT 67 strains were tested by the Stx-ELISA. These strains also produced very low levels of toxin and showed similar increases on induction with mitomycin C. Therefore, low production of Stx_2c may be more strongly associated with these PTs rather than strictly with the LSPA6 genotype. Production of low levels of Stx2c by O157:H7 strains of PT 23 and 67 may explain the infrequent association of these PTs with human disease, and further studies should be performed to examine this relationship.

• Open in new tab
• Download powerpoint

FIG. 3.

Stx₂ or Stx_2c production by selected E. coli O157:H7 strains.

Association between plasmid profiles and LSPA6 genotype.Thirteen different plasmid profiles were observed among the E. coli O157:H7 strains in set 1 (data not shown). Nine plasmids ranging in size from 70 to 2.2 MDa were identified in set 1 strains. All strains possessed the large 60-MDa pO157 virulence plasmid and were PCR positive for the enterohemolysin gene (hlyA). The two predominant profiles among these strains were the presence of pO157 alone and pO157 combined with a 2.2-MDa plasmid. These two profiles were observed in 56.9 and 15.3% of strains, respectively. The common occurrence of the 2.2-MDa plasmid is in agreement with another study that found this plasmid in 57% of O157 strains (51). In the present study, only PTs 21, 31, 32, and 33 contained the 2.2-MDa plasmid, and there appeared to be no bias in distribution among human and nonhuman strains. Previously, a 2.2-MDa plasmid, designated p4821, was characterized by Haarmann et al. (18) and found to be similar to the antibiotic resistance plasmid of Salmonella enterica serovar Typhimurium strains. There did not appear to be any significant association between specific plasmid profiles and LSPA6 genotype or isolate source.

Association between colicins and LSPA6 genotype. E. coli O157:H7 strains positive for the colicin D gene cda were found in 26 of 112 (23.2%) LSPA6-LI strains, 1 of 14 (7.1%) LSPA6-LI/II strains, and 31 of 61 (50.8%) LSPA6-LII strains (Table 4). The difference in cda carriage rates between LSPA6-LI and LII strains was highly significant (Fisher exact test; P = 3.427 × 10⁻⁴). Interestingly, the cda gene was observed in only three out of the eight LSPA6-LII-associated PTs (PTs 23, 45, and 67) (Fig. 4), and carriage of cda was largely confined to certain LSPA6-LI PTs (PTs 1, 4, and 8) as well. Therefore, it appears that possession of cda may be more strongly associated with PT than with the LSPA6 genotype.

• Open in new tab
• Download powerpoint

FIG. 4.

Relationship between PT and presence of cda for the 187 E. coli O157:H7 strains in Table 1.

Other colicins were found at a much lower frequency among E. coli O157:H7 strains; e.g., one of 69 (1.4%) strains had colicins B and M and two strains (2.8%) had colicin Ia/Ib (data not shown). The higher prevalence of cda among the E. coli O157:H7 strains is in agreement with the results of previous reports (15, 51). As expected, all strains that contained the 4.6-Mda plasmid were found by PCR to possess the cda gene (8, 21).

Colicins may offer a selective advantage to their host (45). In one study, pColD157 was identified in the majority of E. coli O157:H7 strains associated with human disease (8). Another study found that pColD157 was less prevalent in recent E. coli O157:H7 isolates than those obtained several years previously (21). Although interesting, the significance of a higher frequency of cda in lineage II strains observed in the present study is unknown. Since cda is not widespread among E. coli O157:H7 strains, it does not appear to be essential, or perhaps its function (or that of genetically linked genes) has been replaced by some other attribute of the organism.

Association between AMR and LSPA6 genotype.Only 7.2% (5 of 69) of the E. coli O157:H7 strains (set 1) selected from the collection based on PT frequency displayed resistance to one or more of the antibiotics tested. All of the resistant strains from this initial set of strains had combined resistance to streptomycin, sulfisoxazole, and tetracycline (AMR-SST) and one AMR-SST bovine strain was also resistant to spectinomycin. Other studies have also found that resistance to streptomycin, sulfisoxazole, and tetracyclines is common in E. coli O157 strains (27, 34, 53, 56). Only 1 of the 37 human strains from set 1 displayed resistance to the antimicrobials tested, whereas 4 of the 32 (12.5%) bovine strains showed AMR (data not shown). Four of the five E. coli O157:H7 strains with AMR were of LSPA6-LII genotypes and also of PTs 23, 45, and 67.

To determine whether any association exists between AMR and LSPA6 genotype and/or PT, additional strains from each PT were tested for AMR (Table 1). The results for all strains indicated that within LSPA6-LII strains, 44 of 61 (72.1%) were AMR-SST positive (Table 4) compared to only 1 of 112 (0.9%) of the LSAP6-LI strains (Fisher exact test; P = 7.536 × 10⁻²⁶). For strains of LSPA6-LII-associated PTs, 24/25 PT 23, 11/11 PT 45, and 8/8 PT 67 strains were AMR-SST, whereas strains of the LSPA6-LII-associated PTs, PT 54 and PT 74, had no AMR. Based on these data it would seem that AMR-SST is largely confined not only to LSPA6-LII strains but also to specific PTs within this group (Fig. 5). Although it can be argued that the strains in the collection may not be representative of these PTs from other geographical regions, Mora et al. (34) have also recently reported an association between PT 23 E. coli O157:H7 strains and AMR-SST in Spain (34). It is possible that AMR may play a role in the survival and persistence of LSPA6-LII E. coli O157:H7 strains in the bovine reservoir.

• Open in new tab
• Download powerpoint

FIG. 5.

Relationship between PT and combined antimicrobial resistance to streptomycin, sulfisoxazole, and tetracycline for the 187 E. coli O157:H7 strains in Table 1.

At 34.5%, the AMR prevalence in bovine strains in the present study is similar to the rate observed in the United States (33) but lower than the 65% AMR rate reported for E. coli O157 strains from cattle in feedlots in Saskatchewan, Canada (53), and the 53% AMR rate of E. coli O157 strains from cattle in Spain (34). However, comparisons between studies should consider differences in study design, strain selection, geographic region, and scope of the study.

Conclusions.In our comparison of characteristics of LSPA6 genotypes and PTs of E. coli O157:H7 strains isolated from Canadian sources, the majority of strains were genotype LSPA6-LI, and smaller proportions belonged to the LSPA6-LI/II and LSPA6-LII genotypes. One of the key findings was that PTs 23, 45, 54, 67, and 74 were genotype LSPA6-LII specific. As has previously been observed in O157:H7 strains from the United States, LSPA6-LII genotypes were significantly more likely to be isolated from cattle than from humans.

Certain genes, such as eaeA, were conserved across all lineages and PTs (data not shown), whereas the stx_2c variant toxin was strongly associated with LSPA6-LII strains. In addition, LSPA6-LII strains of PTs 23 and 67 produced lower levels of toxin than strains from other PTs and LSPA6 genotypes within lineage I and lineage II. E. coli O157:H7 strains of LSPA6-LII were also significantly more likely to possess cda and have AMR than LSPA6-LI and LSPA6-LI/II strains. Collectively, our results imply that the two lineages of E. coli O157:H7 indeed have many unique physiological and ecological characteristics that could influence their propensity to cause disease or their ability to propagate in bovine production environments. Recent studies on E. coli genomes have highlighted the importance of genomic island content in differentiating among various E. coli serotypes and pathotypes (9, 20, 39, 41). Specific sets of these large genetic elements presumably explain the phenotypic differences among these E. coli groups in host adaptation, pathogenesis, virulence, epidemiology, and ecology (19, 24). Generation of new genotypes of E. coli O157:H7 is therefore likely to be driven by the acquisition, rearrangement, and loss of these genetic elements. However, it is interesting that despite the apparent instability of these elements there is also evidence of stable gene linkage (e.g., as parts of functional units such as the LEE type III secretion system or genes of related function such as multiple-antibiotic resistance) (17, 31). The apparent stability within E. coli groups of certain of these mobile multigene elements suggests that they have been retained through the selection for one or more phenotypic traits encoded by these genes. In addition, certain of these elements may act to exclude the uptake of other mobile genetic elements through mechanisms such as plasmid incompatibility, changes in phage receptors, and phage-related immunity to superinfection with other bacteriophages (13, 22) and thereby contribute to genotypic and phenotypic stability of the organism. Linkages of LSPA6 typing markers to genes encoding phenotypic characteristics observed in the present study appear to be indirect. It is important to note that the markers targeted by the LSPA6 assay were selected because they are expected to be selectively neutral (54). The differential association of the various phenotypic traits that were tested in the present study (colicin production, AMR, toxin production, and PT) with neutral lineage-specific markers would suggest that certain combinations of these traits arose early in the divergence of the two lineages and that these combinations have remained stable over time. It is possible that geographical or niche isolation may also have played a role in maintaining relatively fixed constellations of genes in certain populations of E. coli O157:H7 and helped create some of the lineage-specific phenotypes observed in the present study.

Further study of E. coli O157:H7 strains from other geographical regions is needed to determine whether the associations between LSPA6 genotype, PT, and other phenotypic characteristics observed in the present study are widely applicable.