Enteroaggregative Escherichia coli (EAEC) is emerging as a significant diarrheal pathogen in multiple population groups. Although most commonly associated with pediatric diarrhea in developing countries, EAEC is also linked to diarrhea in adults including HIV‐positive patients and travelers and has been a cause of food‐borne outbreaks in the industrialized world. Current data suggest that one set of virulence elements is not associated with all EAEC strains, but that combinations of multiple factors prevail. Pathogenesis is believed to be initiated with adherence to the terminal ileum and colon in an aggregative, stacked‐brick‐type pattern by means of one of several different hydrophobic aggregative adherence fimbriae. Some strains of EAEC may then elaborate cytotoxins including the plasmid‐encoded toxin and the enterotoxins, EAST1 and ShET1. An AraC homolog termed AggR regulates several genes contributing to fimbrial biogenesis in ‘typical EAEC strains’. AggR has now also been shown to regulate genes on a chromosomal island. Sequencing of the EAEC type strain 042 completed at the Sanger Center has revealed two other chromosomal islands that are being explored for their pathogenetic potential. This article reviews these virulence elements and presents on‐going areas of research in EAEC pathogenesis.
The majority of Escherichia coli are harmless commensals of the mammalian gastrointestinal tract, yet some strains have deviated from this ancestral niche, adapting specific virulence traits that enable them to cause disease in otherwise healthy individuals. The resulting clinical syndromes may include gastroenteritis, urinary tract infection, septicemia and meningitis. Escherichia coli enteritis may itself be caused by at least six distinct E. coli pathotypes: enterotoxigenic (ETEC), enteropathogenic (EPEC), enterohemorrhagic (EHEC), enteroinvasive (EIEC), diffusely adherent (DAEC) and enteroaggregative E. coli (EAEC). These pathotypes exhibit distinct clinical, epidemiologic and pathogenic profiles. Strains within each pathotype are characterized by shared virulence traits and can typically be further distinguished by O (lipopolysaccharide) and H (flagellar) antigens (Kaper , 2004).
EAEC was first described in 1985, recognized by its distinctive adherence to HEp‐2 cells in an aggregative, stacked brick‐like pattern. This adherence pattern, distinguishable from the adherence patterns manifested by EPEC and DAEC, was first significantly associated with diarrhea among children in Santiago, Chile in 1987 (Nataro , 1987). Laboratory studies on prototype strains revealed that the aggregative adherence (AA) phenotype was encoded on 55–65 MDa plasmids, collectively called pAA (Vial , 1988). In a subsequent report, a cryptic fragment of the plasmid was found to be useful as a specific and variably sensitive DNA probe, which could distinguish EAEC from other E. coli (Baudry , 1990). Many epidemiologic studies now use the AA phenotype and the plasmid‐encoded probe (termed CVD 432 or simply the AA probe) to identify EAEC (Huppertz , 1997; Germani , 1998; Okeke , 2003). Although the probe only identifies a fraction of strains that exhibit the AA adherence pattern, epidemiologic and pathogenetic studies suggest that these strains may be the true EAEC pathogens.
Since its discovery, an increasing number of studies have implicated EAEC in diarrhea in a variety of settings. These now include endemic diarrhea of infants in both industrialized and developing countries (reviewed in Okeke & Nataro, 2001; Huang & DuPont, 2004; Huang , 2004), persistent diarrhea among human immunodeficiency virus/acquired immunodeficiency syndrome patients and traveler's diarrhea. Several outbreaks of EAEC diarrhea have also been described. Prospective studies suggest that in well‐nourished populations in industrialized countries, EAEC may be a cause of endemic sporadic diarrhea. For example, a large prospective surveillance study in the UK implicated EAEC among the major pathogens at all ages (Wilson , 2001).
The implication of EAEC in diarrheal outbreaks confirmed the fact that at least some strains exhibiting the AA phenotype were true human pathogens, yet in not all studies were EAEC significantly associated with diarrheal illness. Much recent work focuses on identifying required virulence genes to better identify the true EAEC pathogens.
The clinical presentation of EAEC infection comprises watery diarrhea, occasionally with blood and mucus. Several studies have suggested that patients infected with EAEC manifest intestinal inflammation, marked by the presence of fecal lactoferrin and proinflammatory cytokines, notably interleukin (IL)‐8 (Steiner , 1998; Greenberg , 2002). Steiner (1998) have suggested that even asymptomatic carriage of EAEC strains can result in evidence of low‐level enteritis.
Pathogenesis studies of EAEC experienced a breakthrough in 1994, when the senior author published the first EAEC volunteer study (Nataro , 1995). In this report, we showed that only one of four AA probe positive EAEC strains elicited human diarrhea, confirming that not all EAEC strains were equally pathogenic. Investigations since that time in our laboratory have utilized strain 042, which elicited diarrhea in the majority of volunteers. Pathogenesis studies have suggested three major features of EAEC pathogenesis (Fig. 1): (1) abundant adherence to the intestinal mucosa, (2) elaboration of enterotoxins and cytotoxins, and (3) induction of mucosal inflammation. Evidence for each of these features is discussed below.
Studies using explants of viable colonic tissue harvested during intestinal biopsy suggest that EAEC adheres to both human ileal and colonic tissues (Knutton , 1992; Hicks , 1996). EAEC is adherent to intestinal tissues in an aggregative manner in association with a thick mucus layer (Tzipori , 1992; Hicks , 1996). Adherence is followed by mucosal toxicity, including crypt dilation, microvillous vesiculation, and epithelial cell extrusion (Fig. 2) (Hicks , 1996; Nataro , 1996). This histopathology has not yet been observed in natural human infections. Notably, no polymorphonuclear cell infiltrate has been reported for EAEC infection (Vial , 1988).
Virulence factors mediating pathogenesis
Adhesins and other plasmid‐encoded virulence factors
Cloning of the factor mediating the AA phenotype in prototype strain 17‐2 revealed the presence of bundle‐forming fimbriae termed aggregative adherence fimbria (AAF) (Nataro , 1992). The AAF conferred the AA phenotype on E. coli K12 strains, along with strong hemagglutination of human erythrocytes. Genes required for biogenesis of the AAF were found to be localized on the pAA plasmid of this strain. However, the strain that caused disease in volunteers, 042, was found to be nonreactive to gene probes derived from the AAF operon, and we subsequently reported that this strain adheres to the human colonic mucosa by virtue of a distinct allele, termed AAF/II (the fimbriae of 17‐2 were renamed AAF/I) (Nataro , 1995). Importantly, the major pilin subunit of AAF/II (called AafA) exhibited less than 25% amino acid identity with the major pilin of AAF/I (Czeczulin , 1997). Since the description of AAF/II, three other allelic variants have been described (Nataro , 1992; Czeczulin , 1997; Bernier , 2002), and notably, some strains manifest the AA phenotype, but do not harbor any of the known fimbriae, suggesting the presence of additional adhesins not yet characterized (Czeczulin , 1997; Bernier , 2002). Nonfimbrial adhesins are also likely (Monteiro‐Neto , 2003). It is important to note that in almost all cases, a particular EAEC strain expresses only one of the possible AAF alleles. Thus, like ETEC, EAEC apparently adheres and colonizes by virtue of antigenically heterogeneous adhesins, which will hamper any efforts to develop immunologic prophylaxis strategies.
The AAF biogenesis genes feature an organization similar to genes of the Dr family of fimbrial adhesins, which are adhesins of uropathogenic E. coli and DEAC (Bilge , 1989; Savarino , 1994; Elias , 1999; Bernier , 2002). In fact, the accessory genes of the Dr family are more closely related (30–70% amino acid identity) to those of the AAFs than are the major AAF pilin subunits to each other. In addition to the usher and chaperone proteins that are required for assembly of the Dr structures, AAFs also include a gene that has been suggested to encode a minor pilin subunit (aafB), which may mediate epithelial cell invasion (Bernier & Le Bouguenec, 2002). Some investigators have suggested that EAEC may be an invasive pathogen, but proven pathogenic strain 042 is notably noninvasive in in vitro studies.
Expression of AAF genes requires AggR, a transcriptional activator of the AraC class (Nataro , 1994). This protein regulates the genes involved in fimbrial biogenesis for both AAF/I and AAF/II. Genomic studies of the pAA plasmid revealed a small open reading frame (ORF) just upstream of the aggR gene in most EAEC strains. This gene was found to encode a 10 kDa secreted protein that was recognized by the sera of volunteers fed strain 042. A null mutant of this gene revealed a unique hyperaggregative phenotype, and scanning electron microscopy of the bacterial strains showed collapse of the AAF fimbriae onto the bacterial cell surface. The protein product of this gene was therefore termed antiaggregation protein (Aap), nicknamed dispersin (Sheikh , 2002). Dispersin is secreted to the surface of EAEC strains and binds noncovalently to the lipopolysaccharide (LPS) of the outer membrane. Recent data suggest that the mechanism of dispersin's effect may be mediated predominantly through its ability to neutralize the strong negative charge of the LPS, so that the AAFs, which carry a strong positive charge, are free to splay out from the surface and bind distant sites (J. Velarde, unpublished). Interestingly, the secretion of dispersin is dependent on the presence of an ABC transporter complex also encoded on plasmid pAA (Nishi , 2003). The genes encoding this transporter (the Aat complex) correspond to the site of the previously cryptic AA probe. As both dispersin and the Aat complex are under the control of AggR, the latter protein is emerging as the central regulator of virulence functions in EAEC. In recognition of this fact, we have begun to use the term typical EAEC to refer to strains that carry the AggR regulon.
The AggR regulon is not restricted to the pAA plasmid. Dudley et al. have recently found that in addition to these plasmid‐encoded genes, AggR regulates a chromosomal operon inserted on a pathogenicity island at the pheU locus (E. Dudley and J. Nataro, unpublished).
In an effort to identify cytotoxins and enterotoxins, Navarro‐Garcia (1998) analyzed culture supernatants from strains that caused outbreaks of EAEC diarrhea in Mexican hospitals. Two major proteins of 108 and 116 kDa, respectively, were secreted. Both proteins are autotransporter proteins and, thus, undergo two cleavage steps before a functional protein is released into the extracellular space. The passenger domains contain the protease motif (GDSGSP) characteristic of all proteins of the serine protease autotransporter of Enterobacteriaceae (SPATE) group (Henderson , 1998). The 108 kDa protein is termed plasmid‐encoded toxin (Pet) (Eslava , 1998). Strain 042, which expresses Pet, induced dilation of crypt openings and intercrypt crevices and rounding and extrusion of enterocytes in intestinal explants, whereas 042pet did not affect tissue in this way (Henderson , 1999b). Data suggest that the toxic effects of Pet are because of cytoskeletal rearrangements following Pet‐mediated cleavage of the membrane cytoskeletal protein spectrin (Villaseca , 2000; Canizalez‐Roman & Navarro‐Garcia, 2003). Although Pet may play a role in EAEC pathogenesis, it is only present in a minority of strains (Czeczulin , 1999) and was not correlated with damage to T84 enterocytes in culture (Henderson , 1999b).
Rises in short‐circuit current and potential difference have been reported for two other EAEC virulence factors using the Ussing chamber system (World Precision Instruments, Sarasota, FL). The first of these, EAST1, is a plasmid‐encoded enterotoxin with homology to the heat‐stable enterotoxin (STa) of ETEC and the endogenous signaling peptide, guanylin (Savarino , 1993). However, EAST1 is present in only a subset of EAEC and has a broad distribution among other pathogenic and nonpathogenic E. coli (Savarino , 1995; Zhou , 2002).
Chromosomally encoded virulence factors
The third enterotoxin that may contribute to EAEC diarrhea is the chromosomally encoded ShET1. The setBA genes, encoding ShET1, are contained entirely within and antisense to pic, which encodes the 116 kDa secreted protein discovered by Navarro‐Garcia (1998). The setA and setB genes are predicted to encode a holotoxin with an A1 : B5 stoichiometry (Fasano , 1995). ShET1 caused fluid accumulation in the rabbit ligated loop model in addition to activity in the Ussing chamber system. The mechanism mediating ShET1 enterotoxic activity is independent of cyclic adenosine monophosphate, cyclic guanosine monophosphate and Ca++, and may involve signaling through nitric oxide (Fasano , 1997) (A. Fasano, personal communication). While ShET1 is a putative enterotoxin, Pic has mucinolytic and hemagglutinin activities (Henderson , 1999a).
This locus is not unique to EAEC. The pic/setBA genes have also been found in Shigella flexneri 2a (Fasano , 1995) and uropathogenic E. coli (UPEC) CFT 073 (Heimer , 2004) with nucleotide sequence identities of 99% and 96%, respectively, compared with 042. Limited prevalence studies suggest that this locus may play a role in pathogenesis. In EAEC the prevalence of pic/setBA varies between 4.6% and 57% of strains (Czeczulin , 1999; Okeke , 2000; Vila , 2000; Piva , 2003; Sarantuya , 2004) and was associated with diarrhea in Brazilian children and adult travelers from Spain (Vila , 2000; Piva , 2003). In one study, set1A was amplified from 22 of 22 S. flexneri 2a strains tested, and rarely in other Shigella species and serotypes (Noriega , 1995). Heimer (2004) demonstrated a prevalence of pic in 31% of pyelonephritis isolates compared with 7% of fecal strains. However, using the CBA mouse model of ascending urinary tract infection, no colonization advantage was observed.
Basic local alignment search tool analysis of the 042 sequence assembled at the Sanger Center in the UK (http://www.sanger.ac.uk) reveals that pic/setBA resides on the 117 kb pathogenicity island inserted at the pheU tRNA locus (E. Dudley, personal communication). Sequences flanking pic/setBA include insertion sequence (IS) elements IS911 and IS629. An IS‐like orf, perD, and a cryptic prophage gene, L0015, are downstream of pic (Henderson , 1999a). Al‐Hasani (2001) have determined that the pic/setBA locus is contained within a pathogenicity island (PAI) termed she in S. flexneri YSH6000T inserted adjacent to the pheV tRNA gene. The IS elements flanking pic/setBA in EAEC differ from those within the she PAI, suggesting that gene acquisition was distinct for these two species.
Also inserted in the pheU island are gene clusters SciI and SciII that are about 20 kb each, the latter of which is under AggR control (Ed Dudley and J. Nataro, unpublished). SciII comprises a cluster of 16 predicted ORFs, which we have designated as aaiA‐P (for AggR‐activated island). Molecular studies have revealed an AggR‐dependent promoter upstream of aaiA, and have suggested that these 16 predicted genes may be transcribed as an operon. Except for aaiP, which putatively encodes a ClpB chaperone, no other ORF encodes a protein of well‐defined function. The contribution of the aai genes to EAEC pathogenesis is currently unknown, but these genes do not appear to play a role in adherence to abiotic surfaces or intestinal epithelial cells, as do previously described AggR‐regulated genes.
Other researchers have noted the presence of genes similar to aaiA‐P in other organisms, notably gram‐negative pathogenic and symbiotic bacteria (Folkesson , 2002; Rao , 2004). Rao (2004) identified such a group of genes in the fish pathogen Edwardsiella tarda, and designated them as evpA through evpH. Mutations in evpA or evpB increased the LD50 of E. tarda in blue gourami by nearly 2 log, and abolished the secretion of EvpC. Likewise in EAEC, we have found that AaiC is a secreted protein (E. Dudley and J. Nataro unpublished data). However, while sequence comparison between the EAEC and E. tarda systems identifies multiple orthologs, there is no detectable sequence similarity between EvpC and AaiC. One reasonable hypothesis is that each system encodes related proteins comprising a secretion apparatus, but that these systems secrete unrelated effector proteins, as is well documented with well‐defined secretion systems of gram‐negative bacteria such as the type III secretion system.
As sequencing and annotation of the genome of strain 042 are completed, additional potential virulence factors are being revealed. At least two additional pathogenicity islands are present. One of these, inserted at glyU, encodes a homolog of the ETT‐2‐type III secretion system of 0157:H7. A second island contains genes for a transcriptional regulator and type III effector proteins and is inserted in selC (J. Sheikh and J. Nataro, personal communication). Additionally, the Yersinia high‐pathogenicity island, containing the Yersiniabactin siderophore gene, is present in a majority of strains (Schubert , 1998).
Potential proinflammatory role of EAEC factors
In a search for factors that may be responsible for the inflammatory presentation of EAEC‐infected patients, Steiner (2000) reported that the flagellin of EAEC strains induced the release of IL‐8 from Caco‐2 cells in culture. IL‐8 release was subsequently linked to binding to the Toll‐like receptor 5 on the target epithelium (Hayashi , 2001). Steiner and others have also suggested a role for plasmid‐encoded factors in IL‐8 induction (Steiner , 1998; Jiang , 2002). Jiang (2002) reported that IL‐8 levels were higher in feces of patients infected with aggR‐ or aafA‐containing strains compared with those infected with strains negative for these factors. Recently, it was also shown that EAEC strains harboring aggR, aggA and aap were more likely to cause IL‐8 induction of >100 pg mL−1 from nonpolarized HCT‐8 IECs than EAEC negative for those genes (Huang , 2004). Our laboratory sought to identify additional factors that could account for IL‐8 release from epithelial cells infected with EAEC strain 042. In this study, polarized T84 intestinal cells were found to release IL‐8 even when infected with 042 mutated in the major flagellar subunit FliC. IL‐8 release from polarized T84 cells was found to require the AggR activator and the AAF fimbriae, and IL‐8 release was significantly less when cells were infected with mutants in the minor fimbrial subunit AafB (Harrington , 2005).
The AafB protein had previously been reported to be a homolog of a class of putative invasion proteins in uropathogenic and DEAC (Bernier & Le Bouguenec, 2002; Plancon , 2003; Cota , 2004). Plancon (2003) have suggested that the invasion phenotype may involve binding to β1 integrins on the cell surface. Further characterization of the role of the AAF adhesin and its subunits in inflammation is clearly warranted.
Gene regulation in EAEC
The central role of AggR in coordination of EAEC virulence genes confers a strong priority to understand how the expression of this protein is itself regulated. Recent studies in the authors' laboratory have revealed that AggR is essential for its own expression. The protein binds to a specific binding site in the vicinity of the promoter −35 site. A second AggR binding site is suggested by DNA footprinting, though no role for a second site can be shown in gene expression studies (R. Sohoni and J. Nataro unpublished observations). The presence of essential autoactivation raises the important question: how is aggR expression initiated and terminated, in light of the fact that the AggR regulon is preferentially expressed in logarithmic phase and in rich media. The answer to this question came with the direct implication of the E. coli global regulator FIS in aggR activation. FIS, which is itself strongly growth phase dependent, binds to a promoter upstream region close to or overlapping that of AggR, and FIS mutants, which do not contact RNA polymerase, are not able to activate aggR expression. Thus, though AggR may initiate a positive feedback loop amplifying its own expression, the requirement for FIS in a putative nucleoprotein‐activating complex at the aggR promoter constrains the positive feedback loop. The absolute requirement for AggR in its expression also suggests that some degree of promoter leakiness must be tolerated. Our data show that basal levels of aggR expression may be controlled by the repressive actions of the nucleoid‐associated protein H‐NS acting upstream of the aggR promoter. Thus, the aggR promoter is subject to a regulatory scheme that assures its rapid and abundant expression shortly after entering the gastrointestinal tract.
Summary and Conclusions
EAEC has emerged as an important pathogen causing diarrheal disease in multiple epidemiologic and clinical settings. Strain heterogeneity has impeded recognition of this pathogen for many years, but pathogenesis studies are illuminating virulence factors and an unusual degree of heterogeneity among strains in their carriage of putative virulence factors. Better understanding of the roles of individual factors, and of the possibility that different factors could confer distinct clinical presentations, will provide further insights into the true contribution of EAEC as a human pathogen.