Pathogenesis and Diagnosis of Shiga Toxin-Producing Escherichia coli Infections

An important feature of STEC strains that may impact upon their capacity to colonize the human gut, particularly at low infectious doses, is resistance to the acidity of the stomach. It is now known that exposure of certain enteric bacteria, including E. coli, to low pH induces an acid tolerance response (110), and this has been shown to increase the survival of O157:H7 STEC in mildly acidic foods (187). A distinct phenotype referred to as acid resistance has also been described and is mediated by rpoS, which encodes a stationary-phase sigma factor. This factor regulates genes enabling stationary-phase E. coli organisms to survive for extended periods below pH 2.5 (111). Interestingly, Waterman and Small (365) have recently reported heterogeneity in acid resistance phenotype among STEC strains which correlated with mutations in rpoS. Such differences may contribute to apparent differences in infectivity of STEC strains.

Having survived the harsh conditions of the stomach, STEC must establish colonization of the gut by adhering to intestinal epithelial cells. It is generally assumed that the colon and perhaps also the distal small intestine are the principal sites of STEC colonization in humans, although this has not been demonstrated directly. In vitro adherence of STEC has been examined by using several different epithelial cell lines under a range of experimental conditions, and several adherence phenotypes have been described. However, interpretation of the significance of these studies is complicated by the fact that adherence to nonpolarized epithelial cells in tissue culture (even those of human colonic origin) may not be an accurate reflection of molecular interactions that occur between STEC and human colonic epithelium in vivo. Even within STEC strains belonging to serotype O157:H7, there is heterogeneity in adherence, and this may reflect differences in mechanisms. Indeed, Sherman et al. (313) reported marked quantitative differences (up to 250-fold) in the adherence of five O157:H7 strains to both HEp-2 (human laryngeal epithelioma) and Henle 407 (human colonic carcinoma) cell lines. Some strains adhered in a diffuse fashion, with bacteria distributed evenly over the surface of the epithelial cells (diffuse adherence [DA]). Other strains formed tight clusters or microcolonies at a limited number of sites on the epithelial surface (localized adherence [LA]). Moreover, a given strain did not necessarily exhibit the same pattern of adherence on both cell lines (313). A LA phenotype is also exhibited by EPEC and is mediated by type IV fimbriae (bundle-forming pili), which are encoded by a cluster of 14bfp genes carried by the EAF plasmid (319, 324). However, STEC strains do not carry this plasmid and lack bfpgenes (32, 371, 373). McKee and O’Brien (210) also described a distinct pattern of adherence of O157:H7 STEC to HCT-8 (human ileocecal) cells, which they termed “log jam.” Since adherence occurred principally at junctions between the cells, it is possible that this is a consequence of interaction with the basolateral surface. Moreover, the log jam phenotype was also observed in some commensal E. coli strains. The best-characterized STEC adherence phenotype, however, is intimate or attaching and effacing (A/E) adherence. This property is exhibited by a subgroup of STEC strains and is discussed in more detail below.

STEC strains have also been examined for their capacity to invade epithelial cells. STEC strains differ from certain other enteric pathogens (e.g., salmonellae, shigellae, and EPEC) in that they are unable to efficiently invade HEp-2 and Henle 407 cells (239, 313). However, O157:H7 STEC strains were taken up by T24 bladder cells and HCT-8 cells, although individual STEC strains varied in their invasive capacity. The invasion process was dependent upon both bacterial protein synthesis and host cell microfilaments (239). However, McKee and O’Brien (210) reported that the level of uptake of O157:H7 STEC by HCT-8 cells was significantly lower than that of EPEC or Shigella flexneristrains and no greater than that of a commensal E. colistrain. Thus, the clinical relevance of this property is questionable.

The involvement of the 60-MDa STEC plasmid, referred to as pO157, in the adherence of O157:H7 STEC was initially suggested by Karch et al. (150). These investigators found that the presence of the plasmid correlated with expression of fimbriae and adherence to Henle 407 but not HEp-2 cells. However, subsequent studies have produced conflicting results (98, 148, 347), and there is no consistent in vitro evidence for a role for pO157 in STEC adherence. The discordant findings may be attributable to differences in growth or assay conditions, as well as to differences between O157:H7 strains, and possibly also to the fact that the large plasmid itself appears to be heterogeneous, even within serotype O157:H7 (21, 86). The potential contribution of pO157 to pathogenesis has also been examined in animal models. Tzipori et al. (353) found that the presence or absence of the plasmid had no effect on the capacity of STEC strains to colonize the colon or to cause A/E lesions in gnotobiotic piglets. On the other hand, Wadolkowski et al. (361) demonstrated that both O157:H7 strain 933 and its plasmid-cured derivative 933cu could individually colonize the gut of streptomycin-treated mice but that 933cu could not establish colonization when used together with 933. Although the same strains were used in the above piglet experiments, competitive colonization studies were not performed. Therefore, it is not possible to determine whether the apparent contribution of pO157 is influenced by host species. Moreover, the degree to which either of these animal models reflects colonization mechanisms in humans is uncertain. Clearly, more research is needed to determine whether pO157 or related plasmids play a role in adherence of STEC to colonic epithelium.

It has been known for more than a decade that certain strains of STEC are capable of causing A/E lesions on enterocytes (96, 312). A/E lesions involve ultrastructural changes, including loss of enterocyte microvilli and intimate attachment of the bacterium to the cell surface. Beneath the adherent bacteria, there is accumulation of cytoskeletal components, resulting in the formation of pedestals; this is recognizable by electron microscopy and by fluorescence microscopy after staining with phalloidin-fluorescein isothiocyanate (172). The capacity to produce A/E lesions was initially recognized in EPEC strains, and recent studies have elucidated the molecular events involved in their generation, as reviewed by Donnenberg et al. (81). All of the genes necessary for generation of A/E lesions in EPEC are located on a 35.5-kb “pathogenicity island” termed the locus for enterocyte effacement (LEE), which is inserted at 82 min in the E. colichromosome. Binding of EPEC to epithelial cells (initially via the bundle-forming pili) triggers intracellular signals including release of inositol triphosphate, phosphorylation of myosin light chains, and tyrosine phosphorylation of certain proteins in the epithelial cell membrane (81). In contrast to earlier reports, generation of A/E lesions does not require changes in intracellular Ca²⁺levels (17). LEE includes a cluster of genes (sepA to sepI) which encode a type III secretion system. This machinery is responsible for secretion of other LEE-encoded proteins, including EspA, EspB, and EspD, which are necessary for initiation of the signal transduction events referred to above. LEE also includes the eaeA gene, which encodes intimin, a 939-amino-acid outer membrane protein (OMP) which mediates intimate attachment to the enterocyte (81, 181). Interestingly, Kenny et al. (168) have recently reported that the receptor for intimin is also encoded by LEE. This protein was previously referred to as Hp90 but has now been renamed Tir (translocated intimin receptor). Tir is secreted from EPEC as a 78-kDa species, and efficient delivery into the host cell is dependent upon the type III secretion system and other LEE-encoded secreted proteins. Tyrosine phosphorylation of Tir after insertion into the epithelial cell membrane increases its apparent size to 90 kDa, a phenomenon which can be reversed by alkaline phosphatase treatment. However, tyrosine phosphorylation is not essential for intimin binding, at least in vitro (168).

The mechanism whereby STEC strains generate A/E lesions is less well characterized but is essentially analogous to that for EPEC. STEC strains displaying the A/E phenotype have a LEE homolog (208), which, although not yet fully characterized, contains a copy of eaeA, whose 934-amino-acid product has 83% amino acid identity to EPEC intimin (27, 380). The STEC LEE also encodes a Tir homolog (93a), as well as a type III secretion system. Jarvis and Kaper (144) demonstrated that the latter was responsible for the secretion of proteins with masses of 100 to 110, 37, and 24 kDa, which reacted with sera from HUS patients. N-terminal amino acid sequencing identified the 37-kDa protein as an EspB homolog. Ebel et al. (90) demonstrated that secretion of a range of proteins by STEC was influenced by both growth temperature and culture medium, and N-terminal sequence analysis also identified homologs of EspA and EspB. Production of EspB was induced at 37°C and in serum-free tissue culture medium. Interestingly, sequence analysis indicated that EspB homologs from O157:H7 and O26:H⁻ STEC strains had only 80% amino acid homology.

However, there are some differences between EPEC and STEC, as well as gaps in our knowledge. For example, STEC strains do not usually exhibit the LA pattern of adherence to enterocytes to the same degree as EPEC strains. They lack the bfp genes found in EPEC, and the factors which mediate initial interaction with enterocytes are not yet fully characterized, as discussed below. Like EPEC, interaction of STEC with the host cell triggers increases in intracellular levels of inositol triphosphate but does not appear to result in tyrosine phosphorylation of Tir (137).

Studies with eaeA-negative O157:H7 STEC mutants have shown that, like EPEC intimin, STEC intimin is essential for the generation of cytoskeletal rearrangements in HEp-2 cells in vitro (82, 196). Donnenberg et al. (82) demonstrated that such mutants had also lost the capacity to adhere intimately to the colonic epithelium of piglets. These properties were reconstituted by transformation with a plasmid carrying either STEC or EPECeaeA, indicating that the two genes were functionally homologous. Interestingly, however, further studies with the same constructs in a gnotobiotic piglet model demonstrated that the source of eaeA significantly influenced the nature and distribution of the A/E lesions in the piglet intestine (352). TheeaeA mutant STEC strain reconstituted with EPECeaeA colonized and caused A/E lesions in the distal half of the small intestine, as well as on the surface cells of the large intestine, a pattern typical of EPEC infection (351). This strain also caused more severe diarrhea than the wild-type STEC, which colonized the cecum and colon only but caused A/E lesions on both crypt and surface epithelial cells. The STEC eaeA gene was only partially capable of reconstituting these properties in theeaeA mutant STEC strain, a possible consequence of polar effects of the mutation. It seems likely that the marked differences in tissue tropism displayed by these otherwise isogenic strains, as well as the difference in severity of symptoms induced is a consequence of heterogeneity of the primary amino acid sequences of intimin from EPEC and STEC. The two molecules are virtually identical for the first ca. 700 amino acids, but the C-terminal portion (about 25% of the total length) is quite divergent, displaying only about 50% homology; this region is involved in binding to the epithelial cell (96a).

Similar studies have also been conducted by McKee et al. (209), using a derivative of the same O157:H7 STEC with an in-frame eaeA deletion, which eliminates possible complications due to polar effects of the mutation. Complementation of this eaeA mutant STEC with plasmids encoding an intact copy of eaeA demonstrated unequivocally that eaeA is essential for the LA adherence phenotype in HEp-2 cells. This gene was also essential for colonization of the piglet cecum and colon, generation of A/E lesions on enterocytes, and mediation of colitis, as judged by histological testing. However, the fact that the same plasmids could not complement the in vitro HEp-2 adherence phenotype in the eaeA insertion-deletion STEC mutant or confer adherence upon a wild-type eaeA-negative STEC strain indicated that an additional gene(s) downstream from eaeA was essential (209). This region of the STEC LEE is now known to includeesp homologs. In a subsequent study, McKee and O’Brien (211) reconstituted HEp-2 adherence by exogenous addition of either of two purified His₆-EaeA fusion proteins. Both proteins were N-terminally truncated, and the smaller comprised only the carboxyl two-thirds of the protein. Interestingly, the fusion proteins also enhanced the adherence of an eaeA-negative wild-type STEC to HEp-2 cells but without conferring the capacity to generate cytoskeletal rearrangements. No enhancement of adherence ofE. coli K-12 was observed.

There is no doubt that there is a strong association between carriage of eaeA and the capacity of STEC strains to cause severe human disease such as HC and HUS. Several studies have shown that the proportion of eaeA⁺ strains from such sources is much higher than among STEC isolates from animals. Moreover, the presence of eaeA in animal isolates is most commonly associated with known human-virulent strains such as those belonging to serogroups O157, O26, O111, etc. (21, 31, 32, 195, 290, 372). An additional potential complication in the elucidation of the role of intimin in the pathogenesis of human disease is introduced by the significant sequence heterogeneity of the C-terminal portion of the protein. Heterogeneity between STEC and EPEC intimin accounts for marked differences in tissue tropism, as discussed above, but heterogeneity also occurs within STEC strains. For example, there was approximately 25% amino acid sequence divergence over the last 250 residues of intimin from O157:H7 and O111:H8 STEC strains, and additional sequence variation between eaeA genes from O111:H8 and O111:H11 STEC strains was detected by PCR (195). In another study, Wieler et al. (371) found thateaeA probe-positive STEC strains from only 8 of 17 O serogroups tested were PCR positive with primers based on the 3′ portion of O157:H7 eaeA. Such differences have been used as the basis for serotype-specific assays for STEC, as discussed below, but it is not known whether the variations affect the biological activity or receptor specificity of intimin.

Notwithstanding the above, a significant minority of human STEC isolates, including those from patients with HC and HUS, do not containeaeA, indicating that intimin is not essential for human virulence (21, 195). These strains do not produce cytoskeletal rearrangements and A/E lesions in vitro, although at least some are capable of microvillus effacement (89). The possibility remains that these strains produce additional, as yet uncharacterized virulence factors to compensate for the absence ofeaeA. Interestingly, Wieler et al. (371) found that only 65% of eaeA probe-positive bovine STEC isolates were positive by fluorescent actin staining of infected HEp-2 cells and, furthermore, that only 19% were positive for espB by PCR or even by low-stringency hybridization. Thus, the presence ofeaeA does not necessarily imply that a given STEC strain is capable of production of functional intimin and generation of A/E lesions.

Factors implicated in the adherence of other enteric pathogens include fimbriae, OMPs, and lipopolysaccharide (LPS). Studies by Sherman and Soni (311) showed that antibodies to whole cells or outer membranes, but not to H7 flagella, significantly inhibited the adherence of O157:H7 STEC to HEp-2 cells. Moreover, exogenous addition of OMP extracts inhibited adhesion in a concentration-dependent manner but addition of isolated flagella and LPS did not. Subsequent studies demonstrated that polyclonal antiserum raised against a purified 94-kDa OMP also blocked adhesion (310). This protein was subsequently shown to be distinct from intimin (88). Recently, an 8-kDa O157:H7 OMP has also been implicated in adherence. A TnphoA insertion mutant deficient in production of this protein had a significantly reduced adherence to Henle 407 cells in vitro and was less able to colonize chicken ceca than was the wild-type O157:H7 STEC. Furthermore, preincubation with a monoclonal antibody specific for the 8-kDa OMP blocked subsequent in vitro adherence of the bacteria (382). In another recent study, Tarr et al. (335) described the isolation of a chromosomal E. coli O157:H7 gene, designatediha, which appears to encode the capacity to adhere to HeLa cells. The gene was found in all 20 O157:H7 STEC strains tested and in 4 of 5 eaeA⁺ non-O157:H7 human STEC isolates but was not found in 10 eaeA-negative meat isolates. Interestingly, however, iha was not part of the LEE, and the 696-amino-acid product of this gene is a surface protein with approximately 40% homology to IrgA, an iron-regulated protein ofVibrio cholerae. It is not yet known whether ihaencodes the capacity of STEC to adhere to epithelial cells of intestinal origin. Maneval et al. (204) have also recently reported the isolation of 21-kDa fimbrial subunits from O157:H7 and O26:H11 STEC strains. N-terminal sequence analysis indicated a degree of homology to Bordetella pertussis and E. coliF17 fimbriae. However, there was evidence of antigenic variation between fimbriae from the two STEC strains, and their role in adherence remains to be determined.

Two further studies have directly examined the role of LPS O-antigen side chains in adherence of O157:H7 STEC strains. In both studies, TnphoA mutagenesis was used to construct STEC strains deficient in O-antigen biosynthesis, and these were found to be hyperadherent to HEp-2 cells in vitro (41, 74). The enhancement of adherence might be due to increased exposure of one or another of the above-mentioned OMPs on the bacterial surface, although it is possibly an artifact of the gross disturbance of cell surface hydrophilicity due to loss of O antigen.

The mechanism of adherence of the class of STEC responsible for piglet edema disease to intestinal cells has been studied extensively (133). These strains do not generate A/E lesions, and adhesion to isolated porcine intestinal villi is mediated by a specific fimbrial adhesin referred to as F107. The gene cluster encoding F107 biosynthesis has been cloned, and the structural gene encoding the 15-kDa fimbrial subunit (designated fedA) has been sequenced (134). fedA has been found in the majority of edema disease STEC isolates but is also present in a small number of ETEC strains associated with postweaning diarrhea in piglets (132).

There are comparatively few studies in the literature, however, which have examined the adherence of non-O157 STEC strains from humans. Willshaw et al. (373) found that 13 of 48 non-O157/O26 human isolates exhibited a LA phenotype on HEp-2 cells; all of these wereeaeA⁺ but a further 5eaeA⁺ STEC strains were LA negative. Nishikawa et al. (229) examined the effect of growth conditions on adherence of O157 and O111 STEC to HEp-2, Henle 407, and CaCo-2 (human colonic carcinoma) cells and concluded that although adherence was mannose resistant, prior growth in metabolizable sugars resulted in catabolite repression of adherence. Dytoc et al. (89) have studied the adhesion phenotype of an eaeA-negative STEC strain belonging to serotype O113:H21. This piliated strain adhered to rabbit ileal brush border membranes and to both Hep-2 and Henle 407 cells in a diffuse pattern; adherence was resistant tod-mannose. Although this strain was capable of microvillus effacement in vivo, it did not cause the cytoskeletal rearrangements and intimate A/E lesions typical of eaeA⁺ STEC. In a recent study (261), the adherence of a range of STEC isolates from patients linked to an outbreak of HUS and diarrhea (caused by contaminated fermented sausage) was compared with that of apparently nonvirulent STEC strains also isolated from the implicated food source in a quantitative Henle 407 model. The adherence of STEC strains from HUS patients was significantly greater than that of STEC strains found in the contaminated food source but not in any patients. Other STEC strains from sporadic HUS cases, which included aneaeA-negative O48:H21 strain, also displayed enhanced adherence. These studies support the hypothesis that an enhanced capacity to adhere to intestinal cells is one of the factors which distinguishes human-virulent STEC strains from those of lesser clinical significance.

Studies with rabbits have shown that STX has direct enterotoxic properties which result from selective targeting of Gb₃-containing absorptive villus epithelial cells in the ileum. Interestingly, this susceptibility of rabbit intestinal cells is age related and correlates with upregulation of net Gb₃biosynthesis in the third week of life (218, 219). It is possible that the diarrhea seen in human STEC infections is due at least in part to direct exposure of enterocytes to STX in the gut lumen. However, the presence of Gb₃ in human enterocytes has yet to be demonstrated. Other studies suggest that many of the gastrointestinal pathological findings may be caused by systemic toxin. Intravenous injection of Stx1 into rabbits caused diarrhea with edematous and hemorrhagic lesions in the mucosa and submucosa of the cecum (280). Tashiro et al. (336) also demonstrated that local intra-arterial injection of Stx1 or Stx2 caused hemorrhagic lesions in the rat small intestine. In both studies, microvascular endothelium appeared to be the principal cytotoxic target. Since STEC strains appear to be unable to invade gut epithelial cells to any significant extent, the generation of systemic sequelae must presumably involve translocation of STX produced by colonizing bacteria from the gut lumen to underlying tissues and the bloodstream. One possible route might be through lesions in the mucosal barrier caused either by the direct effects of STX or other factors such as intimin or perhaps through gaps between adjacent epithelial cells. An alternative route from gut lumen to tissues might be through intact epithelial cells. This possibility has been examined by using polarized human colonic carcinoma cells (CaCo-2A and T84) grown on collagen-coated polycarbonate membranes (4). When grown for extended periods, these cells form tight junctions and the monolayers exhibit high transepithelial electrical resistance. For both cell lines, a significant proportion of active Stx1 added to the culture medium on the apical side was translocated to the medium on the basolateral side over 24 h; during this time, there was no toxin-induced damage to the epithelial barrier as judged by electrical resistance. This process appeared to be energy dependent, since it was blocked by low temperature or an uncoupler of oxidative phosphorylation. The total amount of Stx1 that could be translocated appeared to be saturable, suggesting the involvement of a cellular receptor, but this is unlikely to be the specific STX receptor Gb₃, since T84 cells lack this glycolipid. Moreover, induction of Gb₃ synthesis in CaCo-2A cells by treatment with sodium butyrate actually reduced toxin translocation. Thus, at least Stx is capable of translocation across intestinal epithelial cells without apparent cellular disruption via a transcellular pathway (4).

Once having crossed the epithelial barrier and presumably entered the bloodstream, STX targets tissues expressing the appropriate glycolipid receptor. The specificity of this interaction and the distribution of receptors among various cell types has a major impact on the pathogenesis of disease, both in humans and in various animal models, as recently reviewed by Lingwood (193). High levels of Gb₃ are found in the human kidney, particularly in the cortical region, the principal site of renal lesions in patients with HUS (47). However, overall levels were higher in adult kidneys than in infant kidneys, which contrasts with the age susceptibility to the disease. A subsequent study involving overlaying human renal sections with fluorescence-tagged Stx1 indicated that in tissue from adults, the toxin bound principally to distal convoluted tubules, particularly those closely apposed to a glomerulus, whereas in infants, overall Stx1 staining was less intense, in parallel with Gb₃ content. However, toxin bound to glomeruli as well as to distal tubules, which is consistent with the increased susceptibility to HUS in this age group (192). Tesh et al. (342) found similar levels of Gb₃ in the cortex and medulla of human, baboon, and mouse kidneys, as well as comparable binding of Stx1; both receptor and bound toxin were associated primarily with tubular epithelial cells. Although the pathological findings of HUS in humans are strongly indicative of significant endothelial cell (EC) disturbance (discussed below), there is also evidence for the involvement of other renal cell types in pathogenesis of disease. For example, Takeda et al. (332) reported that HUS patients have elevated levels of markers of tubular injury during the early stages of disease. Tubular necrosis is also seen in a proportion of patients with HUS (120, 279), and a significant number of patients with HUS present with acute anuric renal failure consistent with tubular necrosis but without obvious signs of coagulopathy. In rabbits, intravenous administration of Stx1 results in vascular damage (thrombotic microangiopathy), particularly in the cecum, colon, and central nervous system (280, 383). EC in these tissues were also the principal sites of uptake of¹²⁵I-labelled Stx1. However, rabbit renal tissue was unaffected and did not bind toxin (280), which is consistent with its lack of Gb₃ (383).

The most dramatic demonstration of the contribution of STX receptor specificity to the disease process involved site-directed mutagenesis of the gene encoding the Stx2e B subunit, such that Gln64 and Lys66 were changed to Glu and Gln, respectively (the analogous residues in Stx2). These amino acid substitutions altered the predominant in vitro binding specificity of the mutant toxin from Gb₄ to Gb₃, that is, to the same receptor binding phenotype as Stx2. They also changed the relative cytotoxicity of the mutant Stx2e for various cell lines, in accordance with their Gb₃ and Gb₄ content (350). When the mutant Stx2e was injected intravenously into pigs, the distribution of toxin to the various organs was different from that obtained with wild-type Stx2e; the former was targeted to tissues containing Gb₃, while the latter bound extensively to erythrocytes (RBCs) (which contain Gb₄) and delivery to specific tissues was influenced by regional blood flow in addition to Gb₄ content. Differences in the clinical characteristics of toxin-induced disease were also observed, but there was no obvious effect on the nature of the histological lesions (49).

Notwithstanding the above, the susceptibility of a given cell type to STX is not determined solely by its total Gb₃ content, and recent studies indicate that the lipid moiety of Gb₃ has a significant influence on the interaction of the oligosaccharide head group with toxin (170, 193, 265). Stx1 exhibits optimum binding to Gb₃ with a fatty acyl chain length of 20 to 22 carbons, while for Stx2c the optimum chain length is 18. For both toxins, receptor binding is increased when the fatty acid is unsaturated. As discussed previously, three distinct sites on the Stx1 B subunit have now been postulated to interact with Gb₃(23, 71, 193, 230, 231). Thus, if the lipid moiety of Gb₃ influences the conformation of the oligosaccharide component, this could in turn alter the binding affinity for one site on the B subunit relative to the others. Moreover, variations in the B subunit among members of the STX family may result in a preference for one receptor binding site over another, which would then result in differential specificity for Gb₃ receptor subsets on the basis of the lipid moiety (193). This may result in different specific activities for STX toxins against given cell types, as well as in vivo variations in tissue specificity, affecting the pathological findings and the 50% lethal dose (LD₅₀), as discussed below.

Effects of the lipid moiety on toxin-receptor interactions may also explain why some Gb₃-containing cell lines are refractory to STX cytotoxicity. As discussed above, receptor-bound STX is internalized and undergoes retrograde transport via the Golgi apparatus and endoplasmic reticulum and is ultimately released into the cytosol in sensitive cells. However, the intracellular trafficking is different and Stx-receptor complexes are internalized and confined in lysosomes in the insensitive cells (294). Interestingly, treatment of such cell lines with sodium butyrate induces retrograde transport and sensitivity to toxin and also results in changes in the fatty acyl component of Gb₃ (293). Thus, the lipid moiety of Gb₃ may influence susceptibility of cell types to STX (193).

It has also been suggested that interaction of STX with glycolipid receptors on the surface of RBCs may play a role in the pathophysiology of HUS. The human P blood group antigens are glycolipids and include Pk (which is Gb₃), P (which is Gb₄), and P1 (a neolactoceramide which also has a terminal Galα1-4Gal moiety); the relative exposure of these antigens on the RBC surface varies with the blood group. On the basis of their observed association between particular P phenotypes and the outcome of HUS, Taylor et al. (338) hypothesized that STX binding by RBCs might remove toxin from the circulation, thereby protecting sensitive Gb₃-containing tissue types. Bitzan et al. (44) have demonstrated by immunofluorescence that purified Stx1, Stx2, Stx2c, and Stx2e all bind in vitro to RBCs, with the affinity varying in accordance with the P phenotype and toxin type. Enhanced binding of Stx1, Stx2, and Stx2c to RBCs of the P1 relative to the P2 phenotype was attributed to increased surface exposure of Gb₃, as well as direct binding of toxin to the P1 antigen. However, direct binding of STX to RBCs has yet to be demonstrated in cases of human STEC disease, and a more recent study by Orr et al. (243) did not report a HUS-protective association with the P1 phenotype. The significance of toxin-RBC interactions is likely to be dependent upon the relative affinity for receptors on RBCs and those on target tissues; low-affinity receptors on RBCs might mediate the transport of STX to tissues (e.g., the kidney) with high-affinity receptors. The importance of the lipid moiety of receptor glycolipids is also suggested by the report that RBCs of patients with HUS have a lower nonhydroxylated fatty acyl Gb₃ content than do those from healthy controls or patients with STEC diarrhea without HUS (226). However, it is not known whether this apparent association with HUS susceptibility is a direct consequence of altered affinity of STX for either RBC or other target tissues.

It is now generally agreed that the major portion of the histopathological lesions associated with both HC and HUS is a consequence of the interaction of STX with endothelial cells. The typical features of HUS include swollen and detached glomerular EC, and deposition of fibrin and platelets in the renal microvasculature (particularly in the glomerulus) (279). Capillary occlusion results in reduced blood flow to the kidneys and hence to renal insufficiency and may also cause physical damage to RBCs. Thrombotic lesions are also observed systemically, particularly in the microvasculature of the bowel, brain, and pancreas. Analogous histopathological lesions in the brain and gastrointestinal tract are seen in animal models of STX-induced disease (280, 336), although there is evidence that in both mice and rabbits, neurological damage by Stx2-related toxins involves direct neuronal injury in addition to microangiopathy (100, 101, 217). Although the precise molecular mechanisms whereby microangiopathic lesions are generated are not fully elucidated, STX may prevent the production of molecules critical for maintenance of the procoagulant-anticoagulant balance of the endothelium and/or an imbalance between vasodilators and vasoconstrictors produced by these cells (149).

Initial in vitro studies were performed with cultured EC derived from large vessels (umbilical or saphenous veins). These cells were shown to be susceptible to Stx/Stx1 and Stx2, as judged by inhibition of protein synthesis, detachment of the cells from the substratum, and loss of viability (237, 343). These EC were much less sensitive to Stx than were Vero cells, which was attributed to their relatively low levels of Gb₃ (343). However, the susceptibility of large-vessel EC to Stx was significantly increased if tumor necrosis factor alpha (TNF-α) was coadministered or if EC were preincubated with the cytokine (198, 343). Subsequent studies demonstrated similar enhancement of Stx cytotoxicity by preincubation of EC with LPS or interleukin-1β (IL-1β) (163, 199, 356). All three factors increase the binding of Stx1 to EC, and for TNF-α and IL-1β, this was shown to be a consequence of upregulation of Gb₃ on the cell surface (163, 356). Subsequent studies demonstrated that TNF-α induces the expression of a galactosyltransferase involved in the biosynthesis of Gb₃ (355). However, there appear to be differences in the intracellular signalling pathways involved in sensitization, since Louise et al. (201) have recently shown that the effects of LPS and TNF-α on EC are mediated by protein kinase C of class I/II and III, respectively, whereas the sensitization due to IL-1β is protein kinase C independent. Interestingly, Louise et al. (197) have shown that sodium butyrate can also sensitize human umbilical vein EC to Stx through up-regulation of Gb₃ receptors. They suggested that during STEC infection, intestinal damage might result in increased exposure of EC to butyrate, which is present in high concentrations in the colon, and thereby contribute to the pathogenesis of EC damage.

Interpretation of in vitro studies of EC in tissue culture is complicated by the fact that cells isolated from different human tissues exhibit different properties. For example, Keusch et al. (169) have reported differences in the responses of human umbilical and saphenous vein EC to Stx1, cytokines, and butyrate. More significantly, Obrig et al. (238) found that levels of Gb₃ in renal microvascular EC were 50 times higher than in umbilical vein EC and that the former were also 1,000 times more sensitive to Stx. Furthermore, the Gb₃ content of renal EC and their sensitivity to Stx were not further increased by preincubation with either TNF-α or LPS. A similar high sensitivity to Stx and lack of enhancement of susceptibility by IL-1β and TNF-α has also recently been reported for human intestinal microvascular EC (2). However, human cerebral EC appear to behave like umbilical vein EC, exhibiting a low baseline sensitivity to Stx, which is stimulated by preincubation with IL-1β and TNF-α (130). Notwithstanding the above, caution must be exercised when comparing studies. Differences in experimental procedures used for isolation, purification, and establishment of EC cultures, and possible contamination with other cell types, may affect the results. Monnens (220), for example, has recently reported that highly purified human glomerular microvascular EC are not susceptible to Stx1 unless they are preincubated with TNF-α. In a separate study, Kohan et al. (173) also reported increased susceptibility of glomerular EC to Stx1 after preincubation with either TNF-α, IL-1β, or LPS.

From the above, it seems probable that development of maximum EC damage during STEC infection requires both STX as well as host and/or bacterial inflammatory mediators to upregulate receptor expression. Endotoxemia has been reported in association with HUS due to S. dysenteriae (176) but may be of lesser magnitude in cases caused by STEC strains, which are less invasive than shigellae. Karpman et al. (162) have reported a higher incidence of systemic symptoms and glomerular pathological changes after intragastric inoculation of O157:H7 STEC in LPS-responder mice than in LPS-nonresponder mice, but the relevance of this to human disease is again uncertain, since many of the animals were bacteremic. Although elevated IL-1β and IL-8 levels have been found in the plasma of HUS patients, TNF-α has only occasionally been detected in plasma (94, 136). However, Karpman et al. (161) reported marked elevation of IL-6 and TNF-α levels in the urine of HUS patients during the acute phase of illness relative to those in healthy controls. Lack of correlation with cytokine levels in paired samples of plasma and urine suggested that the kidney itself was the site of production. Inward et al. (136) have also reported higher levels of IL-8 in urine than plasma in HUS patients. These findings are consistent with studies with transgenic mice carrying a reporter for TNF-α synthesis; injection of Stx resulted in induction of TNF-α in the kidneys but not in any other tissues (125). Kohan et al. (173) have reported Stx1-induced production of TNF-α, IL-1β, and IL-6 by human glomerular EC and proximal tubular epithelial cells. Monocytes and macrophages are also a potential source of proinflammatory cytokines, and production of TNF-α, IL-1β, IL-6, and IL-8 is induced by treatment with Stx1 (272, 341, 358). Levels of monocyte chemoattractant protein 1 (MCP-1) are also significantly elevated in the urine of HUS patients, and immunohistochemical studies have demonstrated MCP-1 expression, as well as infiltration of monocytes, in glomeruli of HUS patients (359). Local production of TNF-α and IL-1β by such cells, as well as by other renal cell types, is likely to ensure maximal induction of Gb₃ expression in glomerular EC, enhancing susceptibility to STX. Robinson et al. (285) have also reported that Stx1 binds to human glomerular mesangial cells and inhibits mitogenesis. These cells are involved in modulation of the glomerular filtration rate and elaborate a range of hormones, cytokines, and growth factors. Thus, it is possible that direct effects of STX on mesangial cells also contribute to the acute renal failure associated with HUS.

The presence of elevated IL-8 levels in the plasma of HUS patients (94) may also be significant, because IL-8 it is a powerful selective activator and chemoattractant of polymorphonuclear leukocytes (PMN). HUS patients have increased PMN counts on presentation, and high levels correlate with poor clinical outcome (mortality or permanent nephropathy) (73, 215, 363). Fitzpatrick et al. (94) reported that levels of α₁-antitrypsin-complexed elastase (a marker of PMN activation and degranulation) correlated with but lagged slightly behind those of IL-8 and that both of these parameters were highest in fatal cases. Moreover, there is evidence of oxidative damage and PMN in the glomeruli of HUS patients at autopsy (95, 337). Thus, cytokine-mediated PMN activation may play an important role in the pathophysiology of HUS. To cause inflammatory damage to the endothelium, however, PMN must be in close proximity to it. IL-1β and TNF-α upregulate the expression of adhesion molecules on the EC surface, and IL-8 and platelet-activating factor also mediate flattening of rolling PMN, promoting strong adherence to the EC (337). Products such as elastase released from adherent PMN could then degrade the extracellular matrix, resulting in detachment of EC from the basement membrane, a common histopathological feature of HUS. Interestingly, Morigi et al. (222) have also shown that purified Stx1 is a strong promoter of adhesion of leukocytes to cultured EC. The adhesive response was comparable to that achieved with IL-1β and was even greater when EC were preexposed to TNF-α. Thus, local production of TNF-α in the kidneys might increase the susceptibility of EC to inflammatory damage from adherent leukocytes as well as enhancing the direct cytotoxic actions of Stx.

Epidemiological studies have indicated that STEC strains producing Stx2 only are more commonly associated with serious human disease, such as HUS, than those producing Stx1 alone or Stx1 and Stx2 (171, 244). One possible explanation for this is that the level of transcription ofstx₂ in vivo is higher than that ofstx₁. Transcription ofstx₁ is known to be iron repressible in vitro, and its promoter region includes a recognition site for thefur gene product (56, 366). However, iron levels in the gut are very low, and so it is likely that transcription ofstx₁ will be fully derepressed in vivo. Sung et al. (329) have shown that in vitro transcription ofstx₂-related genes is constitutive and at a level commensurate with that of derepressedstx₁. Mühldorfer et al. (225) also found that stx₂ promoter activity was unaffected by osmolarity, pH, oxygen tension, acetates, iron level, or carbon source but that there was a slight effect of growth temperature. However, treatment with mitomycin, which induces the lytic cycle ofSTX-converting bacteriophages, has been shown to significantly increase both Stx1 and Stx2 production in lysogenizedE. coli strains (6, 126). For Stx2, this increase appears to be due to a combination of amplification ofstx₂ as the phage DNA is replicated and an increase in stx₂ promoter activity mediated by a phage-encoded positive regulatory factor (225). This increase in Stx2 production may be clinically significant, since mitomycin is used to treat certain neoplastic disorders and these patients are at increased risk of HUS (185).

The link between Stx2 production and HUS may be a direct consequence of increased in vivo toxicity of Stx2, or, alternatively, carriage ofstx₂ may simply be a clonal marker of STEC strains producing some additional virulence factor. The former alternative is consistent with in vitro studies which demonstrated that human renal microvascular EC were approximately 1,000 times more sensitive to the cytotoxic action of Stx2 than of Stx1 (200). Binding studies suggested that there were differences in the number of potential binding sites for Stx1 and Stx2 on the renal EC surface, presumably a function of heterogeneity of the lipid moiety of Gb₃. Thus, preferential binding of Stx2 to Gb₃ receptor subpopulations which are more efficiently internalized and processed would account for the massively increased cytotoxicity.

Increased in vivo toxicity of Stx2 is also supported by studies involving a streptomycin-treated mouse model of toxin-induced renal tubular damage. Streptomycin treatment reduced the normal facultative intestinal flora of the mice and facilitated colonization with orally administered E. coli strains (361). Oral infection of mice with an O157:H7 STEC strain producing Stx1 and Stx2 resulted in fatal cortical tubular necrosis. Death could be prevented by passive immunization with monoclonal anti-Stx2 but not with anti-Stx1 (361, 362). Oral challenge with E. coliK-12 strain DH5α carrying cloned stx₂ but notstx₁ was also capable of inducing fatal tubular damage. The level of toxin production was clearly important, since challenge with E. coli K-12 strain DH5α carryingstx₂ on a low-copy-number vector was not lethal (362). Tesh et al. (339) subsequently demonstrated that purified Stx2 had an approximately 400-fold lower LD₅₀ for mice than did Stx1 by both the intravenous and intraperitoneal routes. Interestingly, immunofluorescent staining showed that both toxins were capable of binding to Gb₃receptors on renal tubular cells, and in a solid-phase in vitro binding assay, Stx1 actually exhibited a 10-fold greater affinity for Gb₃ than did Stx2.

Assessment of the impact of STX type on the capacity of STEC to cause severe disease is complicated by the fact that there is a significant degree of naturally occurring amino acid sequence variation, particularly among Stx2 types from different strains (106, 138, 189, 214, 242, 258, 259, 305). Lindgren et al. (190) compared the oral virulence of a number of Stx2-producing clinical isolates for streptomycin-treated mice. In this study, two strains belonging to serotype O91:H21 were exquisitely virulent (LD₅₀ < 10 CFU, compared with >10¹⁰ CFU for other strains tested). Although the capacity to grow in mouse small intestinal mucus appeared necessary for maximal virulence, it was not the sole determinant, since at least one other strain tested grew well under these conditions yet was of low virulence; the production of a particularly mouse-virulent Stx2-related toxin appeared to be a likely explanation for this observation. Assessment of the contribution of specific toxins to the virulence of Stx2-producing STEC strains is further compounded because many isolates produce more than one Stx2-related toxin (305). Indeed, one of the highly virulent O91:H21 isolates referred to above carried threestx₂-related genes, while the other isolate carried two genes (190); strains carrying onestx₁ gene and twostx₂-related genes have also been described (260). One strength of the streptomycin-treated mouse model is that the oral virulence of E. coli K-12 strains expressing different cloned stx₂-related genes can be compared. Interestingly, clones expressingstx₂ genes from the O91:H21 STEC strains were no more virulent than those expressing classicalstx₂ in this mouse model; in addition, the purified toxins had indistinguishable intraperitoneal LD₅₀(191). Moreover, sequence analysis indicated that the B subunits of the O91:H21 toxins were identical to that of Stx2c, which contains amino acid differences in the B subunit, resulting in a 100-fold-lower cytotoxicity for Vero cells compared with that of Stx2 (191).

Studies involving otherwise isogenic E. coli DH5α derivatives expressing a different set of naturally occurring Stx2 variants have, however, detected differences in virulence in the streptomycin-treated mouse model (251, 252). The least virulent (and least verocytotoxic) clone produced an Stx2-related toxin with an amino acid substitution (Arg176→Gly) in a region of the A subunit known to be important for the catalytic activity of the toxin. Interestingly, clones producing two other toxins (designated Stx2/O48 and Stx2/OX3b) had high cytotoxicity for Vero cells, but the latter was more virulent when fed to streptomycin-treated mice, as judged by the median survival time (but not by the overall survival rate). Subsequent studies demonstrated that the increased virulence of clones producing Stx2/OX3b was a function of the combination of A-subunit residues Met4 and Gly102 (252). A clone producing another Stx2 variant (designated Stx2/OX3a), which had the B-subunit amino acid sequence variations characteristic of the Stx2c subgroup, had similar oral virulence to clones producing Stx2/O48, even though it was approximately 500-fold less verocytotoxic (251), in agreement with the earlier finding of Lindgren et al. (191). However, when the model was modified by withdrawal of streptomycin selection and reintroduction of the normal gut flora after 3 days, the mortality rate of mice challenged with the more verocytotoxic clone was significantly greater (262).

Recently, Melton-Celsa et al. (212) reported that the Stx2-related toxins purified from one of the highly mouse-virulent O91:H21 strains referred to above exhibited 10- to 1,000-fold greater cytotoxicity for Vero cells after incubation in the presence of mouse proximal small intestinal mucus or human colonic mucus. In marked contrast, no stimulation in cytotoxicity was observed when extracts containing Stx1, Stx2, Stx2c, or Stx2e were similarly treated. Activation of the Stx2-related O91:H21 toxins by mucus also appeared to be a function of the A subunit, because the B subunit is identical to Stx2c, which was not activatable. However, the A-subunit amino acid differences (with respect to Stx2c) common to both the activatable toxins (Ser291 and Glu297) are not, on their own, sufficient for activation, since they are also found in Stx2e, which is not activatable. At present, the significance of activation of some Stx2-related toxins by intestinal mucus to pathogenesis of human disease is uncertain. Clearly, it is not essential, since serious human disease frequently results from infection with STEC strains producing nonactivatable toxins. However, Melton-Celsa et al. (212) have suggested that production of an activatable toxin may increase virulence and/or compensate for the absence of another separate virulence factor, since both the O91:H21 STEC strains studied lackeaeA. Interestingly, Baker et al. (18) have reported differences in the virulence for gnotobiotic piglets among O157:H7 STEC strains, all of which were positive forstx₁, stx₂, andeaeA. However, whether this is a consequence of differences in toxin sequence or accessory virulence factors is unknown (discussed below).

It should also be emphasized that the link between Stx2 production and the capacity of an STEC strain to cause HUS is not absolute. STEC strains producing only Stx1 are capable of causing HUS, as are strains of S. dysenteriae type 1 which also produce only Stx. Indeed, Stx has recently been shown to cause medullary tubular injury and proteinuria in isolated perfused rat kidneys (314), although the toxicity of Stx2 in this model system was not investigated.

Beutin et al. (33) observed that a high proportion of STEC strains (89% of those tested) had a novel hemolytic phenotype which was distinct from that associated with the E. coli alpha-hemolysin (Hly). Strains producing this enterohemolysin (subsequently designated EHEC-Hly) are not hemolytic on standard blood agar but produce small, turbid hemolytic zones on washed sheep RBC agar (supplemented with Ca²⁺) after overnight incubation. Unlike alpha-hemolysin which is chromosomally encoded, EHEC-Hly was found to be encoded by the 60-MDa “virulence plasmid” (pO157) of the O157:H7 strain EDL933 (302). Subsequent sequence analysis of DNA cloned from pO157 demonstrated the presence of an operon, consisting of four open reading frames, with approximately 60% homology to the E. colialpha-hemolysin operon (hlyCABD) and operons encoding other members of the RTX family of pore-forming cytolysins. The O157 enterohemolysin operon was therefore designated EHEC-hlyCABD(300, 303). Within the alpha-hemolysin operon,hlyA is the structural gene for the hemolysin, which is synthesized as an inactive precursor and converted to its active form by the product of the hlyC gene. The secretion of HlyA is signal peptide independent and mediated by both a specific membrane translocator system encoded by the hlyB and hlyDgenes and an unlinked OMP, TolC (369). EHEC-HlyA, with a predicted size of 107 kDa, is not efficiently secreted by E. coli EDL933. However, trans complementation withhlyBD significantly increases export, resulting in an alpha-hemolysin-like phenotype (300). Schmidt et al. (303) have suggested that the poor secretion of EHEC-HlyA by wild-type strains could be due to a specific amino acid difference in the ATP binding cassette (ABC) domain of EHEC-HlyB compared with HlyB (Asn and Gly, respectively).

Schmidt et al. (300), using a PCR assay to detect EHEC-hlyA, surveyed diarrheagenic E. coliisolates and found that all 22 STEC O157:H7 strains and 12 of 25 STEC strains of other serotypes were positive. All non-STEC diarrheagenic strains were negative for EHEC-hlyA. Although PCR-positive strains hybridized with the CVD419 probe (a commonly used marker for the large virulence plasmid [186]), analysis of this probe indicated that part of the EHEC-hlyA and EHEC-hlyB genes were included. Thus, there seems little doubt that there is a correlation between STX and enterohemolysin production (particularly in O157 STEC strains). Schmidt and Karch (301) have also shown that an EHEC-hlyA homolog from an O111:H⁻ STEC strain had 99.4% homology to O157 EHEC-hlyA. Within this STEC serotype, 16 of 18 isolates from patients with HUS were both enterohemolytic and positive for EHEC-hlyA. In contrast, only 4 of 18 isolates from patients with diarrhea had this phenotype and genotype, suggesting the possibility of a direct association between enterohemolysin production and the capacity of a given STEC strain to cause more serious disease. Interestingly, regardless of the presence or absence of EHEC-hlyA, all the O111:H⁻ STEC strains tested carried large plasmids (301). A possible role for EHEC-Hly in pathogenesis is also consistent with the finding that 19 of 20 convalescent-phase sera from patients recovering from HUS due to O157 STEC contained antibodies to EHEC-Hly, compared with only 1 of 20 sera from age-matched controls (300).

The manner in which EHEC-Hly might contribute to the pathogenesis of STEC disease is not understood. One possibility is that hemoglobin released by the action of EHEC-Hly provides a source of iron, thereby stimulating the growth of STEC in the gut (184). However, the biological effects of members of the RTX toxin family are not confined to lysis of RBCs, and cytotoxicity has been reported for a range of cell types (369). Moreover, sublytic concentrations of E. coli alpha-hemolysin have been shown to mediate the release of leukotrienes from granulocytes and IL-1β from cultured monocytes and to reduce the binding of chemotactic factors by neutrophils (36, 299). Of potential interest regarding the pathogenesis of HUS is the finding that alpha-hemolysin is cytotoxic for endothelial cells (330) and increases superoxide production by rat renal tubular cells (164). Intravenous administration of alpha-hemolysin was also reported to induce granulocytopenia and thrombocytopenia in monkeys (354). Notwithstanding this, there appear to be substantial differences in the target cell specificities of alpha-hemolysin and EHEC-Hly. Both toxins lysed sheep and human RBCs and a bovine lymphoma cell line, but, unlike alpha-hemolysin, EHEC-Hly did not lyse two human lymphocyte cell lines (26). Interestingly, a major area of sequence divergence between HlyA and EHEC-HlyA is the 70 residues at the N terminus that includes the region believed to be important for target cell specificity (26). Thus, although it is tempting to speculate that EHEC-Hly might be involved in modulation of inflammatory processes associated with EC damage or might act in synergism with STX or LPS, direct studies with purified EHEC-Hly are needed to define the tissue specificity and biological properties of this toxin.

Another putative virulence factor also encoded on pO157 is the recently described extracellular serine protease EspP (extracellular serine protease, plasmid encoded) (52). The espP gene encodes a 1,300-amino-acid protein, which is subsequently subjected to N- and C-terminal processing during the secretion process. The mature form has an apparent size of 104 kDa. The amino acid sequence of the N terminus of the mature protein indicated that EspP may be the same as a 104-kDa secreted protein found in an O26:H⁻ STEC strain by Ebel et al. (90). The deduced amino acid sequence of the complete O26 protein was subsequently shown to be virtually identical to the sequence of EspP (80). EspP has significant homology (approximately 70% overall) to EspC, a 110-kDa EPEC secreted protein, and to a lesser degree to immunoglobulin A1 (IgA1) proteases ofHaemophilus influenzae and Neisseria spp. The region of homology includes the serine-containing proteolytic active site. EspP did not exhibit IgA1 protease activity in vitro but was capable of cleaving pepsin, an activity which was inhibited by the serine protease inhibitor phenylmethylsulfonyl fluoride. EspP was also able to cleave human coagulation factor V. Thus, Brunder et al. (52) have suggested that secretion of EspP by STEC colonizing the gut could result in exacerbation of hemorrhagic disease. Djafari et al. (80) also reported that EspP is cytotoxic for Vero cells. A role for EspP in pathogenesis is also consistent with the presence of antibodies to the protease in sera from five of six children with STEC infection but not in sera from age-matched controls. Nevertheless, EspP was produced by only four of six O157 and one of two O26 isolates but by none of two O103 STEC clinical isolates tested, indicating that it is not a universal virulence factor of STEC (52).

A final virulence factor that might contribute to the pathogenesis of the watery diarrhea often seen during the early stages of STEC infection is the enterotoxin EAST1 (encoded by astA). This is a 39-amino-acid enterotoxin that was initially recognized in certain strains of enteroaggregativeE. coli (EAggEC) and is distinct from the heat-stable toxins produced by ETEC (296). Recently, Savarino et al. (297) demonstrated that an astA-specific probe hybridized with all 75 O157:H7 STEC strains tested, as well as with 8 of 9 O26:H11 strains and 12 of 23 STEC strains of other serotypes. This probe also hybridized with 41, 41, and 22% of EAggEC, ETEC, and EPEC strains, respectively. Enterotoxicity for rabbit ileal tissue was confirmed by testing culture ultrafiltrates (<10 kDa) of three of theastA⁺ O157:H7 strains in Ussing chambers. In these three strains, there were two copies of astA, which were located on the chromosome (297).

The profound sensitivity of Vero cells to STX was first observed by Konowalchuk et al. (175), and cytotoxicity for this cell line remains the “gold standard” for confirmation of putative STX-producing isolates. Since Vero cells have a high concentration of Gb₃and Gb₄ (the preferred receptor for Stx2e) in their plasma membranes, they can be used to detect all known STX variants. HeLa cells have also been used, but this cell line lacks Gb₄ and therefore is less sensitive to Stx2e. It is not practical or economical to test filtrates of individual colonies from a fecal culture plate to isolate STEC, because these organisms frequently comprise less than 1% of the aerobic flora. However, the Vero cell cytotoxicity assay is an effective means of demonstrating the presence of STX in cell extracts of feces or fecal culture filtrates. The assay involves treatment of Vero monolayers (usually in 96-well trays) with sterile extracts or filtrates (serially diluted in some cases) and examining cells for cytopathic effect after 48 to 72 h of incubation (156, 318). Historically, this assay has played an important role in establishing a diagnosis of STEC infection, particularly where subsequent isolation of the causative organism has proven to be a difficult task. When testing such crude samples, the sensitivity is influenced by the abundance of STEC, the total amount and potency of the STX produced by the organism concerned, and the degree to which the particular STX is released from the bacterial cells. Karmali et al. (157) found that treating mixed fecal cultures with polymyxin B to release cell-associated STX improved the sensitivity of the Vero cell assay to the point where it could reliably detect STEC at a frequency of 1 CFU per 100. Clearly some STEC strains produce very high levels of toxin, and these can be detected at even lower frequencies. However, the converse also applies. Since the presence of cytoxicity in a crude filtrate could be due to other bacterial products or toxins, positive samples should always be confirmed (and typed) by testing for neutralization of cytotoxicity by specific (preferably monoclonal) antibodies to Stx1 or Stx2 (317).

Although a valuable diagnostic method, detection of STX by tissue culture cytotoxicity is labor-intensive, time-consuming, and cumbersome. Not all microbiology laboratories perform tissue culture work with Vero cell monolayers available on demand. Moreover, rapid diagnosis is important, and the results of cytotoxicity testing are generally not available for 48 to 72 h. Development of a Vero cell line carrying the gene encoding firefly luciferase might potentially overcome the latter drawback (122). Since luciferase has a very short in vivo half-life, it is a sensitive marker of STX inhibition of protein synthesis. By using luciferase, equivalent or lesser amounts of STX were detected by diminution of chemiluminescence after 6 h of incubation compared to detection by the conventional 72-h cytotoxicity assay. Another potentially useful modification that eliminates the cost of filter-sterilizing bacterial extracts before testing involves overlaying confluent Vero monolayers with tissue culture medium containing 0.9% agarose (249). Putative STEC colonies can then be inoculated directly onto the surface of each well and cytotoxicity in the underlying monolayer can be assessed after 48 h. Although this technique worked well with pure STEC isolates, its sensitivity has yet to be assessed with crude fecal cultures.

During the past decade, a number of enzyme-linked immunosorbent assays (ELISAs) have been developed for the direct detection of Stx1 and Stx2 in fecal cultures. Like verocytotoxicity, these play a potentially important role in diagnosis because they can detect the presence of STEC (or other STX-producing species), regardless of serogroup. Such assays can also be used to confirm toxin production by putative STEC isolates where tissue culture facilities are unavailable. Most of the published ELISA methods involve a sandwich technique with immobilized monoclonal or affinity-purified polyclonal antibodies to the toxins as capture ligands. After incubation with cultures, bound toxin is detected with a second STX-specific antibody followed by an appropriate anti-Ig–enzyme (usually alkaline phosphatase) conjugate (84, 87). Some assays involve an STX detection antibody directly conjugated to the enzyme (174) or a biotinylated detection antibody which is used with a streptavidin-enzyme conjugate (19, 321). Another variation in the ELISA protocol involves the use of purified Gb₃ (or a deacylated derivative) or hydatid cyst fluid (containing P₁ glycoprotein, which also binds STX) to coat the solid phase (7, 14, 15, 24, 182, 183). Two STX ELISAs are commercially available in kit form (Premier EHEC [Meridian Diagnostics, Inc., Cincinnati, Ohio] and LMD [LMD Laboratories, Carlsbad, Calif.]).

Tests of pure isolates show that the specificities of the various STX ELISAs are in close agreement with the results of verocytotoxicity assays (15, 24, 87, 321). Also, Law et al. (183) reported a specificity of 99.7% when fecal cultures were tested by STX ELISA and the results were compared with isolation of STEC. On the other hand, Ball et al. (19) found that 6 of 13 ELISA-positive pure isolates (4 E. coli strains, 1Citrobacter freundii strain, and 1 Enterobacter cloacae strain) from fecal cultures were negative when subsequently tested for verocytotoxicity. Beutin et al. (34) also reported false-positive reactions with several strains ofPseudomonas aeruginosa using the Premier EHEC kit; this included one ATCC strain which had previously been tested as negative by the manufacturer.

The sensitivity of the various ELISAs is affected by a number of variables including the avidity of the antibodies used and the type and amount of STX produced by a given strain. ELISAs are generally less sensitive than the verocytotoxicity assay. Downes et al. (87) concluded that ELISA sensitivity was inadequate to reliably detect low levels of STX found in direct fecal extracts. However, the amount of free STX present in primary fecal cultures is generally greater, particularly when broths are supplemented with polymyxin B and mitomycin to enhance the release of Stx1 and Stx2, respectively. Under such circumstances, ELISAs were reported to be capable of detecting the presence of Stx1-producing organisms comprising less than 1% of total flora (15, 182) and Stx2-producing organisms at an abundance of less than 0.1% (182). Thus, under optimal conditions, STX ELISAs can provide a reliable primary screen for the presence of STEC strains (including non-O157 strains) in fecal cultures as long as the specimen is obtained fairly early in the course of infection. Studies of comparative specificity and sensitivity carried out to date indicate that commercially available STX ELISA kits are likely to be of considerable utility for laboratories without access to more specialized diagnostic procedures, particularly for detection of non-O157 STEC strains (5, 167, 202, 245, 247). However, reports of false-positive ELISA reactions indicate that independent confirmation of STX production or the presence of STX genes would be prudent.

A reverse passive agglutination test for the detection of STX production is also commercially available (Oxoid, Unipath Ltd., Basingstoke, United Kingdom). The test involves incubating serially diluted polymyxin B extracts of putative STEC cultures with Stx1- and Stx2-specific antibody-coated latex particles and observing for agglutination after 24 h. Beutin et al. (35) found that this assay detected toxin production (of the appropriate type) in strains containing stx₁,stx₂, and stx_2c but did not detect toxin produced by the strains carryingstx_2e. Since a number of Stx2 and Stx2c producers gave positive reactions only when undiluted extracts were tested (29), this method is not sufficiently sensitive for screening primary fecal cultures for the presence of STX-producing organisms.

The availability of cloned stx₁ andstx₂ genes enabled the development of DNA probes for the detection of STEC (227, 374). Initially, probes labelled with ³²P or ³⁵S were used for testing large numbers of fecal E. coli isolates or the direct screening of colonies on primary isolation plates for the presence ofSTX genes by colony hybridization (307, 345). These procedures were both highly sensitive and specific, and when stringent washing conditions were used, strains carryingstx₁, stx₂, or both could be differentiated. However, radioactively labelled probes had disadvantages for clinical laboratories, such as delays due to the need for long autoradiographic exposures, short probe half-life, and the problems associated with handling and disposal of radioisotopes. These problems have been largely overcome by the introduction of nonradioactive labels such as digoxigenin and biotin, andSTX probes that use these have been used for detection of STECs without loss of sensitivity or specificity (345). The availability of nucleotide sequence data for STX genes has also permitted the design of synthetic oligonucleotide probes for detection of STEC (51, 152). Some oligonucleotide probes were based on sequences which are highly conserved among the various toxin genes and hence permitted detection of all types. Other probes were directed against less highly conserved regions, which, under the appropriate hybridization and washing conditions, distinguished betweenstx₁, stx₂, andstx_2e genes (51). Although hybridization with DNA or oligonucleotide probes is not a particularly sensitive means of screening broth cultures or fecal extracts for the presence of STEC, it is a powerful tool for distinguishing colonies containing STX genes from commensal organisms (discussed below).

Access to sequence data for the variousSTX genes has also permitted the design of a variety of oligonucleotide primer sets for amplification of STX genes by PCR. Crude lysates or DNA extracts from single colonies, mixed broth cultures, colony sweeps, or even direct extracts of feces or foods can be used as templates for PCR. STX-specific PCR products are usually detected by ethidium bromide staining after separation of the reaction mix by agarose gel electrophoresis. To date, someSTX PCR assays have combined different primer pairs forstx₁ and stx₂, and in some cases stx₂ variants, in the same reaction, thereby directing the amplification of fragments which differ in size for each gene type (29, 50, 104, 147, 270). OtherSTX PCR assays use a single pair of primers based on consensus sequences. These primers are capable of amplifying allSTX genes, with subsequent identification of the gene type requiring Southern or dot-blot hybridization with labelled oligonucleotides directed against type-specific sequences within the amplified fragment (153, 257, 276). Apart from increasing the sensitivity, secondary hybridization steps act as independent confirmation of the identity of the amplified product. Restriction fragment length polymorphism analysis of amplified portions ofstx₂ genes has also been used to discriminate between stx₂ and stx₂variants (188, 287, 349). In addition, PCR can be used for preparation of labelled DNA probes for use in hybridization reactions by amplification in the presence of, for example, digoxigenin-labelled nucleotides (140, 141).

The use of PCR technology permits the detection of STX genes from samples which are microbiologically complex (such as feces or foodstuffs), including samples containing nonviable organisms. PCR assays are potentially extremely sensitive; using serially diluted DNA extracted from an STEC isolate, Brian et al. (50) showed that amplification of less than 1,000 genomes resulted in visiblestx₁ and stx₂ PCR products after ethidium bromide staining of agarose gels. When secondary Southern hybridization with a labelled probe was used to detect the PCR products, fewer than 10 STX-containing bacterial genomes per assay could be detected (50). In this study, the sensitivity was about 100-fold lower when the DNA template was prepared by direct extraction from feces seeded with known numbers of STEC. This was a consequence of the presence of inhibitors ofTaq polymerase in the sample, which necessitated dilution before assay. Other studies have also shown suboptimal sensitivity when PCR is carried out directly on fecal extracts (274). Inhibitors of Taq polymerase are also present in meat. Begum and Jackson (28) showed that ground-beef homogenates had to be diluted 1,000-fold before assay. For both feces and food samples, the sensitivity of PCR assays is vastly increased if template DNA is extracted from broth cultures (28, 104, 257). Broth enrichment (which can involve as little as 4 h of incubation) serves two purposes; inhibitors in the sample are diluted, and bacterial growth increases the number of copies of the target sequence. Optimization of sensitivity is of paramount importance, because the number of STEC in the feces of patients with serious STX-related diseases or in suspected contaminated foodstuffs may be very small. Another consideration that may affect the performance of some PCR assays is the DNA sequence polymorphisms that are known to exist amongSTX genes; this is particularly so forstx₂-related genes, for which significant variation has been reported (106, 138, 189, 214, 242, 258, 259, 305). Sequence divergence between the primer and its target (particularly at the 3′ end of the primer) will significantly reduce the efficiency of annealing, with potentially dramatic effects on the sensitivity of the PCR reaction. When selecting or designing primers, care must be taken to avoid regions where sequence heterogeneity has been reported. PCR assays which use a single primer pair to amplify both stx₁ and stx₂ may be less susceptible to this potential complication. Target sequences which are conserved between otherwise widely divergent genes are likely to encode structurally important domains, and so random mutations will be strongly selected against.

The speed of diagnosis of STEC infection is also an important consideration in the clinical setting. The precise time required for a PCR assay varies with the amplification protocol itself (the number of cycles and incubation times at each temperature), the method used for DNA extraction, and the procedure for detection of the PCR products. The minimum time required for direct PCR analysis of an unenriched fecal sample analyzed by agarose gel electrophoresis could be as little as 4 to 6 h. Inclusion of a broth enrichment step and the use of a more sophisticated DNA purification procedure would increase this time to 12 to 24 h, while hybridization of PCR products withSTX probes could add another day. The cumulative increase in sensitivity resulting from each additional step must be balanced against the increase in time; this equation will vary depending on the particular clinical or epidemiological context.

It has often been argued that PCR should be confined to reference laboratories, because it is labor-intensive and requires highly skilled staff. However, an increasing number of clinical laboratories are now routinely using PCR for a range of applications. The risk of cross-contamination of samples leading to false-positive PCR results can also be minimized by the physical separation of pre- and postamplification procedures and other precautions such as the use of plugged micropipette tips. Unlike the STX-specific antibodies and other specialized reagents needed for ELISAs, custom-made oligonucleotide primers are inexpensive and universally available and have a very long shelf life. Modern versatile PCR thermal cyclers are no more expensive than ELISA plate readers and can handle assays in 96-well format for laboratories that have a high specimen throughput. Moreover, a variety of alternatives to agarose gel electrophoresis have been developed for high-volume, sensitive, semi-automated detection of PCR products (for example, Q-PCR and TaqMan [Perkin-Elmer Applied Biosystems, Foster City, Calif.] and PCR-ELISA [Boehringer, Mannheim, Germany]).

PCR has also been used for the detection of genes encoding accessory virulence factors, such as eaeA and EHEC-hlyA, in STEC isolates (105, 300). This information may be of significance, because as discussed previously, there is a link between the presence of these genes and the capacity of an STEC isolate to cause serious human disease (21, 300, 301). Thus, a child presenting with acute diarrhea who is infected with a STEC isolate that is also positive for eaeA and EHEC-hlyA is likely to be at increased risk of developing complications such as HUS. Fratamico et al. (99) combined previously describedSTX-specific (153) and eaeA-specific (105) PCR primer pairs with those specific for a portion of the 60-MDa virulence plasmid from an O157:H7 STEC in a multiplex assay. They concluded that this assay was suitable for the identification of STEC strains belonging to serogroup O157. However, the O157 virulence plasmid primers actually recognize a portion of the EHEC-hlyA gene, which is not confined to serogroup O157. Thus, this particular multiplex PCR will also be capable of identifying a significant proportion of potentially virulent non-O157 STEC strains. Gannon et al. (103) have recently described two other multiplex PCR formats based on STX-specific and two distincteaeA-specific primer sets, as well as a primer pair specific for a portion of the fliC_h7 gene, which encodes the H7 antigen. The two eaeA-specific primer sets recognized either the highly conserved 5′ portion of eaeA or a region at the (variable) 3′ end of the gene which was specific for O157 strains and a small number of other serogroups (105). Collectively, these assays could distinguish between EPEC and STEC strains, identify STEC strains that were likely to be of increased virulence, and identify those likely to belong to serotype O157:H7 (103). Another multiplex PCR to detect O157:H7 STEC usesstx₁- and stx₂-specific primer pairs in combination with primers that recognize a single-base mutation in the β-glucoronidase-encoding uidA gene (60). This mutation in uidA results in a β-glucoronidase-negative phenotype, a feature strongly associated with O157:H7 strains (discussed below).

While O157:H7 strains are the most common cause of serious STEC disease in Europe and North America, O111:H⁻ STEC strains are also a common cause of disease and in some regions (e.g., Australia) are the most frequent cause of HUS. Louie et al. (195) also exploited the heterogeneity of the 3′ end of eaeA to design two PCR assays, one of which was specific for eaeA from O157 STEC and O55 EPEC strains. The other primer pair gave a positive reaction with 16 of 22 O111 STEC strains but not with O111 EPEC strains or 72 STEC strains belonging to six other serogroups, including O157. Sequence heterogeneity of eaeA within O111 STEC is presumably responsible for the fact that six of these strains (all of which were O111:H11) were PCR negative even though they hybridized to a probe specific for the conserved region of eaeA(195). A more specific O111 PCR assay has recently been described which directly targets a group-specific portion of the O-antigen-encoding rfb locus of E. coli O111 (260); this region (designated ORF 3.4) is thought to encode the biosynthesis of GDP-colitose (25). We have also recently described two multiplex PCR assays for the detection and characterization of STEC. The first uses four primer pairs and detects the presence of stx₁,stx₂ (including variants ofstx₂), eaeA, and EHEC-hlyA. The second assay combines the above O111rfb-specific primers with primers directed at a type-specific portion of the O157 rfb locus (therfbE gene); the latter primer pair does not give a false-positive reaction with E. coli O55 sequences (255).

For many years, sorbitol-MacConkey agar culture (SMAC) has been the most commonly used method for isolation of STEC, because of the predominance of O157:H7 and O157:H⁻ strains as etiological agents of human disease in North America and Europe. Most of these strains are unable to ferment sorbitol, which distinguishes them from the majority of fecal E. coli belonging to other serotypes (178, 205, 370). Sorbitol-MacConkey agar plates are inoculated with the fecal specimen and examined after 18 to 24 h of incubation for the presence of colorless, sorbitol-negative colonies. Individual colonies are tested by slide or tube agglutination with O157 and H7 antisera. Several commercial latex reagents for O157 antigen and one for H7 antigen are also available and have been shown to be both accurate and sensitive compared with standard serological tests (320). It is still necessary to confirm STX production in tissue culture or ELISAs (as discussed above), since not all O157 strains are toxin producers.

The sensitivity of SMAC is limited by the capacity to recognize nonfermenting colonies against the background of other organisms on the plate; this is particularly difficult when the O157 strain forms less than 1% of the flora. However, Chapman et al. (65) improved the isolation rate of O157 STEC by supplementing SMAC with cefixime, to inhibit Proteus spp., and rhamnose, which is fermented by most sorbitol-negative non-O157 E. coli strains (O157 strains generally do not ferment rhamnose). Zadik et al. (381) have reported a further improvement in O157 isolation rates by using SMAC supplemented with cefixime and potassium tellurite (CT-SMAC). Although screening fecal cultures on SMAC is inexpensive and involves minimal labor and equipment, it will primarily detect STEC belonging to serogroup O157. Serious STEC disease has been associated with many other serogroups (147), and although some are sorbitol negative (241), most are sorbitol positive. Thus, the efficacy of SMAC will vary in accordance with the local STEC serotype prevalence. In one study, SMAC resulted in the isolation ofE. coli O157 from 80% of fecal samples which were positive for STX by direct cytotoxicity (283), whereas in another study, SMAC was positive for only 30% of verocytotoxin-positive samples (273). An additional consideration is that Stx2-producing, sorbitol-positive E. coliO157:H⁻ isolates have been associated with both outbreaks and sporadic cases of HUS in Germany. In one study, 17 of 44 HUS patients had Stx-producing E. coli O157 strains that were sorbitol fermenters (116). Such strains have also recently been isolated from patients with HUS in the Czech Republic (40). Moreover, these strains are very sensitive to tellurite, which argues strongly against the use of CT-SMAC for isolation of STEC.

Another biochemical feature that distinguishes O157:H7 STEC strains from other E. coli strains is failure to produce β-d-glucuronidase (178, 275). This enzyme can be readily detected fluorigenically with the substrate 4-methylumbelliferyl-β-d-glucuronide or colorimetrically on plates supplemented with 5-bromo-6-chloro-3-indolyl-β-d-glucuronide. Thompson et al. (346) examined 188 serogroup O157 strains by this method; 166 isolates were glucuronidase negative and STX positive, while the remaining 22 isolates were all glucuronidase positive and STX negative. Again, this method is not useful in cases of STEC-related disease where the causative organism is of a serogroup other than O157. Moreover, the 17 sorbitol-positive, Stx2-producing O157 isolates from HUS patients described by Gunzer et al. (116) were also glucuronidase positive.

Recently, a commercial agar medium (Rainbow Agar O157 [Biolog Inc., Hayward, Calif.]) containing selective agents for E. coliand chromogenic substrates for β-d-glucuronidase and β-galactosidase has become available for the isolation of STEC. Glucuronidase-negative, galactosidase-positive O157 strains appear as black colonies on this medium, whereas commensal E. colistrains are pink. It has also been claimed that some non-O157 STEC strains overproduce β-galactosidase relative to β-d-glucuronidase on this medium, giving the colonies a distinctive intermediate color. To date, analyses of the efficacy of this medium for the detection of either O157 or non-O157 STEC have been confined to testing of foods, particularly raw-meat samples and fruit juices (135, 295), and its utility for clinical diagnosis has yet to be reported. Again, it should be emphasized that isolation of a putative O157 strain is not a definitive diagnosis in itself, since isolates must be tested to confirm STX production.

Park et al. (246) have used direct immunofluorescent staining of fecal specimens with polyclonal anti-O157–fluorescein isothiocyanate as an alternative to SMAC for the detection of E. coli O157. This was at least as sensitive as SMAC and moreover involved only a 2-h turnaround time. Commercial ELISAs for rapid detection (less than 1 h) of the presence of O157 antigen in fecal specimens are also available (LMD and Premier E. coli O157 [Meridian Diagnostics, Inc.]). Both of these ELISAs have similar sensitivity to SMAC (202, 248) and, importantly, are capable of detecting sorbitol-fermenting O157 STEC strains, should they be present. Both the immunofluorescence assays and ELISAs, however, require confirmation by culture or by demonstration of STX in the sample.

As discussed above,E. coli strains belonging to a large range of serotypes, as well as other bacterial species, can produce STX and cause disease in humans. Regrettably, there is no definitive biochemical characteristic which distinguishes STEC strains belonging to serogroups other than O157 from commensal fecal E. coli strains, a fact which significantly complicates the isolation of such organisms. As discussed, however, nearly all O157 STEC strains and a significant proportion of non-O157 STEC strains produce EHEC-Hly. These organisms can be distinguished phenotypically from alpha-hemolytic strains. Strains producing EHEC-Hly are not hemolytic on standard blood agar but produce small, turbid hemolytic zones on washed sheep RBC agar (supplemented with Ca²⁺) after 18 to 24 h of incubation at 37°C. Alpha-hemolytic strains form large, clear zones of hemolysis on standard blood agar or washed sheep RBC agar plates after only 4 to 6 h of incubation at 37°C. Production of EHEC-Hly has a high positive predictive value, since Beutin et al. (33) found no enterohemolytic strains among 267 fecalE. coli isolates which did not produce STX. However, the predictive value of a negative EHEC-Hly result is low. Schmidt and Karch (301) investigated the enterohemolytic genotypic and phenotypic profiles of 36 O111:H⁻ STEC strains isolated from patients with HUS and diarrhea. Twenty strains were positive for EHEC-Hly; two additional strains were positive for EHEC-hlyAby DNA hybridization and PCR but were not hemolytic. The remaining 14 O111:H⁻ STEC strains were negative by both DNA hybridization and plating. The fact that a significant proportion of disease-causing STEC strains (44% in this particular study) are EHEC-Hly negative diminishes the usefulness of washed sheep RBC agar as a “generic” screen for primary isolation of STEC.

The only comprehensive means of isolating STEC or other STX-producing organisms involves direct analysis of colonies on agar plates with either nucleic acid probes or antibodies, and a variety of protocols for this purpose have been described (129, 146, 216, 260, 266). This is a labor-intensive process and should be carried out only for specimens which have tested positive in screens for STX (by cytotoxicity or ELISA) or for STX (by PCR). Colonies from agar plates are directly blotted onto a suitable membrane (e.g., nitrocellulose for immunoblots or positively charged nylon for hybridization). A carefully aligned replicate of the filter is made and then processed and reacted with an antibody or nucleic acid probe by standard procedures. Theoretically, up to several hundred discrete colonies can be tested on a single filter, although this may require dilution and replating of primary cultures. Alternatively, colonies from primary isolation plates can be selected and inoculated into 96-well microtiter trays containing broth. This is a time-consuming step (15 to 20 min per tray), but the 96-well format enables the subsequent use of semi-automated machinery to make replicate copies of trays and, after incubation, to transfer aliquots onto appropriate filters; the trays are also convenient for preservation of the isolates at −70°C. Hull et al. (129) have directly compared the sensitivity and specificity of immunoblotting and DNA probing for the detection of STEC colonies. Cytotoxicity testing indicated that if the results disagreed between the two methods, it was almost always a consequence of a false-positive immunoblot reaction. This was more prevalent when polyclonal rather than monoclonal antibodies were used. When measured against DNA probing, the immunoblot had a sensitivity of 88% and a specificity of 89% for the identification of STX-positive colonies. Occasional false-positive results with colony immunoblotting have also been reported by Milley and Sekla (216). Immunoblot techniques have the additional disadvantage of having to grow colonies on special media in order to optimize the production and/or release of Stx (129, 216).

Immunomagnetic separation (IMS) techniques have been developed to assist in the isolation of STEC (principally O157) from low-abundance specimens. The procedure involves coating magnetic beads with anti-LPS antibody and mixing them with broth cultures or suspensions of feces or suspect food homogenates. The beads and bound bacteria are then trapped in a magnetic field, the unbound suspension is decanted, and the beads are washed. After additional binding and washing cycles, the beads are plated and the resultant colonies are tested for reactivity with the appropriate O antiserum and, more importantly, for STX production. Several studies have examined the utility of a commercial magnetic bead reagent (Dynabeads anti-E. coli O157 [Dynal, Oslo, Norway]) for the isolation of E. coli O157. IMS was 100 times more sensitive than direct culture on either cefixime-rhamnose SMAC or CT-SMAC for isolation of E. coli O157 from bovine feces (66). It was also more effective than either plate medium for isolation from human fecal samples (64). As might be expected, the enrichment procedure is not 100% efficient, and a number of sorbitol-nonfermenting, non-O157 bacterial strains (of several species) also adhered to the beads. The most comprehensive investigation of the efficacy of IMS for isolation of E. coli O157 is that of Karch et al. (151). IMS allowed the detection of O157 strains at 10² CFU per g of feces in the presence of 10⁷ coliform background bacteria. When tested on fecal samples from HUS patients, IMS followed by culture on CT-SMAC resulted in the isolation of E. coli O157 in 18 of 20 samples, compared with only 12 by colony hybridization and 7 by direct culture on CT-SMAC. The principal drawback of IMS is, of course, its serogroup specificity. However, it is an extremely valuable enrichment technique, particularly in circumstances where STEC strains of known serogroups are being deliberately targeted, for example when testing foods suspected of being the source of an outbreak. During an outbreak of HUS, we isolated STEC strains belonging to serotype O111:H⁻ from the feces of 13 HUS patients and from cultures of two samples of fermented sausage suspected of being the source of the outbreak (260). Cultures from an additional five HUS patients and three sausage samples were weakly positive by PCR for both STX and O111-specific sequences, suggesting the presence of small numbers of the causative organism. However, attempts to isolate O111:H⁻ STEC by colony hybridization were unsuccessful even after up to 1,000 colonies from each sample were tested. Magnetic beads were therefore coated with anti-O111 LPS and used to enrich the cultures by IMS, which enabled the isolation of the causative O111:H⁻ STEC from three of the previously negative fecal cultures and two more of the sausage cultures (260).

Methods for the detection of STEC in foods have focused upon serogroup O157, because this was the cause of all reported major outbreaks until the early 1990s. The infectious dose of the O157:H7 STEC strain responsible for the large outbreak in the western United States in 1993 has been estimated to be of the order of 10 to 100 CFU (114).E. coli O157:H7 has justifiably been declared an adulterant by regulatory authorities in the United States, and as such there is a zero tolerance for this particular STEC organism in foods.

However, food-borne outbreaks of Stx-related disease (including HUS) have now been unequivocally attributed to STEC strains belonging to other serogroups, including O111 (57, 260) and O104 (63), and to Citrobacter freundii(348). Furthermore, outbreaks of STEC disease are not necessarily caused by contamination with a single “outbreak” strain and may indeed be microbiologically complex (260). There is also no reason to believe that sporadic cases of STEC disease, which have been associated with a vast array of STEC serogroups, are not principally food-borne infections. Both the food industry and regulatory authorities need to be cognizant of the potential hazards posed by all these bacteria, particularly as contaminants of foods destined for human consumption without further cooking or processing. When such foods are being analyzed for the presence of STEC, it is essential to use the most sensitive methods. Analysis of the food source responsible for the outbreak of HUS in South Australia in 1995 referred to above suggested that there may have been as little as 1 O111:H⁻ STEC organism per 10 g, implying that the infectious dose for a susceptible child may be of the order of 1 to 10 organisms. Of note, the food source was identified only by PCR analysis of enrichment cultures.

There is, however, a problem with interpretation of the significance of the presence of any STX-producing organism in a food. While any such organism is likely to be of clinical significance when detected in feces of symptomatic patients, foods may contain STEC strains that have yet to be associated with human disease. The properties which distinguish these strains from human-virulent STEC are not completely understood. Clearly, STEC strains which are positive foreaeA (and perhaps also EHEC hlyA) should be regarded as highly significant, but a proportion of serious human infections, including HUS, are caused by STEC strains which areeaeA and/or EHEC hlyA negative. Until definitive criteria are established, we propose that all STEC-contaminated foods be considered potentially hazardous.

Antibiotic therapy might be expected to satisfy all three of the above goals. However, doubts have been raised as a consequence of retrospective studies of its efficacy in preventing the progression of STEC infection from diarrhea or bloody diarrhea to HUS. Such analyses have been compounded by variations in the types of antibiotics used, the timing of commencement of therapy in relation to onset of symptoms, and the possibility that the severity of disease may have influenced the decision to implement therapy. Nevertheless, the bulk of these studies suggested either that there was no significant benefit associated with administration of antibiotics or that therapy (either during or preceding infection) actually increased the risk of developing HUS (59, 70, 264, 282). However, in one study, HUS patients who had been given antibiotics during the diarrheal prodrome had milder illness (206). Examination of antibiotic use in two recent large O157:H7 STEC outbreaks in Scotland and Japan have also produced conflicting findings. Stewart et al. (323) found a significant association between prior antibiotic usage and subsequent development of HUS. On the other hand, Takeda et al. (333) found that the proportion of patients who progressed from bloody diarrhea to HUS was significantly lower when antibiotics had been administered within 3 days of the onset of symptoms, compared with untreated patients or those given antibiotics later in the course of infection. Very few prospective studies have been performed, but Proulx et al. (271) found that administration of trimethoprim-sulfamethoxazole to patients infected with O157 STEC (albeit late in the course of infection) did not prevent progression to HUS.

Apart from the lack of unequivocal evidence for clinical benefit, there are theoretical arguments against the use of antibiotics. First, although STX is extracellular, much of the toxin remains associated with the STEC cell surface. Thus, antibiotics which result in cell lysis might actually increase the amount of free STX in the gut lumen available for systemic absorption. Moreover, in vitro studies have shown that treatment of O157:H7 STEC with subinhibitory concentrations of antibiotics results in a significant increase (up to 50-fold) in the amount of free STX in the culture medium (364, 376). The effect was most pronounced with antibiotics such as trimethoprim-sulfamethoxazole and ciprofloxacin, which interfere with bacterial DNA synthesis, and correlated with increased induction of toxin-converting bacteriophages (376). Second, Cordovez et al. (75) noted a high rate of antibiotic resistance amongst STEC, and so empirical treatment with an inappropriate drug might confer a selective advantage on the STEC over other members of the gut flora and cause overgrowth. The same risk/benefit considerations are also relevant when considering whether to administer antibiotics either to asymptomatic STEC carriers to limit the spread of infection or to uninfected close contacts of patients with proven infection to prevent acquisition. Indeed, the case for prophylaxis is weakened by reports of patients becoming infected with O157:H7 STEC while undergoing therapy for an unrelated condition with an antibiotic to which the STEC was sensitive (334). More extensive randomized controlled trials are required to determine whether there is a role for antibiotic prophylaxis or therapy in STEC disease.

There are also sound reasons for not administering antimotility agents to patients with STEC diarrheal disease, since these would be expected to impede the elimination of STEC from the gut and thereby extend the exposure to STX. Indeed, retrospective analyses have shown that administration of these agents to patients with O157:H7 infection extended the duration of bloody diarrhea and increased the risk of developing HUS and central nervous system lesions (30, 69). At present, the risks or benefits of administration of other antidiarrheal agents such as kaolin or bismuth are not known.

As discussed above, STX is principally responsible for the clinical manifestations of STEC infection, particularly the serious systemic complications. Thus, in vivo binding or neutralization of STX is an attractive therapeutic alternative, particularly in view of the possible risks associated with the use of antibiotics. Such strategies have the potential to limit the severity or duration of disease but would not, of course, be expected to reduce STEC transmission. Therapeutic strategies exploiting the high degree of specificity and the strength of the interaction between STX and its glycolipid receptor are currently being developed. One particularly promising agent is Synsorb-Pk (Chembiomed, Edmonton, Canada), which consists of the oligosaccharide component of Gb₃(Galα[1→4]Galβ[1→4]Glc-) covalently linked via an 8-carbon spacer to silica particles derived from diatomaceous earth (11). Synsorb-Pk was capable of binding and neutralizing Stx1 and Stx2 in polymyxin B extracts of STEC cultures and in feces from patients with HC and/or HUS. A phase I clinical trial did not detect any adverse effects associated with oral administration, and Synsorb-Pk retained its STX binding capacity even after passage through the human gastrointestinal tract (13). Armstrong et al. (13) have proposed that if treatment with this agent is started early enough in the course of STEC diarrhea, the major portion of free STX in the gut lumen will be bound, thereby limiting further systemic absorption and subsequent development of HUS. Compounds capable of binding STX in the gut might also play a role as an adjunct to antibiotic therapy. Preliminary studies with streptomycin-treated mice have shown that while oral Synsorb-Pk did not protect the mice from oral challenge with highly virulent STEC, it did delay death by 1 day (286). Randomized, double-blind trials of Synsorb-Pk versus placebo, aimed at determining the clinical efficacy of Synsorb-Pk in humans with STEC infection, are in progress.

A likely limitation of Synsorb-Pk or other orally administered agents is that they would not be expected to be active against Stx that has already been absorbed from the gut. Although absorption of toxin into the blood has never been directly demonstrated in humans or animals, it is inferred by the generation of vascular lesions at remote sites. Thus, parenterally administered agents may have advantages, particularly since the total amount of absorbed toxin is only a minute fraction of that present in the gut lumen. Recent studies indicate that water-soluble receptor analogs may be suitable for this purpose. Boyd et al. (46) synthesized a galabiosyl analog dimerized via the meta position of benzene, which is capable of blocking both Gb₃ binding sites on the Stx1 B subunit. The analog also protected Vero cells against Stx1 cytotoxicity and to a lesser extent against Stx2.

As described above, parenteral administration of monoclonal anti-Stx2 antibody protects mice from fatal renal damage following oral challenge with STEC. Clearly, murine Igs are not suitable for parenteral administration to humans. However, anti-Stx1 and anti-Stx2 monoclonal antibodies have now been “humanized” by cloning variable-region cDNA from the mouse hybridoma cell lines into a mammalian expression vector, such that chimeric mouse variable region/human IgG1/kappa antibodies are produced (92). In vitro characterization of these antibodies is in progress. An alternative to the humanized monoclonal antibodies that could be considered is hyperimmune human Ig from donors who have been vaccinated with STX toxoid, once such antigens become licensed for use.

A common feature of all these proposed therapeutic agents is that efficacy is likely to depend upon early commencement of treatment, i.e., before significant amounts of STX have been absorbed from the gut. Indeed, preliminary analysis of clinical trials of Synsorb-Pk suggests that therapy may have to begin within 3 days of onset of symptoms (12). This underscores the need for rapid and sensitive diagnostic procedures. However, such agents may also be valuable in the outbreak setting, where persons are known to have been exposed to an implicated vehicle, or in settings such as child care centers or nursing homes, where there is a high risk of secondary transmission from a proven sporadic or index case. In these circumstances, a high proportion of the at-risk population could receive treatment even before any symptoms become apparent and thereby gain maximum benefit from the therapeutic agent.

Vaccines based on STX are likely to be effective in preventing HUS in persons infected with STEC and may also have a significant impact on the severity of gastrointestinal symptoms. This confidence is underpinned by the results of studies of the protective efficacy of vaccines based on Stx2e for protection of pigs from edema disease. MacLeod and Gyles (203) demonstrated that immunization with Stx2e toxoided with glutaraldehyde protected piglets from intravenous challenge with an otherwise lethal dose of purified toxin. Gordon et al. (112) found that Stx2e toxoided with formaldehyde was immunogenic in piglets but that there appeared to be residual in vivo toxicity. However, adverse reactions were not observed when piglets were immunized with an immunogenic derivative of Stx2e with a defined amino acid substitution (Glu167→Gln) in the active-site region of the A subunit. Immunization with this “genetic toxoid” has recently been shown to protect piglets from edema disease after oral challenge with a Stx2e-producing strain of E. coli (45).

Clark et al. (71) described a derivative of Stx1 with a Phe30→Ala substitution in the B subunit, which exhibits reduced binding to Gb₃ and is 10⁵-fold less cytotoxic for Vero cells than is native Stx1. This potential vaccine antigen has recently been shown to be highly immunogenic in rabbits, but marked systemic toxicity was observed when larger doses of antigen were administered, indicating that the Phe30→Ala substitution had a much smaller effect in vivo than in vitro (22). The authors suggested that differences between lipid moieties of Gb₃from Vero cells and rabbit tissues might account for their finding and concluded that additional mutations would be required to reduce the toxicity to acceptable levels for use in human vaccines. Interestingly, a similar discrepancy between effects on Vero cell cytotoxicity and LD₅₀ was previously observed for amino acid substitutions in a similar region of the Stx2 B subunit (191).

An alternative means of circumventing the problem of residual toxicity is to use just the B subunit as an immunogen. Several B subunit-specific Stx/Stx1-neutralizing monoclonal antibodies have been prepared (326), and neutralizing polyclonal antibodies were isolated from hyperimmune sera by affinity chromatography with immobilized B subunit (85). Boyd et al. (48) demonstrated that immunization of rabbits with purified Stx1 B subunit confers protection against challenge with an otherwise lethal intravenous dose of purified Stx1.

Recently, much interest has centered upon the vaccine potential of live-attenuated enteric bacteria expressing foreign antigens. These vaccine strains are capable of limited invasion of the gut mucosa, and they deliver the foreign antigens directly to Peyer’s patches, resulting in strong humoral and secretory immune responses. Su et al. (328) described the construction of fusions of the completestx B-subunit gene, or N-terminal segments thereof, with the gene for the E. coli OMP LamB. The stx sequences were inserted at a point in the lamB gene which encoded a surface-exposed loop of the protein. These constructs directed the expression of the hybrid LamB/StxB protein on the surface of E. coli K-12 and wild-type Salmonella typhimurium. However, when attenuated S. typhimurium aroA vaccine strains were used, the recombinant antigen was confined to the cytoplasm, presumably a consequence of membrane export defects in these strains. In spite of this, oral immunization of mice with these latter constructs resulted in the production of significant levels of Stx B subunit-specific antibodies, both in serum and in intestinal washings. Acheson et al. (3) have also expressedstxB₁ in a Vibrio cholerae vaccine strain. Sera from rabbits immunized orally with this strain contained significant levels of Stx1 B-subunit-specific antibodies, which were capable of neutralizing Stx1 holotoxin in vitro. There was also evidence of protection against enterotoxic effects of purified Stx1 in ileal loops constructed from immunized animals.

It is presumed that human anti-STX vaccines will need to be bivalent to provide protection against both Stx1- and Stx2-producing STEC strains. Polyclonal antisera raised against Stx/Stx1 do not neutralize the verocytotoxicity of Stx2 and vice versa (308, 327). Interestingly, however, Bielaszewska et al. (38) recently reported evidence for cross-neutralization in vivo in a rabbit model of STX toxemia. As expected, sera from rabbits immunized with either Stx1 or Stx2 toxoids did not neutralize the heterologous toxin but there was a degree of cross-reaction as judged by ELISA. Moreover, when rabbits were challenged with homologous or heterologous¹²⁵I-labelled holotoxins, both toxins were cleared by the liver and spleen rather than being directed to the usual Gb₃-containing target tissues (the brain and the gastrointestinal tract), which occurs in unimmunized animals. This cross-protective property appeared to be mediated by the A subunit. Thus, inclusion of both Stx1 and Stx2 antigens may be more important for B-subunit vaccine formulations.

While vaccines based on STX would be expected to prevent the serious systemic complications of STEC disease, they would not be expected to prevent colonization of the gut by STEC and so would have little effect on the transmission of disease in the community. Vaccines directed against colonization factors would be expected to be more effective, but at present our knowledge of these factors is incomplete. STEC intimin (EaeA) may be an appropriate vaccination target, at least for the majority of human pathogenic STEC strains which produce it. Indeed, Butterton et al. (53) constructed a V. choleraevaccine strain carrying a chromosomally integrated copy ofeaeA which expressed intimin on its surface. Oral immunization of rabbits with this strain elicited a serum antibody response to intimin. However, the C-terminal portion of intimin exhibits marked amino acid sequence heterogeneity among different STEC strains (195), and it remains to be seen whether this will limit cross-protection.

Vaccines directed against LPS may also be effective against STEC, since experience with other enteric pathogens suggests that LPS-specific serum IgG is likely to leak into the gut lumen in quantities sufficient to block colonization. Phase I clinical trials of O157 O antigen conjugated to a protein carrier (to enhance immunogenicity) have shown that such vaccines are safe and immunogenic (331). It should be emphasized, however, that LPS-based vaccines will provide only serotype-specific protection, and so the overall efficacy will be limited by the prevalence of O157 STEC in a given region. It may not be economically viable to include more than three or four additional serotypes in an LPS conjugate vaccine formulation. One way of partially addressing this problem is to broaden coverage by using STX B subunits as the carrier for the O157 O antigen; such a conjugate has also recently been constructed (331).