Though many potential virulence markers have been identified in VTEC, the quality of the evidence supporting their role can vary greatly. The following short review focuses on virulence markers for which there is well established evidence, and distinguishes between markers identified through a causal or correlative relationship with human illness as follows:
causal, the relationship between a virulence marker and illness or the likelihood of severe patient outcomes (BD and HUS) is based on a known mechanism.
correlative, an epidemiological relationship between the virulence marker and the likelihood of severe patient outcomes has been established, but a biological mechanism has yet to be identified. The virulence marker may not be involved in the mechanisms of illness. The epidemiological association with severe patient outcomes may represent correlation with one or more unidentified mechanisms.
It is important to identify whether a virulence marker has been identified on the basis of a causal or correlative relationship as this allows appropriate weight to be put on specific pieces of evidence when making decisions, and indicates issues to be resolved by future research.
The genes for verotoxin (stx or vt) are encoded by a lysogenic phage, the stx-phage, which integrates into the bacterial chromosome (Krüger and Lucchesi, 2015). Thus, stx genes may be inherited by daughter cells or acquired by infection by a bacterium with stx-phages, which permits continual creation of new VTEC strains by horizontal gene transfer. E. coli cells can be potentially infected by multiple stx-phages, and VTEC strains which carry genes for two or three different verotoxins are not unusual.
VTEC and Shigella dysenteria are established human pathogens for which verotoxin/Shiga toxin is a virulence factor, but there have been reports of isolates of other genera (Acinetobacter, Aeromonas, Citrobacter, Enterobacter) which have acquired verotoxin genes (Probert et al., 2014; Grotiuz et al., 2006; Alperi and Figueras, 2010; Schmidt et al., 1993; Paton and Paton, 1996). Strains of Escherichia albertii positive for the genes for verotoxin variant 2f have also been reported (Ooka et al., 2012). There is currently no consensus on the public health significance of verotoxin positive bacteria other than E. coli.
On the basis of amino acid sequence and serological reaction, verotoxins can be divided into two main types, VT1 and VT2 (Strockbine et al., 1986). VT2 has a greater epidemiological association with BD and HUS than VT1 (Ostroff et al., 1989; Boerlin et al., 1999). VT2 also displays greater cytotoxicity and is more likely to produce a HUS like pathology in experimental models (Fuller et al., 2011; Donohue-Rolfe et al., 2000; Siegler et al., 2003).
VT1 and VT2 can be further subdivided into subtypes. The toxin subtypes were originally identified on the basis of serological properties but subtypes are now defined by amino acid sequence. Currently, there is an established taxonomy of three VT1 (a,c,d) and seven VT2 (a,b,c,d,e,f,g) subtypes (Scheutz et al., 2012). Since 2012, at least five additional VT subtypes have been proposed: VT1e (Probert et al., 2014), VT2h (Bai et al., 2018), VT2i (Lacher et al., 2016), VT2k (Meng et al., 2014) and VT2l (Lacher et al., 2016) (Table 2). Epidemiological evidence indicates that some VT subtypes have a higher correlation with severe disease than others (Friedrich et al., 2002; Persson et al., 2007; Buvens et al., 2012; Marejková et al., 2013; Mellmann et al., 2008). Epidemiological data indicates that two subtypes, VT2a and VT2d, are associated with a greater likelihood of BD and HUS, and two subtypes, VT1a and VT2c, are associated with an increased likelihood of BD (FAO/WHO, 2019; NACMCF, 2019). The association of stx-subtypes with severe disease also involve correlation with the presence of other virulence factors, particularly those involved in colonisation (FAO/WHO, 2019; NACMCF, 2019). The remaining VT subtypes may have a greater correlation with mild disease and lower association with BD and HUS, but VTEC strains with these VT subtypes cannot be excluded as potential pathogens and as a cause of serious illness. Though rare and typically involving individuals with a vulnerable health status, there have been reports of BD and HUS involving VTEC carrying VT1c (Lienemann et al., 2012), VT2b (Stritt et al., 2013), VT2e (Fasel et al., 2014; Thomas et al., 1994) and VT2f (Friesema et al., 2014; Friesema et al., 2015; Grande et al., 2016). VTEC strains carrying VT1d and VT2g have been isolated from cases of relatively mild enteric illness, but not from patients with BD or HUS (Nüesch-Inderbinen et al., 2018; Scheutz 2014; Prager et al., 2011). The significance of the 5 additional subtypes proposed since 2012 is unclear, VT2h, VT2i, VT2k, and have only been reported as isolates from animals, and though VT1e was reported in a clinical isolate, it was a strain of Enterobacter cloacae (Probert, et al., 2014; Bai et al., 2018; Lacher et al., 2016; Meng et al., 2014).
Caution should be taken in interpreting studies correlating severe disease with specific verotoxin subtypes as there are two important potential sources of bias. Firstly, studies conducted prior to the establishment of the current subtype taxonomy may have categorized subtypes differently. Secondly, molecular and antibody-based screening methods for VTEC may not detect all verotoxin subtypes (Feng et al., 2011; Staples et al., 2017; De Rauw et al., 2016). Additionally, the VTEC isolated from a patient may not necessarily be the causal agent of the symptoms observed. Coinfection of a patient suffering from HUS with two strains of VTEC which possessed different serotypes and virulence gene profiles has been reported (Gilmour et al., 2007a).
Differences in cytotoxicity may explain the apparent relationship between the VT subtype and the likelihood of severe disease. A study with purified toxins found that VT2a and VT2d were 25 times more toxic than VT2b and VT2c to Vero monkey kidney cells and human renal proximal tubule epithelial cells and 40-400 times more toxic to mice than VT2b, VT2c and VT1 (Fuller et al., 2011). However, factors other than cytotoxicity may play an important role. VTEC may carry multiple copies of VT genes (Ashton et al., 2015), strains carrying the same VT-subtype may differ in their levels of toxin expression (Kimmitt et al., 2000) and, as recently reported, in the rate of translocation across the intestinal epithelium (Tran et al., 2018). A Dutch study of clinical VTEC isolates reported that the presence of specific VT subtypes was correlated positively and negatively with other virulence associated genes (Franz et al., 2015). The authors proposed that the association of VT-subtype with illness may be a consequence not of the properties of the toxin, but of the assemblage of virulence genes associated with that VT subtype among certain VTEC linages (Franz et al., 2015). This hypothesis does not contradict previous observations of an association with severe disease and VTEC with VT2a and LEE, and strains LEE negative strains with VT2d (Friedrich et al., 2002).
Finally, VTEC carrying some VT subtypes appear to be more common in some environments than others. VT2e is associated with VTEC isolates from pigs, in which they cause edema disease (Tseng et al., 2014). VT2b and VT2c are associated with deer and other wildlife (Hofer et al., 2012; Mora et al., 2012). A study of VT2f positive VTEC isolated from pigeons or human sources in Holland reported that they clustered phylogenetically by isolation source with limited overlap (van Hoek et al., 2019). The VT2h subtype has only been reported in VTEC isolated from Tibetan marmots (Bai et al., 2018). Thus, the probability of human exposure to different verotoxin subtypes may contribute to the epidemiological relationship with disease.
Locus of Enterocyte Effacement (LEE)
The LEE pathogenicity island is a complex suite of genes that function together to regulate colonisation of the intestinal epithelium by the formation of attaching-effacing (A/E) lesions (Stevens and Frankle, 2014). The LEE is encoded upon a mobile genetic element; its presence is typically determined by detection of the presence of the gene eae, which encodes intimin, a protein that is essential for bacterial attachment to gut mucosal epithelial cells, thus serving as an essential element in the establishment of infection. The LEE is the definitive virulence factor of EPEC, but may also be present in VTEC strains (Croxen et al., 2013). EPEC cause acute self-limiting diarrhoea and are estimated to be much less infectious than VTEC, with infection in volunteer exposure studies requiring ingestion of >1 million cells (Todd et al., 2008).
The capacity of E. coli O157:H7 to create A/E lesions was first reported in 1986 (Tzipori et al., 1986) and shortly after Levine (1987) proposed that the capacity to create A/E lesions was diagnostic of E. coli O157:H7 and related E. coli pathogens. Since then the role of LEE and the mechanism of A/E lesion formation has been identified (Stevens and Frankle, 2014). It is now well established that LEE positive VTEC have a higher association with BD and HUS (Boerlin et al., 1999; Brooks et al., 2005, Ethelberg et al., 2004; Naseer et al., 2017). However, the virulence mechanism of LEE is not simple. Integral to the LEE is an encoded Type Three Secretion System by which the bacterium translocates proteins into host cells (Gaytán et al., 2016). The translocated proteins include a core group of LEE encoded effectors which are essential for A/E lesion formation, such as the translocated intimin receptor (tir), as well as other non-LEE encoded effectors. The presence of specific effectors may vary between E. coli with LEE (Santos and Finlay, 2015). This variability in the sápecific effectors present as well as, variability in protein sequence and the translation rates of those effectors present, may contribute to differences in the pathogenicity potential of individual strains.
Aggregative Adhesion (AA)
AA of the intestinal epithelium is typical of infectious colonisation by EAEC. EAEC are a cause of acute diarrhoeal illness in healthy adults and may cause persistent diarrhoea in infants and immune compromised adults (Hebbelstrup Jensen et al., 2014). The definitive EAEC virulence factors are Aggregative Adherence Fimbriae (AAF). The AAF is encoded on a plasmid, pAA, along with a suite of proteins with roles in adhesion, biofilm formation, the regulation of plasmid and chromosomal genes, and toxins (Hebbelstrup Jensen et al., 2014; Boisen et al., 2014). The presence of the plasmid can be detected by the presence of the gene aggR, which encodes a regulatory protein. Since the plasmid may be lost during cell replication, a chromosomally encoded gene aaiC can also be used to identify EAEC (EFSA, 2015).
The potential role of AA in the pathogenicity of HUS associated VTEC was identified in 1998, following molecular analysis of the VTEC O111:H2 strain associated with a 1992 outbreak in France (Morabito et al., 1998). However, AA carrying VTEC was not generally considered a significant pathogen until the European outbreak of VTEC O104:H4 in 2011. Centred in Germany, with over 4000 cases of illness and 53 deaths this outbreak is among the largest outbreaks of VTEC reported (Beutin and Martin, 2012). This strain was VT2a positive and LEE-negative, but was identified as carrying pAA, it is apparently highly infectious and had a very high rate of HUS (22% of reported cases) (Boisen et al., 2015). Genomic analysis revealed that AA VTEC O104:H4 had originated as an EAEC which had been infected with the stx-phage (Rasko et al., 2011). Consequently, such strains are designated VTEC/EAEC or stx positive EAEC.
Subsequent to the 2011 outbreak, it has been proposed the VTEC/EAEC hybrid strains be recognised as a class of high risk VTEC strains (WHO/FAO, 2018). To date there has been no further large-scale outbreaks of VTEC/EAEC, but there have been reports of their isolation from clinical cases (Dallman et al., 2012; Prager et al., 2014). Retrospective analysis of Italian clinical isolates revealed that a VTEC/EAEC O104:H4 strain, related to the 2011 outbreak strain, had been isolated in a sporadic case of HUS in 2009 (Scavia et al., 2011). Though a relatively rare subgroup of VTEC, VTEC/EAEC cases have probably gone unrecognised historically due to the failure to test VTEC isolates for the presence of EAEC virulence genes. Thus, there may be VTEC/EAEC strains currently circulating that have yet to be identified and new strains of VTEC/EAEC will likely arise in the future, due the mobility of the primary virulence genes, stx and pAA.
As with the LEE, AA colonisation and pathogenicity involves a complex of genes. The genes involved are not encoded solely on the pAA but also chromosomally and individual EAEC strains may possess variants of the genes or some may be absent (Hebbelstrup Jensen et al., 2014). This variability may contribute to differences in the pathogenic potential of individual VTEC/EAEC strains.
Locus of Adhesion and Autoaggregation (LAA)
A third suite of virulence genes involved in colonisation of the gastrointestinal tract has recently been identified in LEE-negative VTEC. The LAA is a pathogenicity island that to date has only been reported in strains of LEE-negative VTEC (Montero et al., 2017; Colello et al., 2018; Montero et al., 2019). The complete LAA pathogenicity island is 86 kb, and composed of four modules (Montero et al., 2017). Individual strains of VTEC may carry one or more modules and strains carrying all four LAA modules have an increased association with BD and HUS (Montero et al., 2017). Module IV contains genes that are widely distributed among pathogenic and non-pathogenic E. coli and are therefore not suitable for diagnostic tests, but primer sets for the detection of the modules I, II and III have been published (Montero et al., 2017)
Other Virulence Factors
In addition to the LEE, AA, and LAA suites of virulence genes which have an identified role in host colonisation in VTEC, there are numerous other putative virulence genes whose role in determining the pathogenicity potential of VTEC is unknown or uncertain. A selection of these putative virulence genes is included in Table 2. These putative virulence factors include adhesins and toxins, and like the established virulence genes, many are encoded within pathogenicity islands and mobile genetic elements. Examples of such pathogenicity islands identified in LEE-negative VTEC are the Locus of Proteolysis Activity (Schmidt et al., 2001), Subtilase-Encoding Pathogenicity Island (Michelacci et al., 2013), Pathogenicity Island I (CL3) (Montero et al., 2019), and the High Pathogenicity Island of Yersinia pestis (Karch et al., 1999). The importance of VTEC with gene assemblages that do not include eae and aaiC/aggR is illustrated by the epidemiological significance in Europe of HUS associated VTEC O91:H21, which is characterised by genes for VT2d, enterohemolysin (ehxA) and cytolethal distending toxin (ctdB) (Bielaszewska et al., 2009).
There is currently no consensus on whether the presence or absence of any of these putative virulence factors and pathogenicity islands are predictive of an increased likelihood of severe disease. Specific examples which have been proposed as markers of increased likelihood of severe disease include: toxB (adhesin) in LEE-positive VTEC (Tozzoli et al., 2005), and in LEE-negative VTEC, saa (STEC agglutinating adhesion), subAB (subtilase toxin), ureC (Urease-associated protein), ehxA and ctdB (Paton et al., 2001; Paton et al., 2004; Khaitan et al., 2007; Bielaszewska et al., 2009; Franz et al., 2015). Though the mechanistic roles in disease processes has not been established for these putative markers, analysis of large numbers of VTEC genomes from differing isolations sources or linked to patient outcome data may provide evidence for their association with human illness (Franz et al., 2015; Montero et al., 2019).
In a 1998 review paper Nataro and Kaper (1998) established a categorisation of E. coli causing enteric illness into enteropathotypes based on the associated pathology and virulence factors. The pathotype categories have evolved and currently accepted pathotypes include EPEC, VTEC, EAEC, Enteroinvasive E. coli (EIEC), Enterotoxigenic E. coli (ETEC), Diffusely Adherent E. coli (DAEC), and Adherent Invasive E. coli (AIEC) (Croxen et al., 2013). Since many E. coli virulence factors are encoded on mobile elements, including the phage encoded verotoxin genes, it should be expected that strains with hybrid sets of virulence factors will exist. LEE-positive VTEC and VTEC/EAEC illustrate the clinical significance of these hybrids, as can Shigella dysenteriae, which can be viewed as essentially a host specialised population of VTEC/EIEC hybrids. Additionally, there have been reports of VTEC strains which carry virulence genes associated with ETEC, including heat stable toxins (Michelacci et al., 2018; Leonard et al., 2016; Nyholm et al., 2015; Bai et al., 2019). Finally, there have been VTEC isolates which do not conform to simple hybrid states as would be produced by the infection of a strain of another pathotype with a stx-phage, including a VT1-positive clinical isolate with features of AIEC, EAEC and EPEC strains (da Silva Santos et al., 2015) and VTEC O80 an emerging pathogen in Europe possessing virulence genes previously associated with extraintestinal pathogenic E. coli (ExPEC)(Cointe et al., 2018). A graphical representation of the relationship between VTEC and other E. coli pathotypes is presented in Figure 1.
VTEC, as with other bacteria, can be subtyped on the basis of phenotypic (i.e., serology, biochemical profile) or genomic methods to establish taxonomic relationships between isolates within a population. An implicit assumption in all subtyping approaches is that the greater the number of phenotypic or genomic traits shared by two isolates, the more recently they shared a common ancestor. Such relationships can be used to identify potential pathogens, though this role has been largely replaced by analysis for specific virulence traits. Subtyping remains a crucial tool for surveillance, outbreak response and source tracking.
The assumption of a relationship based on shared traits may be unreliable if the traits used to determine the relationship are prone to rapid change, such as horizontal gene transfer (HGT) events, mutation, or are phenotypic traits which can be masked by differential gene expression. For example, HGT of O- and H-antigen loci in E. coli is common, which impacts the utility of serotyping for assessing evolutionary relationship among strains (Ingle et al., 2016). When interpreting the results of subtyping, or selecting a subtyping methodology, consideration should be given as to whether the traits selected are appropriate for the classification of isolates for the required purpose. This can be an issue with VTEC as many of the virulence markers, including verotoxin, are encoded on mobile genetic elements.
The second issue relevant to the interpretation of subtyping results is the level of resolution. Subtyping can be too discriminatory or not discriminatory enough, excluding relevant or including irrelevant isolates, depending upon the application. Phenotypic methods for subtyping VTEC, including biochemical tests, serotyping, phage typing and multi-locus enzyme electrophoresis, are well established but resource intensive and provide limited discrimination (Fratamico et al., 2016). There is a diverse variety of genomic methods for VTEC subtyping (i.e., restriction length polymorphism, ribotyping, multi-locus variable tandem repeat analysis (MLVA), pulsed-field gel electrophoresis (PFGE), multi-locus sequence typing (MLST)) of varying discriminatory power, resource requirements and reliability (Fratamico et al., 2016). Following the establishment of PulseNet USA in 1996 and subsequent international networks, PFGE became the established standard for high-resolution subtyping of VTEC, and other bacterial pathogens for outbreak detection (Ribot and Hise, 2016). Currently, PFGE is being supplanted by methods based upon WGS (Nadon et al., 2017).
In principle, methods based on WGS are capable of providing subtyping information at any level of discrimination required. At the highest level of resolution, the entire genome is considered using methods such as whole-genome MLST (wgMLST) based on species-specific pan-genome allele definitions (Nadon et al., 2017). Core-genome MLST (cgMLST) is a high-resolution typing approach based on a subset of conserved genes that can be implemented using standardized allele definitions developed for the species. The wgMLST scheme for E. coli used by PulseNet and implemented in BioNumerics currently includes 14837 accessory gene loci and the 2513 core gene loci used in the Enterobase cgMLST scheme (http://www.applied-maths.com/applications/wgmlst, accessed 2019-05-09)(Zhou et al., 2019). PulseNet Canada identifies possible outbreak clusters on the basis of allelic differences between strains characterized by wgMLST on the BioNumerics platform (Nadon et al., 2017). For some applications, such as long-term or global epidemiological studies, lower resolution approaches based upon variability within a smaller subset of genes may be useful. Ribosomal MLST (rMLST) uses 53 conserved ribosomal protein genes using a scheme that is universally applicable to all bacteria (Jolley et al., 2012) and identification of rMLST alleles can be used to assess sequence quality (Low et al., 2019). Historic E. coli MLST typing schemes include the 7 gene Achtman scheme (Wirth et al., 2006) and the 8 gene Pasteur scheme (Jaureguy et al., 2018) which can be easily derived from the WGS data.
Single Nucleotide Variant (SNV) (also referred to as Single Nucleotide polymorphism (SNP)) analyses can also be conducted to assess variability within a set of strains at a similar level of resolution to wgMLST (Petkau et al., 2017; Nadon et al., 2017). Care must be taken in the interpretation of SNV results as only core genome sequences conserved within the set of strains are included in the analysis, and mobile elements such as phage or plasmids that may be missing in one or more strains within a set of closely-related strains would be excluded from the analysis. Furthermore, inclusion of unrelated isolates will reduce the size of the core genome and thus the discriminatory power of the approach. While WGS-based approaches are now routinely used for identification of clusters of related illnesses, standards for analysis and the interpretation of results, including the number of SNVs or allelic differences necessary to distinguish strains, have yet to be firmly established. Biological considerations, such as differences in rates of mutation in among strains of E. coli, could confound cluster identification based on WGS data (LeClerc et al., 1996; Grad et al., 2012).
Another consideration for the implementation of WGS-based approaches is the need to maintain interoperability with historic data which can be met by computational (in silico) prediction of typing data. MLST profiles can be reliably determined using a number of tools (Zhou et al., 2019; Maiden et al., 2013; Joensen et al., 2014). E. coli serotype can also be predicted based on WGS data using tools such as ECTyper (Le et al., 2018), SeroType Finder (Joensen et al., 2015) or from raw reads with the EcOH database (Ingle et al., 2016). In cases where high-quality closed genomes have been produced using long-read sequencing data, PFGE patterns can be predicted (Babenko and Toleman, 2016); however, for public health purposes WGS assemblies are typically produced using short-read sequencing technologies and PFGE patterns are difficult to predict from the fragmented assemblies produced by these methods. VTEC WGS analysis pipelines implemented in Canadian public health organizations include modules for subtyping at different levels of resolution including wgMLST, SNV analysis, rMLST, MLST, serotype prediction, virulence profiling and AMR prediction.
Due to its historical importance the interpretation of serotype is a topic that needs to be addressed specifically in regards to hazard characterisation of VTEC. Though serotyping is unreliable for hazard characterisation, it remains significant due to the use of serology to define VTEC of legislative or regulatory concern, and the need to maintain compatibility with historical classification schemes.
Serological typing of E. coli is based on the scheme originally developed in the 1940's by Kauffman (1947), and subsequently modified (Ørskov et., 1977), to allow E. coli strains to be distinguished by the binding of antibodies to antigenic structures on the cell surface. An E. coli serotype is defined in the Kauffman scheme by 3 antigen types, O, H, and K. The O-antigen is the polysaccharide of the lipopolysaccharide of the outer membrane. The H-antigen is the protein flagellin, which comprises the filament of the bacterial flagellium. The K-antigens are diverse acidic capsular polysaccharides (Scheutz and Strockbine, 2005). E. coli isolates are often designated only by the O- and H-antigens, and approximately 186 O and 53 H antigens have been identified (Stenutz et al., 2006). Serotyping is relatively time-consuming and expensive due to the need to test multiple serological reactions and cross reactions. Additionally, a significant proportion of strains cannot be serotyped, either because they produce antigens that do not conform to the Kauffman scheme (untypable) or they do not express the lipopolysaccharide O chain (rough) or flagellin (non-motile). For these reasons serological typing is being superseded by molecular subtyping based on PCR (polymerase chain reaction) or sequencing (Gilmour et al., 2007b; Iguchi et al., 2015; Joensen et al., 2015). Molecular subtyping may not be fully analogous to serological typing, especially for O type. The H antigen is a structural protein encoded by a specific gene, but the O-antigen is a polysaccharide which is the output of multiple synthesis genes, and so indirect molecular markers must be used (DebRoy et al., 2016).
Historically, serotyping has been an important tool for the identification of pathogenic E. coli because of the limited phenotypic characteristics which allow pathogenic strains to be distinguished from commensal E. coli. The identification of virulence factors and methods for their rapid detection has made serotyping redundant for pathogen identification. Moreover, horizontal transfer of O- and H- loci highlights potential issues with reliability of the approach for assessing strain similarity (Ingle et al., 2016). However, serotype has continued to be used as a factor in the identification of VTEC of public health significance, based on association with outbreaks or severe illness. This type of analysis has been developed into formal schemata, such as the seropathotype classification proposed by Karmali et al. (2003).
E. coli O157:H7 and nonmotile (NM) is the most common serotype of pathogenic E. coli reported in Canada, constituting 87.6% of 20,926 clinical isolates from 1999 to 2016 (Supplement 1*). The serotype of clinical isolates of non-O157 VTEC is diverse, with 71 O-types among confirmed VTEC isolates reported to the PHAC National Microbiology Laboratory between 1998 and 2012 (Table 3). Among 498 non-O157 VTEC isolates the most common serotypes were O26 (14.1%), O121 (12.4%), O103 (11.0%), rough or untypable (10.8%), O111 (8.8%), and O145 (3.2%). The proportion of reported E. coli isolates of the O157:H7/NM serotypes has been declining consistently since 2009, in 2015 and 2016 they accounted for only around 70% of isolates (Figure 2 The decline in VTEC O157 as a proportion of reported VTEC could be a consequence of changes within the beef processing industry (Pollari et al., 2017), the VTEC strains that Canadians are exposed to from other sources and changes in clinical testing methods.
In the US, as in Canada, VTEC serotypes O157:H7/NM are the most commonly reported, accounting for approximately 50% to 36% of cases annually reported in the period of 2010 to 2015 (CDC FoodNet, 2017). As in Canada, there has been a decline in the proportion of O157:H7/NM reported in recent years. In the US a group of 7 VTEC serogroups (O157, O26, O45, O103, O111, O121, O145) are considered a public health priority and have been identified legislatively as adulterants of raw ground beef and its precursor material (Gould et al., 2013; USDA-FSIS, 2012). It should be noted that the National Enteric Surveillance Program data indicates an overlap with this group, though serogroup O45 does not fall within the top ten most commonly reported serogroups in Canada (Table 3).
As a low discrimination method of subtyping, serotype indicates a potential clonal relationship between isolates. As such the presence of a VTEC serotype associated with outbreaks or severe illness can be viewed as a marker for the potential presence of currently unidentified virulence, infectivity or ecological factors which can result in increased likelihood of illness. Unfortunately, such an interpretation is not reliable, as the O- and H-antigens are not virulence factors, nor are they known to be genetically linked to any factors that promote pathogenicity in VTEC. Additionally, it should be recognised that the presence of a serotype not previously associated with VTEC causing BD and HUS provides no assurance that the strain cannot cause life threatening illness, as demonstrated dramatically by VTEC/EAEC O104:H4 (Beutin and Martin, 2012).
As the preceding review indicates, an extensive set of makers related to the potential pathogenicity of VTEC strains have been proposed. Interpretation of the relative relevance of these markers is complex due to knowledge gaps and the limitations of experimental models. The following summary identifies the primary points of agreement which form the basis for the current scientific consensus (FAO/WHO STEC Expert Group, 2019; NACMCF, 2019).
- VTEC are coli with the potential to produce verotoxin. E. coli strains which do not possess the verotoxin gene, stx, are not VTEC; even if they possess the serotype or accessory virulence factors associated with VTEC capable of causing severe illness, such as BD and HUS.
- There are 10 established subtypes of verotoxin. Subtypes VT2a and VT2d have a greater epidemiological association with BD and HUS. VT1a and VT2c are associated with an increased likelihood of BD.
- An increased likelihood of BD and HUS is indicated by the presence of virulence factors involved in adherence of VTEC cells to the epithelium of the gut. Virulence factors with an established role in gut colonisation include:
- The Locus of Enterocyte Effacement (LEE), gene eae.
- Aggregative Adhesion (AA), genes aggR and/or aaiC
- A third set of virulence factors involved in VTEC colonisation of the gut, The Locus of Adhesion and Autoaggregation (LAA) has recently been identified.
- A potential clonal relationship, as indicated by subtyping, between a VTEC isolate and previously reported VTEC strains isolated from cases of BD or HUS can be used as an indicator of increased likelihood of severe illness. Serotyping provides low discrimination subtyping and is not reliable for hazard characterisation of VTEC in the absence of information on the virulence gene profile.
- Numerous additional gene markers have been proposed as indicators of increased pathogenic potential for VTEC. Due to the limitations of experimental models of VTEC disease process, the relationship between gene markers and the likelihood of severe illness is dependent upon the ability to link genomic data from clinical isolates to metadata on patient outcomes.