Enteropathogenic and enterohaemorrhagic Escherichia coli : more subversive elements


Gad Frankel. E-mail g.frankel@ic.ac.uk; Tel. (0171) 594 5253; Fax (0171) 594 5255.


Enteropathogenic (EPEC) and enterohaemorrhagic Escherichia coli (EHEC) constitute a significant risk to human health worldwide. Both pathogens colonize the intestinal mucosa and, by subverting intestinal epithelial cell function, produce a characteristic histopathological feature known as the ‘attaching and effacing’ (A/E) lesion. Although EPEC was the first E. coli to be associated with human disease in the 1940s and 1950s, it was not until the late 1980s and early 1990s that the mechanisms and bacterial gene products used to induce this complex brush border membrane lesion and diarrhoeal disease started to be unravelled. During the past few months, there has been a burst of new data that have revolutionized some basic concepts of the molecular basis of bacterial pathogenesis in general and EPEC pathogenesis in particular. Major breakthroughs and developments in the genetic basis of A/E lesion formation, signal transduction, protein translocation, host cell receptors and intestinal colonization are highlighted in this review.


An emerging theme in the interaction between microbial pathogens and mammalian species is the ability of bacteria to subvert host cell signal transduction pathways and exploit host cytoskeletal/membrane components to allow colonization and invasion of the target host. A particularly good example is demonstrated by enteropathogenic Escherichia coli (EPEC), which has the ability to induce a characteristic ‘attaching and effacing’ (A/E) histopathology on gut enterocytes characterized by localized destruction of brush border microvilli, intimate bacterial adhesion and gross cytoskeletal reorganization (reviewed by Donnenberg et al., 1997a) (Fig. 1). A/E lesion formation is essential for full EPEC pathogenicity, and a similar histopathology has been associated with several other bacterial mucosal pathogens, including the closely related enterohaemorrhagic E. coli (EHEC).

Figure 1.

. A. Electron micrographs illustrating key features of A/E histopathology including intimate bacterial attachment and pedestal formation. B. Localized destruction of microvilli by cytoskeletal breakdown and membrane vesiculation. C. Gross localized perturbation of brush border architecture. Original magnification: A and B, × 45 000; C, × 8000. (A and B reproduced with permission from Knutton et al . (1987) Infect Immun 55: 69–77).

EPEC refers to certain serotypes of E. coli that were first incriminated in epidemiological studies in the 1940s and 1950s as causes of epidemic and sporadic infantile diarrhoea. EHEC, on the other hand, were first identified as a human pathogen in 1982 in the USA. Whereas EHEC is regarded as an emerging zoonotic pathogen that can cause acute gastroenteritis and haemorrhagic colitis and produce severe/fatal renal and neurological complications as a result of the translocation of Shiga toxins (Stx 1 and Stx 2) across the gut, EPEC, the prototype A/E organism, is an established aetiological agent of human diarrhoea and remains an important cause of infant mortality in developing countries (Nataro and Kaper, 1998). Population genetic surveys using sequence variation in housekeeping genes show that EPEC and EHEC strains can be divided into two major groups of related clones, designated EPEC clones 1 and 2 and EHEC clones 1 and 2. Within each group, a variety of O antigens are present, while the somatic flagellar (H) antigens are conserved; EPEC clone 1 typically expresses flagellar antigen H6, whereas EPEC clone 2 expresses flagellar antigen H2 or is non-motile. EHEC clone 1 includes the O157:H7 serotype, whereas EHEC clone 2 contains other Stx-producing E. coli serogroups (including O26 and O111). Interestingly, EHEC strains belonging to clone 2 are closely related to the typical EPEC strains, while EHEC O157:H7 is closer to the atypical EPEC O55:H7 (reviewed by Kaper et al., 1998).

Bundle-forming pili and EPEC colonization of the intestinal mucosa

EPEC colonize the small intestinal mucosa and produce the characteristic A/E histopathology (Fig. 1). Previous studies using cultured epithelial cells of non-intestinal origin have implicated bundle-forming pilus (BFP) (Giron et al., 1991) as an initial EPEC attachment factor. BFP is a type IV pilus encoded by a cluster of 14 genes that encompass 11.5 kb of the EPEC adherence factor (EAF) plasmid (reviewed by Donnenberg et al., 1997b). In addition to these plasmid-encoded genes, the formation of a functional BFP structure requires the presence of the chromosomal dsbA gene, which encodes a periplasmic oxidoreductase necessary for disulphide bond formation. The bfpA gene encodes the major pilin subunit, and a number of other bfp genes encode products that are homologous to proteins involved in the biogenesis of other type IV pili. Expression of BFP is regulated by growth phase, temperature, calcium and ammonium ions, and transcription of bfpA is regulated by the bfpTVW genes (Tobe et al., 1996), which are in fact allelic with the perABC genes described previously by Gomez-Duarte and Kaper (1995) (see below).

Recent evidence using in vitro organ culture (IVOC) of paediatric small intestine, a model that resembles the in vivo situation more closely, indicates that adhesin(s) other than BFP initiate colonization of the mucosal surface and that BFP allows the formation of three-dimensional bacterial microcolonies via interbacterial interactions (Hicks et al., 1998). In contrast to wild-type strains, this study demonstrated that EPEC lacking the EAF plasmid or a functional bfpA gene still infect the mucosa and cause A/E lesions but did not form microcolonies. Thus, BFP is not required for initial bacterial attachment to human intestinal mucosa in vitro, and other candidates should be sought. However, BFP remains a potentially important virulence factor, as shown by volunteer challenge studies in which an EAF-positive strain produced more clinical illness than an EAF-negative strain. Indeed, a recent study by Bieber et al. (1998) has shown that a mutant in one of the bfp genes (bfpF ) failed to disperse from microcolonies on infected HEp-2 cells; this phenotype severely impaired the ability of the bfpF mutant to cause diarrhoea in volunteers. One explanation for this phenotype is that BFP-mediated interbacterial interactions may allow detachment of individual bacteria and colonization of other mucosal sites, possibly contributing to the spread of infection along the gut. Although BFP is an established virulence factor in EPEC, it should be noted that BFP is not present in EHEC. The absence of BFP may reflect the evolutionary origin of EHEC strains from atypical EAF-minus EPEC, and/or EHEC may possess an as yet undescribed pilus that also functions in colonization.

The LEE pathogenicity island of EPEC and EHEC

The genes encoding the A/E histopathology are contained on a 35.6 kb pathogenicity island called the locus of enterocyte effacement (LEE) (Fig. 2). The LEE was first described in EPEC E2348/69 by McDaniel et al. (1995) but is also present in EHEC, Hafnia alvei, Citrobacter rodentium and other A/E E. coli that are pathogenic in a variety of animal species. The entire LEE of E2348/69 has been cloned on a single plasmid which, when present in E. coli K-12, is sufficient to produce the A/E phenotype (McDaniel and Kaper, 1997). The G + C content of the LEE (38.3%) is significantly lower than that of the E. coli chromosome (50.8%), suggesting that the LEE arose by horizontal gene transfer of this pathogenicity island from another species.

Figure 2.

. The locus of enterocyte effacement (LEE) of E. coli E2348/69 showing the arrangement of ORFs (large arrows) and their nomenclature as esp , esc , sep , ces , eae and tir . Remaining ORFs are designated orf or rorf , depending on the direction of transcription relative to eae , and those ORFs homologous to previously described genes are shaded. Also noted (as thin arrows) are putative polycistronic operons, as determined by the close spacing of co-directionally transcribed genes. Forty-one ORFs are shared by EPEC strain E2348/69 and EHEC O157:H7 strain EDL933 ( Perna et al., 1998 ), and the percentage of predicted amino acid identity between the shared ORFs is indicated. ERIC, enterobacterial repetitive intergenic consensus sequences. Further details in the text.

The LEE from EPEC E2348/69 contains 41 open reading frames (ORFs) of more than 50 amino acids (Elliott et al., 1998). These genes are organized into three major regions with known functions. The middle region contains the genes eae (encoding the adhesin intimin) and tir (encoding a receptor for intimin that is translocated into host cells via a type III secretion system), the products of which are involved in epithelial cell adherence, as discussed below. Upstream of eae and tir are genes encoding the type III secretion system (Elliott et al., 1998). Nine of these genes are homologous to ysc genes encoding type III secretion system components in Yersinia, Shigella, Salmonella and other pathogens possessing this important secretion mechanism.

The third major region of the LEE, located downstream of eae, encodes several proteins that are secreted via the type III secretion system; most prominent of these proteins are EspA, EspB and EspD. Another secreted protein, EspF, is encoded downstream of espABD at the far end of the LEE (McNamara and Donnenberg, 1998). The 21 kDa EspF plays an unknown role in EPEC pathogenesis, as mutation of the espF does not affect the ability to form A/E lesions or to induce host cell signalling events. One protein encoded in the LEE, cesD, has been shown to act as a secretion chaperone for the EspD protein and is also required for complete extracellular secretion of EspB (Wainwright and Kaper, 1998).

The complete sequence for the LEE of EHEC O157:H7 strain 933 has been determined recently (Perna et al., 1998). The 41 genes found in the EPEC E2348/69 LEE are also present in the O157:H7 LEE in the same order, and the average nucleotide identity between these strains is 93.9% across the 41 shared genes. Interestingly, the rate of divergence is heterogeneous across the LEE. The esc genes encoding the type III secretion system are highly conserved, sharing 98–100% identity, but other genes are much less conserved. The eae genes encoding intimin only share 87.23% identity, and the secreted proteins are also quite divergent: espB (25.99% difference); espA (15.37%); espD (19.64%); and tir (33.52%) (Perna et al., 1998) (Fig. 2). The limited amount of sequence data for LEE genes from other EPEC and EHEC strains shows a similar pattern of greater divergence in espA and eae genes (see below). The variability in genes encoding proteins that interact directly with the host suggests that these variable proteins may be subject to natural selection for evasion of the host immune system.

The LEE of EPEC E2348/48 is inserted at minute 82 of the E. coli K-12 chromosome, adjacent to the selC locus (McDaniel et al., 1995), which is also the site of insertion of a pathogenicity island in uropathogenic E. coli. The insertion site of the LEE can vary according to the clonal phylogeny of the strain, whereby EPEC or EHEC that were grouped together had the identical LEE insertion site. A second insertion site for the LEE was shown recently to be at minute 94 of the K-12 chromosome at the pheU locus, and there is at least a third, as yet unidentified, insertion site for the LEE (reviewed by Kaper et al., 1998). These results indicate that the LEE has inserted at multiple times during the evolution of the EPEC/EHEC family of pathogens and subsequently acquired additional virulence factors encoded on bacteriophage and on large approximately 90 kb plasmids found in EPEC and EHEC.

Many genes on the EPEC LEE are regulated in trans by a transcriptional activator called Per, which is encoded by a cluster of three ORFs designated perA, perB and perC (Gomez-Duarte and Kaper, 1995). The PerA protein belongs to the AraC family of regulators, but all three ORFs are required for full activation. The Per regulator, which is encoded on the EAF plasmids present in most EPEC strains, was first discovered through its ability to increase expression of the eae gene (Gomez-Duarte and Kaper, 1995). It has been shown recently that operons encoding components of the LEE-encoded type III secretion system, specifically operons containing escR and escJ, were transcriptionally activated by Per (J. Mellies, S. J. Elliott and J. B. Kaper unpublished data). As already indicated, Per also regulates the expression of BFP. Although details are lacking, it is becoming apparent that Per serves as a global regulator affecting transcription of both chromosomal and plasmid-encoded genes (Donnenberg et al., 1997a), thereby allowing EPEC to respond to different environmental conditions and different phases of growth. It should be pointed out that, as EHEC do not possess per gene homologues, the regulation of expression of LEE-encoded proteins in EHEC is at present unknown.

Another potential regulator of A/E lesion formation, BipA, has been reported recently by Farris et al. (1998). BipA is encoded outside the LEE and is a tyrosine-phosphorylated GTPase that appears to represent a new class of virulence regulators, although the mechanisms involved have yet to be defined.

The LEE-encoded type III secretion/translocation system

Several animal and plant pathogens use type III protein secretion systems to secrete key virulence factors, some of which are injected (translocated) directly from the pathogens' cytoplasm into the host cell cytosol (Hueck, 1998). EPEC and EHEC use the LEE type III secretion system to secrete several LEE-encoded EPEC-secreted proteins (Esps) including EspA, EspB and EspD (reviewed by Kaper et al., 1998; Nataro and Kaper, 1998). Secretion of the Esp proteins is essential for signal transduction in host cells and A/E lesion formation. Esp secretion is induced when bacteria are grown in eukaryotic cell culture medium and in response to conditions similar to those found in the gastrointestinal tract. Recent work has provided unexpected new insights into the function of the Esps.

One unexpected recent finding was the observation that EspA is not, as previously thought, a protein involved directly in the activation of host cell signalling, but rather a structural protein and a major component of a large filamentous organelle that is transiently present on the bacterial surface and that interacts with the host cell during the early stage of A/E lesion formation (Fig. 3). EspA filaments, because they form interactions with host cells, may contribute to bacterial adhesion but, more importantly, they appear to be a component of a translocation apparatus and are essential for the translocation to host cells of EspB (Knutton et al., 1998; Wolff et al., 1998) and also of Tir (Kenny et al., 1997). Each approximately 50-nm-diameter and up to 2-μm-long filament is composed of a number of smaller approximately 7-nm-diameter fimbrial-like structures arranged as a cylindrical structure. Thus, EspA filaments appear to have a hollow cylindrical structure that could form a channel through which proteins are delivered into the host cell. EspD, whose function is currently unknown, may also be a component of the EspA filaments, because an EspD mutant was found to secrete only low levels of EspA and produced barely detectable filaments (Knutton et al., 1998). It was reported recently that EspA, in common with many other proteins secreted by type III systems, is predicted to contain a coiled-coil domain (Pallen et al., 1997). This suggests that assembly of the EspA filament might involve coiled–coil interaction between helices of different subunits. However, although widespread in type III secretion systems (Pallen et al., 1997), the importance of coiled-coil domains in protein–protein interactions has been demonstrated experimentally only for YopN (Iriarte et al., 1998).

Figure 3.

. Electron micrographs showing EspA filaments (arrows) expressed on the bacterial surface (A) and forming a bridge between bacteria and epithelial cells during early stages of A/E lesion formation (C and D). The cylindrical EspA filaments are composed of smaller, approximately 7-nm-diameter, fimbrial-like structures (B). Original magnification: A, × 40 000; B, × 80 000; C, × 60 000; D, × 30 000. (Reproduced with permission from Knutton et al . (1998) EMBO J 17: 2166–2176).

EspB is not a component of the EspA filament structure because an antiserum made to this Esp did not stain EspA filaments and an EspB mutant still produced normal EspA filaments. Instead, using immunofluorescence, immunoblotting of fractionated host cells and a calmodulin-dependent adenylate cyclase reporter system, it has been shown that, after bacterial attachment, EspB is translocated into the host cell, where it is distributed in both membrane and cytosol, and that EspA filaments are essential for EspB translocation (Knutton et al., 1998; Wolff et al., 1998). Furthermore, on contact with host cells, it was shown that there is an immediate burst of EspB secretion, suggesting that, in the absence of epithelial cells, Esp secretion by EPEC probably represents basal levels and that, in a manner similar to protein secretion mediated by other type III secretion systems, EspB secretion is cell contact dependent. Translocation of EspB was not dependent on but was strongly enhanced by intimate bacterial attachment (Wolff et al., 1998). As EspB is translocated to the host cell membrane and has weak structural similarity to YopB of Yersinia (a protein involved in forming the translocation channel in the host cell membrane) (Hakansson et al., 1996), it is possible that EspB serves a similar function in EPEC; the cytoplasmic localization of a subpopulation of EspB may indicate additional functions.

The structural basis for protein translocation has yet to be fully elucidated for any type III secretion system (Hueck, 1998). EscC is a major component of the EPEC type III secretion system exposed at the bacterial surface (Elliott et al., 1998). YscC, a Yersinia homologue of EscC, has recently been shown to exist as a ring-shaped oligomeric complex in the outer bacterial membrane with an approximately 20-nm-diameter central pore (Koster et al., 1997). Thus, EscC may represent a novel type of channel-forming protein in the EPEC outer membrane. In further support of EspB-generated pores in the host cell membrane, high-resolution fluorescence microscopy of translocated EspB revealed discrete foci of fluorescent staining in the host cell membrane adjacent to adherent bacteria and with a diameter similar to that of the EspA filaments (S. Knutton and G. Frankel, unpublished results). Integrating these data suggests the presence of pore-forming proteins of bacterial origin in both the bacterial outer membrane and the plasma membrane of the infected host cells. However, a structure connecting two such membrane pores to provide a continuous pathway from bacterial cytosol to host cell cytosol has not been identified previously. We hypothesize that EspA filaments provide this missing link that serves to join the bacterial membrane pore formed by EscC to the eukaryotic membrane pore formed by EspB. Of course, it remains to be determined whether these separate components of the proposed EPEC translocon are functionally integrated as suggested but, if such a translocon exists, it could function like a ‘molecular syringe’ that, powered by the type III secretion system EscN ATPase, injects proteins directly from the bacterium into the host cell cytosol (Fig. 4). It is now of interest to know whether the mechanisms responsible for protein translocation in EPEC are common to other pathogens that exploit type III secretion systems. In this regard, it should be noted that EspA and EspB homologues exist in Salmonella, and filamentous structures similar to EspA filaments produced from another type III secreted protein, HrpA, have been described in Pseudomonas syringae (Roine et al., 1997).

Figure 4.

. Model of EPEC/EHEC protein translocation and A/E lesion formation. Cell contact stimulates the expression of LEE-encoded proteins and the assembly of a protein translocation apparatus (translocon). The proposed translocon consists of pores in the bacterial envelope (EscC-generated pore) and in the host cell membrane (EspB-generated pore) with the pores connected by a hollow EspA filament, thereby providing a continuous channel from the bacterial to the host cell cytosol. Energy for protein translocation is thought to be provided by the EscN protein. The translocon is used to translocate Tir(EspE) into the host cell, where it becomes inserted into the host cell membrane; EPEC Tir, but not O157:H7 Tir, becomes phosphorylated on tyrosine residues following translocation. Translocated Tir and/or other yet to be identified effector proteins transduce signals that induce breakdown of the brush border microvillous actin cytoskeleton with consequent vesiculation of the microvillous membrane. Although the mechanisms responsible are not known, localized translocation of effector proteins results in localized cytoskeletal changes. Intimate adhesion and pedestal formation results from the interaction of intimin and Tir and the accumulation of actin (and other cytoskeletal proteins) beneath intimately attached bacteria following microvillous effacement.

Intimate EPEC and EHEC adhesion

The hallmark of EPEC and EHEC adhesion to host cells is a very intimate attachment (Fig. 1). The genes involved in intimate adhesion, eae and tir, are mapped within the LEE upstream of the esp genes. The product of the eae gene, the intimate EPEC adhesin intimin, is homologous to invasins, proteins that promote eukaryotic cell invasion by Yersinia. Studying the intimin family of proteins has shown that, like invasin, their cell binding activity is localized to the C-terminal 280 amino acids (Int280) (Frankel et al., 1994) and that within this domain lies a 76-amino-acid loop formed by a disulphide bridge between two cysteines at positions 862 and 937 (Kelly et al., 1998). This loop is required for intimin-mediated intimate attachment and invasion into cultured mammalian cells. Recent studies have shown that introducing small in frame mutations at the C-terminus of EPEC intimin could reduce intimin-mediated cell invasion dramatically without affecting A/E lesion formation detectably. In particular, deletion of the last amino acid (Lys-939) from the intimin C-terminus segregated intimin-mediated A/E lesion formation from intimin-mediated HEp-2 cell invasion (Frankel et al., 1998).

The expression of intimin, which is regulated by the per locus, is influenced by growth phase and temperature and is maximal during late exponential–early stationary phase at 37°C when the protein can be found distributed uniformly over the bacterial surface. During A/E lesion formation, some surface intimin binds to translocated Tir to produce an irreversible intimate cell attachment. On infected mammalian cells in culture, surface intimin not involved in Tir interaction is downregulated after A/E lesion formation (Knutton et al., 1997). However, reduced intimin expression is not observed using fixed epithelial cells, implying that EPEC may have responded to some change in the host cell. The nature of the signals transduced from infected mammalian cells to the adherent bacteria are not known.

Previously published data from Agin and Wolf (1997) have provided evidence for the existence of at least three immunologically distinct intimin types. More recently, a study using antisera made against Int280 and the polymerase chain reaction (PCR) to investigate antigenic variation and classify the cell binding domain of intimin expressed by A/E lesion-forming bacterial pathogens found the presence of four distinct intimin subtypes: intimin α, intimin β, intimin γ and intimin δ (Adu-Bobie et al., 1998). Importantly, intimin α was specifically expressed by strains that belong to the EPEC clone 1, and intimin β was mainly associated with EPEC and EHEC strains belonging to their respective clone 2. Intimin γ was associated with EHEC O157:H7, EPEC O55:H7 and O55:NM, while intimin δ was expressed only by EPEC O86:H34. Thus, it appears that there is a direct correlation between an intimin type, LEE insertion site and a specific EPEC/EHEC clone.

Antigenic variation within the cell binding domain of intimin raises the question whether this may contribute to tissue tropism exhibited by EPEC (mainly colonizing the small bowel) and EHEC O157 (colonizing the large bowel). Intimin exchange studies have been performed in piglets using wild-type EHEC (expressing intimin γ) or EHEC expressing the EPEC-derived intimin α. In conventional animals (Donnenberg et al., 1993), no differences were seen in the intestinal distribution of the A/E lesions but, in gnotobiotic piglets (Tzipori et al., 1995), EHEC expressing intimin α produced A/E lesions in both small and large intestine, whereas EHEC expressing intimin γ produced lesions only in the large intestine. These results suggest that different intimin types might determine the tissue tropism, but it should be pointed out that intimin-mediated intimate adhesion occurs after initial bacterial attachment and protein translocation. Thus, other factors, e.g. initial attachment factors, are possibly more likely to be responsible for tissue tropism.

One of the more significant and unexpected recent findings is the realization that the host cell intimin receptor is, in fact, a LEE-encoded bacterial protein first described by Kenny et al. (1997) in EPEC as Tir and shortly afterwards by Deibel et al. (1998) in an O26:NM EHEC strain as EspE. The 78–80 kDa Tir/EspE proteins were shown to be secreted by the type III secretion system and translocated into the host cell, where they are localized to the cytoplasmic and plasma membrane fractions (Deibel et al., 1998). It is believed that Tir consists of at least three functional regions, an extracellular domain that interacts with intimin, a transmembrane domain and a cytoplasmic domain that can induce focusing of the polymerized actin and other cytoskeletal proteins beneath intimately attached bacteria to produce the characteristic pedestal-like structure. From these new data, the essence of EPEC infection now appears to be an interaction between the carboxy-terminal region of intimin and translocated Tir, which in turn results in intimate EPEC attachment, actin accretion and A/E lesion formation (Fig. 4). An intriguing unanswered question associated with Tir/EspE translocation is the mechanism and pathway whereby this soluble bacterial polypeptide becomes an integral host cell membrane protein.

Tir/EspE become tyrosine phosphorylated upon translocation, causing the translocated polypeptide to appear as a 90 kDa protein on SDS–PAGE (hence Tir was initially referred to as Hp90 and thought to be a host cell-derived intimin receptor) (Rosenshine et al., 1992). The physiological significance of Tir phosphorylation is at present not clear for several reasons. Intimin can bind the unphosphorylated form of Tir, and Tir from O157:H7 EHEC does not appear to become tyrosine phosphorylated in the host cell (Kenny et al., 1997; Deibel et al., 1998). Experiments with tyrosine protein kinase inhibitors did not inhibit A/E lesion formation (Rabinowitz et al., 1996). Furthermore, several reports have shown that purified intimin can bind mammalian cells in the absence of Tir (Frankel et al., 1994). In contrast, a recent study by Rosenshine et al. (1996), using HeLa cells, showed that cloned intimin and purified Int280 would not bind HeLa cells unless the cultures were preincubated (‘activated’) with EPEC CVD206, which is intimin negative but able to translocate Tir into the cell. Although the results appear contradictory, the differences may be methodological and merely reflect the fact that intimin binds to more than one receptor. In vitro studies have shown that purified Int280, like Yersinia invasin, has the ability to bind to β1 integrins (Frankel et al., 1996). However, direct evidence for the presence of integrins on the apical surface of intestinal enterocytes is not available, and so any contribution of intimin–integrin interactions in colonization and A/E lesion formation or the identity of any other intimin receptor of host cell origin have yet to be determined. As most bacterial pathogens exploit existing host cell components as receptors, it is interesting to speculate why EPEC/EHEC have found it necessary to insert their own mucosal receptor, especially when it appears that there may be host cell proteins (e.g. integrins) that can bind intimin. Appropriate receptors might not be present on intestinal enterocytes or existing receptors might promote invasion, but this would be of little value without the capacity for intracellular survival. Being able to inject a tailor-made receptor has the advantage of providing a concentration of receptor exactly where it is required. The additional property of Tir in being able to interact directly or indirectly with the host cell cytoskeleton presumably provides other advantages in terms of adhesion, motility and signal transduction.

Signal transduction and A/E lesion formation

Several signal transduction pathways appear to be stimulated in epithelial cells after infection with EPEC and EHEC. New data on specific signalling events, in particular calcium mobilization, together with the new data on protein translocation, have necessitated a re-evaluation of the signalling events and mechanisms responsible for EPEC/EHEC-induced microvillous effacement and A/E lesion formation.

The early studies on signal transduction attempted to explain the mechanisms whereby EPEC induce cytoskeletal breakdown, membrane vesiculation and effacement of brush border microvilli. Based on the similarity between the effects on brush border cells of EPEC and calcium, it was proposed that EPEC produce these effects via a calcium signal and activation of calcium-dependent actin-severing proteins, which break down actin in the microvillous core. Demonstration of elevated levels of calcium in EPEC-infected epithelial cells supported such a mechanism, and studies showing elevated levels of inositol 1,4,5-triphosphate (IP3) and, more recently, tyrosine phosphorylation of phospholipase C-γ1 suggested EPEC stimulation of the classical phospholipase C pathway leading to production of IP3 and release of calcium from IP3-sensitive intracellular stores (reviewed by Kaper, 1998). However, such a global calcium signalling mechanism does not explain the very localized nature of the EPEC/EHEC-induced cytoskeletal rearrangements. Now, new data have raised serious doubts about any role for calcium signalling in A/E lesion formation. Bain et al. (1998) used calcium imaging fluorescence microscopy, which allows both temporal and spatial measurements of intracellular calcium concentration ([Ca]i) in EPEC/EHEC-infected epithelial cells to be determined. Not only were these workers unable to demonstrate localized increases in [Ca]i at sites of A/E EPEC and EHEC adhesion, they were unable to detect any significant alterations in cell calcium in infected compared with uninfected cells. Furthermore, buffering intracellular calcium did not, as previously reported, prevent A/E lesion formation. Thus, A/E lesion formation occurs in the absence of any detectable calcium signal. One important difference between this and previous studies was a very low level of EPEC/EHEC-induced epithelial cell death, suggesting that the sustained and variable increases in calcium reported previously possibly reflect cytotoxic effects of EPEC on cells rather than specific signalling events associated with A/E lesion formation.

Localized translocation of specific effector proteins now provides a more obvious mechanism whereby EPEC and EHEC generate localized signal transduction responses (Fig. 4). The signal(s) and the mechanisms responsible for cytoskeletal disruption are currently unknown, although Rac, Rho and Cdc42-dependent pathways do not appear to be involved (Ben-Ami et al., 1998). Yersinia, Salmonella and Shigella each translocate into host cells via a type III secretion system virulence proteins that induce cytoskeletal rearrangements. Shigella, for example, translocate the Ipa invasins to induce highly localized and dramatic cytoskeletal rearrangements, which result in membrane ruffling and bacterial uptake. It now seems likely that translocated EspB, Tir/EspE or other LEE-encoded effector proteins induce the highly localized and unique cytoskeletal alterations involved in A/E lesion formation in a calcium-independent manner.

Also associated with A/E lesion formation is stimulation of protein kinase activity, including serine/threonine phosphorylation of myosin light chain (MLC), tyrosine phosphorylation of Tir and tyrosine phosphorylation and dephosphorylation of several other proteins (reviewed by Kaper, 1998). Serine/threonine phosphorylation of MLC and tyrosine phosphorylation of Tir are independent of intimin–Tir interaction, as these events occur in cells infected with an intimin-minus mutant. The significance of MLC phosphorylation is not obvious, but as dephosphorylation of MLC results in loss of actin microfilament bundles, and phosphorylation of MLC by MLC kinase results in the association of myosin with the actin cytoskeleton, the phosphorylation state of myosin probably plays a key role in the control of actin filament polymerization–depolymerization and, thus, in brush border effacement during pedestal formation when myosin co-localizes with actin. As discussed earlier, the significance of Tir phosphorylation is similarly unclear, particularly as infection with O157:H7 does not result in phosphorylation of Tir. Intimin–Tir interaction appears to trigger signal transduction events including tyrosine phosphorylation and dephosphorylation of several proteins, although it is not yet clear whether these events are directly related to intimin–Tir interaction or whether they are a consequence of enhanced translocation of effector proteins resulting from intimate bacterial attachment.

Several recently reported host cell responses to EPEC infection could also have important consequences for an understanding of how EPEC cause diarrhoea. A further physiological consequence of EPEC-induced phosphorylation of MLC could be to contribute to diarrhoea by increasing the permeability of mucosal enterocyte tight junctions (Yuhan et al., 1997). EPEC also induce the activation of NFk-B in intestinal epithelial cells, which is associated with increased interleukin (IL)-8 production and transmigration of polymorphonuclear cells (PMNs). Increased paracellular permeability and stimulation of chloride secretion might be a consequence of this EPEC-induced neutrophil infiltration (Savkovic et al., 1997). EPEC infection of Caco-2 intestinal cells has been shown to stimulate chloride secretion (Collington et al., 1998), and this secretory response is dependent on the LEE-encoded Esps and a functional type III secretion system (Collington et al., 1997).

Concluding remarks

The recent intensive investigations on the pathogenic mechanisms used by the EPEC/EHEC family are now beginning to bear fruit. Work on the LEE-encoded type III secretion system, which has general implications for other bacterial pathogens containing similar pathogenicity islands, is providing important insights into how these multifactorial genetic systems contribute to virulence. However, there is still much work to be performed before we have a complete understanding of EPEC/EHEC pathogenesis or have identified all the gene products that contribute to colonization of the host and virulence. In particular, our understanding of how EPEC induce diarrhoea is still very limited. Nevertheless, we can expect more important breakthroughs in this important area in the near future. For example, comparison of the LEE-encoded effector proteins in different EPEC/EHEC isolates has allowed an initial classification of some of the proteins, in particular intimin, into antigenic types. In addition, several reports have indicated that intimin and the Esp proteins are highly immunogenic in different hosts and that these proteins are potential vaccine targets. This work has important implications for any future attempts to generate EPEC, or more probably, EHEC vaccines.

Note added in proof

Since the submission of this review, the global fold of Int280 in solution was determined by multidimensional nuclear magnetic resonance (NMR) (G. Kelly et al., submitted to Nature Structural Biology). The structure shows that Int280 comprises three separate domains, two immunoglobulin-like domains and a C-type lectin-like module. These data define a new family of bacterial adhesion molecules and suggest a carbohydrate–protein recognition role in intimin-mediated cell adhesion.


We thank Simon Elliott for generating Fig. 2 and Michael Donnenberg for allowing us to quote unpublished results. This work was supported by a Leverhulme Grant to G. Frankel and S. Knutton, by a Wellcome Trust Project Grant to S. Knutton, by a Wellcome Trust Programme Grant to G. Dougan, by NIH grants (AI 21657 and AI 41325) to J.B. Kaper, and by grants to I. Rosenshine from the Israeli Academy of Science, the Israel–United States Binational Foundation and the Israeli Ministry of Health.