Virulence Profiling and Disease Association of Verocytotoxin-Producing Escherichia coli O157 and Non-O157 Isolates in Belgium

Introduction

V erocytotoxin (VT)–producing Escherichia coli (VTEC), also called Shiga toxin–producing E. coli (STEC), are foodborne pathogens associated with sporadic cases and outbreaks of bloody diarrhea that may be complicated by hemorrhagic colitis and hemolytic uremic syndrome (HUS), mainly in young children and the elderly (Karch et al., 2005). The production of one or more types of bacteriophage-encoded VT is the cardinal virulence trait involved in HUS development. Over 400 VTEC serotypes have been identified (www.usc.es/ecoli/SEROTIPOSHUM.htm), but only a subset has been frequently associated with severe disease and outbreaks. This suggests that additional virulence markers are necessary for disease. Karmali et al. (2003) have classified VTEC serotypes into five seropathotypes (SPT) (A through E) according to their association with serious disease and outbreaks. SPT A comprises O157:H7/H-, which are most frequently associated with outbreaks and HUS; SPT B strains are associated with outbreaks and HUS, but less frequently than SPT A; SPT C strains are associated with sporadic HUS but not outbreaks; SPT D strains are associated with uncomplicated diarrhea; and SPT E strains are not associated with human disease and appear to be limited to animals.

Sequencing of VTEC O157:H7 EDL933 revealed a 4.1-Mb genomic backbone, which is homologous to E. coli K-12, and 1.34 Mb of O157:H7-specific DNA organized in genomic “O islands” (Perna et al., 2001). One class of genomic islands, pathogenicity islands, represents a flexible pool of virulence-associated genes, which can be transferred horizontally in the bacterial population contributing to pathogen evolution (Hacker et al., 2000) and possibly in the emergence of new pathogenic VTEC serotypes. One such pathogenicity island is the locus of enterocyte effacement (LEE), which encodes a type III secretion system involved in the formation of attaching and effacing lesions. While other O islands encoding putative virulence factors were identified in EDL933, O island 122 (OI-122) has received increasing attention. OI-122 is a 23-kb pathogenicity island containing 26 open reading frames (ORF), of which six show significant homology to known virulence genes. ORF Z4321 is homologous to pagC of Salmonella enterica serovar Typhimurium; ORF Z4326 is homologous to sen of Shigella flexneri; ORFs Z4328 and Z4329 are homologous to non-LEE effector (nle) genes nleB and nleE of Citrobacter rodentium; and ORFs Z4332 and Z4333 are homologous to the EHEC Factor for Adherence gene cluster efa1 and efa2 found in VTEC O157:H7. Thirteen ORFs are associated with mobile genetic elements, and the function of the remaining seven genes is unknown. While nearly all O157:H7/H- strains carry a complete OI-122 (COI-122), a progressive decrease in the prevalence of OI-122 genes in non-O157 VTEC belonging to SPT B through E with a concomitant decreasing pathogenicity was observed (Karmali et al., 2003). Wickham et al. (2006) have described a modular arrangement of OI-122 genes based upon their association with each other across HUS-associated non-O157 VTEC strains: module 1 contains Z4318, pagC, and Z4322; module 2 contains Z4323, sen, nleB, and nleE; and module 3 contains the efa gene cluster. The presence of putative transposases in OI-122 has led to the hypothesis that its elements are acquired or lost in a modular manner. It was shown that, while pagC, Z4322, sen, nleB, nleE, and efa1 individually were more prevalent in non-O157 VTEC associated with HUS, the simultaneous presence of all of these genes strengthened the association with serious disease. Thus, after acquiring OI-122 modules from VTEC O157:H7 via horizontal transfer, less pathogenic non-O157 strains could cross a virulence threshold, resulting in sufficient pathogenicity to cause HUS.

Since no specific data on this pathogenicity island are available in Belgium, we have investigated the prevalence and distribution of OI-122 elements and other virulence factors by PCR in human VTEC isolates belonging to SPT A through D and correlated these data with clinical manifestations.

Methods

Results

O island 122 as a marker for virulence in VTEC

We first considered the OI-122 content in association with disease for all studied isolates. The prevalence of a COI-122 was highest in VTEC recovered from HUS patients (22/33; 66.7%) and decreased progressively in patient groups with bloody diarrhea (15/32; 46.9%), diarrhea (43/145; 29.6%), disease other than HUS or diarrhea (5/27; 18.5%), and abdominal pain (2/28; 7.1%; Fig. 1). A COI-122 was significantly more frequent in isolates recovered from patients suffering from HUS or bloody diarrhea (37/65; 56.9%) as compared to isolated from patients with diarrhea, abdominal pain, or non-gastrointestinal disease (50/200; 25.0%) (p<0.0001).

OPEN IN VIEWER

FIG. 1.

Distribution of O island 122 in verocytotoxin-producing Escherichia coli in association with clinical manifestations. HUS, hemolytic uremic syndrome; BD, bloody diarrhea; HC, hemorrhagic colitis; D, diarrhea; AP, abdominal pain; O, disease other than HUS, diarrhea, or abdominal pain.

Since nearly all O157 VTEC contained the complete array of virulence genes that were searched for, we focused on the variable virulence profiles of non-O157 serotypes. Of 19 non-O157 isolates recovered from HUS patients, eight (42.1%) carried a COI-122, nine (47.4%) carried an incomplete OI-122, and two were negative for OI-122. Among those not associated with HUS, a COI-122 was detected in 27 (27/194; 13.9%) isolates, 80 (41.2%) carried an incomplete OI-122, and OI-122 was absent in 87 (44.8%). The difference between the occurrence of COI-122 in both patient groups was highly significant (p=0.001). We investigated whether there were epidemiological differences in the association of OI-122 modules with HUS after non-O157 infection (Table 3). Isolates carrying modules 2 (p=0.0006) or 3 (p=0.0001) were significantly associated with HUS, yet module 1 was not (p=0.11). The combined presence of modules 2 and 3 strengthened the association with HUS (p<0.0001).

OPEN IN VIEWER

Table 3.

Prevalence of O Island 122 Modules in Non-O157 Verocytotoxin-Producing Escherichia coli in Association with Hemolytic Uremic Syndrome (HUS)

Module	Prevalence in HUS-associated isolates (%)	Prevalence in non-HUS-asociated isolates (%)	p
1	8/19 (42.1)	49/194 (25.3)	0.11
2	17/19 (89.5)	94/194 (48.4)	0.0006
3	17/19 (89.5)	84/194 (43.3)	0.0001
2+3	17/19 (89.5)	82/194 (42.3)	<0.0001

Significant associations of individual virulence markers were observed when comparing HUS-associated non-O157 strains to those not associated with HUS (Table 2). While vtx1 was significantly more frequent in VTEC not associated with HUS (70.6%; p=0.0002), 17 out of 19 (89.5%) non-O157 VTEC associated with HUS were vtx2 positive (p=0.0001). eae (p=0.013), ehxA (p=0.016), and espP (p=0.014) were significantly more prevalent in HUS than in non-HUS strains. Among the OI-122 genes, sen, nleB (both p=0.001), nleE (p<0.0001), and the efa gene cluster (p<0.0001) were significantly more frequent in non-O157 VTEC strains associated with HUS as compared to those not associated with HUS. No significant differences were observed for saa, subA, katP, etpD, and pagC.

The combined presence of vtx2-nleE-efa in individual non-O157 isolates showed the strongest association with HUS (p<0.0001), suggesting an additive effect of these virulence genes in the ability of VTEC to cause disease. For the collection of isolates in this study, the use of the vtx2-nleE-efa profile to detect non-O157 isolates causing HUS after infection had a sensitivity of 78.9%, a specificity of 94.8%, and positive and negative predictive values of 60.0% and 97.9%, respectively.

Discussion

We have investigated the distribution of pathogenicity island OI-122 and other virulence genes in O157 and non-O157 VTEC associated with human disease in Belgium. Because of the lack of distinct biochemical properties characteristic of non-O157 VTEC, many clinical laboratories only search for VTEC O157:H7, leaving most non-O157 infections undiagnosed. The absence of a widespread surveillance system of non-O157 VTEC opens a window of opportunity for the emergence of other serotypes with a pathogenic potential that is possibly comparable to that of O157:H7. In addition to careful evaluation of key genes associated with serious disease and outbreaks, there is a need for rapid diagnostic tools that allow clinical laboratories to quickly identify emerging pathogenic VTEC strains.

While the clinical importance of non-O157 VTEC is still underestimated, determining the virulence profile may help in the identification of potentially pathogenic strains. The overall prevalence of COI-122 (87/265; 32.8%) in our Belgian collection was lower as compared to previous data. Karmali et al. (2003) reported COI-122 in 40% of 70 strains of a North American collection. In that study, COI-122 was present in all strains of SPT A and in 60% of SPT B strains. Karama et al. (2009), on the other hand, showed that COI-122 was less frequent (31.8%) in a collection of O103:H2 VTEC strains, but that its prevalence was higher in North American human strains than in European human strains of the same serotype.

The present study confirms the strong correlation of COI-122 with VTEC strains belonging to SPT associated with outbreaks (A and B) and SPT associated with HUS (A, B, and C), as noted previously by Karmali et al. (2003) and Wickham et al. (2006). Moreover, all COI-122–positive strains in this study were also positive for eae, a marker for LEE. The association between OI-122 and LEE has been recently reported by Morabito et al. (2003). This provides more evidence for the assumption that differences in virulence between serotypes might be based on the presence or absence of specific pathogenicity islands rather than on the basis of individual genes. In this study, non-O157 VTEC associated with HUS were significantly more frequently positive for vtx2, eae, ehxA, espP, and the OI-122–associated genes sen, nleB, nleE, and the efa gene cluster. Hedican et al. (2009) argued that testing of samples to identify the presence of vtx2 versus vtx1 may not be reliable in predicting whether the infecting VTEC strain is capable of causing severe illness. Here, the combined presence of the genes vtx2-nleE-efa showed the strongest association with HUS. Identifying these virulence markers could aid in distinguishing non-O157 VTEC strains that pose a serious risk to public health from those strains that are associated with less severe disease. For the collection of isolates in this study, the use of the vtx2-nleE-efa profile to detect non-O157 isolates causing HUS after infection had a sensitivity of 78.9%, a specificity of 94.8%, and positive and negative predictive values of 60.0% and 97.9%, respectively. Obviously, this test is not completely appropriate for clinical practice. Testing for three genes limits its practical use, but this may be overcome by developing a multiplex PCR. In future experiments, subtyping of vtx, especially vtx2, may improve the diagnostic capabilities to detect highly pathogenic non-O157 strains.

The concept of detecting virulence markers to identify highly pathogenic VTEC is called “molecular risk assessment” and was recently introduced by Coombes et al. (2008). While vtx2-positive, LEE-positive, and OI-122–positive VTEC are associated with outbreaks and serious disease, this and previous studies have shown that strains lacking these virulence genes are also capable of causing important human illness (Karmali et al., 2003; Wickham et al., 2006). These observations suggest that additional genes, possibly organized in pathogenicity islands, may also enhance the virulence potential of VTEC strains. The precise contribution of all mobile genetic elements to VTEC virulence is currently unknown in most cases, and the complete virulence repertoire has not been determined for most non-O157 VTEC (Coombes et al., 2011). Recently, Ogura et al. (2009) compared the whole genome of three non-O157 strains to that of O157, but additional genome-wide studies are necessary to untangle the key players in non-O157 VTEC virulence.

The virulence potential of other E. coli pathotypes may be enhanced by the acquisition of vtx genes via mobile bacteriophages. This was recently evidenced by the German enteroaggregative, VT-producing E. coli O104:H4 strain, which had gained a vtx2 gene as well as additional virulence genes and a CTX-M-15 extended β-lactamase through horizontal gene transfer (Bielaszewska et al., 2011; Rasko et al., 2011; Scheutz et al., 2011). To our knowledge, however, there are no data available on the presence of OI-122 in this O104:H4 outbreak strain.

In conclusion, virulence profiling of clinically relevant non-O157 VTEC may contribute to a better understanding of their virulence potential and could be used for the risk assessment of individual isolates. The World Health Organization has called the rapid identification of virulent non-O157 VTEC a public health priority (WHO, 1998). A continued surveillance of known and newly discovered putative virulence genes and pathogenicity islands seems therefore warranted. Virulence profiling by evaluation of genomic island content should be performed as quickly as possible to ensure prompt identification of highly pathogenic VTEC strains. To this aim, the development of multiplexed real-time PCR or microarray techniques for the detection of selected virulence markers could be useful. This study provides more evidence that a subset of OI-122 genes could represent possible gene candidates.