Characteristics of Clinical Shiga Toxin-Producing Escherichia coli Isolated from British Columbia

Shiga toxin-producing Escherichia coli (STEC) are significant public health threats. Although STEC O157 are recognized foodborne pathogens, non-O157 STEC are also important causes of human disease. We characterized 10 O157:H7 and 15 non-O157 clinical STEC derived from British Columbia (BC). Eae, hlyA, and stx were more frequently observed in STEC O157, and 80 and 100% of isolates possessed stx 1 and stx 2, respectively. In contrast, stx 1 and stx 2 occurred in 80 and 40% of non-O157 STEC, respectively. Comparative genomic fingerprinting (CGF) revealed three distinct clusters (C). STEC O157 was identified as lineage I (LI; LSPA-6 111111) and clustered as a single group (C1). The cdi gene previously observed only in LII was seen in two LI O157 isolates. CGF C2 strains consisted of diverse non-O157 STEC while C3 included only O103:H25, O118, and O165 serogroup isolates. With the exception of O121 and O165 isolates which were similar in virulence gene complement to STEC O157, C1 O157 STEC produced more Stx2 than non-O157 STEC. Antimicrobial resistance (AMR) screening revealed resistance or reduced sensitivity in all strains, with higher levels occurring in non-O157 STEC. One STEC O157 isolate possessed a mobile bla CMY-2 gene transferrable across genre via conjugation.

In 2003, Karmali et al. [6] grouped STEC into seropathotypes based on serogroup occurrence in human disease, the capacity to cause outbreaks, and the association with hemolytic uremic syndrome (HUS). STEC O157:H7 and O157:NM were identified as the most significant public health risk (seropathotype A), whereas non-O157 STEC are progressively of lower risk from groups B to E. Recent estimates suggest that STEC O157 causes 50 to 70% of human infections, meaning that 30 to 50% are caused by non-O157 STEC [7][8][9][10]. A US study examining high-risk non-O157 STEC in beef revealed that 1006 of 4133 samples were positive for STEC by PCR, though only 10 isolates possessed virulence gene combinations of known pathogens [11]. Outside North America, non-O157 STEC are well-recognized sources of human illness [12][13][14]. In line with increasing concern for non-O157 STEC, the US Department of Agriculture has amended food safety regulatory policy, declaring STEC of O26, O45, O103, O111, O121, and O145 serogroups as beef adulterants [15].
Reports of hypervirulent clades or lineages of STEC O157 linked to human disease have been made. SNP analysis of 96 loci amongst 500 strains identified nine clades, including clade 8 linked with severe disease and designated hypervirulent [16]. Octamer-based genome scanning and length polymorphisms have identified three lineages (L) of O157; LI strains are represented in both cattle and human clinical isolates, and LII are predominantly from cattle [17][18][19][20]. Subsequent research examining Stx2 production showed differences across and within lineages. Sequencing of the 2 flanking region demonstrates that LII has the transcriptional activator gene Q of L1 strains replaced by a pphA homologue. Further, LI and LI/II strains produced more Stx2 than LII, and LI strains of human origin produced more Stx2 than bovine LI [21].
In British Columbia (BC), Canada, the rate of STEC infections have remained above the Canadian average since 2004, ranging between 2.4 and 4.3 cases/100,000 individuals [22]. However, no data exist describing the salient genetic features of STEC causing disease in BC or examined levels of antimicrobial resistance (AMR) in clinical isolates. To this end, we examined clinical STEC originating from BC using molecular and phenotypic methods to examine lineage, subtype, toxin production, presence of other virulence loci and plasmids, and AMR.

Pulsed-Field Gel Electorphoresis.
Pulsed-field gel electrophoresis (PFGE) patterns were used to compare the genetic relatedness of isolates belonging to the same serogroup. Whole-cell isolation of DNA for PFGE analysis was prepared according to standard procedures [23]. Briefly, DNA-containing agarose plugs were subjected to endonuclease restriction using XbaI. Resulting fragments were separated using a CHEF DR III system (BioRad, Hercules, California). In each run, Salmonella Branderup standards were included every four lanes. PFGE pattern analysis was performed using Bionumerics v.6.0 software using standard comparison criteria [24].
Escherichia coli EDL933 was used as a control. Plasmid size markers included the BAC-Tracker Supercoiled DNA Ladder (Epicentre, Markham, ON) and the Supercoiled DNA Ladder (Invitrogen, Burlington, ON).

Comparative Genomic Fingerprinting (CGF).
PCRs of 30 loci spanning the E. coli O157:H7 genome were used to fingerprint isolates as previously described [28]. Control strains included Sakai, ECI-272 and ECI-1717 for E. coli LI, I/II, and II, respectively, and K-12 (MG1655). Seven additional loci were used to increase genomic coverage and resolution (Table 1). For each locus, PCRs were repeated twice, with a positive result in either replicate being scored as presence ("1") and a negative result in both replicates as absence ("0"). PCR reactions were performed as previously described [28]. Amplicons were visualized on a QIAxcel using the QIAxcel DNA Screening Kit (Qiagen). Binary PCR data were analyzed by constructing an Euclidean distance matrix and hierarchically clustering strains using complete linkage. Analyses were performed in R (http://www.r-project.org/) using the heatmap.2 method of the gplots package (http://cran.r-project.org/web/packages/gplots/index.html). The image was colored using the GNU Image Manipulation Program v2.6.11.

Lineage
Typing of STEC O157. STEC O157:H7 EDL933 and Sakai and FRIK 2001 and ECI-1717 were used as LI and II controls, respectively. The LSPA was carried out according to Yang et al. [19] and analyzed as detailed by Sharma et al. [29].

Genotypic Characterization of AMR.
Isolates were screened for the presence of class I, II, and III integrons [31,32]. A multiplex PCR assay was used to detect the presence of CMY-2, CTX-M, OXA-1, SHV, and TEM -lactamases [33]. DNA sequencing was used to confirm CMY-2 identity.
Plasmid mobility in MDR strains was evaluated by conjugation with E. coli K802 (NAL R ), S. Typhimurium MSC001 (NAL R ), and Citrobacter rodentium MRS0026 (AMP R , NAL R ). Due to intrinsic resistance to ampicillin, a nonpolar bla deletion was generated in C. rodentium MRS0026 (lambda-red system), rendering it AMP sensitive. Donors/recipients were grown for 18 h in LB broth (37 ∘ C) containing appropriate antimicrobials. For both, 400 L was centrifuged (2000 rpm, 10 min), washed, and resuspended in 400 L of fresh LB broth. Matings were incubated for 5 h at 37 ∘ C on a 400 L agar slant within a 1.5 mL microfuge tube. The mixture was resuspended in 100 L LB broth, plated on LB agar with antimicrobials, and incubated (24 h, 37 ∘ C). Transconjugants were streaked on LB agar and screened for AMR.

Quantification of Stx.
The production of Stx2 was quantified using polymixin lysis as described in Ziebell et al. [34], with some modification. Bacterial overnight cultures grown at 37∘C with shaking (150 RPM) were diluted 1 : 250 and used to inoculate 5 mL of fresh brain heart infusion (BHI) broth in 50 mL Falcon tubes. Subsequently, 0.5 mg/mL of polymixin was added and incubated at 37 ∘ C for 1 h. Three experimental replicates were used to assess Stx2 toxin production. Total Stx was assessed using polymixin lysis [34], with slight modification. Cells were incubated with 0.5 mg/mL polymixin and incubated for 1 h at 37 ∘ C. Stx2 production differences between clusters were assessed using the t-test function of R.

STEC Serotypes, Clonality, and Virulence
Profiles. In total, 10 serogroups and 12 unique serotypes were represented in the STEC panel. All O157 serogroup isolates displayed the H7 flagellar antigen. Non-O157 serogroups included O26, O121, and O165 NM variants; the remaining isolates were unique serotypes ( Table 2). PFGE of STEC O157 isolates showed two indistinguishable isolates (BC-20 and -21), with all strains being distinguishable despite having ≥88% similarity ( Figure 1). In spite of similar pulsotypes, BC-20 and -21 were distinguishable based on differing plasmid profiles and AMR phenotypes. Plasmid profiling revealed that 96% of isolates possessed plasmids which varied in size and number. Overall, 21 of 25 (84%) strains possessed a plasmid similar in size to pO157, with all STEC O157 possessing it. Two non-O157 STEC (O26:H11 and O111:NM) were shown to have seven plasmids.

AMR.
All isolates were sensitive to AMK, CIP, GEN, IMP, NAL, RIF, and TMP (Tables 2 and 3 (Table 4). These were observed less frequently (0 to 40%) in the 10 O157:H7 isolates. Whilst four of 10 O157:H7 STEC were resistant/RSC to one antibiotic, only three of 15 non-O157 were singularly resistant. Interestingly, BC-20 possessed resistance to AMP, CAZ, and TIO, suggesting the presence of a beta-lactamase affording resistance to extended spectrum cephalosporins (ESC).

Molecular AMR Characterization and Mobility.
No integrons were detected in any isolate. In BC-20, the presence of CMY-2 was confirmed by PCR and DNA sequencing. Transformants were shown to possess a similar 70 kb plasmid and resistance profile and were positive for CMY-2 . Mating experiments showed that resistance was transferable to E. coli, S. Typhimurium, and C. rodentium through conjugation. Transformations and conjugations were performed using other MDR STEC (Table 5). With the exception of E. coli O118:H16 (BC-6), all strains readily transferred resistance.

Discussion
Boerlin et al. [35] reported an association between clinical EHEC serotypes and 2 and eae and, to a lesser extent, hlyA. More recently, it was reported that lineage and isolation origin correlate with Stx2 production [21]. Specifically, human LI isolates produce more toxin than cattle LI and LII strains. Also, high-toxin producing LI strains encode 2 whereas LII strains possess 2c , and LI/II have both. In this study, all E. coli O157:H7 isolates belonged to LI (LSPA-6 111111) and carried the 2 subtype. This is consistent with observations made by Sharma et al. [29] who reported 91.6% of clinical strains in Alberta typed as LSPA-6 111111 and elsewhere [17][18][19]. However, it was recently shown by Franz et al. [36] and Mellor et al. [37] that the majority of clinical STEC O157 in The Netherlands, Argentina, and Australia, respectively, are LI/II strains. As such our study provides further evidence demonstrating that disease-causing STEC O157 in North America differ from STEC causing disease on other continents. When Shiga toxin production was examined, while levels of Stx2 associated with O157 strains were variable, these strains clustered together by CGH and generally produced more Stx2 than non-O157 STEC strains possessing 2 . Interestingly, STEC O157 BC-17 produced higher levels of toxin than the Sakai strain.
Seropathotyping [6] has been useful in assigning risk but is broad in scope. Here O26:H11 (BC-14) was found to differ at 10% of CGF loci from the other O26:H11 strains (BC-5, BC-10). Thus, similar to O157:H7, non-O157 STEC may contain distinct lineages differing in genomic content and their capacity to cause disease. While the resolution provided by seropathotyping and the O157 CGF are helpful in differentiating among strains, complete genome sequence analysis of non-O157 STEC will be required to identify lineage-specific loci among them and the existence of unique "genopathotypes. " Previous research found only LII O157:H7 STEC possessed cdi [42], which is common in uropathogenic E. coli  (3) CHL, STR, TET + + − n/a n/a O157:H7 (5) AMP, CAZ, AMC, FOX, TIO + c + c n/a + + O157:H7 (6) CHL, STR, TET + + − n/a n/a a C. rodentium MCS026 was constructed by deleting bla in C. rodentium DBS100. b Not applicable. c Reduced susceptibility. [43]. In this collection, we observed two LI strains to be positive for cdi. This implies that these isolates have either recently acquired this gene or the original study was not sufficiently broad to capture LI diversity. At this time, antibiotic administration for human STEC infection is contraindicated in the treatment of STEC infections in North America [44], though conflicting reports of clinical outcomes and antibiotic administration have been made. Antibiotic usage for treatment has been linked to diminished clinical outcomes [45,46] or had little influence on patient outcomes [47,48]. In contrast, early administration of fosfomycin was reported to improve clinical outcomes [49]. Recent data from the E. coli O104:H4 outbreak support administration of antibiotics, with fewer seizures, deaths, and surgeries required for antibiotic-treated patients [50]. Further, treated individuals experienced shorter symptom duration and shed the pathogen for significantly less time, thus posing a lower risk of secondary disease transmission [51]. For these reasons, AMR data observed in clinical STEC strains is needed to provide appropriate therapeutic guidance to physicians should the current contraindication be rescinded.
With the recognition of STEC as a significant source of human disease, increased reports of AMR have been made in recent years [14,[52][53][54][55][56][57][58]. In this study, though clinical STEC were sensitive to many drugs, RSC in all examined strains examined or resistance in 11 of 25 strains to at least one antibiotic was observed. Similar sensitivity to AMK, various -lactams, CIP, and TMP has been reported in STEC of diverse origin [53,54,56]. Although the sample size in our study is small, levels of AMR were high considering the clinical origins of the strains. For example, in the US higher AMR levels were reported in cattle (34%) and food (20%) compared to clinical (10%) isolates [59]. Similarly, in Spain STEC O157 resistant to one or more antimicrobials were recovered in 53% of bovine and 57% of beef isolates, but only 23% of clinical isolates were resistant [52]. In Alberta, Canada, whilst 34% of bovine isolates were resistant to one of more antimicrobials, only 10% of clinical isolates were resistant with the most commonly observed resistances being to STR, sulfisoxazole, and TET [29]. Although resistance to sulfisoxazole was not screened in our study, NEO, STR, and TET were the most frequently occurring AMR phenotypes.
Notably, BC-20 possessed a plasmidic CMY-2 . Considering the highly promiscuous nature of plasmid harbouring observed CMY-2 in this study and previously reported plasmids encoding it [60,61], it is not surprising that generic E. coli and STEC O157 possessing a similar plasmid have been recovered from cattle hides, carcasses, processing environments, and ground beef in Canada [62,63]. Our observation of a clinical STEC possessing RSC to critically important therapeutic agents suggests caution in the administration of ceftriaxone and other therapeutic ESCs that may be considered for treatment of STEC infections.

Conclusions
We observed a small collection of clinical STEC in BC to be of variable genomic content. STEC O157 were LI strains producing significant amounts of Stx2. Based on CGF, increased genomic variation was observed in non-O157 STEC strains, with isolates clustering into two distinct groups. CGF and Stx2 assays suggest that serogroups O121 and O165 were more similar to STEC O157 in genetic content than other non-O157 and that these isolates may produce high levels of Stx. However, further work examining clinical O121 and O165 serogroup strains is required to substantiate this assertion. Although this may make these serogroups of greater public health concern than other non-O157 STEC, surveillance data is required to examine the frequency and disease severity that these serogroups cause in human STEC infections. Despite the clinical origins of the non-O157 strains, the genetic variability revealed by our CGF strategy highlights the need for more detailed genetic information, such as that offered by whole genome sequencing. Lastly, we also observed high levels of AMR and RSC in this clinical collection, including a highly mobile CMY-2 -encoding plasmid conferring resistance to clinically relevant treatment options. Considering recent evidence suggesting that antimicrobial therapy may lead to reduced severity of clinical outcomes, further data examining AMR in STEC seem prudent.