Subtyping Escherichia coli Virulence Genes Isolated from Feces of Beef Cattle and Clinical Cases in Alberta

Bacterial strains

A total of 507 EC strains were screened, 465 isolated from feedlot cattle feces (207 × O157, 200 × O26, 43 × O103, and 15 × O111) during a 2-year Alberta study (Stanford et al., 2016) and 42 from human isolates (28 × O157, 8 × O26, 1 × O103, and 5 × O111) from the Provincial Laboratory of Public Health, Alberta, Canada. EC strains were confirmed for O-serogroup by 7-plex polymerase chain reaction (PCR) (Conrad et al., 2014).

PCR before pyrosequencing

Toxin gene subtyping was performed using a PCR-pyrosequencing application (Goji et al., 2015). Two multiplex PCR assays were used to amplify (1) stx₁ and sxt₂ or (2) eae and rfbE genes. Briefly, 2 μL DNA template—extracted using NucleoSpin Tissue (Macherey-Nagel, Düren, Germany)—was added to 23 μL 1 × PyroMark Master Mix (Qiagen, Toronto, Canada) and primers (5′-end biotinylated). PCR products were visualized using QIAxcel Advanced (Qiagen). Amplicons of virulence genes were further characterized by pyrosequencing for subtype determination following current nomenclature (Scheutz et al., 2012).

Virulence gene subtyping by pyrosequencing

Pyrosequencing was performed using PyroMark Q24 (Qiagen) according to manufacturer's instructions. Sequencing primers for stx₁ , stx₂ , eae, and rfbE subtyping have been described (Goji et al., 2015). Briefly, 7 μL PCR amplicons, stx₁ /stx₂ or 10 μL eae/rfbE, were mixed with 2 μL streptavidin-coated sepharose beads (GE Healthcare, Piscataway), 40 μL binding buffer (Qiagen), and water to a total of 80 μL. After denaturing, PCR amplicon/sepharose beads conjugate was washed, eliminating nonbiotin-labeled DNA. Labeled DNA strands were resuspended and heated at 80°C for 2 min in an annealing buffer containing 7.5 μM (stx₁/eae/rfbE) or 12.5 μM (stx₂ ) sequencing primer. Pyrosequencing was initiated using PyroGoldQ24 reagents. The subtypes stx_2b , stx_2c , and stx_2d , were analyzed by conventional PCR (Scheutz et al., 2012). Target amplicons were size confirmed by QIAxcel. The stx_1a subtype was classified into stx_{1a -I}, stx_{1a -II}, and stx_{1a -III} based on single-nucleotide polymorphisms (SNPs). For eae subtyping, λ-eae and γ1-eae were grouped together (λ/γ1-eae), as were θ-eae and γ2-eae (θ/γ2-eae), due to sequence similarity. Flagellar H7 and nonmotile (NM) variants of O157 were investigated using pyrosequencing, based on an SNP in the rfbE gene common to all O157:H7/NM, but absent in non-H7/NM O157 serotypes (Goji et al., 2015).

Pyrosequencing analysis

Data were analyzed using PyroMarkQ24 software (v2.0; Qiagen) and sequences were exported to Geneious (v8.1; Biomatters) and aligned to reference sequences (Goji et al., 2015).

Statistics

Prevalence of toxin genes among human and cattle isolates was compared using generalized linear mixed models (Proc Glimmix, SAS 9.3; SAS Institute, Inc., Cary), with p-values <0.05 deemed significant.

Toxin gene prevalence

Among 465 bovine EC strains tested in this study, ∼79% were PCR positive for at least one virulence gene (stx₁ , stx₂ , and eae), whereas 55% had at least two genes and 21% tested negative for all genes. Prevalence of stx₁ , stx₂ , and eae among cattle is shown in Table 1.

The PCR-positive strains were subsequently subtyped by pyrosequencing, except five O103 strains. The majority (129/207) of O157 strains were stx₁ positive/stx₂ positive/eae positive. Of two hundred O26 isolates, 85 (42.5%) harbored solely eae and 41 (20.5%) were stx₁ positive/eae positive. Strains belonging to O103 and O111 serogroups represented ∼12% of the total strains studied, and although the majority of O103 strains did not harbor target toxin genes, the most common genotype among these serogroups was stx₁ positive/eae positive (prevalence of 21% and 100% for O103 and O111, respectively).

Among human isolates from 42 patients with STEC infection, 100% harbored eae and at least one stx gene (Table 1). Prevalence of stx₁ , stx₂ , and eae in human isolates was 95.2%, 73.8%, and 100%, respectively, higher (p < 0.05) than cattle isolates. Most O157 (26/28) and O111 (3/5) clinical isolates were stx₁ positive/stx₂ positive/eae positive. Two O111 strains, all O26 and one O103 strain, were stx₁ positive/eae positive. Although fewer human isolates were available than cattle (465 vs. 42), similar genotypes among serogroups were present in both sources.

Shiga toxin subtypes among cattle and human isolates

Following current nomenclature (Scheutz et al., 2012), we identified among cattle isolates four distinct Shiga toxin subtypes (1a, 2a, 2c, and 2e; Table 2). All stx₁ -positive strains (248/465) belonged to stx_1a . Following the stx_1a classification suggested by Goji et al. (2015), stx_{1a -II} was most prevalent (81.9%), whereas stx_{1a -I} and stx_{1a -III} were less frequent (15.3% and 2.8%, respectively). Most O157 and O111 strains harbored stx_{1a -II} (96.8% and 100%, respectively), whereas O26 and O103 strains had similar prevalence of stx_{1a -II} and stx_{1a -I}. The stx₂ gene was present in 183 (39.4%) isolates and identified in all serogroups except O111. Since five O103 strains were not subtyped and six O157 strains harbored two distinct types of stx₂ genes, a total of 178 strains, or 184 genes (178 + 6), were eligible for stx₂ subtyping. Three distinct stx₂ subtypes were identified with stx_2a (165/184), with highest prevalence among O157 isolates (Table 2).

All human strains positive for stx₁ (40/42) harbored stx_1a (Table 1). Of these strains, 70% carried stx_{1a- II} and 30% carried stx_{1a- I}. In addition, stx_{1a -II} was carried exclusively by O157 (23/26) and O111 (5/5), whereas stx_{1a -I} was detected in O26 (8/8) O103 (1/1) and a few O157 strains (3/26). The stx₂ gene was not detected in serogroups O26 and O103; however, it was present in three O111 (60%) and all O157 strains (n = 28).

eae subtype distribution among isolates

Among cattle isolates, 70.5% harbored the eae gene (Table 2) and four subtypes were identified: λ/γ1-eae, ɛ-eae, θ/γ2-eae, and β-eae. With the exception of two strains, all eae-positive isolates belonging to O157 (158/160) carried λ/γ1-eae, whereas β-eae was mostly carried by O26 strains (130/136). Subtypes ɛ-eae and θ/γ2-eae were predominant among O103 (11/17) and O111 (15/15) strains, respectively.

The human isolates were 100% eae positive and belonged to one of the four main eae subtypes (Table 2). Subtype λ/γ1-eae was present exclusively in O157 strains, whereas θ/γ2-eae, β-eae, and ɛ-eae were exclusive to O111, O26, and O103, respectively.

Genotype prevalence in cattle and human isolates

Among all four serogroups from cattle isolates, there were 27 distinct virulence profiles (Table 3). While O111 strains had one genotype (stx_1a +θ/γ2-eae), over 11 virulence profiles were detected among other serogroups. The majority of O157 strains (56.5%) carried stx_1a +stx_2a +λ/γ1-eae. Of O26 strains, 42% carried β-eae alone and 20% carried stx_1a +β-eae. In contrast, O103 isolates were more heterogeneous, and unlike other serogroups, were not characterized by a dominant genotype.

We observed similar genotypes between human (Table 4) and cattle (Table 3) isolates belonging to the same serogroup. Between these two sources, O157 isolates had a higher genotypic correspondence, whereas non-O157 isolates had lesser similarity. Like most cattle isolates (56.5%), human O157 strains (78.5%) carried stx_1a +stx_2a +λ/γ1-eae.

SNP analysis of O157-specific rfbE gene

Out of O157 strains, ∼80% of cattle and 100% of human isolates belonged to the H7/NM group (Table 5). Generally, the most common genotype among H7/NM strains from both sources was stx_1a +stx_2a +λ/γ1-eae (prevalence of 70.7% cattle and 78.6% human).

stx subtypes and virulence profiles

Apart from host factors (e.g., age, immunodeficiency), the ability of EC to cause severe disease is attributed to virulence factors (Gyles, 2007). Strains harboring stx₂ are most prevalent in severely ill patients and enteric infection with stx₂ -producing EC is more likely to cause hemolytic-uremic syndrome (HUS) than those harboring only stx₁ (Siegler et al., 2003; Kawano et al., 2008; Luna-Gierke et al., 2014).

All O157 clinical isolates harbored stx₂ , whereas cattle isolates had a lower prevalence (∼80%). Similarly, Cookson et al. (2002) reported ∼90% stx₂ prevalence in O157:H7 from clinical isolates in Canada during a 7-year interval and Elder et al. (2000) reported ∼98% of O157:H7/NM harboring stx₂ in cattle feces, hide, and carcass. Our lower prevalence of stx₂ in O157 cattle isolates might be due to O157:non-H7/NM isolates (40/207), which commonly lack stx genes (Rump et al., 2015).

Both Shiga toxins have a distinct toxigenic activity, including differences in receptor-binding affinity (Tesh et al., 1993). Even though renal microvascular endothelial cells bind 10 times more strongly to stx₁ than stx₂ , stx₂ is 1000 times more cytotoxic than stx₁ and poses a higher risk for development of HUS (Louise and Obrig, 1995). STEC infection has been largely attributed to specific virulence gene subtypes (Friedrich et al., 2002; Orth et al., 2007). While stx_2a and stx_2c have been associated with more severe cases of human illness and higher risk of developing HUS, stx_2e and stx_1a have been correlated with low pathogenicity or asymptomatic infections (Eklund et al., 2002; Friedrich et al., 2002; Persson et al., 2007; Fuller et al., 2011). In this study, stx_1a and stx_2a were the major subtypes among all strains tested. The stx_2c was carried exclusively by O157 strains (cattle, n = 17; human, n = 5). Two O26 cattle strains carried stx_2e , which was surprising as stx_2e is typically associated with pig edema disease (Sonntag et al., 2005; Zweifel et al., 2006; Baranzoni et al., 2016) and rarely isolated from cattle (Brett et al., 2003).

STEC can harbor multiple stx genes and strains possessing three different stx genes (e.g., stx_1a +stx_2a +stx_2c ) may have higher pathogenicity than those harboring only one or two stx genes, assuming all genes are expressed (Furst et al., 2000; Bertin et al., 2001). However, the extent to which the coexistence of three stx genes can modulate pathogenicity is not well defined and such strains are often associated with mild diarrhea (Furst et al., 2000).

Strains harboring stx₂ , alone or in combination with another stx gene, are more likely to cause severe illness than those harboring stx₁ alone (Eklund et al., 2002; Brooks et al., 2005). The stx₁ +stx₂ +eae profile was present in 139/465 of cattle and 29/42 of human strains. Previous reports indicate this profile as most prevalent among STEC isolated from cattle (Bertin et al., 2001; Badouei et al., 2010; Shen et al., 2015) and human disease cases (Brooks et al., 2005; Haugum et al., 2014; Ashton et al., 2015); however, genotype frequency differs among studies. In some reports, stx_1a +eae was the most prevalent cattle genotype (Badouei et al., 2010; Shen et al., 2015), whereas stx_2a +stx_2c +eae was most prevalent in human disease (Eklund et al., 2002; Friedrich et al., 2002; Ashton et al., 2015). Others analyzed stx genes of clinical isolates and reported differences in the virulence profiles of O157 and non-O157 strains, with stx_2a +stx_2c and stx₁ alone being dominant among O157 and non-O157, respectively (Eklund et al., 2002; Brooks et al., 2005). These apparent genotype discrepancies may be due to STEC serogroup examined (O157 and non-O157) and sample origin (geographic variation, age group, clinical condition). Unfortunately, association of virulence gene subtypes with clinical disease is poorly understood, and additional similarities not identified in this study may exist among human and bovine STEC genotypes. Regardless, some strains isolated from cattle feces had identical virulence profiles to those isolated from clinical cases.

SNP analysis of stx_1a

Distribution analysis of stx_1a among “top-6” non-O157 strains showed that 90% harbored stx_{1a -I}, while ∼94% of O157:H7/NM harbored stx_{1a -II} (Goji et al., 2015). In this study, we reported similar prevalence of stx_{1a -II} among human O157 (88%), but lower prevalence of stx_{1a -I} among human non-O157 strains (64%). Among cattle isolates, ∼97% of O157 strains harbored stx_{1a -II} and non-O157 isolates had a lower prevalence of stx_1a variants than previously reported, with stx_{1a -II} being most common, (55%) followed by stx_{1a -I} (40.5%) and stx_{1a -III} (4.5%). Apparent discrepancies in non-O157 isolates were likely due to larger numbers of samples tested in our study (i.e., 10 and 7 times more O26 and O103 strains, respectively) than by Goji et al. (2015), which likely increased chances of finding stx_1a variants. Nonetheless, these SNPs may indicate future sites of mutation conferring distinct toxin profiles to strains, although those SNPs did not alter the amino acid composition of stx_1a .

eae subtype analysis

Overall, 100% of human and ∼70% of cattle isolates were eae positive. Although eae is a key factor in STEC infection, eae-negative STEC strains have also been isolated from HUS cases (Paton et al., 1999). Consequently, other adherence factors may trigger STEC infection, and strains lacking eae may efficiently colonize the human gut. In addition, strains harboring eae combined with one or more stx genes (100% human; 50.8% cattle) were identified in this study. A similar association between eae and stx genes was reported in isolates from ruminants (Ramachandran et al., 2003; Cernicchiaro et al., 2013) and humans (Ramachandran et al., 2003; Beutin et al., 2004); however, prevalence varied among studies, possibly due to factors including sample sources, isolation/detections assays, and geographic distribution.

Four eae subtypes (β-, ɛ-, γ1-, and θ-eae) are commonly associated with more virulent STEC, and all eae-positive strains tested in this study belonged to these subtypes. Furthermore, the “top-7” STEC serogroups are often associated with certain eae types (e.g., O157 and O145 harbor γ1-eae; O111 harbors θ- or β-eae, O103 and O121 harbor ɛ-eae; and O26 harbor β-eae). In this study, most strains followed this pattern, although we observed “uncommon” relationships between serogroup and eae subtype (e.g., O26 harboring θ/γ2- and λ/γ1-eae instead of the typical β-eae). These variations might be attributed to potential differences in flagellar H type of strains. An association between eae subtypes and distinct H-type has been reported (Ramachandran et al., 2003; Beutin et al., 2004; Shen et al., 2015), for instance, flagella-type H11 and H2 have been associated with β- and ɛ-eae, respectively, regardless of their serogroup (Shen et al., 2015). In this study, human isolates were serotyped (data not shown) and eae subtypes did not vary within serotypes (e.g., all O26 strains had flagellar-type H11 and harbored β-eae; whereas all O157 strains had flagellar-type H7 and harbored λ/γ1-eae). Conversely, cattle isolates presented a variety of eae subtypes within serogroups, which could be due to differences in flagellar H-type (not all tested). For instance, two O26 strains harboring atypical eae were O26:H32 (λ/γ1-eae) and O26:H18 (θ/γ2), whereas O26:NM or O26:H11 harbored typical β-eae (data not shown).

Absence of stx genes in cattle isolates

O157 strains lacking stx genes have been isolated from patients (Schmidt et al., 1999; Mellmann et al., 2005; Bielaszewska et al., 2007; Friedrich et al., 2007) and cattle (Nielsen and Scheutz, 2002; Lee and Choi, 2006; Wetzel and Lejeune, 2007). Production of stx may not be essential for EC pathogenicity and stx-negative strains may cause diarrhea and HUS (Schmidt et al., 1999; Karch and Bielaszewska, 2001). In this study, 12.6% of O157 cattle strains did not harbor stx genes, and of these, only three strains possessed eae. Strains of O157 lacking both stx and eae genes, although uncommon, have been clinically reported (Schmidt et al., 1999; Mellmann et al., 2005). Our non-H7/NM strains represented ∼91% of cattle O157 lacking stx. These findings agree with previous reports suggesting the absence of stx genes in non-O157 and O157:non-H7 to be more common than in O157:H7 strains (Schmidt et al., 1999; Karch and Bielaszewska, 2001; Mellmann et al., 2005).

Among non-O157, all O111 bovine strains of this study harbored stx₁ , but not stx₂ possibly as stx₁ in some O111 strains are expressed in a defective prophage immobilized in the EC genome, preventing stx₁ loss by phage excision (Creuzburg et al., 2005). Conversely, loss of stx₂ by O111 isolates during in vitro subculture has been reported (Watahiki et al., 2014). However, we did not confirm if stx absence was inherent or a result of potential gene loss. Overall, absence of stx genes was highest among O26 isolates (70%) followed by O103 (58%). Similar results have been reported in which the absence of stx genes was more frequent in O26 than in O103 strains (Bielaszewska et al., 2007). Others reported loss of stx genes to be more frequent among EC O26 (∼30%) than other serogroups (Mellmann et al., 2005), suggesting that O26 prophages were less stable than those of other strains. Accordingly, absence of stx genes is not uncommon among serogroups of potentially pathogenic E. coli, although mechanisms by which this occurs are not well understood. Hence, more studies are necessary to determine roles of stx-negative strains in human disease, as well as how often and under which conditions loss or gain of stx genes occurs.

In conclusion, EC strains isolated from Alberta beef cattle had often similar virulence profiles then isolates from infected human; although a number of additional virulence gene combinations were identified. The vast identification of virulent EC mandates constant and efficient prevention and surveillance systems along the beef farm-to-fork production chain to prevent risks for food safety and human health.