OUP user menu

Enteropathogenic Escherichia coli: unravelling pathogenesis

Huiwen Deborah Chen, Gad Frankel
DOI: http://dx.doi.org/10.1016/j.femsre.2004.07.002 83-98 First published online: 1 January 2005


Enteropathogenic Escherichia coli (EPEC) is a gram-negative bacterial pathogen that adheres to intestinal epithelial cells, causing diarrhoea. It constitutes a significant risk to human health and remains an important cause of infant mortality in developing countries. Although EPEC was the first E. coli strain to be implicated in human disease in the 1940s and 1950s, the mechanisms by which this pathogen induced diarrhoea remained a complete mystery throughout most of the 40 years since its description. It was only during the late 1980s that major advances were made in unravelling the mechanisms behind EPEC pathogenesis. Ever since, progress has been made at a stunning pace and there have been major breakthroughs in identifying the bacterial factors involved in attaching and effacing (A/E) lesion formation, host signal transduction pathways in response to EPEC infection and the genetic basis of EPEC pathogenesis. The rapid pace of discovery is a result of intensive research by investigators in this field and portends that EPEC will soon be among one of the most understood diarrhoea-causing infectious agents. This review aims to trace the progress of EPEC research since its existence was first reported by John Bray in 1945, highlighting the major findings that have revolutionised our understanding of EPEC pathogenesis.

  • Enteropathogenic Escherichia coli
  • Enterohaemorrhagic Escherichia coli
  • Attaching-and-effacing lesion

1 Introduction

As the pre-dominant species among the facultative anaerobic normal flora of the intestine, Escherichia coli has an important role in maintaining intestinal physiology [13]. This organism was first described by German paediatrician Theobald Escherich in 1885, under the name ‘Bacterium coli commune’ as a short, plump rod that had initially been isolated from normal infant faeces [4]. However, he regarded them to be harmless saprophytes.

For more than half a century, E. coli was considered a major commensal in faeces which was avirulent. This view has changed progressively over the years with accumulation of evidence indicating that E. coli is capable of causing disease in man. Infections due to pathogenic E. coli may be limited to colonisation of a mucosal surface or can disseminate throughout the body and have been implicated in urinary tract infection, sepsis/meningitis and gastro-intestinal infections [5]. E. coli have also been associated with a variety of pathological conditions in farm animals and occur most frequently in young animals [6, 7]. In particular, neonatal colibacillary diarrhoea, caused by lethal infection of newborn farm animals by enterotoxigenic E. coli (ETEC), is one of the most common enteric diseases of farm animals in the UK [8].

Due to the ease of access of pathogens ingested with food, the human gastro-intestinal tract is susceptible to diarrhoeagenic E. coli infections. Several E. coli pathogens have been implicated with diarrhoeal illness, a major public heath problem worldwide, with over 2 million deaths occurring each year (http://www.who.int); this underscores the importance of research into the pathogenic mechanisms of E. coli. The major categories of diarrhoeagenic E. coli strains include enteropathogenic E. coli (EPEC), enterotoxigenic E. coli (ETEC), enteroinvasive E. coli (EIEC), enterohemorrhagic E. coli (EHEC) and enteroaggregative E. coli (EAEC).

In particular, EPEC was the first strain of E. coli incriminated as the cause of outbreaks of infantile diarrhoea in the 1940s and 1950s [9]. These outbreaks of ‘summer diarrhoea’ were frequent in developed countries until the 1950s and had a high mortality [1012]. For unknown reasons, EPEC strains are no longer important causative agents of infant diarrhoea in developed countries [5]. However, EPEC is still responsible for occasional outbreaks in paediatric wards and day-care centres [13, 14]. In developing countries, EPEC remains a major cause of infant diarrhoea, with recent outbreaks reporting a mortality rate of 30%[15]. Studies in Brazil [16], Mexico [1719], South Africa [20] and Bangladesh [21] have shown that 30–40% of infant diarrhoea is due to EPEC infection and is estimated to cause the deaths of several hundred thousand children per year [5].

The hallmark of EPEC infection is the ‘attaching-and-effacing’ (A/E) histopathology observed in small bowel biopsy specimens from infected patients [5]. Decrease in the number and height of microvilli, blunting of enterocyte borders, loss of glycocalyx and presence of a mucous pseudomembrane coating the mucosal surface were also observed in many patients. Clinically, EPEC illness is characterised by acute diarrhoea, fever, malaise and vomiting [22].

Intriguingly, though major advances toward defining EPEC pathogenesis have been made over recent years as a result of cell biology, genetics and host intestinal physiology studies, the pathogenic mechanism of this important microbe remains elusive after almost 40 years. This review aims to highlight the important research breakthroughs which have helped advance our understanding of EPEC as a pathogen and will focus mainly on (1) epidemiological studies in the 1940s and 1950s that identified the diarrhoeagenic potential of certain E. coli strains; (2) challenge studies carried out in the 1970s which unambiguously confirmed that EPEC possessed virulent properties; (3) basis of host–pathogen interactions and (4) current research directed at establishing the molecular and genetic basis of EPEC pathogenesis.

21940 s: The rise of EPEC

The key event that led to the recognition of Escherichia coli as the causative agent of summer diarrhoea in infants were reports by Bray and by Bray and Beaven [9, 23]. As E. coli is also a component of the normal intestinal flora, standard bacteriological examination in use at that time to detect Shigella and Salmonella was unable to distinguish between infected and healthy individuals. By raising antibodies against an E. coli strain isolated from an infant with summer diarrhoea, Bray and Beaven were able to distinguish between pathogenic strains that caused human diarrhoea and strains that were part of the normal intestinal flora. Agglutinating E. coli was detected in cultures from 92% of infected infants but only 6% of the controls and they named this particular strain of E. coliBacillus coli nepolitanum.’

In the ensuing period, several outbreaks of infantile gastroenteritis led to the epidemiological incrimination of a series of E. coli strains throughout the world as a primary gut pathogen. Using the same general approach, Giles et al. [24] isolated a single serotype (α-type) of E. coli from over 90% of the infected individuals in an outbreak in Aberdeen; Taylor et al. [25] found that one serological type (D433) was implicated in outbreaks of infantile diarrhoea in England. As more pathogenic E. coli strains were isolated and identified, each strain was given a unique epithet from its discoverer, resulting in much confusion [12].

This was resolved by Friz Kauffmann who developed a defined serotyping system for E. coli in 1944, allowing a systemic characterisation of isolates based on their serological characteristics and making it possible to compare strains accurately [26]. According to the modified Kauffmann scheme, E. coli are serotyped based on their somatic O, flagella H and capsular K surface antigens. The usefulness of this scheme was demonstrated by the fact that the various strains of E. coli that had been associated with infant gastroenteritis in the 1940s belonged to a surprisingly small number of O serogroups. According to Kauffmann at the International Congress of Microbiology at Rome in 1953, it proved beyond doubt the ‘etiological importance of certain serological types of E. coli for epidemic infantile enteritis.’ The serotyping scheme has since been expanded to include the analysis of fimbriae [27], though it is still not commonly included in the serological formula.

In 1955, Neter et al. [28] coined the term enteropathogenic E. coli (EPEC) to describe strains of E. coli that were primary intestinal pathogens but were rarely encountered in the faeces of healthy individuals and in infections other than diarrhoeal diseases. Presently, more than 180 different O serogroups and more than 60 H serogroups are recognised [29]. A specific combination of O and H antigens defines the O:H serotype of an isolate. The major O serogroups considered to contain EPEC serotypes are O55, O86, O111, O119, O125, O126, O127, O128ab and O142 [30, 31]. For some serogroups like O55 which are rarely isolated from healthy individuals, O serotyping is sufficient to indicate the presence of an enteropathogen; however, serotypes within other O serogroups are not equally pathogenic and only certain H types are incriminated within the O serogroup [30].

Experimental challenge studies in the early 1950s established the pathogenicity of certain specific strains of E. coli isolated from cases of infantile gastroenteritis. Young adult volunteers dosed orally with varying amounts of E. coli cultures belonging to O111 and/or O55 serogroups developed diarrhoea. In general, the severity of symptoms corresponded to the size of the dose [3234]. It was further shown by Koya et al. that intestinal specimens from the adult volunteers who developed diarrhoea contained large numbers of the experimental E. coli strain (O111:B4) 18–24 h after the oral dose. E. coli strains from healthy babies did not cause a diarrhoeal response [35].

Although epidemiological evidence and studies with volunteers had implicated the classical serotype of EPEC as causes of diarrhoea, up till the mid 1960s it was still impossible to differentiate EPEC from the strains of normal flora using biochemical, microbiological or animal model assays. Hence, serotyping remained the only diagnostic tool for EPEC up till the 1970s [22].

3 The 1970s: challenge studies

In the 1970s, two other diarrhoeagenic E. coli strains came into prominence. Strains designated enterotoxigenic E. coli (ETEC), were shown to produce heat-labile enterotoxins (LT) and heat-stable enterotoxins (ST); strains that possessed a Shigella-like invasiveness and caused an inflammatory dysentery-like disease were termed enteroinvasive E. coli (EIEC). With the advent of laboratory tests to assess heat-labile and heat-stable enterotoxin production and enteroinvasiveness of E. coli, the classic serotype enteropathogenic E. coli strains were found to lack those particular virulence properties. These observations led some to question their pathogenicity. Some investigators concluded that strains that had been designated EPEC were not pathogenic. They considered the EPEC isolates to be simply ETEC strains that had lost their enterotoxin plasmids during subculture and storage [3638]. Moreover, they suggested that serotyping of E. coli was of little value to investigations of strains involved in enteric diseases since tests for the production of enterotoxin allowed recognition of pathogenic E. coli. The issue was further confused when it was seen that LT and ST-producing E. coli from adults and children with diarrhoea rarely were the classic infant diarrhoea EPEC serotypes that had been epidemiologically incriminated in challenge studies in the 1950s [39]. Tests carried out by Gross et al. [40] proved that EPEC strains from infantile diarrhoea did not produce enterotoxins, indicating that E. coli strains responsible for virulence were not ETEC. They contended that EPEC strains were pathogens whose virulence depended on other factors.

The confirmation that EPEC strains were pathogenic came from human volunteer studies carried out by Levine et al. in 1978 [41]. Classic EPEC strains (O127 and O142) associated with infant diarrhoea that had been stored for 7–9 years and gave negative results in sensitive tests for LT and ST enterotoxins and invasiveness were fed to healthy young adult volunteers. The challenge study caused a notable diarrhoeal illness in the volunteers. LT and ST enterotoxins were absent in E. coli stool isolates from affected individuals and confirmed that EPEC were indeed diarrhoeal-inducing pathogens. This hallmark study stimulated renewed research on the pathogenesis of EPEC diarrhoea and rapidly yielded results.

4 The 1980s: the basis of host-pathogen interactions

4.1 Attaching-and-effacing lesions

The hallmark of EPEC infections is the attaching-and-effacing (A/E) histopathology which can be observed on epithelial cells at ultrastructural levels (Fig. 1). This distinctive ultrastructural histopathologic lesion was noted by Staley et al. [42] and Polotsky et al. [43] and was later described in biopsies from infants infected by EPEC by Ulshen and Rollo [44] and Rothbaum et al. [45]. However, it was only until the report by Moon et al. [46] in 1983 that the phenotype became widely associated with EPEC and the term ‘attaching-and-effacing’ was introduced. Numerous reports have since confirmed the characteristic A/E phenotype in epithelial cells in animal models [47], tissue culture cells [48, 49] and in humans infected with EPEC [50].

Figure 1

(a) Gross localised perturbation of brush border architecture. Original magnification: 5000×. (b) Electron micrograph of cultured human intestinal mucosa infected with EPEC, illustrating the key features of A/E histopathology. Bacteria are intimately attached to cup-like projections of the apical enterocyte membrane. At regions of attachment, localised destruction of microvilli (MV) and disruption of cytoskeleton were observed. Magnification: 45,000×. Reproduced from Knutton et al. (1987), with permission from ASM press.

Examination of affected epithelial cells under the electron microscope show that EPEC induce profound cytoskeletal alterations, disrupting the brush border cytoskeleton and leading to a proliferation of filamentous actin beneath adherent bacteria. Effacement of microvilli and intimate adherence between the bacterium and the epithelial cell membrane are also observed. The epithelial membrane beneath the adherent bacteria is raised locally in a characteristic pedestal shape which may extend up to 10 μm outwards from the cell to form pseudopod-like structures [51]. In severe infections, there is almost complete destruction of the absorptive surface of the intestinal specimen, with extensive villus atrophy and thinning of the mucosal lining. These observations of lesion formation were crucial in identifying adherence as an important factor for EPEC pathogenesis.

Since actin is the major component of the brush border cytoskeleton, Knutton et al. [49] proposed that the dense concentration of microfilaments in the apical cytoskeleton beneath attached bacteria was composed of actin and was a result of attachment-and-effacement by EPEC. This was confirmed by staining filamentous actin with fluorescein-labelled phallotoxin [52]. Examination by fluorescence microscopy showed that infected cells exhibited intense spots of fluorescence which corresponded in size and position with each adherent bacterium; cells that were infected with adherent E. coli strains not known to form A/E lesions did not display such a pattern of actin accumulation. The results indicated that such site-specific concentration of cytoskeletal actin was characteristic of A/E histopathology and that fluorescence actin staining (FAS) formed the basis of a simple, highly specific diagnostic test for EPEC and other organisms capable of causing such a histopathology. Clinical E. coli isolates that showed diffuse or aggregative adherence tested negative in the FAS test, proving that they did not exhibit A/E activity and were not EPEC [53]. Prior to the development of this test, the A/E phenotype could only be detected through electron microscopy and was not useful for extensive screening of clones and mutants. The development of the FAS test was a major breakthrough, providing an assay to begin genetic studies into the A/E phenotype [54].

In addition to filamentous actin, other cytoskeletal proteins including α-actinin, talin, erzin, myosin-light chain, VASP, WASP and the Arp2/3 complex were identified by immunofluorescence microscopy in A/E lesions [5557]. Adherence of EPEC to cultured Hep-2 cells also lead to phosphorylation of proteins on tyrosine residues located at the tip of pedestals, just beneath the plasma membrane [58]. Interestingly, video microscopy has shown pedestal formation to be dynamic. Furthermore, propelled by attached actin stalks, some EPEC can actually move along the surface of the epithelial cell, though the significance of this motility to pathogenesis is unknown [5, 59].

4.2 Localised adherence

In 1979, Cravioto et al. [60] showed that 80% of the EPEC strains as defined by serotype could adhere to HEp-2 cells in vitro while most non-EPEC E. coli could not. For many years, EPEC had been identified solely on the basis of serotyping which was tedious, expensive, displayed limited sensitivity and specificity and could only be performed reliably by a small number of reference laboratories [5]. The HEp-2 adherence assay [61] involves inoculating the test strain on a semi-confluent HEp-2 monolayer and incubating it for 3 hr at 37 °C under 5% CO2. The monolayer is then washed, fixed, stained and examined by oil-immersion light microscopy. This phenotypic assay soon became one of the most useful phenotypic assays for the detection of diarrhoeagenic E. coli.

Later, Scaletsky et al. [62] showed that E. coli strains attach to HeLa cells in two different patterns – localised adherence (LA) in which bacteria adhere in discrete microcolonies (Fig. 2(a)) and diffuse adherence in which bacteria adhere uniformly over the cell surface. Localised adherence was highly correlated with specific EPEC serogroups in strains isolated from patients with diarrhoea and most EPEC serogroups O55, O86, O111ab, O119, O125, O128ab and O142 showed LA [63]. These results were confirmed by Nataro et al. [64] using HEp-2 cells. Furthermore, these investigators showed that many pathogenic E. coli strains and some EPEC stains adhered to HEp-2 cells, displaying the diffuse adherence phenotype. This led to the realisation that some strains that had been classified as EPEC were not true EPEC strains and to the characterisation of two other categories of diarrhoeagenic E. coli–enteroaggregative E. coli (EAEC) and diffusely adhering E. coli (DAEC). These distinct adherence patterns had not been observed by Cravioto et al. mainly due to differences in incubation times; the longer incubation periods used by Cravioto et al. [62] made the distinction impossible.

Figure 2

(a) EPEC bacteria showing the localised adherence phenotype on cultured epithelial cell. (b) Immunostaining with BFP-specific polyclonal rabbit antiserum revealed that clustering of bacterial cells within the microcolonies is mediated by BFP (figures are courtesy of Stuart Knutton).

The HEp-2 adhesion assay was used by Baldini et al. [65] in 1983 to show that the ability of EPEC stain E2348/69 (O127:H6) to adhere in a localised adherence pattern was associated with the presence of a 60 MDa plasmid denoted pMAR2. E2348/69 that had been cured of pMAR2 resulted in the loss of the LA phenotype, a property later confirmed by McConnell et al. [66]. Moreover, transfer of the plasmid to non-adherent E. coli K12 conferred adherence to HEp-2 cells [65] although adherence was poor compared to that of EPEC [66]. The term EPEC adherence factor (EAF) was suggested to refer to the plasmid-mediated adhesin that conferred HEp-2 adherence [65]. In further support of the role of the EAF plasmid in EPEC pathogenesis, E. coli strains isolated from outbreaks of infantile gastroenteritis in the United States and from stools of infants with diarrhoea in Brazil almost invariably possessed the EAF plasmid [67].

An insertion mutation which inactivated localised adherence of EPEC identified a plasmid region of unknown function that was necessary for this phenotype [67]. The isolated 1 kb fragment from the plasmid proved to be a highly sensitive and specific DNA probe and has been used extensively in epidemiological studies to identify EPEC that contain the plasmid [6769]. In addition, the probe revealed that HEp-2 adherence was more frequent in some O serogroups of EPEC than in others, designated Class I and Class II respectively by Nataro et al. [67]. These classes are also more commonly termed typical and atypical EPEC [70, 71].

The importance of the EAF plasmid in human disease was shown through human challenge studies carried out by Levine et al. [72]. Diarrhoea occurred in 9 out of 10 volunteers who ingested the wildtype E2348/69 strain possessing the EAF plasmid; in contrast, only two of the nine volunteers who took the cured derivative showed mild symptoms of diarrhoea. All volunteers who fell ill from the wildtype strain mounted an antibody response to a 94 kD membrane protein. Notably, the individual among the ten challenged who failed to develop diarrhoea after ingestion of E2348/69 was shown to possess antibody before the challenge. A surprise finding was that although the plasmid was highly stable in vitro, considerable spontaneous cure was observed from volunteer stool specimens; this might provide important clues towards better understanding of EPEC pathogenesis.

A study by Knutton et al. [73] showed that in tissue culture cell adhesion assays, EPEC display LA and A/E phenotypes. Though the EAF plasmid promoted cell adhesion, it was of the non-intimate type. For A/E lesion formation, there was no absolute requirement for the plasmid and plasmid-cured derivatives retained A/E activity. This indicated that other chromosomally encoded factors were required for EPEC to display the A/E phenotype. These observations led to the proposal that mucosal adhesion by EPEC involved two distinct stages: (i) initial attachment of EPEC promoted by plasmid-encoded adhesins and (ii) effacement of brush border microvilli and intimate EPEC attachment [73]. Although the second stage could occur without the first, the presence of plasmid-encoded adhesin enhanced EPEC ability to colonise the mucosa.

Despite the recognition of the importance of EAF in pathogenesis, the molecular nature of localised adherence remained elusive for many years. Candidates for the adhesin were proposed and rebuffed. Among previous candidate adhesins were a 32 kDa outer membrane protein (OMP) [74], later reported to be OmpF [75] and a 94 kDa OMP [66, 72] later shown to be intimin [76].

Finally, in 1991, Girón et al. [77] described a 7 nm fimbriae produced by EPEC strains growing on epithelial surfaces which tended to aggregate and bind individual organisms together. When EPEC strains were cured of the plasmid, they failed to express the fimbriae and did not grow as adherent colonies. Moreover, an antiserum against the fimbriae reduced the ability of EPEC to colonise cultured epithelial cells. Expression of the fimbrial gene product could be induced in tissue culture medium and under these conditions, the bacteria tended to aggregate and clump [78] (Fig. 2 (b)). These fimbriae were termed ‘bundle-forming pilus’ (BFP) and were only produced under certain culture conditions, accounting for the failure of previous investigators to identify them [77].

5 Recent progress

Although EPEC was the first E. coli to be associated with human disease, research into its pathogenic mechanism had been hampered by the lack of precise characterisation and classification. Hence, the consensus definition of EPEC at the First International Symposium on EPEC in 1982 defined EPEC as ‘diarrhoeagenic E. coli belonging to serogroups epidemiologically incriminated as pathogens but whose pathogenic mechanisms have not been proven to be related either to heat-labile enterotoxins (LT) or heat-stable enterotoxins (ST) or to Shigella-like invasiveness’ [79]. The remarkable advances in EPEC research in the 1980s and early 1990s allowed EPEC to be defined on basis of the characteristics that distinguish it from other diarrhoeagenic E. coli, instead of on a definition of exclusion. Based on pathogenic data available, the participants of the Second International Symposium on EPEC in 1995 reached a consensus definition for EPEC which included the production of the characteristic A/E pathology on epithelial cells and the absence of Shiga toxin production. Furthermore, it made a distinction between EPEC containing the EAF plasmid (typical EPEC) and EPEC strains that did not possess the plasmid (atypical EPEC) [70].

A three-stage model of EPEC pathogenesis, comprising of localised adherence, signal transduction and intimate attachment, was first proposed by Donnenberg and Kaper in 1992 [69]. There has been substantial progress in elucidating the complexities of each stage; bacterial factors – including the bundle-forming pilus (BFP) proteins, EPEC secreted proteins (Esp), a Type III secretion apparatus, an OMP intimin and its host cell receptor Tir-involved in A/E lesion formation, host signal transduction pathways in response to EPEC infection and the genetic basis of EPEC pathogenesis have been characterised to some extent.

5.1 Initial adherence to epithelial cells – a role for BFP?

The first stage in EPEC pathogenesis involves the initial adherence of bacteria to epithelial cells. Previous studies have implicated the BFP as the initial EPEC attachment factor [77]. The first structural gene encoding the major pilin subunit of BFP (bfpA) was identified through the isolation of EPEC strain E2348/69 TnphoA mutants which no longer conferred localised adherence [80]. A bfpA probe was later developed and proved to be more sensitive than the EAF probe, probably because it consisted primarily of a sequence encoding an important EPEC virulence factor while the EAF probe contained a sequence of unknown function [81]. The complementation studies also indicated that localised adherence was a complex process that required a large stretch of contiguous and non-contiguous EAF plasmid [80]. Subsequent genetic studies revealed that BFP was encoded by a cluster of 14 genes on the EAF plasmid [82, 83] and that most genes were involved in pilus biogenesis [84, 85]. These proteins showed some homology to those involved in the biogenesis of type IV pili in other bacteria [83] and some functions for these proteins have been proposed [85]. Genes external to the bfp gene cluster were also necessary for full expression of BFP. This included the global regulator element of EPEC pathogenesis perABC (bfpTVW) [86, 87] and the chromosomal dsbA gene encoding for a disulphide isomerase [88].

Although BFP are definitely involved in interbacterial adherence in the localised adherence pattern [8991] (Fig. 2 (b)), there is no definite proof that BFP mediate actual adherence to epithelial cells. Tobe and Sasakawa found that BFP-expressing cells preferentially bound to cultured epithelial cells rather than joining established microcolonies already on the surface, pointing to a role in initial attachment [92]. Moreover, recent studies by Cleary et al. [93] have shown that an EPEC intimin mutant strain (CVD206) which cannot form A/E lesions but expresses BFP, adhered to human intestinal explants ex vivo, providing further support for a cell adhesion role for BFP. In contrast, evidence from earlier studies, using an in vitro organ culture (IVOC) of paediatric small intestine, suggested otherwise. EPEC mutants that lacked the EAF or bfpA gene could still adhere and cause A/E lesions on mucosa tissue but microcolonies were not observed, while CVD206 did not colonise the intestinal explant. Hence, these investigators concluded that adhesins other than BFP initiated colonisation of mucosal surfaces although BFP was required for the formation of complex, three-dimensional colonies via interbacterial interactions [94]. Taken together, these results demonstrate that there is redundancy in colonisation factors expressed by EPEC and that differences in experimental design might favour one mechanism or another.

Despite these conflicting reports, recent data have confirmed that BFP remains an important virulence factor. Human challenge studies by Bieber et al. [95] showed that inactivation of the bfpA gene encoding the pilus subunit or the perA gene encoding the bfp operon transcriptional activator caused significantly less diarrhoea. A parallel study with bfpF mutants caused increased pilation, enhanced localised adherence and failure of microcolonies on HEp-2 cells to disperse. Moreover, an EPEC strain with a modified bfpF gene had to be administered as a 200-fold greater bacterial inoculum to produce the same diarrhoeal response as the parent. One explanation for this is that BFP-mediated interbacterial interactions may allow the dispersal of individual bacteria from autoaggregates and colonisation of other epithelial sites, contributing to the spread of infection within the gut. This hypothesis was supported by in vitro tissue culture adhesion assays where the dispersal of EPEC from microcolonies was shown to be associated with the transformation of BFP from thin to thick bundles [96].

In addition to BFP, additional fimbrial structures have been characterised and could have roles in bacterium–host cell adhesion [97, 98]. Girón et al. [99] reported an extensive characterisation of rod-like fimbriae and fibrillae produced by EPEC strain B171; this suggests that bacterial-host cell interaction is a multifactorial process.

More recently, flagella have been implicated in EPEC adherence to epithelial cells and fliC insertional mutants of two EPEC strains showed marked impairment in adherence and microcolony formation on cultured cells [100]. Furthermore, it was observed that medium pre-conditioned by growth of cultured epithelial cells induced flagella expression, leading to the proposal that epithelial cells may produce a signal that bacteria recognise. However, recent studies by Cleary et al. raised some doubt regarding the role of flagella in EPEC adhesion as a flagellated strain that lacked BFP, intimin and EspA failed to adhere to epithelial cells in vitro [93].

5.2 Pathogenicity islands: the locus of enterocyte effacement

Unlike the non-pathogenic strains of E. coli which are part of the normal intestinal flora, EPEC contains a 35.6 kb pathogenicity island called the locus of enterocyte effacement, LEE [101]. The entire LEE of E2348/69 has been cloned into E. coli K12 and the resulting A/E histopathology provided unequivocal evidence that this pathogenicity island is a functional cassette both necessary and sufficient for EPEC virulence [102]. It also reinforces the notion that avirulent bacteria can be transformed into pathogenic ones through a single genetic step. Analysis of the G + C content of the LEE (38%) showed that it was strikingly lower than that of the E. coli chromosome (50.8%), implying horizontal gene transfer from a foreign gene [101].

The complete sequence of the LEE of EPEC strain E2348/69 showed that the LEE contains 41 open reading frames (ORFs) of more than 50 amino acids arranged in five polycistronic operons LEE1 to LEE5 [103]. These genes are separated into three functional domains – a region encoding intimate adherence (Tir and intimin), a region encoding the EPEC secreted proteins (including EspA, EspB, EspD and EspF) and their putative chaperones and the region encoding a type III secretion system.

5.3 A novel type III secretion system

The second stage of EPEC infection is characterised by signal transduction. One key finding which has facilitated the study of EPEC pathogenesis was that EPEC secreted a number of proteins (EPEC secreted proteins Esp) via a type III secretion pathway [104106]. The type III secretion apparatus is utilised by many enteric pathogens; it acts as a macromolecular syringe to inject effector proteins directly into host cells and is almost exclusively involved in virulence [107, 108]. Based on experimental data and homology with other type III secretion systems (TTSS), 12 LEE-encoded genes (esc and sep) involved in TTSS biogenesis have been identified [103]. Though the structural basis for protein translocation via the type III secretion apparatus remains unclear, it appears that EscC may be a novel type of channel-forming protein in the EPEC outer membrane, EscN an ATPase that powers the type III secretion system and EscF the putative needle component [90, 109, 110]. A model of the TTSS of EPEC is shown in Fig. 3.

Figure 3

Type III protein translocation in EPEC. EPEC elaborates a classical needle complex that spans the bacterial cell wall. Polymerisation of EspA from the EscF-needle tip leads to formation of a hollow filaments through which translocator effector proteins are injected into the host cell. Once translocated, the effectors target different cellular compartments. Tir is targeted to the plasma membrane, Map to the mitochondria and EspI/NleA to the Golgi. Subversion of cellular functions lead to A/E lesion formation and disease.

In EPEC, the type III secretion system appears dedicated to the secretion of specific proteins, including Tir, EspA, EspB and EspD which are essential for the subversion of host cell signal transduction pathways and the formation of A/E lesions [105, 106, 111113]. In addition, mutational analyses have shown that the type III secretion system, together with the secreted translocated proteins is essential for the delivery of proteins into the host cell [114, 115].

EspA is a major constituent of a filamentous organelle found on the bacterial surface in the early stages of A/E lesion formation and it was postulated that EspA formed a translocation tube through which bacterial effectors could be delivered directly into infected host cells [116, 90]. Electron micrographs of the supermolecular structure later gave definite proof that EspA filaments were hollow extensions of the type III secretion system needle complexes, hence defining a new class of filamentous type III secretion system (FTTSS) [117119] (Fig. 4). Furthermore, EspA filaments have been shown to have a role in mediating initial adhesion of EPEC to host cells; however, adhesion was much weaker than that mediated by BFP [93].

Figure 4

(a) Transmission electron micrograph showing EspA filaments (black arrows), labeled by immunogold with EspA specific rabbit polyclonal antiserum, forming a bridge between bacteria and epithelial cells during early stages of A/E lesion formation. (b) Scanning electron micrograph showing EspA filaments (black arrows) connecting EPEC bacteria to the plasma membrane enabling injection of translocated effector proteins. Reproduced from Knutton et al. (1998) with permission from the EMBO J.

Based on their homology to the Yersinia YopB/D proteins and their ability to lyse red blood cells, EspB and EspD are believed to be involved in formation of a translocation pore, enabling proteins to be delivered into the host cell [120]. Hence, the primary function of EspB and EspD may be to deliver other virulence factors into the host cell rather than acting as translocated effectors. However, transfection of espB into HeLa cells alters actin stress fibres and cell morphology, suggesting an effector function as a cytoskeletal toxin [121, 122].

Chaperone proteins have been discovered in type III secretion systems of other related species and seem to prevent premature interaction with other proteins during secretion and translocation. Such proteins were identified in EPEC and were essential for the proper secretion of EspD (CesD and CesD2) [123, 124] and EspA and EspB (CesAB) [125]. Efficient transfer of Tir and Map into the host cell was demonstrated to be dependent on the protein CesT [126129] while EspF was reported to require a chaperone CesF to be translocated into host cells [130].

5.4 Intimate adherence – the roles of intimin and Tir

The third stage of EPEC infection is characterised by enterocyte effacement, pedestal formation at the apical enterocyte–cell membrane and intimate bacterial attachment to the host cell. This is mediated by a 94 kDa OMP, intimin. The gene encoding intimin (named eae for E. coli attaching-and-effacing) was detected by Jerse et al. [48] in 1990 through screening programmes to identify TnphoA mutants defective in pedestal formation using the FAS test. Previously, Levine et al. [72] had reported that a 94 kDa OMP engendered a strong antibody response in volunteers during a challenge study. This immunogenic OMP was later shown to be intimin. Challenge studies were also carried out to test the role of intimin in infection; diarrhoea was seen in all volunteers who ingested the wildtype EPEC strain as compared to the four of 11 volunteers ingesting an isogenic eae null mutant. This demonstrated that intimin was essential for full virulence of EPEC [131].

Molecular and serological studies of intimin have established the presence of at least four distinct intimin subtypes (α, β, γ, δ) and there appears to be a direct correlation between an intimin type and a specific EPEC/EHEC clone [90, 132]. IVOC infection experiments were conducted to investigate the role of intimin subtypes in tissue specificity of A/E lesion-forming microbial pathogens. Complementing an EPEC eae mutant with EPEC eaeα (encoding intimin-α) restored the ability to colonise the small intestinal mucosa in a manner similar to that of the parent strain; in contrast, complementation with EHEC eaeγ (encoding intimin-γ) resulted in adherence of the strain to Peyer's patches, similar to that observed in EHEC bacteria. These results confirmed that different initmin subtypes contributed to tissue tropism [133].

Sequence homology studies have shown that while the N-terminal domain which is inserted in the bacterial outer membrane is highly conserved, the C-terminal region which binds to receptors on epithelial cells is highly divergent [134]. This cell-binding activity is localised to the C-terminal 280 amino acids (Int280) [134, 135]. The global fold of Int280 has been determined by multidimensional NMR in solution [136] and by X-ray crystallography just recently [137]. It was suggested that Int280 comprised of three domains: two immunoglobulin-like domains (D1 and D2) and a C-type lectin domain (D3) which defined a new family of bacterial adhesion molecules.

Initial work by Rosenshine et al. [138] found a correlation between EPEC adherence with the tyrosine phosphorylation of a 90 kDa protein (Hp90) within host membranes. Subsequent studies revealed that intimin-mediated adherence was dependent on prior tyrosine phosphorylation of Hp90 and that Hp90 served as the receptor for intimin [51]. One significant and unexpected finding was that Hp90 was, in fact, a bacterial effector protein translocated from the bacteria to the host cell via the type III secretion system [115]. This was the first documented, unprecedented instance of a pathogenic bacterium inserting its own receptor into the host cell membrane. Hp90 was hence renamed translocated intimin receptor (Tir).

The binding of intimin to Tir focuses or clusters Tir beneath adherent EPEC, directly linking extracellular EPEC to the epithelial membrane and anchoring it to the host cell actin and cytokeratin cytoskeleton networks [139]. By doing so, EPEC can initiate pedestal formation and mediate its pathogenic effects on the host cell while remaining on the extracellular surface [139, 140]. The fact that EPEC can insert its own intimin receptor is difficult to reconcile with the concept of tissue tropism. This raises the possibility that intimin can bind to more that one receptor and is supported by reports by Frankel et al. [134, 135] that the C-terminal domain can bind HEp-2 cells in the absence of Tir, possibly by binding to β1-integrins [141]. More recently, Sinclair and O'Brien [142] showed that different intimin types could bind to the host cell protein nucleolin and that nucleolin is colocalised with adherent bacteria. However, the question of which receptor is relevant for intimate adherence remains unanswered. Other important issues arising from this discovery are how a bacterial pathogen can take a secreted soluble protein (Tir) and insert it into a host membrane and how Tir evolved and where it came from.

5.5 The cell responds: signal transduction and subversion of host the cytoskeleton

One of the most striking changes in the host cell after EPEC infection is the formation of the characteristic A/E lesion, leading to the localised loss of microvilli and the formation of actin-rich pedestals [46, 49, 51]. Presumably, these structures provide strong attachment of EPEC to the cell surface, preventing EPEC from being dislodged in the ensuing diarrhoeal response [143]. Since EPEC remains predominantly extracellular during the infection process, the subversion of the host cell cytoskeleton must involve signal transduction within the host epithelial cells [144].

Based on the observation that isolated brush border cells treated in vitro with Ca2+ ionophores caused morphologically similar changes to that observed during EPEC infection, Baldwin et al. [145] proposed a role for calcium signalling and activation of Ca2+-dependent actin severing proteins in brush border effacement. This was supported by early studies on signal transduction which showed an increase in intracellular Ca2+ in infected cells [145, 146]. These results were consistent with later reports describing elevated levels of inositol-3-phosphate (IP3) and tyrosine phosphorylation of phospholipase C-γ1 [147, 148]. Thus, it was generally accepted that EPEC stimulation of the phospholipase C-γ1 pathway resulted in release of Ca2+ from IP3-sensitive intracellular stores which then mediated cytoskeletal rearrangement.

However, in an attempt to analyse the spatial and temporal distribution of Ca2+ during EPEC infection using calcium-imaging fluorescence microscopy, Bain et al. [149] failed to detect any increase in Ca 2+ levels at the site of A/E lesion and throughout the infected cell. The recent data strongly indicated that EPEC-induced pedestal formation was Ca2+-independent. Yet, the conflicting results could be methological and the controversial role of Ca2+ in signalling and cytoskeletal reorganisation remains unresolved.

Progress in studies of EPEC-induced cytoskeletal rearrangement have begun to yield other insight into the mechanisms EPEC use to subvert its hosts’ cytoskeleton and signalling pathways. It is becoming apparent that the localised translocation of specific effector proteins is essential for the triggering of signalling pathways in EPEC-infected cells [90]. One such pathway results in the phosphorylation of substrates that co-localise with accumulated actin underneath adherent bacteria. This is dependent on the type III secretion system, on the EspA/EspB translocator proteins and other LEE-encoded effector proteins [5].

Notably, Tir appears to be a key bacterial factor driving pedestal formation. Its location at the tip of the actin pedestal and the requirement of a tyrosine phosphorylation event for A/E lesion formation makes Tir a likely candidate to link EPEC to the host cytoskeleton and direct accumulation of actin and cytokeratins [139, 150]. To accomplish this, Tir might exploit the regulators of actin and cytokeratin dynamics to initiate polymerisation and reorganisation. Surprisingly, several groups were unsuccessful in finding a role in pedestal formation for members of the Rho family, which are usually involved in actin-based processes [151]. Recent studies have implicated members of the Wiskott–Aldrich syndrome (WAS) family of proteins and the actin nucleating factor (Arp2/3) [57]. This is significant as it hints at how EPEC might initiate signalling cascades at the cell surface to rearrange the host cytoskeleton. Additionally, many proteins involved in focal adhesion such as α-actinin and vinculin were found to be recruited to sites of A/E lesions, independent of Tir phosphorylation [152]. Hence it is hypothesised that the concommittant effacement of microvilli together with pedestal formation could have originated from the subversion of fundamental host cell functions by EPEC to build structures which anchor it to the host epithelium [143].

5.6 Endgame: diarrhoea

Despite impressive advances in our understanding of EPEC pathogenesis on cellular and genetic levels, the pathophysiology of the resultant diarrhoea is still uncertain. Dramatic loss of microvilli and the subsequent malabsorption due to brush border enzyme deficiency certainly contributes to diarrhoea. However, this alone is insufficient to account for the rapid onset of diarrhoea reported in human volunteer studies [5, 153].

Several investigations have implicated the alteration of electrolyte transport and the opening of channels during EPEC infection with diarrhoea. Using a whole-cell current clamp method, Stein et al. [154] examined the direct interaction of live EPEC with individual host cells. They found a significant decrease in transmembrane potential in infected cells, suggesting that EPEC could directly alter the relative distribution of ions across membranes. Furthermore, EPEC infection of Caco-2 cells was shown to stimulate a rapid increase in short circuit current which was dependent on intimate attachment and a functional type III secretion system [155157]. This effect could be due to changes in chloride ion secretion, a common mechanism that leads to secretory diarrhoea.

Alteration of epithelial permeability could contribute to diarrhoea. EPEC is capable of inducing a signalling cascade within the host cell, resulting in the phosphorylation of myosin light chains on Ser/Thr residues [158]. Yuhan et al. [159] have shown that this might contribute to diarrhoea by increasing the permeability of tight junctions and disrupting tight junction integrity. Recently, another EPEC effector molecule, EspF, was identified [160]. This protein is translocated by the type III secretion system into host cells where it disrupts host intestinal tight junction functions, suggesting a possible role of this protein in inducing diarrhoea in susceptible hosts [161]. Also, attachment of EPEC to cultured epithelial monolayers was shown to induce the activation of NF-κB in host cells, leading on to increased interleukin-8 production and transmigration of polymorphonuclear (PMN) cells through the monolayers [162, 163]. Increased paracellular permeability and stimulation of chloride secretion could be a consequence of this EPEC-induced PMN infiltration. Another consequence is tissue damage through the release of inflammatory mediators and recruitment of other inflammatory cells to the site of infection.

Investigations focused on elucidating how EPEC cause diarrhoea have mostly been carried out on in vitro model systems such as Caco-2 or T84 cells. However, the interplay of in vivo factors that result in diarrhoea is very complex and in vitro experiments only yield limited results [164]. Regardless the mechanism, diarrhoea is evidently beneficial to the A/E-lesion-causing EPEC as it disrupts normal host-prokaryote equilibrium, providing EPEC a competitive advantage over normal intestinal flora [143].

6 Concluding remarks

This review has aimed to chart the advances and breakthroughs in EPEC research. Remarkable progress has been made in our understanding of EPEC since it was first convincingly associated with human diarrhoea by John Bray in 1945 [9], including information on its clinical features, epidemiology, diagnosis and most fundamentally, pathogenesis. The integration of the diverse fields of cell biology, genetics and molecular biology have provided new insight into how pathogen-specific virulence factors could potentially exert adverse effects on a wide range of eukaryotic host cell processes like signal transduction, transcription and ion secretion. Research has also revealed features common to other enteric pathogens as well as strategies unique to A/E lesion-causing bacteria, such as possession of a filamentous type III secretion system (FTTSS) and the translocation of a bacterial receptor into the host membrane. Besides contributing to advances in the field of bacterial pathogenesis, EPEC has proven to be a useful system for cell biologists to study actin rearrangement and dynamics at the plasma membrane.

Models for EPEC pathogenesis have been proposed and refined over the years [69, 94, 116]. Nevertheless, much remains to be done to substantiate or refute these models before a comprehensive picture of EPEC pathogenesis emerges. As highlighted in this review, some controversial issues like the role of BFP in initial adherence or the contribution of Ca2+ signalling to A/E lesion formation have yet to be resolved. Based on experimental evidence, there have been suggestions that intimin binds to a host-cell receptor as well as Tir. However, the relevance of the putative host-cell intimin receptors for infection in vivo remains unknown. Recently, new LEE-encoded effector proteins have been discovered, including Map (mitochondrion-associated protein), EspG and EspH [165168]. Moreover, a large number of type III effectors that are encoded on loci other than the LEE have also been reported [169172]. Although functional roles have been assigned to some of these effector proteins, further research characterising these molecules as well as their role in causing diarrhoea within the host cell are essential and undoubtedly, there are many more bacterial effector proteins that remain to be identified.

Diarrhoea is the primary symptom of an EPEC infection but our knowledge of how A/E pathogen infection leads to the onset of diarrhoea is significantly lacking. Research in this area has proven that there is no one unifying mechanism of pathogenesis nor is there a single effector molecule responsible for EPEC-induced diarrhoea. Rather, several EPEC virulence factors perturb different facets of the host physiology, including the subversion of the host cell cytoskeleton and signalling pathways, resulting in the cumulative effect of diarrhoea.

This points the way forward in future EPEC research. in vitro models should be used to identify candidate virulent genes that deserve further study in vivo. EPEC has a narrow host specificity and is a human pathogen; however, the number of human challenge studies is very limited. There are some related bacteria that share similar virulence factors and mechanisms with EPEC and cause analogous diseases in animals. These include rabbit EPEC and the Citrobacter rodentium in rodents and provide convenient in vivo models to study EPEC pathogenesis.

Ultimately, the goal of EPEC research would be to come up with better strategies for the prevention and treatment of life-threatening diarrhoeal diseases. Many virulence properties of EPEC have been elucidated over the years and are potential targets for vaccination. Volunteer studies by Levine et al. [72] suggested that antibodies to the 94 kDa protein (intimin) correlated with protection against EPEC infection and that it was highly immunogenic. Stimulation of the intestinal immune response against this antigen could elicit protective immunity against EPEC infection. Alternatively, factors involved in bacterial adhesion, such as the components of the type III secretion system or translocation apparatus (EspB/EspD) could be targeted.

EPEC research has come a long way since its first description as a pathogen 60 years ago. With enhanced understanding comes the opportunity for the development of new therapeutic strategies and diagnostics that will help overcome the worldwide scourge of EPEC-mediated infantile diarrhoea.


We thank EMBO J. and ASM press for permission to reproduce published material, Stuart Knutton for providing un-published images and the Wellcome trust for the financial support.


  1. [1].
  2. [2].
  3. [3].
  4. [4].
  5. [5].
  6. [6].
  7. [7].
  8. [8].
  9. [9].
  10. [10].
  11. [11].
  12. [12].
  13. [13].
  14. [14].
  15. [15].
  16. [16].
  17. [17].
  18. [18].
  19. [19].
  20. [20].
  21. [21].
  22. [22].
  23. [23].
  24. [24].
  25. [25].
  26. [26].
  27. [27].
  28. [28].
  29. [29].
  30. [30].
  31. [31].
  32. [32].
  33. [33].
  34. [34].
  35. [35].
  36. [36].
  37. [37].
  38. [38].
  39. [39].
  40. [40].
  41. [41].
  42. [42].
  43. [43].
  44. [44].
  45. [45].
  46. [46].
  47. [47].
  48. [48].
  49. [49].
  50. [50].
  51. [51].
  52. [52].
  53. [53].
  54. [54].
  55. [55].
  56. [56].
  57. [57].
  58. [58].
  59. [59].
  60. [60].
  61. [61].
  62. [62].
  63. [63].
  64. [64].
  65. [65].
  66. [66].
  67. [67].
  68. [68].
  69. [69].
  70. [70].
  71. [71].
  72. [72].
  73. [73].
  74. [74].
  75. [75].
  76. [76].
  77. [77].
  78. [78].
  79. [79].
  80. [80].
  81. [81].
  82. [82].
  83. [83].
  84. [84].
  85. [85].
  86. [86].
  87. [87].
  88. [88].
  89. [89].
  90. [90].
  91. [91].
  92. [92].
  93. [93].
  94. [94].
  95. [95].
  96. [96].
  97. [97].
  98. [98].
  99. [99].
  100. [100].
  101. [101].
  102. [102].
  103. [103].
  104. [104].
  105. [105].
  106. [106].
  107. [107].
  108. [108].
  109. [109].
  110. [110].
  111. [111].
  112. [112].
  113. [113].
  114. [114].
  115. [115].
  116. [116].
  117. [117].
  118. [118].
  119. [119].
  120. [120].
  121. [121].
  122. [122].
  123. [123].
  124. [124].
  125. [125].
  126. [126].
  127. [127].
  128. [128].
  129. [129].
  130. [130].
  131. [131].
  132. [132].
  133. [133].
  134. [134].
  135. [135].
  136. [136].
  137. [137].
  138. [138].
  139. [139].
  140. [140].
  141. [141].
  142. [142].
  143. [143].
  144. [144].
  145. [145].
  146. [146].
  147. [147].
  148. [148].
  149. [149].
  150. [150].
  151. [151].
  152. [152].
  153. [153].
  154. [154].
  155. [155].
  156. [156].
  157. [157].
  158. [158].
  159. [159].
  160. [160].
  161. [161].
  162. [162].
  163. [163].
  164. [164].
  165. [165].
  166. [166].
  167. [167].
  168. [168].
  169. [169].
  170. [170].
  171. [171].
  172. [172].
View Abstract