Carriage of stx2a Differentiates Clinical and Bovine-Biased Strains of Escherichia coli O157

Background Shiga toxin (Stx) are cardinal virulence factors of enterohemorrhagic E. coli O157:H7 (EHEC O157). The gene content and genomic insertion sites of Stx-associated bacteriophages differentiate clinical genotypes of EHEC O157 (CG, typical of clinical isolates) from bovine-biased genotypes (BBG, rarely identified among clinical isolates). This project was designed to identify bacteriophage-mediated differences that may affect the virulence of CG and BBG. Methods Stx-associated bacteriophage differences were identified by whole genome optical scans and characterized among >400 EHEC O157 clinical and cattle isolates by PCR. Results Optical restriction maps of BBG strains consistently differed from those of CG strains only in the chromosomal insertion sites of Stx2-associated bacteriophages. Multiplex PCRs (stx1, stx2a, and stx2c as well as Stx-associated bacteriophage - chromosomal insertion site junctions) revealed four CG and three BBG that accounted for >90% of isolates. All BBG contained stx2c and Stx2c-associated bacteriophage – sbcB junctions. All CG contained stx2a and Stx2a-associated bacteriophage junctions in wrbA or argW. Conclusions Presence or absence of stx2a (or another product encoded by the Stx2a-associated bacteriophage) is a parsimonious explanation for differential virulence of BBG and CG, as reflected in the distributions of these genotypes in humans and in the cattle reservoir.


Introduction
Enterohemorrhagic E. coli O157:H7 (EHEC O157) is an important cause of food-and water-borne illnesses in developed nations. Diseases associated with EHEC O157 infections include hemorrhagic colitis (HC) and the hemolytic uremic syndrome (HUS) [1]. HUS following EHEC O157 infection is associated with 3% to 5% mortality [2]. Cattle are a major reservoir of EHEC O157 and consumption of contaminated bovine-origin foods is frequently linked to EHEC O157 outbreaks [3,4].
Shiga toxin (Stx) expression is an important virulence factor of EHEC O157. Stx are encoded in late genes of lambdoid bacteriophages; specific bacteriophages, associated with specific Stx variants, are typically inserted at one or two preferred chromosomal locations. Bacteriophages play a major role in generating genetic diversity in the EHEC O157 genome [5,6,7,8,9,10,11,12,13]. Molecular epidemiological studies using Stx-associated bacteriophage insertion (SBI) sites for strain differentiation have shown that EHEC O157 can be classified into several genotypes [5,10,14]. Bovine-biased genotypes (BBG) are significantly over-represented among cattle isolates compared to clinical isolates, whereas clinical genotypes (CG) are typical of clinical isolates and are also frequently isolated from cattle [15,16]. Other genotyping methods also demonstrate a similar bias in distribution of EHEC O157 lineages/clades among cattle and human host [6,8,12]. Both BBG and CG usually carry both EHEC O157 cardinal virulence factors: production of one or more Stx and expression of the Locus of Enterocyte Effacement (LEE) pathogenicity island required for intestinal colonization [1,17]. The pronounced differences in their frequency of isolation from human disease and from cattle suggest that BBG may have reduced virulence. This possibility was supported by the recent demonstration that CG strains cause more severe clinical signs, more severe histopathologic lesions, and higher mortality than BBG strains in two animal models of human disease [15].
We hypothesize that specific bacteriophage-associated genetic factors underlie the differential virulence of CG and BBG EHEC O157. To test this hypothesis, we: 1) identified consistent bacteriophage-mediated genetic differences associated with four BBG and three CG strains using optical mapping; 2) developed a 12-plex PCR to efficiently SBI genotype EHEC O157 isolates and formatted a biologically relevant SBI genotype nomenclature; and 3) applied SBI typing to evaluate the consistency of the genetic differences discovered in the optically mapped strains in approximately 200 additional isolates each from cattle and human clinical sources. The presence of specific stx2 variants and/or their associated bacteriophages were strongly associated with specific SBI genotypes and with their relative frequency of isolation from human illness and cattle reservoir.

Bacterial Strains
For optical mapping, seven EHEC O157 strains representing the four most common SBI genotypes (one isolate CG-1 and two isolates each CG-3, BBG-5 and BBG-6) isolated from cattle and banked at the Field Disease Investigation Unit at Washington State University [5] were selected by random number table (Table 1). An additional larger set of EHEC O157 strains were randomly selected from the strain bank, stratified to include no more than one isolate per cattle farm per calendar year, for comparison with the optically mapped strains. The bovine isolates included 227 EHEC O157 isolates from four countries including Canada, Japan, Scotland and several states within the USA (WA,  OR, ID, CO, AZ, TX, KS, MN, SD, IA,

Optical Mapping
Optical maps of seven EHEC O157 isolates were constructed and provided by OpGen technologies, Inc (Madison, WI, USA) using the procedure described previously [21,22]. We compared the whole genome optical maps of each strain with the in silico BamHI restriction enzyme (RE) map of strain Sakai (BA000007) using Map Viewer software (OpGen). An inventory of genomic differences involving restriction fragments .2 kb (the resolution of the optical mapping technology) was prepared for each strain in comparison to the reference Sakai sequence [22].

Multiplex PCR Method for SBI Genotyping
To confirm the genetic differences between CG and BBG strains detected by optical mapping, a multiplex PCR was developed and standardized ( Table 2). The PCR targets included were the left and right bacteriophage-bacterial backbone junctions of three different bacteriophages potentially inserted at four sites (yehV, wrbA, argW and sbcB) in the EHEC O157 genome, a previously identified variable bacteriophage-yehV right junction [5,10], and stx1, stx2a, and stx2c. Each of the forward primers was 59-labeled with a fluorescent dye (PET, 6-FAM VIC or NED) as indicated ( Table 2). This multiplex PCR is capable of amplifying 16 different genetic regions; the twelve listed above and in addition the four intact chromosomal insertion sites.
The method used to differentiate stx2a and stx2c in the current study did not differentiate stx2c and stx2d-activatable. Therefore, Table 1. Summary of genetic differences among three clinical genotypes (CG) and four bovine-biased genotypes (BBG) of EHEC O157 identified using optical mapping technology. we further tested all stx2c strains (n = 131) to detect possible presence of stx2d-activatable using one or both of the two subtyping methods [23,24]. These analyses confirmed that no strains in this study carried stx2d-activatable. Additionally, two of the stx2c carrying strains, one each from human and cattle sources, carried an insertion sequence within stx2c coding region as demonstrated by partial DNA sequencing (data not shown).

Capillary Electrophoresis
For capillary electrophoresis, 2 ml of PCR product was mixed with 12.5 ml of Hi-Di formamide (Applied Biosystems, Foster City, CA) and 0.5 ml of Liz 1200 size standard (Applied Biosystems). The capillary electrophoresis was performed using ABI-3730 DNA analyzer at WSU Genomic Core. GeneMarker software (Soft-Genetics, LLC, State College, PA, USA 16803) was used to identify the electropherogram peaks corresponding to each PCR product according to their molecular size and associated dye colors.

Statistical Analysis
Association of the Stx-associated bacteriophage and stx2 gene variant content with the bovine-biased and clinical genotypes of EHEC O157 was analyzed using the Fisher's exact test.

Optical Mapping
A total of 112 polymorphisms were identified in the optical scan restriction maps of seven representative SBI genotype strains compared to the in silico BamHI restriction map of the genome sequence of Sakai (Table 1 and Table S1). Of these polymorphisms, 32 unique to one or more CG strains (and therefore uninformative about BBG -CG strain differences) were not further investigated. Of remaining 80 polymorphisms, 30 affected one or more strains belonging to both CG and BBG, while 50 were unique to one or more BBG strains.
Only one polymorphism differentiated all BBG strains from all CG strains, an insertion detected in BBG strains but absent in CG  [9,25,26]. This was confirmed when PCR (Table 2) on the optically mapped strains amplified Stx2c-associated bacteriophage sequences adjacent to sbcB sequences in all four BBG strains (Table 3 and Figure S1). In contrast, optically mapped CG strains lacked this insertion and amplified PCR products consistent with an intact sbcB locus.
Corresponding to the Stx-associated bacteriophage differences detected in optically mapped strains, stx2a was detected by PCR in all CG-1 and CG-3 strains, whereas stx2c (but not stx2a) was detected in all BBG-5 and BBG-6 strains. The absence of stx2a in BBG-6 strains that contain Stx2-associated bacteriophage sequences within argW was associated with altered BamHI restriction fragments consistent with a partial Stx2a-associated bacteriophage at this locus ( Figure S2): E2325 (CG1, with an intact stx2a encoding bacteriophage) has four restriction fragments totaling 59,809 bp, whereas BBG-6 strain E6996 has seven restriction fragments totaling 58,185 bp and BBG-6 strain E2309 has five restriction fragments totaling 50,780 bp.

Distribution of Stx2 Gene and Associated Bacteriophages
Stx2a-associated bacteriophage sequences detected adjacent to either wrbA and/or argW and the detection of stx2a were significantly associated with CG compared to BBG strains (P#0.001): all CG have Stx2a-associated bacteriophage sequences inserted in wrbA and/or argW and carry stx2a. In contrast, the presence of Stx2c-associated bacteriophage sequences inserted in sbcB and the presence of stx2c were significantly more common in BBG compared to CG isolates (P#0.001): all BBG had both these traits, while only 11.8% of CG belonged to genotypes (ASY2 and ASY22c) with Stx2c-associated bacteriophage sequences inserted in sbcB, and only 7.1% of CG isolates (ASY22c) carried stx2c (Table 5). Many BBG (42.9%) had Stx2a-associated bacteriophage sequences adjacent to wrbA or argW but lacked detectable stx2a (Table 5).
Corresponding differences were observed in the frequency of carriage of stx2 variants in human-and bovine-origin isolates. A large majority (85.9%) of the 192 human-origin isolates carried stx2a only and an additional 7.8% carried both stx2a and stx2c, while only 5.2% carried stx2c only and 1.0% carried neither stx2 variant. Of 227 bovine isolates 49.8% carried stx2a only, 44.1% carried stx2c only, 2.6% carried both, and 3.5% carried neither variant.

Discussion
Optical mapping technology accurately identified genetic changes consistent with bacteriophage insertions or deletions in EHEC O157 genotypes. This technology was used previously by Kotewicz et al. to identify differences in EHEC O157 strains representing SNP clusters and spinach outbreak strains [22,30]. While Kotewicz et al. provided valuable information on genetic diversity among EHEC O157 strains of clinical origin; their clinical isolate strain set lacked defined BBG isolates. Our study was specifically designed to identify bacteriophage-related differences between the predominant CG (CG-1 and CG-3, now AY2 and WY12) and BBG (BBG-5 and BBG-6, now SY2c and ASY12c) groups that previously had been associated with differential virulence potential [15].
SBI has proven to be a useful genotyping technique to identify strains of EHEC O157 differing in distribution [5,10,14], gene expression [16] and virulence [15]. Although based on the content and insertional locations of bacteriophages (i.e., mobile genetic elements), SBI genotypes are empirically stable characteristics of EHEC O157 strains, as recently confirmed by Bono et al. [33] in their demonstration of close correlation between SBI genotypes and chromosomal backbone SNP-defined lineages of EHEC O157. Specifically, CG-1 (referred in current paper as AY2/  The SBI Genotypes that are significantly over-represented among cattle isolates as compared to human isolates are referred as bovine-biased genotypes (BBG); the SBI Genotypes that show significantly over-representation/similar-representation among human isolates as compared to cattle isolates are referred as clinical genotypes (CG); the SBI genotypes that could not be evaluated for a bovine association due to small numbers (1-4 isolates) are designated ''unclassified'' genotypes in this manuscript (Fisher's exact test P#0.05).
c Percent of isolates belonging to most common previously defined SBI genotype [5,10,14]. doi:10.1371/journal.pone.0051572.t004 ASY2/ASY22c) and CG-3 (WY12) correspond to SNP lineage II and I, respectively, and BBG-5 (SY2c) and BBG-6 (SY12c/ ASY12c) corresponds to lineages III, IV, V and VII [33]. Therefore, the SBI multiplex PCR described here provides a means for presumptive identification of EHEC O157 lineages as defined by SNP typing. SBI genotype designations were modified to incorporate data from additional targets and to provide a biologically-based nomenclature. The additional targets include independent detection of stx2a and stx2c variants and two additional chromosomal -bacteriophage insertion sites, argW and sbcB. Even before addition of these targets well over 20 different SBI genotypes had been reported, and this cumbersome number as well as the arbitrary nature of the ordinal genotype designations threatened the usability of the system. Therefore, here we propose identification of SBI genotypes by the concatenation of letters representing specific chromosomal insertion sites followed by numbers indicating the stx variants detected. Isolates classified by this revised SBI designation system are more strongly associated with specific hosts, suggesting that the system is consistent with the biological behavior of the bacterium.
The findings of this study clarify the bacteriophage related genetic differences characterizing BBG and CG, and confirm the strong association of stx2 subtypes in these two genogroups.
Using a panel of 419 EHEC O157 geographically and temporally diverse cattle-and human-origin isolates, consistent patterns of Stx-associated bacteriophage insertions and stx2 variant carriage defined by the optically mapped strains was consistently observed: CG strains were characterized by the presence of the Stx2a-associated bacteriophage sequences inserted into argW or wrbA and by carriage of stx2a, while BBG were characterized by Stx2c-associated bacteriophage sequences adjacent to sbcB and carriage of stx2c. Some CG strains additionally have Stx2cassociated bacteriophage sequences adjacent to sbcB, with or without stx2c carriage, and about half of BBG isolates additionally have Stx2a-associated bacteriophage sequences adjacent to argW but lack stx2a. The close association of stx2a with the more clinically relevant CG strains is the most striking finding of our study.
We recently demonstrated that CG strains (carrying stx2a with or without stx2c) induced more severe clinical symptoms, earlier and higher mortality, and more severe histopathological lesions compared to BBG strains (carrying stx2c only) in two animal models [15]. That study also demonstrated that CG strains produced higher amounts of ELISA-reactive Stx than BBG strains. A recent microarray study [16] demonstrated differential expression profiles of BBG and CG strains suggesting that the increased prevalence of EHEC O157 illness caused by certain strains is associated with over-expression of several virulence factors including ehx, genes on the LEE pathogenicity island, and pO157 genes, although no difference in stx expression was identified.
Other comparative genomic studies on EHEC O157 have demonstrated similar variation in the frequency of stx variants among cattle and human isolates. Using octamer-based genome scanning and LSPA-6 genotyping [12,13,34], EHEC O157 isolates were differentiated into three lineages: I, I/II, and II. Recent studies demonstrated concordance between LSPA-6 and SBI genotyping, indicating that the lineage I, I/II, II strains predominantly belonged to SBI genotypes CG-3, CG-1 and BBG-6, respectively [35,36]. Ziebell et al. [37] and Zhang et al. [34] demonstrated that nearly all lineage II strains carried stx2c, nearly all of the lineage I strains carried stx2a and 50.0% of lineage I/II strains carried stx2c. Similarly a recent study on Japanese isolates also suggests stx2a is the most distinctive feature in clinical isolates compared to cattle isolates. The human isolates, predominantly LSPA-6 lineage I or I/II, typically carried stx2a alone or in combination with stx1 or stx2c while LSPA-6 lineage II cattle isolates carried stx2c alone or in combination with stx1 [38]. These investigators also demonstrated that lineage I (stx1-stx2) isolates showed multiple stress resistances when compared to the lineage II (stx1-stx2c) genotypes [39].
Other studies have evaluated association of stx2a and stx2c content with human virulence [23,40,41,42,43,44,45,46,47,48,49]  demonstrated that EHEC O157 carrying stx2a (with or without stx2c) rather than stx2c were more frequent among clinical isolates, in agreement with the results presented here. Recently, Kawano et al. [48] along with others [43,45,46] suggested that EHEC O157 carrying stx2c were associated with asymptomatic individuals and mild disease however it is also isolated from patients with severe disease (bloody diarrhea, HC and HUS) as shown by others [23,40,41,44,47,49]. Furthermore, Fuller et al. recently demonstrated that purified Stx2a is more potent than Stx2c against primary human kidney cell lines and in mouse models [42]. It is possible that carriage and expression of stx2a alone is sufficient to confer increased virulence in animal models and increased expression of human disease, but alternatively these phenotypes may result in whole or in part on other genetic factors that are correlated with stx2a. Knock-out and complementation studies to evaluate the role of stx2a in otherwise isogenic strains may be required to clarify this question.
In conclusion, optical scanning proved effective at identifying large genetic indels associated with insertion or deletion of Stxassociated bacteriophages among major SBI genotypes of EHEC O157. The multiplex PCR described in this study provides a simple and rapid method to efficiently determine SBI genotypes of EHEC O157 isolates, now known to correlate with lineages within the EHEC O157 clade. We conclude that differential virulence of EHEC O157 genotypes may be due to variation in the presence of stx2 variants and/or bacteriophages associated these variants. Unraveling the roles of these genetic elements in EHEC O157 virulence will improve our understanding of pathogenesis and may facilitate development of strategies for the prevention and control of this important food-borne pathogen. Figure S1 Differences in insertion of Stx2c-associated bacteriophage in sbcB. The differences in insertion of Stx2cassociated bacteriophage in sbcB are shown by the hatched fragments. The yellow arrows indicate the phages (Sp) or prophage-like elements (SpLE) in sequenced strain Sakai (names shown below the map). Red marks indicate the insertion sites (sbcB and yegQ) for Stx-associated bacteriophage, and left junction (YL) for Stx1-associated bacteriophage inserted in yehV (names shown above the map). The restriction enzyme map of sequenced strain EC4115 (GenBank accession # CP001164) shows the known insertion of Stx2c-associated bacteriophage in sbcB locus (hatched fragments). The restriction enzyme map for Sakai and EC4115 are in silico maps and the other seven maps are optical maps of test strains. Plus (+) and minus (2) signs represent presence and absence of stx gene in the strains. (TIF) Figure S2 Differences in insertion of Stx2a-associated bacteriophage in argW. The differences in insertion of Stx2aassociated bacteriophage in argW are shown by the hatched fragments. The yellow arrows indicate the phages (Sp) in sequenced strain Sakai (names shown below the map). Red marks indicate the insertion sites (argW/IntS) for Stx-associated bacteriophage (names shown above the map). The restriction enzyme map of the sequenced strain EC4115 (GenBank accession # CP001164) shows the known insertion Stx2a-associated bacteriophage in argW locus (hatched fragments). The restriction enzyme map for Sakai and EC4115 are in silico maps and the other seven maps are optical maps of test strains. Plus (+) and minus (2)