Relationship between Phylogenetic Groups, Genotypic Clusters, and Virulence Gene Profiles of Escherichia coli Strains from Diverse Human and Animal Sources

Abstract

Escherichia coli strains in water may originate from various sources, including humans, farm and wild animals, waterfowl, and pets. However, potential human health hazards associated with E. coli strains present in various animal hosts are not well known. In this study, E. coli strains from diverse human and animal sources in Minnesota and western Wisconsin were analyzed for the presence of genes coding for virulence factors by using multiplex PCR and biochemical reactions. Of the 1,531 isolates examined, 31 (2%) were found to be Shiga toxin-producing E. coli (STEC) strains. The majority of these strains, which were initially isolated from the ruminants sheep, goats, and deer, carried the stx_1c and/or stx_2d, ehxA, and saa genes and belonged to E. coli phylogenetic group B1, indicating that they most likely do not cause severe human diseases. All the STEC strains, however, lacked eae. In contrast, 26 (1.7%) of the E. coli isolates examined were found to be potential enteropathogenic E. coli (EPEC) strains and consisted of several intimin subtypes that were distributed among various human and animal hosts. The EPEC strains belonged to all four phylogenetic groups examined, suggesting that EPEC strains were relatively widespread in terms of host animals and genetic background. Atypical EPEC strains, which carried an EPEC adherence factor plasmid, were identified among E. coli strains from humans and deer. DNA fingerprint analyses, done using the horizontal, fluorophore-enhanced repetitive-element, palindromic PCR technique, indicated that the STEC, potential EPEC, and non-STEC ehxA-positive E. coli strains were genotypically distinct and clustered independently. However, some of the potential EPEC isolates were genotypically indistinguishable from nonpathogenic E. coli strains. Our results revealed that potential human health hazards associated with pathogenic E. coli strains varied among the animal hosts that we examined and that some animal species may harbor a greater number of potential pathogenic strains than other animal species.

Escherichia coli strains are commonly present in the gastrointestinal tracts of warm-blood animals, including humans (31). Currently, E. coli is used as an indicator of fecal contamination in freshwater systems (52). There may be several sources of E. coli that contribute to high counts of this bacterium in waterways and on beaches, including humans, farm and wild animals, waterfowl, pets, and several environmental reservoirs (27, 28). Identification of the source of E. coli impacting beaches and waterways is important to establish proper risk assessment and abatement procedures (e.g., chlorination of wastewater or animal manure management). Several phenotypic and genotypic methods have been examined as microbial source tracking tools to determine the source(s) of E. coli and other indicator organisms in waterways (14, 29, 49). However, the distribution and prevalence of pathogenic E. coli among diverse animal sources are not well known.

While most E. coli strains are considered to be harmless commensals of the mammalian gastrointestinal tract, some strains can cause human diseases. E. coli strains that produce Shiga toxins (Stx1 and Stx2), referred to as Shiga toxin-producing E. coli (STEC), can cause bloody diarrhea as well as potentially fatal diseases in humans, including hemolytic-uremic syndrome (HUS) and hemorrhagic colitis (31, 35, 37, 42). Several variants of Shiga toxin genes exist in E. coli. The stx₁ variants (subtypes stx₁, stx_1c, and stx_1d) and stx₂ variants (subtypes stx₂, stx_2c, stx_2d, stx_2e, and stx_2f) are located on chromosomally integrated lambdoid bacteriophages and are transmissible (35, 37). The ability of STEC to cause human diseases depends on the Shiga toxin gene subtype (20, 21). For example, STEC strains with the stx_2c subtype can cause HUS, whereas stx_2d or stx_2e subtype STEC strains are found in patients with uncomplicated diarrhea and in asymptomatic humans (20). Therefore, subtyping of the Shiga toxin genes is a useful tool for identifying virulent strains causing human infections and to predict the potential health risks associated with STEC. Many STEC strains harbor large (60- to ∼121-kb) virulence plasmids encoding additional virulence factors, including enterohemolysin (encoded by ehxA) (5, 6), extracellular serine protease (encoded by espP) (8), and/or catalase-peroxidase (encoded by katP) (7).

Some pathogenic STEC strains also carry a chromosomally localized pathogenicity island, referred to as the locus of enterocyte effacement (LEE), and these strains are often called enterohemorrhagic E. coli (EHEC) (31). The LEE region contains genes coding for a type III secretion system, translocated proteins, and proteins required for attaching and effacing lesions, such as intimin (encoded by eae) (31, 37). The LEE is thought to play an important role in the initial stage of E. coli infection, and most STEC strains isolated from patients with HUS are generally LEE positive (EHEC) (42). However, this region is not essential for strains to cause human diseases, and some STEC strains lacking an LEE have been isolated from human patients, including those with HUS (41, 42). Since genes in the LEE region are homogeneous in genetically, ecologically, and geographically diverse strains, detection of eae can be used to indicate the presence of the entire LEE region (26).

Cattle and other ruminant animals (predominantly sheep and goats) may serve as reservoirs of STEC strains that are potentially pathogenic to humans (2, 35, 42). Serologically diverse STEC strains have been isolated from healthy and diarrheic cattle and sheep (3, 12, 13, 26, 35, 53, 54). Several recent outbreaks of STEC have been reported in association with agricultural fairs and petting zoos in the United States (10) and in Europe (24), suggesting that sheep and goats may be major sources of STEC. Wild deer, which are also ruminants, have also been reported to carry STEC (1, 17) and can also contribute to the human diseases (45). It should be noted, however, that STEC strains are not confined to ruminants but have been isolated from cats, chickens, dogs, and pigs (2, 19, 36).

Diarrheagenic E. coli strains that harbor an LEE but do not produce Shiga toxins are referred to as enteropathogenic E. coli (EPEC) (31, 37, 50). The EPEC strains are a leading cause of infantile diarrhea, especially in developing countries (37, 50). A mutagenesis study has shown that intimin is the central virulence factor in EPEC (37). Some EPEC strains also contain an EPEC adherence factor (EAF) plasmid and are referred to as typical EPEC, whereas EPEC strains lacking the EAF plasmid are called atypical EPEC (50). Although the EAF plasmid is not necessary to cause human diseases, some genetic elements on this plasmid, such as plasmid-encoded regulatory genes (perABC) and genes encoding bundle-forming pili (bfp), enhance virulence (23, 37, 50). Typical EPEC strains have been isolated only from humans, whereas several animal hosts, including birds and dogs, harbor atypical EPEC (33, 37, 50). While there have been several reports about EPEC hosts, the distribution of this pathogenic E. coli type among diverse animal hosts is largely unknown.

Phylogenetic grouping analyses have been used to investigate the evolutionary origins of pathogenic E. coli strains (11, 15, 22, 48). Most E. coli strains fall into one of four phylogenetic groups: A, B1, B2, and D (11). While the majority of the STEC and EHEC strains comprise phylogenetic group B1, other strains can be found in phylogenetic groups A and E (15). Recently, Girardeau et al. (22) reported that some STEC virulence factors were distributed unevenly among the phylogenetic groups. Based on this result, along with comparisons with the seropathotype classification originally proposed by Karmali et al. (32), these authors concluded that STEC strains in phylogenetic group A most likely do not cause human diseases.

The analysis of phylogenetic groups, along with detection of virulence factor genes, may provide a useful tool for predicting potential health risks associated with E. coli strains found in the environment. Consequently, the objectives of this study were (i) to investigate the distribution of potentially pathogenic E. coli strains isolated from diverse human and animal hosts in Minnesota and western Wisconsin by using multiplex PCR as a screening tool, (ii) to characterize the virulence profiles of the identified pathogenic E. coli strains, (iii) to examine the genetic relatedness of these potentially pathogenic strains, and (iv) to assess potential health risks associated with E. coli strains with broad host origins.

MATERIALS AND METHODS

E. coli strains.

A total of 1,531 E. coli strains used in this study were isolated between 1999 and 2002 from feces of healthy humans and three wild and nine domesticated animal hosts from multiple locations throughout Minnesota and western Wisconsin (Table 1) (14, 29). Our strategy was to obtain three to six E. coli isolates from each geographically isolated animal and to avoid isolation from herds. However, isolates from turkeys and chickens were obtained from only two and four counties, respectively, in Minnesota. Isolation and confirmation procedures were described previously (14, 29). Briefly, fecal samples were streaked onto mFC agar (Difco, Detroit, MI) plates and incubated at 44.5°C for selective isolation of E. coli. Well-isolated blue colonies were confirmed to be E. coli by using a series of biochemical tests, including the indole and methyl red tests, the inability of isolates to grow on citrate agar, and the presence of β-d-glucuronidase activity using EC-MUG broth (Difco) (14, 29). Isolates giving atypical responses for any of these tests were examined further using API 20E test strips (bioMerieux, St. Louis, MO). Strains identified as E. coli with “good,” “very good,” or “excellent” identification scores by API 20E tests were used in this study. This scheme allowed isolation of sorbitol-negative or β-glucuronidase-negative E. coli strains. Sorbitol utilization and β-glucuronidase activity are often used to isolate and identify E. coli serotype O157:H7 strains (37, 42). The isolated strains were previously shown to have unique DNA fingerprints, as determined by using horizontal, fluorophore-enhanced, repetitive-element, palindromic PCR (HFERP), and are therefore not considered clones (29).

TABLE 1.

E. coli isolates used in this study and their animal sources

Sourcea	No. of E. coli strains	No. of host individuals sampled
Humans	210	197
Cats	48	37
Dogs	106	71
Horses	78	44
Deer	96	64
Sheep	61	37
Goats	42	36
Cows	189	115
Pigs	215	111
Chickens	144	86
Turkeys	126	69
Ducks	81	42
Geese	135	73

^aE. coli strains from humans were isolated from healthy students at the University of Minnesota and from healthy patients at VA Medical Center, Minneapolis, MN. E. coli strains from healthy cats and dogs were isolated at the University of Minnesota Veterinary Medical Center. E. coli strains from horses, sheep, goats, cows, pigs, chickens, and turkeys were isolated from fresh feces of healthy animals brought to the Minnesota State Fair and from multiple farms in Minnesota and Wisconsin. Samples from whitetail deer (Odocoileus virginianus), Canada geese (Branta canadensis), and Mallard ducks (Anas platyrhynchos) were collected at Minnesota and Wisconsin wildlife management areas.

Biochemical tests and serological analyses.

All E. coli strains were incubated in LB broth at 37°C in 96-well microtiter plates. Fresh overnight cultures of isolates were stamp inoculated onto sorbitol MacConkey agar (Difco) and CHROMagar O157 (CHROMagar Microbiology, Paris, France). Positive (E. coli O157:H7 strain ATCC 43895) and negative (E. coli strain ATCC 25922) controls were used in all tests. Serological analysis of STEC strains was done by the E. coli Reference Center (formerly the Gastroenteric Disease Center) at Pennsylvania State University.

Screening for potentially pathogenic E. coli.

The collection of 1,531 E. coli strains was screened for the presence of potential pathogens by using a multiplex PCR procedure that amplifies four E. coli virulence factor genes: stx₁ and stx₂ variants, eae, and ehxA (38). DNA was extracted from cell cultures as previously described (28) and diluted 10-fold in distilled H₂O, and 1 μl of each solution was used as the template for PCR. Electrophoresis was performed using 1.5% SeaKem LE agarose gels (FMC, Rockland, ME) at 90 V for 3 h, and the gels were stained with 0.5 μg/ml ethidium bromide and visualized under UV light. Positive (E. coli O157:H7 strain ATCC 43895) and negative (no DNA) controls were used each time that PCR was performed. Based on the multiplex PCR results obtained, potentially pathogenic E. coli strains were classified into three groups: STEC (E. coli strains positive for stx₁ and/or stx₂ variants), potential EPEC (strains carrying eae but not stx₁ and stx₂ variants), and non-STEC ehxA-positive E. coli (strains that did not harbor stx₁ and stx₂ variants and eae but possessed ehxA). The latter group is currently not considered to be pathogenic to humans, but we included it in this study to investigate the distribution of ehxA among animal-associated E. coli strains.

Subtype analysis of potentially pathogenic E. coli.

The E. coli strains that were positive for one or more of the virulence factor genes as determined by multiplex PCR were further characterized by molecular subtyping analysis. Subtyping of stx₁ variants was performed by restriction fragment length polymorphism (RFLP) analysis of PCR products amplified with primers VT1-A and VT1-B (22). Subtyping of the stx₂ variants was performed by PCR-RFLP analysis using primers VT2c and VT2d for the stx₂, stx_2-vha, stx_2-vhb, and stx_2-NV206 subtypes (43, 51) and primers VT2 cm and VT2f for the stx_2d subtypes (43). Primers VT2ea and VT2eb were used for stx_2e-specific PCR (30, 55). Subtyping of the eae gene was done by using PCR-RFLP analysis with the EaeVF, EaeVR, EaeZetaVR, and EaeIotaVR primers to distinguish among 14 eae subtypes, including the recently proposed μ, ν, and ξ subtypes (47). PCRs were performed as previously described (22, 30, 43, 47, 51, 55). The reaction mixtures (15 μl) for all RFLP analyses consisted of 1.5 μl of 10× reaction buffer provided by the manufacturer (New England Biolabs, Ipswich, MA), 10 U of the required restriction endonuclease, 0.1 μg/μl bovine serum albumin, and 10 μl of the PCR products. Reaction mixtures were incubated at 37°C for 4 h, and DNA fragments were separated by electrophoresis on 2% agarose gels.

The presence of additional virulence factor genes in E. coli strains was also determined by using PCR. The espP and katP genes were amplified by PCR using the primer pairs esp-A/esp-B (6) and wkat-B/wkat-F (7), respectively. The EAF plasmid and bfpA were detected by using the primer pairs EAF1/EAF25 (18) and EP1/EP2 (23), respectively. Primers SAADF and SAADR (39) were used to amplify saa. Agarose gel electrophoresis was performed as described above, and PCR and/or RFLP products of the expected sizes were recorded as positive for the presence of the target genes and/or subtypes. E. coli O157:H7 strain ATCC 43895 served as a positive control for detection of the stx₁, stx₂, and eae γ subtypes and the espP and katP genes. Negative (no DNA) controls were used in all studies.

Classification of E. coli strains into the four major phylogenetic groups (A, B1, B2, and D) was done by using the multiplex PCR protocol described by Clermont et al. (11). This method is not designed to classify E. coli strains into the minor phylogenetic groups C and E, although most EHEC strains, including the E. coli O157:H7 serotype, belong to phylogenetic group E (15, 48). With this approach, most of the phylogenetic group E strains were classified into phylogenetic group D (11).

Statistical analysis.

A chi-square goodness-of-fit test was used to test if STEC, potential EPEC, and non-STEC ehxA-positive E. coli strains were evenly distributed among different animal hosts (null hypothesis). The genetic relatedness of the STEC, potential EPEC, and non-STEC ehxA-positive E. coli strains was determined based on HFERP DNA fingerprint analysis (29) using BioNumerics software (version 2.1) (Applied Math, Kortrijk, Belgium). Dendrograms were constructed using the curve-based Pearson product-moment correlation coefficient, and clustering was done using the unweighted-pair group method with arithmetic means (29). Multivariate analysis of variance (MANOVA), a form of discriminant analysis, was performed to cluster E. coli strains (14).

RESULTS AND DISCUSSION

Distribution of potentially pathogenic E. coli strains.

Figure 1 shows the distribution of STEC, potential EPEC, and non-STEC ehxA-positive E. coli strains among diverse human and animal hosts in Minnesota and Wisconsin. The χ² goodness-of-fit test (P < 0.005) indicated that the distribution of STEC was not even among the isolates obtained from different animal hosts. Relatively large proportions of the E. coli strains isolated from sheep and goats (23 and 26%, respectively) carried stx₁ and/or stx₂ variants and therefore were considered to be STEC. While wild deer and pigs also harbored STEC, animals other than ruminants and pigs did not contain STEC strains (Fig. 1). Sheep, goats, and deer were previously reported to harbor STEC strains (1, 2, 3, 12, 13, 17, 35, 36, 53, 54), supporting the contention that ruminants may be major reservoirs for STEC. In addition, pigs have also been shown to harbor STEC (2, 19). Contrary to other reports, however, we did not detect any STEC among the cattle that we evaluated, and cattle are believed to harbor many STEC, including E. coli serotype O157:H7 (35, 42). This negative finding may be due in part to our use of mFC agar, which contains bile salts and rosolic acid as selective agents, and incubation of plates at 44.5°C. These growth conditions likely influenced our ability to isolate E. coli serotype O157:H7 strains, which have been shown to be sensitive to high temperature (44.5 to 45.5°C) in the presence of bile salts (16).

FIG. 1.

Frequency of occurrence of STEC, EPEC, and non-STEC ehxA-positive E. coli strains in diverse human and animal hosts. The number of E. coli strains from each source is indicated above the animal host. Filled bars, STEC; striped bars, potential EPEC; open bars, non-STEC, ehxA-positive E. coli strains. ND, not detected.

The ehxA gene was also distributed unevenly among the E. coli strains isolated from the animal hosts examined in our study (Fig. 1). Goats and deer harbored non-STEC, ehxA-positive E. coli strains more frequently than the other animals tested, whereas the sheep that we examined were found not to carry these strains. In contrast, eae-positive and stx-negative E. coli strains (potential EPEC) were found to be present in many animal hosts (humans, cats, dogs, deer, cows, horses, ducks, and geese) (Fig. 1). The χ² goodness-of-fit test did not reject the null hypothesis at P = 0.05, suggesting that potential EPEC may be evenly distributed among the human and animal hosts tested in this study. The broad distribution of potential EPEC in a large number of animal hosts may in part explain the frequent detection of this pathogen in the environment. For example, we previously reported that one potential EPEC strain was found among 3,633 E. coli isolates obtained from a beach site contaminated mainly by wastewater effluent and goose droppings (27). Similarly, Lauber et al. (34) reported the occurrence of potential EPEC strains at a Lake Erie beach, and Higgins et al. (25) reported that the intimin receptor gene tir, another EPEC virulence factor, was detected more frequently than stx genes in water samples from urban streams. In contrast to the previous reports (2, 33), we did not detect potentially pathogenic E. coli strains among the strains that we isolated from chickens and turkeys. This may in part be due to (i) the relatively small number of chickens and turkeys examined in our study, (ii) the limited number of flocks analyzed, or (iii) strain selection bias due to the presence of antibiotics in feed (9).

Phenotypic and genotypic characterization of STEC strains.

A summary of the characteristics of the STEC strains identified in this study is shown in Fig. 2A. The majority (83%) of STEC strains were Sor⁺, and 100% were Gud⁺ (data not shown). Our results are in agreement with those of Djordjevic et al. (12), who reported that the majority of the STEC strains that they studied were Sor⁺. Our results also support previous findings that Sor⁻ and Gud⁻, which are characteristic of most O157 STEC strains, were not common among other STEC strains (42).

FIG. 2.

Genetic relatedness of pathogenic E. coli strains identified in this study. The dendrograms were generated from HFERP DNA fingerprints using Pearson's product-moment correlation coefficient and the unweighted-pair group method with arithmetic means clustering. (A) Dendrogram and characterization of STEC strains. All STEC strains were eae negative and Gud⁺. Phylogenetic groups (PG), virulence factor profiles, and reactions on sorbitol MacConkey agar (Sor) are shown. Strains S70^a and S71^a were isolated from the feces of the same host animal, as were strains S49^b and S50^b and strains G118^c and G119^c. (B) Dendrogram and characterization of EPEC strains. All EPEC strains in this study were negative for stx₁, stx₂, and espP and Gud positive. The virulence factor profiles, reactions on sorbitol MacConkey agar, and phylogenetic groups are shown. (C) Compressed dendrogram of non-STEC ehxA-positive E. coli strains. The number of strains and host animal are shown for each cluster. All strains were negative for stx₁, stx₂, and eae, Sor⁺, and Gud⁺.

Considerable genetic variation was observed among the STEC strains (Fig. 2A), with similarity values ranging from 26.1 to 99.1%. We previously defined E. coli strains having HFERP DNA fingerprints that were >92% similar as being genetically identical (27, 28, 29). Based on this criterion, genetically identical STEC strains were isolated from two different animal hosts, sheep and goats (strains G11, S50, S55, S58, and S71). A single animal host can also harbor multiple STEC strains. For example, while strains G118 and G119 were isolated from the feces of the same host animal, as were strains S49 and S50 and strains S70 and S71, the DNA fingerprints of the strains isolated from the same host animal were not identical, and they exhibited different virulence factor profiles.

The STEC strains identified in this study were serologically diverse (Fig. 2A). While serologically diverse STEC strains were previously identified in sheep (3, 12, 13, 35, 42, 53, 54), results from our studies suggest that E. coli strains with multiple serotypes can also be isolated from goats, deer, and pigs. Although serotype O39:H8, O76:H19, and O91:H14 STEC strains were previously isolated from humans and cows (http://www.microbionet.com.au/frames/feature/vtec/brief01.html; http://www.lugo.usc.es/∼ecoli/), Fig. 2A shows that they are also present in goats, sheep, and pigs. Moreover, here we also report identification of four new serotypes of STEC (O5:H19, O81:H36, O85:H1, and O161:H36) isolated from sheep and goats that were not previously identified in any animal host.

In general, STEC strains having the same serotype clustered together based on HFERP DNA fingerprints (Fig. 2A, dendrogram). However, not all strains having the same serotype were genetically identical. For example, some serotype O5:H19 strains had <80% DNA similarity. Urdahl et al. (54) also reported that STEC strains with the same serotype grouped together when pulsed-field gel electrophoresis was used.

STEC virulence factor profiles.

Over 50% of STEC strains identified in this study (16 of 31) carried only the stx₁ variant gene, whereas two strains (6.4%) were found to have only the stx₂ variant (Fig. 2A). The remaining 42% of the STEC strains (13 of 31 strains) carried both the stx₁ and stx₂ variant genes. These results are similar to previous reports concerning the distribution of stx genes in STEC isolated from sheep (3, 13, 53, 54).

Eighty-three percent of the stx₁ variants were subtype stx_1c (formerly known as stx_1OX3) (Fig. 2A). In contrast, only 7 and 10% of the strains carried subtypes stx₁ and stx_1d, respectively. In our study, the STEC stx₁ subtype was found only in pigs, whereas subtypes stx_1c and stx_1d were found in strains isolated from sheep, goats, and deer. Both the stx_1c and stx_1d subtypes have been previously reported to be common among STEC strains originating from healthy sheep (4, 35, 54). STEC strains with the stx_1c subtype were previously isolated from asymptomatic humans and patients with mild diarrhea without hemorrhagic colitis or HUS symptoms, whereas strains with the stx₁ subtype have been shown to cause severe human diseases (20, 21). Consequently, the majority of the STEC strains identified in our study were most likely not human pathogens.

While stx_2d was the major subtype (93%) found in the stx₂ variant strains obtained from sheep and goats, the stx_2e subtype was found in only one E. coli strain from pigs. Previous studies have shown that the stx_2d subtype is frequently detected in STEC strains from healthy sheep and goats (35, 46, 53, 54) and that the stx_2e subtype is the most frequent stx₂ variant found in STEC from pigs (19). STEC strains carrying the stx_2d or stx_2e subtype have been reported to be less virulent than strains carrying the stx₂ or stx_2c subtype. STEC stx_2d subtype strains have also been isolated from asymptomatic human carriers (20). Taken together, our results suggest that the majority of the STEC strains identified in this study most likely do not cause severe human diseases.

PCR analyses indicated that many of the STEC strains found in this study (Fig. 2A) carried the ehxA gene. Interestingly, while the ehxA gene is often reported to be clustered with the espP and katP genes on virulence plasmids in STEC strains (42), the latter genes were detected in only 11% of the ehxA-positive STEC strains used in the present study. While this suggests that the virulence plasmids of STEC were highly variable among the strains which we analyzed, the genomic location of these genes currently remains unknown. Brunder et al. (6) reported that the virulence factors present on a large virulence plasmid were highly variable, and several STEC strains which they studied also lacked a plasmid-borne espP gene.

None of the STEC strains examined in this study carried the eae gene, whereas 87% of the strains were found to contain the adherence factor gene saa. Blanco et al. (3) reported a low prevalence (5%) of eae-positive STEC strains recovered from sheep in Spain. Studies in Australia also showed that most STEC strains isolated from healthy adult sheep did not carry eae (12, 13). However, many STEC strains isolated from cattle have been reported to possess eae (26), suggesting that eae-positive STEC may be more specific to cattle than to sheep. While Paton et al. (40) reported that the saa and ehxA genes were located on the same large virulence plasmid in the STEC strains which they studied, the ehxA-negative STEC strains found in our study contained saa. Our results are similar to those previously reported by Urdahl et al. (53).

Phylogenetic grouping of STEC strains.

The majority (97%) of the STEC strains which we identified belonged to phylogenetic group B1 (Fig. 2A) and were genetically diverse. One strain (ONT:H4 strain S96), however, was classified as a member of phylogenetic group D/E and served as an outlier in the dendrogram. Our results are similar to those reported by Escobar-Páramo et al. (15) and Girardeau et al. (22) showing that majority of the STEC strains belonged to phylogenetic group B1. The uneven distribution of STEC among phylogenetic groups suggests that STEC virulence factor genes were maintained in E. coli strains with specific evolutionary origins and that a specific genetic background may be necessary to maintain and express Shiga toxin genes (15).

Phenotypic and genotypic characteristics of potential EPEC strains.

Characteristics of the potential EPEC strains identified in this study are shown in Fig. 2B. Similar to what was found for the STEC strains, the majority (92%) of the potential EPEC strains were Sor⁺, and 100% were Gud⁺. Figure 2B shows that most of the potential EPEC strains examined were genetically diverse, with similarity values ranging from 14.5 to 99.5%. However, the cow strains K46, K77, and K78 had HFERP DNA fingerprints that were >92% similar, indicating that they were nearly genetically identical.

EPEC virulence factor profiles.

PCR-RFLP subtype analysis (Fig. 2B) of the eae gene revealed that 38 and 35% of the strains examined were intimin subtype θ and β strains, respectively. The remainder of the potential EPEC strains were eae subtype ζ, κ, ξ, and ι strains. While the intimin β, ζ, θ, γ, and ɛ subtypes were previously reported to be frequently detected in eae-positive E. coli strains isolated from ruminants and humans, subtypes κ, ι, and ξ were infrequently detected (47). In our study, intimin subtypes θ and β were distributed among potential EPEC strains originating from several different animals, whereas subtype κ was detected only in E. coli strains isolated from cats and dogs and subtypes ι and ξ were identified in strains from ducks and geese, respectively. These results suggest that pets and birds may be reservoirs for potential EPEC strains carrying intimin subtypes κ, ι, and ξ. Since our study involved a relatively small number of potential EPEC strains, more strains from diverse animals need to be examined before a clear relationship between intimin subtype and animal host can be established.

PCR analyses indicated that potential EPEC strains DR130 and HU120 originating from a deer and a human, respectively, carried the EAF plasmid and the bundle-forming pilus gene, bfpA. These strains are considered to be typical EPEC strains and previously were found only in humans (37, 50). To our knowledge, this is the first report of the presence of typical EPEC in deer, suggesting that this host may be a potential reservoir for pathogenic E. coli strains. In addition, four of the potential EPEC strains (DR142, P172, P206, and P127) contained the virulence gene ehxA and/or katP, which are typically found on large virulence plasmids in STEC (Fig. 2B). This result suggests that large virulence plasmids may be transferable from STEC to EPEC strains or, alternatively, that plasmid-independent ehxA or katP may be present. It is currently not known whether these genes contribute to the virulence of EPEC.

EPEC phylogenetic groups.

Multiplex PCR and DNA fingerprint analyses indicated that the potential EPEC strains examined in this study were distributed in all four phylogenetic groups and were genetically distinct. Our results are similar to previous reports (15) which together indicate that virulence factors in EPEC strains generally do not require a specific genetic background. However, some of the potential EPEC strains could be assigned to specific phylogenetic groups based on their intimin subtypes (Fig. 2B), and they clustered together based on their HFERP DNA fingerprints (Fig. 2B). For example, while most intimin subtype β strains belonged to phylogenetic groups A and B1, all the intimin subtype ξ strains belonged to phylogenetic group B2. Previously, Escobar-Páramo et al. (15) and Reid et al. (48) reported that the intimin subtype α EPEC strains which they examined belonged to phylogenetic group B2. This is consistent with the finding that intimin subtypes ξ and α have similar amino acid sequences in their C-terminal regions (47).

Non-STEC ehxA-positive E. coli strains.

Analysis of the genotypic relatedness of the non-STEC ehxA-positive strains, presented in the compressed dendrogram in Fig. 2C, showed that the ehxA-positive strains clustered by host animal species. This result suggests that these strains most likely became adapted to their specific host environments. Moreover, the non-STEC ehxA-positive E. coli strains were negative for espP and katP, and 97% did not possess saa. This is in contrast to results from our analysis of the STEC, where most strains had the saa gene. Taken together, these data suggest that the large virulence plasmids present in STEC strains may be different from those in the non-STEC ehxA-positive E. coli strains or that the latter strains may have acquired ehxA via lateral gene transfer events.

Genetic relatedness of the potentially pathogenic E. coli strains.

The genetic relatedness of the STEC, potential EPEC, and non-STEC ehxA-positive E. coli strains identified in this study was examined by using HFERP DNA fingerprint analysis. Results of MANOVA discriminant analysis (Fig. 3A) show that the STEC, potential EPEC, and non-STEC ehxA-positive E. coli strains clustered independently. This result indicates that there is a relationship between the presence of virulence factors and the overall genetic background of strains. While results of our analyses show a tighter clustering of strains than previously reported (15), this may be due in part to the different method which we used for phylogenetic analyses. In our studies the first and second discriminants accounted for 100% of the variation, indicating that the strains were tightly clustered by pathogen type.

FIG. 3.

MANOVA plots based on HFERP DNA fingerprinting analysis of (A) STEC (•), potential EPEC (□), and non-STEC ehxA-positive E. coli (○) strains identified in this study (n = 87) and (B) STEC (•), potential EPEC (□), non-STEC ehxA-positive (○), and nonpathogenic E. coli (▵) strains isolated from sheep, goats, and deer (n = 199).

Figure 3B shows the genetic relatedness of the potentially pathogenic E. coli strains (STEC, potential EPEC, and non-STEC ehxA-positive strains) isolated from sheep, goats, and deer to a subset of nonpathogenic strains isolated from the same animals. The first and second discriminants from MANOVA analysis accounted for 89% of the variation, indicating that the strains were tightly grouped by genetic background. While the STEC and non-STEC ehxA-positive strains clustered separately from nonpathogenic E. coli strains, the potential EPEC strains clustered as a subgroup within the nonpathogenic strains (Fig. 3B). This suggests that the EPEC strains and some of the commensal nonpathogenic E. coli strains have a common genetic background. Similar results were obtained when STEC, potential EPEC, non-STEC ehxA-positive E. coli, and nonpathogenic strains from all animals (n = 1,531) were analyzed (results not shown due to the complexity of the MANOVA plot). Phylogenetic analyses done by Pupo and colleagues (44) using geographically distinct E. coli strains also showed that while the EPEC strains did not cluster together, the STEC strains did. However, in this previous study, only a relatively small number of strains were examined, and while EPEC clustered with some of the nonpathogenic E. coli strains, the association was not as strongly correlated as that seen in our study.

Conclusions.

Results of our studies indicate that 57 of the 1,531 E. coli strains (4%) which we examined were potential human pathogens. The majority of STEC strains which we isolated in this study were from sheep and goats, consistent with outbreaks of pathogenic E. coli from petting zoos in the United States and Europe. Based on Shiga toxin gene subtyping and phylogenetic analyses, our data show that the STEC strains which we identified most likely do not cause severe human diseases. In contrast, the potential EPEC strains which we examined were distributed evenly among the diverse animal hosts and were genetically clustered with nonpathogenic isolates. This suggests that EPEC strains have less exacting requirements for host association than the STEC strains have. This suggestion is also supported by the even distribution of the strains in all four E. coli phylogenetic groups. Results from genotypic and discriminant analyses are consistent with the notion that the potentially pathogenic E. coli strains which we identified most likely did not arise from a single evolutionary event (44). Rather, our results suggest that these strains arose from the acquisition of chromosome- and plasmid-borne genes via lateral transfer events and the integration of these elements into genetically distinct genomic backgrounds.