In most Stx1-producing EHEC strains, stx1 genes are transcribed from p_stx1 which is induced under low iron concentrations by release of Fur-mediated repression [8]. As transcription from p_stx1 does not involve expression of the late phage lysis genes, induction of stx1 expression does not result in bacterial lysis and Stx1 remains within the bacterial cell where it mainly localises in the periplasm [9,10,11].

In contrast to stx1, transcription of stx2 genes is highly dependent on induction of the phage lytic cycle as it is mainly governed by the late phage promoter p_R’ [12]. As a result, stx2 expression is induced by DNA-damaging agents, such as UV radiation, mitomycin C and certain antibiotics, the use of which is contraindicated during EHEC infection because of enhanced Stx production [9,13,14]. DNA damage results in a bacterial SOS response and activation of RecA which in turn leads to autocleavage of the phage repressor cI and initiates sequential transcription of early and late phage genes. Early genes are responsible for replication of the phage genome thereby increasing the number of stx2 gene copies within the bacterial cell. Expression of late genes leads to production of Stx, bacterial lysis and toxin release [15,16]. The strong association between stx2 expression and the phage lytic cycle likely explains why Stx2 is mainly found in supernatants rather than lysates of bacterial cultures [6,10,17]. Increased release of Stx2 compared with Stx1 might contribute to the high HUS risk that has been associated with Stx2-producing EHEC [18,19]: as EHEC are non-invasive and do not cross the intestinal epithelium and cause systemic infection, it is likely that HUS is caused by Stx released into the gut lumen which subsequently gets access to the bloodstream.

So far, no bacterial secretion system has been identified for Stx, and phage-induced bacterial lysis is assumed to be the main mechanism of toxin release into the environment. However, outer membrane vesicles (OMV), which are naturally shed from the outer membrane of gram-negative bacteria, are gaining increased recognition as delivery vehicles for bacterial virulence factors and toxins (reviewed by [20,21]). Interestingly, it has been demonstrated that OMV from EHEC broth cultures contain Stx1 and Stx2 [22,23]. OMV are formed during bacterial infection and can be internalised by eukaryotic cells as has been reported for toxin delivery by enterotoxigenic E. coli [24]. It remains to be investigated whether Stx release in OMV plays a relevant role in Stx delivery across intestinal epithelium and development of HUS.

Regulation of Stx expression has been investigated extensively in bacterial broth cultures but it remains unknown how toxin production is regulated in the complex environment of the human gut. Prominent features of the intestinal milieu include low oxygen levels, the presence of a mucus layer with commensal microflora, and polarised intestinal epithelial cells (IEC).

Shiga toxins have an AB₅ structure with one enzymatically active A subunit non-covalently linked to a pentamer of B subunits responsible for Gb3 binding [43]. After binding of Stx to lipid raft-associated Gb3, Stx is internalised into early endosomes and then travels further to the trans-Golgi network (Figure 1). During this early stage, the Stx A subunit is cleaved by furin into an active A1 subunit and an A2 fragment associated with the B pentamer [44]. StxA1 still remains linked to StxA2-B via a disulfide bond and the activated toxin complex is transported via the retrograde pathway to the Golgi apparatus, endoplasmic reticulum (ER) and nuclear membrane [45]. After reduction of the disulfide bond in the ER, StxA1 is retro-translocated into the cytoplasm where it binds to the large ribosomal subunit and inhibits protein synthesis by cleaving off a single adenine residue from the 28S rRNA [46]. While inhibition of protein synthesis in itself does not kill the host cell, it triggers a ribotoxic stress response which ultimately leads to apoptosis [47].

Retrograde Stx transport to the ER and nucleus has been linked to cytotoxicity as toxin trafficking is different in Stx-resistant cells (Figure 1). In macrophages and dendritic cells, Gb3 is not associated with lipid rafts and Stx is transported along the endosomal/lysosomal pathway and ultimately degraded in lysosomes [38]. A similar mechanism of Stx inactivation has also been demonstrated for bovine lower crypt IEC which bind Stx1 but are resistant to cytotoxicity [48]. Cattle are a natural reservoir of EHEC which harmlessly colonise the distal colon [49]. Resistance of bovine intestinal epithelium to Stx and the absence of Gb3 on endothelium most likely contribute to the asymptomatic status of cattle during EHEC infection [50].

Apart from lysosomal degradation, Stx transport in resistant cells such as epidermoid carcinoma, astrocytoma, and ovarian carcinoma cells has been reported to be arrested in the Golgi apparatus. Cytotoxicity can be induced by addition of the differentiation agent butyrate which restores retrograde transport to the ER/nucleus [42,51]. Both studies suggest that this effect is mediated via alteration of Gb3 fatty acid chain length.

Human intestinal epithelium represents the first point of contact of released Stx with the host and furthermore acts as a barrier by preventing toxin access to the systemic circulation. Interestingly, normal human intestinal epithelium does not express Gb3 or any other Stx receptor [52,53,54,55]. In contrast, Gb3 expression on colonic epithelium is associated with malignancy and metastasis, and the potential use of Stx in cancer therapy is currently under extensive investigation [56,57,58].

Accordingly, most commonly used human colon carcinoma cell lines (Caco-2, HT-29, and HCT-8) express Gb3 on the cell surface and are susceptible to Stx to a variable degree [54,59,60]. Treatment with butyrate has been reported to promote Gb3 expression and Stx sensitivity in Caco-2A and HT-29 cells, indicating that Gb3 expression is regulated by cell maturation [59]. In contrast to these villus-like cell lines, crypt-like colon carcinoma-derived T84 cells do virtually not express Gb3 in the fully differentiated state and are resistant to Stx-induced inhibition of protein synthesis and apoptosis [54,59]. Surprisingly, Stx1 and Stx2 are internalised by T84 cells despite the lack of Gb3 and activated toxin is transported to the ER [54,61]. It still remains to be determined why retrograde toxin transport does not result in T84 cytotoxicity, but it could be speculated that StxA1 release from the ER into the cytoplasm is disrupted. This is supported by recent studies which suggest that retro-translocation is the rate-limiting step in Stx cytotoxicity as only 4% of StxA1 is released into the cytoplasm [62].

Stx interaction with normal (as opposed to cancerous) IEC has been studied by in vitro organ culture of human intestinal mucosal biopsies, and results have indicated that Stx2 but not Stx1 causes damage to crypt epithelial cells [54]. As Stx access was not restricted to the mucosal biopsy surface, no conclusions could be drawn as to whether damage was caused by direct Stx-epithelial interaction or indirectly via effects on cells in the underlying lamina propria [54]. Interestingly, recent studies have demonstrated a direct effect of Stx1 on IEC by enhancing galectin-3 secretion and thereby causing mistargeting of absorptive brush border proteins [63]. Notably, this effect could have an implication in EHEC-induced diarrhoea. In addition, it has been shown that Paneth cells, which are key players in the small intestinal innate immune defense, express Gb3 and bind Stx1 and Stx2, and the functional role of this interaction needs to be determined [64].

Earlier studies have demonstrated translocation of purified Stx1 and Stx2 across polarised monolayers of Stx-resistant Caco-2A and T84 colon carcinoma cells. Neither cell viability nor epithelial barrier function was compromised indicating a transcellular pathway of translocation for both Stx types [70,71]. However, differences in translocation rates and directionality, effects of cellular drugs, and competition experiments indicate that Stx1 and Stx2 transcytose IEC by different pathways [71,72]. Transcytosis of Stx1 was further confirmed by immunoelectron microscopy which detected intracellular toxin in association with endosomes, Golgi, ER, and the nuclear membrane, but showed no association with tight junctions [61]. Stx localisation in these cellular compartments is in agreement with retrograde transport in colon carcinoma cells and could be part of the translocation process. However, use of brefeldin A which disrupts the Golgi apparatus, showed no effect on Stx1 or Stx2 translocation indicating that transcytosis is independent of retrograde transport [71]. This is in agreement with more recent studies showing that Stx1 translocation by T84 cells is an unsaturable, Gb3-independent process which involves actin turnover [73]. These criteria apply to the process of macropinocytosis, and it has recently been demonstrated that induced Src kinase-mediated macropinocytosis enhances Stx1 uptake and translocation. Interestingly, results from this study have also shown that EHEC infection can trigger Src kinase activation, thereby suggesting a promoting role of infection in toxin translocation [74].

EHEC infection causes redistribution of intercellular tight junction proteins such as occludin and ZO-1, and thereby leads to decreased epithelial barrier function. These changes are initiated by type III secretion of the EHEC effector proteins EspF and EspF_U into the host cell which trigger subsequent signal transduction events [75,76]. EHEC-mediated disruption of tight junctions may provide a means for Stx to cross the epithelial barrier and this hypothesis has been tested by infecting polarised T84 cells with Stx-negative EHEC to decrease barrier function and subsequently applying Stx1. No significant difference in Stx translocation rate was observed compared with cells treated with purified toxin alone indicating that tight junctions of infected cells still retain their sieving function for larger molecules such as Stx [61].

Another mechanism whereby disruption of tight junctions is likely to occur during infection is by transmigrating neutrophils. Recruitment of neutrophils to infected mucosa is a prominent feature observed during EHEC diarrhoea, and elevated neutrophil levels in stools indicate neutrophil transmigration from the mucosa into the gut lumen [77,78,79,80]. This event has been simulated in vitro by co-incubating polarised T84 monolayers with neutrophils on the basal side and inducing neutrophil transmigration by adding a chemoattractant to the apical side. During the process of transmigration, epithelial barrier function was diminished and apical to basal Stx1 and Stx2 translocation was enhanced. Importantly, the authors demonstrated induced neutrophil transmigration by EHEC infection indicating that this process might be relevant in the in vivo situation [72].

Previous reports have demonstrated that Gb3 expression and Stx1 and Stx2 cytotoxicity on Gb3-positive endothelial cells can be enhanced by treatment with butyrate, pro-inflammatory cytokines or bacterial lipopolysaccharide (LPS) [81,82,83,84]. Similarly, butyrate treatment of Gb3-positive colon carcinoma cells results in increased Stx1 susceptibility due to enhanced Gb3 expression, but this does not affect Gb3-negative T84 cells [59]. The effect of bacterial LPS on Gb3 expression on human IEC has not been tested yet, but it could be speculated that EHEC infection induces epithelial Gb3 expression (by LPS or other bacterial factors) and thereby leads to toxin uptake, retrograde transport and cell death. In addition, EHEC infection results in severe inflammation of the intestinal mucosa which might lead to Gb3 induction on IEC. This hypothesis has been tested by comparing Stx1/Stx2 binding to normal and inflamed intestinal mucosa. No difference in Stx binding was detected between the two groups and no expression of epithelial Gb3 was observed in inflamed samples [64]. However, Gb3 expression and Stx binding to Paneth cells was observed in 30–60% of samples and it remains to be investigated whether Paneth cells might provide an entry site for Stx into underlying tissues [64].

In vitro organ culture of human intestinal biopsies has shown that EHEC adheres predominantly to the follicle-associated epithelium of small intestinal Peyer’s patches [33]. This area is particularly rich in M (microfold) cells which are specialised in uptake and translocation of antigens from the intestinal lumen to underlying antigen-presenting cells. Because of their transcytotic activity, M cells are often exploited by intestinal pathogens for invasion into underlying tissues and have also been implicated in translocation of bacterial toxins [85,86]. It is tempting to speculate that EHEC and/or Stx can breach the epithelial barrier via M cells and future research in this area is needed to test this possibility.