Adhesins of Diffusely Adherent and Enteroaggregative Escherichia coli

CHANTAL le BOUGUÉNEC^1* AND JAMES P. NATARO²

[SECTION EDITOR: JAMES P. NATARO]

Posted July 25, 2005

Pathogénie Bactérienne des Muqueuses, Institut Pasteur, 28 rue du Docteur Roux, 75724 Paris Cedex 15, France,¹ and Center for Vaccine Development, Departments of Pediatrics, Medicine and Microbiology & Immunology, University of Maryland School of Medicine, 685 West Baltimore Street, Baltimore, MD 21201²

*Corresponding author. Phone: 33 1 40 61 3280, Fax: 33 1 40 61 3640, E-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it

INTRODUCTION

Escherichia coli is a normal inhabitant of the intestinal tract of humans. However, pathogenic E. coli strains cause intestinal or extraintestinal infections. A classification of diarrhea-causing E. coli has been established based on clinical aspects of the disease and on the identification of the virulence factors produced by the strains. Six categories have been defined (58, 71). However, in all cases, the bacteria have to adhere to host cells to allow microbial colonization and invasion of mucosal epithelia. This pathogenic property is associated with the production of adhesins, and different adhesins are associated with different diseases. Historically, testing adherence to HEp-2 and HeLa cells in tissue cultures has been used to differentiate diarrheagenic E. coli isolates that did not produce heat-labile or heat-stable enterotoxins (enterotoxigenic E. coli) or Shiga-like toxins (enterohemorrhagic E. coli), did not belong to standard enteropathogenic O serogroups, and were noninvasive (enteroinvasive E. coli). Such enteroadherent E. coli have been classified into three subtypes according to their adherence patterns upon coincubation with epithelial cells. Enteropathogenic E. coli (EPEC) have a localized adherence pattern, recognized by the presence of microcolonies on the surface of epithelial cells. EPEC are a common cause of diarrhea in infants in developing countries. Enteroaggregative E. coli (EAEC) display an aggregative adherence "stacked-brick" pattern of adherence (Fig. 1B). Epidemiological studies have implicated EAEC strains in acute and persistent diarrhea in children, in food-borne diarrhea outbreaks, and in traveler’s diarrhea, and this group is recognized as an emerging pathotype of enteric disease. Several recent studies have reported an association between EAEC strains and persistent diarrhea in patients infected with the human immunodeficiency virus in developing and industrialized countries (72, 80). Members of the third group exhibit a diffuse adherence pattern on HEp-2 cells (Fig. 1A). These diffusely adherent E. coli (DAEC) that have been implicated as a cause of diarrhea, especially in children more than 2 years old, in both developing and developed countries (32, 33, 60, 62, 92). Although EAEC and DAEC strains appear to have different molecular equipment for attachment to host cell surfaces, identification and characterization of the gene clusters encoding adherence evidenced close relatedness between those determinants most frequently detected in isolates belonging to these two pathotypes of diarrheagenic E. coli.

Fig. 1Adhesion properties and adhesive cell surface structures of DAEC and EAEC strains. Photomicrographs of HeLa cells infected with the DAEC isolate AL845 (A) and the EAEC isolate 55989 (B). Electron micrographs of DAEC and EAEC strain preparations negatively stained with 1% ammonium molybdate and shadowed with platinum. No fimbriae were visualized on the cell surface of the DAEC strain AL845 (C). In contrast, thin fimbriae were detected in the DAEC strain F1845 (E). The EAEC isolate 55989 produces both flagella and long, flexible fimbriae (arrows), which are involved in the cohesion of bacterial aggregates (D and E).

Both EAEC and DAEC are heterogeneous; strains of both groups produce diverse adhesins, and fimbrial and afimbrial adhesins have been described in both pathotypes. EAEC prototype strains that display a characteristic aggregative adhesion (AA pattern) adhere to HEp-2 and HeLa cells; the bacteria adhere to each other to form a "stacked brick wall"-like structure on epithelial cells and glass coverslips (Fig.1B and D). This adhesion is mediated by aggregative adherence fimbriae (AAF) encoded by gene clusters related to the afa operons in both DAEC and uropathogenic E. coli isolates. Although three types of AAF have been characterized, the distribution of the genes encoding these fimbriae in EAEC strains indicates that only a minority of isolates produces these types of adhesive structures. This suggests the existence of other AAF-related structures and of other as yet uncharacterized adhesins. The first adhesin characterized in DAEC strains was the fimbrial F1845 adhesin (19). Four DAEC-specific adhesins have now been described: the fimbrial and afimbrial F1845-related adhesins called "Dr adhesins," the AIDA-I system (8), a 57-kDa adhesion-mediating protein (105), and the CF16K molecule (46).

MAJOR DAEC ADHESINS: THE Dr FAMILY OF PROTEINS

DAEC strains are a heterogeneous group of E. coli isolates, many of which express the related so-called Dr adhesins. The single designation is based on the identification of one similar cellular receptor for all these proteins. Although structurally different, they all recognize the Dr human blood group antigen on the decay-accelerating factor (DAF or CD55) (61, 77). These adhesins are encoded by a family of closely related operons, the first characterized and sequenced being the afa operon. Consequently, it has been suggested that this group of DAEC strains producing such adhesins be named the Afa/Dr DAEC family (40).

Afa Family of Adhesins

The daa (diffuse adherence adhesin) gene cluster was described in 1989 in DAEC strain C1845, the causative agent of a persistent diarrhea in a child (19). It codes for the fimbrial F1845 adhesin (19) (Fig. 1E). The daa operon has the same genetic organization as, and is closely related to, the afa gene clusters, which were described in 1984 in uropathogenic E. coli and were reported to code for afimbrial adhesins (Afa adhesins) (51, 52, 53) (Fig. 1C). The AFA family includes adhesins that are expressed by both extraintestinal pathogenic and diarrhea-associated E. coli strains (61). The Dr hemagglutinin (encoded by the dra operon) produced by uropathogenic strains (79) belongs to this family of determinants that have closely related DNA sequences. Consequently, the probes (2, 3, 19, 79) and PCR assays (59, 61) based on the sequence of the daaC and afaB-C genes that were described also detect the presence of all the operons of the family carried by DAEC strains (61) (Fig. 2). The afa operons are found both on plasmids and on the chromosome (19, 53, 55, 60, 106). It has been demonstrated that the afa-3 operon flanked by two directly oriented IS1 elements can translocate from a recombinant plasmid to the E. coli chromosome (29). In two strains, afa genes were associated with pathogenicity islands (51, 57).

Fig. 2Genetic organization of the afa gene clusters. (A) Schematic representation of the genetic determinants and role of the Afa products. The DNA fragments specifying the operons are indicated (afa and daa probes and afa PCR). (B) Backscattered electron imaging of double immunogold staining of AfaE and AfaD proteins produced by a DAEC isolate. AfaE was probed with 10-nm gold particles; AfaD was probed with 15-nm gold particles.

Genetic Organization.

Sequence analysis of the afa-3 gene cluster (29), encoding the afimbrial AfaE-III adhesin, indicates that the afa operons are composed of six genes (afaA to afaF) organized in two divergent transcriptional units consisting of afaA to afaE and afaF, respectively (Fig. 2). Five of the six Afa products are similar to proteins encoded by other adhesion systems, allowing the attribution of a putative function in the biogenesis of the adhesive sheath to each of them. The biogenesis of the AFA adhesive structure, like that of classical fimbrial adhesive structures (45), requires proteins with specialized functions, including a periplasmic chaperone (AfaB protein), an outer membrane anchor protein (AfaC), and transcriptional regulators (AfaA and AfaF) (29). The afaA and afaF genes encode products that belong to the family of PapI-PapB regulators commonly involved in the regulation of E. coli fimbrial adhesin expression (18, 29). The afaE gene encodes the structural adhesin. No function could initially be attributed to the AfaD product encoded by a gene that is nonessential for the expression of the adhesive properties (54). The role of AfaD was then studied by immunogold and immunofluorescence experiments. Like AfaE, AfaD was found to be a surface-exposed protein as well as an adhesin; and AfaE and AfaD are concomitantly expressed by the bacterial cell (27, 38) (Fig. 2). The roles of AfaE and AfaD during the association with epithelial cells were investigated by ultrastructural and genetic analyses. This revealed that afa-expressing bacteria could be internalized into various cell lines (see Fig. 5). The examination of mutants and polystyrene beads coated with either of the purified proteins demonstrated that AfaE is involved in initial binding to cells and that AfaD is involved in the internalization process (47) (see Fig. 5D). Thus, the afa operon is the first known example of a single operon encoding both adhesion of the bacteria to the epithelial cells and invasion of these cells. Structural and functional conservation of the AfaD invasins was demonstrated by comparing the proteins produced by diverse strains (28).

Fig. 5Invasive properties of DAEC and EAEC isolates. (A to D) DAEC strains; (E) EAEC isolate. (A) Transmission electron micrograph of HeLa cells infected with DAEC isolate A30. (B) Transmission electron microscopy of piglet ileal explants infected with an E. coli strain expressing the afa-3 operon. Some bacteria are internalized. (C) Microscopic examination of Giemsa-stained HeLa cells 2 days after infection with DAEC strain A30. Intracellular bacteria clustered within large vacuoles. (D) Transmission electron microscopy of intestinal Caco-2 cells infected with AfaD-coated beads. Beads enter the cells. (E) Transmission electron micrograph of HeLa cells infected with EAEC isolate 55989. Arrows indicate intracellular bacteria and beads.

afaE Adhesin-Encoding Genes.

Unlike the other afa genes, the afaE gene is highly heterogeneous, leading to the production of antigenically distinct adhesins (53). Gene subtypes afaE1, afaE3 (and draE, which is 99.4% identical with afaE3), afaE5, and daaE have been found in both isolates from diarrhea and uropathogenic human isolates, suggesting that, regardless of the afaE subtype, strains expressing these operons may cause both intestinal and urinary tract infections (61, 106). Indeed, an AfaE-I-producing clone lacking other virulence factors has been reported to be the causative agent of both diarrhea and cystitis (31). The afaE1, afaE3, and afaE5 subtypes are the most frequent of the strains associated with diarrhea (61). The afaE8 subtype that was first identified in animal pathogenic E. coli strains and was then associated with human extraintestinal isolates encodes an adhesin that does not recognize the Dr human antigen (Afa/Dr^- adhesin) (30, 56, 57). However, no DAEC strains belong to this subtype (61).

Transcriptional and Translational Regulation.

A model for the regulation of expression of the afa operons was deduced from studies on the daa operon (Fig. 3). Low temperature, high osmolarity, glucose as a carbon source, and rich medium repressed transcription and H-NS is a central regulator in the response to these environmental cues (104). In addition to this environmental regulation, the daa operon is also regulated by a phase variation mechanism that is under the control of the global regulators deoxyadenosine methylase (Dam) and leucine-responsive regulatory protein (Lrp) as well as operon-specific proteins (18, 100). Posttranscriptional differential regulation of the daa genes has also been observed. It occurs through endoribonucleolytic cleavage of the daaA-E transcript upstream from daaE followed by a rapid degradation of the upstream mRNA (with respect to the cleavage site); the daaE message is stabilized. Translation of a small reading frame, daaP, which flanks the cleavage site, was shown to be required for the daa mRNA processing. The following model was proposed: during translation of daaP, a tripeptide sequence (Gly-Pro-Pro) within the nascent DaaP product interacts with the ribosome, triggering cleavage of the associated mRNA at a fixed distance upstream (Fig. 3) (18, 63, 64, 100).

Fig. 3Model of regulation of expression of the daa operon.

Dr Adhesin Architecture

Two classes of adhesins have been described. Most E. coli isolates produce adhesins that are associated with bacterial cell surface structures known as fimbriae, which can be visualized by electron microscopy (EM). However, in some cases, adhesins have been described as being associated with an amorphous outer membrane-associated structure termed the afimbrial sheath (96). The Afa/Dr family of adhesins is interesting because it contains both afimbrial and fimbrial members (Fig. 1C and E). The afimbrial members of the Afa/Dr family (AfaE-I, AfaE-II, AfaE-III, AfaE-V, Dr-II) were classified as such because surface fimbriae could not be detected by EM (27, 53, 55, 60, 86). Other members, including the Dr hemagglutinin and F1845 fimbriae (19, 79), however, have been subsequently shown to exhibit a fimbrial morphology. Although, substantial similarities between products of afa accessory genes (AfaB and AfaC) and proteins implicated in fimbrial structure biogenesis suggest a similar mechanism of construction, none of the AfaE proteins possess homology with orthodox fimbrial domains. Various studies have tried to elucidate how these adhesins assemble into surface organelles with different morphologies. Microscopic examination of chimeras between the afa-3 and daa operons encoding afimbrial and fimbrial structures, respectively, indicated that the morphology of the adhesive architecture depends on the sequence of the adhesin-encoding gene (27). The debate has been further stimulated by the recent description of an AfaE-I adhesin associated with a fine fibrillar structure in an enteropathogenic E. coli strain (51). It is not clear how the Afa/Dr adhesins assemble into afimbrial and fimbrial structures. An atomic-resolution model, recently proposed (1, 1a), reveals a common structural basis for assembly of AfaE adhesin molecules by a process previously termed donor strand complementation; subunits possess an immunoglobulin fold structure missing one of the beta strands but possessing an additional N-terminal extension strand (of 18 amino acids). The N-terminal extension mediates linking of successive subunits by complementing the missing strand in the immunoglobulin fold. The periplasmic chaperone serves to protect the groove into which the extension will ultimately insert. This complementation is also exploited by the chaperone as its intraperiplasmic role (90). AfaE-III adhesins assemble into a flexible fiber that provides the link between the bacterial usher (AfaC) and the invasin (AfaD) at the tip (1). Consequently, it seems that the afimbrial pattern of adhesins, including AfaE-III, results from a collapse of the fine fibrillar structures on the bacterial surface.

Adhesion and Induction of Signal Transduction Properties

The Afa/Dr adhesins mediate mannose-resistant agglutination of human erythrocytes (MRHA) expressing the Cromer blood group antigen Dr(a) on DAF (77). The adhesin receptor has been identified as the short consensus repeat-3 (SCR-3) domain of the DAF (85, 101). A rare polymorphism of human DAF containing a point mutation in SCR-3 (Ser165-Leu), designated Dr(a-), abolishes binding of Afa/Dr adhesins (78). DAF is also one of the cell membrane proteins that regulate the complement cascade and protect eukaryotic cells from autologous complement-mediated damage by preventing the formation and accelerating the decay of C3 convertases (66). The classical pathway convertase and the Afa/Dr adhesin-binding regions are both localized on the SCR2-SCR3 interface of DAF. The importance of the SCR-3 of DAF in both binding the Afa/Dr adhesins and the regulation of complement was recently investigated by mutagenesis of 11 amino acids close to Ser165. This led to the suggestion that Ser155, and not Ser165, is the key residue of the SRC-3 domain for adhesin binding (41). Furthermore, two close but distinct regions of SCR-3 in DAF, approximately 20 Å apart, seem to be involved in adhesin binding and complement regulation, which raises the possibility that binding of adhesins might interfere with complement regulatory function (41, 84a).

DAF is a 70-kDa glycoprotein that is widely distributed on hematopoietic cells, intestinal and urinary epithelial cells, and endothelial cells. The density of DAF molecules, however, depends on the cell type, and DAF molecules show tissue and host tropism (65). DAF not only has a role in complement regulation but also in cell signaling (74, 95). Afa/Dr adhesin strains mediate a specific attachment to human cell lines expressing DAF molecules, especially intestinal cell lines (12). This adhesion results in a dense accumulation of DAF molecules beneath adherent bacteria (36, 37, 40, 61). A fluorescent DAF-staining test revealing this accumulation of DAF molecules on infected HeLa cells has been proposed as an alternative to the PCR assay for detection of Afa/Dr E. coli strains (39, 61) (Fig. 4A).

Fig. 4Model of AfaE-induced signal transduction. (A) Recruitment of DAF and carcinoembryonic antigen-related molecules. (B) Microvillus injury and proinflammatory responses.

In addition to recognizing the DAF molecule, some members of the Afa/Dr family of adhesins (the Afa/Dr-I subfamily including Dr, F1845, and AfaE-III adhesins) also recognize various carcinoembryonic antigen-related molecules (CEACAM) including biliary glycoprotein (BGP, CEACAM1, CD66a), carcinoembryonic antigen (CEA, CEACAM5, CD66e), and nonspecific cross-reacting antigen (NCA, CEACAM6, CD66c) as receptors (11, 40) (Fig. 4A). Like DAF, CEACAM1, CEACAM5, and CEACAM6 were recruited around adhering Afa/Dr DAEC strains in detergent-insoluble microdomains. It is interesting in terms of signaling that DAF and CEA are recruited during the initial step of adhesion, together with the Afa/Dr adhesin-dependent mobilization of raft-associated molecules, including ganglioside GM1 and VIP/21 caveolin (48).

In differentiated intestinal epithelial cells, the binding of Afa/Dr DAEC strains to DAF is followed by microvillus injury (13) (Fig. 4B). Brush-border lesions result from substantial rearrangements of apical cytoskeleton proteins, in particular, F-actin, villin, and fimbrin, all proteins important to the organization of the brush-border integrity (83). A DAF-associated signal transduction cascade promotes these cytoskeleton rearrangements through a Ca²⁺-dependent mechanism (84). Recent data show that recognition of CEA and CEACAM6 by the Afa/Dr-I adhesins induces signal transduction via activation of the Rho GTPase Cdc42, allowing elongation of cell surface microvilli-like extensions (11). Recognition of DAF by the bacteria also promotes the microvilli-like extensions (Fig. 4B). In the intestinal barrier, lesions in tight junctions have been observed and shown to result from rearrangements in at least two tight-junction-associated proteins, ZO-1 protein and occludin (82). Binding is also associated with a change in the distribution of functional brush-border-associated proteins controlling absorption or secretion functions (81, 83). Finally, although a correlation between Afa/Dr DAEC intestinal infection and inflammatory bowel diseases has not yet been established, recent work indicates that proinflammatory responses follow Afa/Dr infection in cultured human epithelial cells (Fig. 4B). One recent report indicates that the binding of the AfaE adhesin and the SCR-3 domain of DAF on infected human intestinal Caco-2 cells leads to an increased expression of the major histocompatibility complex class I-related MICA, a molecule central to innate immune responses (99). Moreover, the apical recognition of the SCR-3 domain of DAF on colonic T84 cultured epithelial cells by Afa/Dr adhesins leads to adhesin-receptor-dependent cell signaling sufficient to trigger proinflammatory stimuli in vitro. These responses include interleukin 8 (IL-8) production and polymorphonuclear (PMNL) transepithelial migration (16). Production of tumor necrosis factor alpha (TNF-α) and IL-1β follows PMNL migration, and these two cytokines upregulate the expression of DAF, the receptor of Afa/Dr adhesins, thus increasing the brush-border adhesion of DAEC strains (17). Both DAF and MICA molecules are more abundant than normal on the surfaces of epithelial cells in colonic biopsies from patients with Crohn’s disease (15, 99). These various observations suggest that Afa/Dr DAEC-induced MICA expression may be significant in the pathogenesis of inflammatory bowel disease.

One member of the Afa/Dr adhesin family, the Dr hemagglutinin, is unique because it also binds to another receptor, the 7S domain of type IV collagen (103). Binding by the Dr hemagglutinin (to both DAF and collagen type IV) is inhibited by chloramphenicol, a property not shared by other Afa/Dr adhesins (77). Mutagenesis and functional experiments suggest, first, that the collagen type IV-binding site consists of multiple adhesin subunits and is thus unlike the DAF-binding site that is contained within a single Dr adhesin molecule, and second, that residues 63 to 81 may be essential for the function of this binding site (20, 98, 102). Using recent structural data, Pettigrew et al. (84a) proposed that the chloramphenicol activity on Dr hemagglutinin is due to a direct disruption of the DAF-binding site on the adhesin, with the chloramphenicol masking a portion of the Dr molecule involved in the formation of the complex.

Invasion Properties

Pathogenic E. coli expressing afa operons may be associated with persistent diarrhea and chronic or recurrent extraintestinal infections (4, 19, 25, 34). One explanation for the development of such infections is that, after entering epithelial cells, Afa isolates may create special niches that allow bacteria to persist intracellularly, thus forming reservoirs in some tissues. These bacteria are not true invasive pathogens, unlike Salmonella and Shigella. However, several reports demonstrate that various Afa strains are able to enter epithelial cells in vitro (35, 39, 47, 93) and can enter differentiated enterocytes after infection of piglet ileal explants (43) (Fig. 5A and B). Of interest, at least two Afa products (AfaD and AfaE) contribute to different internalization processes, evidence of the complexity of the interactions between afa-expressing bacteria and epithelial cells.

Several experiments implicate AfaD proteins in the internalization processes into both HeLa and undifferentiated Caco-2 cells (28, 47, 87). This entry does not totally depend on an adhesion step mediated by the AfaE adhesin (87). The characteristics of the AfaD-mediated entry are consistent with intracellular Afa bacteria being able to form an intracellular reservoir for subsequent infection cycles. This strategy has previously been described for type 1 fimbriae-producing E. coli strains associated with recurrent cystitis (69). Indeed, only a small percentage of bacteria were internalized, but a viable pool of intracellular bacteria was maintained for several days. The intracellular bacteria are contained within inclusions in which bacteria appeared filamentous, suggesting slow intracellular bacterial multiplication (Fig. 5C). The bacteria-containing vacuoles became so large that it seemed impossible that the host cell could survive, allowing release of bacteria into the extracellular environment (87). Various data might explain the low level of bacterial internalization. AfaD, recently localized at the tip of the flexible fibrillar structure formed by AfaE molecules (1) (see above), was previously shown to be able to detach from the bacterial surface (47). The AfaD concentration is important for the invasion process, and therefore only bacteria having sufficient AfaD necessary for internalization at their surface may be able enter the cells.

The AfaD invasins, like invasin from enteropathogenic Yersinia (44) and intimin produced by enteropathogenic E. coli (26, 49), bind to β1 integrins (39, 87). These adhesion molecules are on the basolateral parts of epithelial cells, raising the question of how they become accessible during the infection. Various observations reporting induction of signal transduction and cellular damage during infection of intestinal cells by Afa bacteria (see above) suggest that bacteria could interact with basolateral surfaces. Furthermore, it was recently shown that β1 integrins are recruited at the site of bacterial-epithelial cell interaction (87).

Entry mediated by the adhesin has also been reported for the Dr hemagglutinin and the AfaE-III proteins (35, 87). Selvarangan et al. demonstrated that the SCR-3 domain of DAF was required for Dr hemagglutinin-mediated internalization and that glycosylphosphatidylinositol anchorage was more efficient than chimeric transmembrane anchorage of DAF in providing entry into epithelial cells (93). In contrast, although AfaE-III was also implicated in entry of clinical isolates, it seems that this protein plays an indirect role in the internalization process (87). AfaE-III and Dr hemagglutinin are 99.4% identical (60). The different abilities of these two adhesins to promote entry are presumably due to this minor divergence (37, 87).

MAJOR EAEC ADHESINS: THE AGGREGATIVE ADHERENCE FIMBRIAE

The phenotype of aggregative adherence (AA) to cultured epithelial cells, involving autoaggregation of bacteria in liquid cultures and adhesion to the intestinal mucosa, requires the presence of 60- to 65-MDa transmissible plasmids (pAA plasmids) (23). Despite the fact that these pAA plasmids are heterogeneous with regard to AAF and toxin expression (EAST1 and Pet toxins), they harbor several conserved genetic loci and belong to a single family (23). An EAEC probe was empirically derived from one of the conserved loci (5). For a long time, no protein product or putative function was known for the genes. However, it was recently shown that this sequence encodes a putative ABC transporter system required for dispersin export (76) (see below).

Three distinct but closely related gene clusters coding for phenotypically and morphologically distinct AAF have been characterized. In each case, electron microscopy revealed that bacterial surfaces were surrounded by long, relatively flexible fimbrial structures (2 to 5 nm thick) (14, 22, 70). Fimbriae seemed to be involved in autoagglutination, facilitating contact between bacteria for aggregate formation (Fig. 1D and F). Unlike AAF-I and AAF-II, which form bundles, AAF-III is mostly observed as individual filaments (Fig. 1F).

AAF-Encoding Gene Clusters

The three AAF-encoding gene clusters (agg, aaf, and agg-3 coding for AAF-I, AAF-II, and AAF-III fimbriae, respectively) display a very similar genetic organization and a high degree of DNA sequence similarity (Fig. 6). They all are carried on pAA plasmids and linked to the astA gene, which encodes the EAST1 toxin. The adhesion-encoding operon is followed by the toxin gene, which in turn is followed by the regulator gene. These genes are tightly linked to sequences reminiscent of the sequences of various insertion elements. Thus, the genetic linkage of adhesion and toxin-encoding gene blocks on pAA plasmids may be the consequence of multiple recombination events, and this in turn suggests that such topological associations of virulence genes may contribute to determining the pathogenic potential of an isolate. In each case, a gene encoding the AggR product has been detected downstream from the operon. AggR is a transcriptional activator of the AraC class (73). Positive regulation by AggR on the agg and aaf gene clusters has been demonstrated but there is no evidence that it is involved in AAF-III production. AggR is one of the most prominent factors on pAA plasmids; it is also present in most EAEC strains that do not express any identified AAF (23). It was recently proposed that the term "typical EAEC" should be reserved for strains carrying AggR and at least a subset of aggR-regulated genes and that the term "atypical EAEC" be used for strains lacking the AggR regulon (50).

Fig. 6Schematic representation showing the similar genetic organization of the three AAF-encoding gene clusters and the afa operons.

Genetic Organization.

Sequence analysis of the three gene clusters indicates that the operons encoding AAFs are similar to afa operons (14, 24, 91) (Fig. 6). They all are composed of two transcriptional units in divergent orientation that are noncontiguous and separated by 9, 6, and 5 kb in agg, aaf, and agg-3, respectively. Region 1 in both agg and agg-3 gene clusters comprises four open reading frames (the D, C, B, and A genes) transcribed in the same orientation, and mapping to the same loci as the afaB, afaC, afaD, and afaE genes in the afa operon. Three of the products (D, C, and B) are significantly similar to Afa products (periplasmic chaperone AfaB, usher protein AfaC, and AfaD invasin, respectively). The fourth gene (A gene) was identified as the adhesin subunit gene. Region 2 located downstream from the adhesin gene (9 and 5 kb in agg and agg-3, respectively) contains aggR encoding the regulator. The aaf gene cluster features a unique organization in which the chaperone (aafD), adhesin subunit (aafA), and transcriptional activator (aggR)-encoding genes lie in one cluster, whereas the second, unlinked cluster comprises a silent chaperone gene (aafD'), the usher gene (aafC), and an AfaD invasin-related gene (aafB) (24). Whereas the genes most similar to the aaf region 1 genes are the agg genes, those most similar to the aaf region 2 are afa genes (24). This suggests that the two regions originated independently and are not the result of evolutionarily recent duplication. In the three gene clusters, the biogenesis-associated products (D and C) are highly conserved and the sequence similarities extend over the full length of the proteins. Agg-3D shows 63 to 67% identity with AggD, AafD, and AfaB; and the Agg-3C product displays 65 to 67% identity with AggC, AafC, and AfaC (14). An AAF probe corresponding to an internal agg-3C fragment was defined from an alignment of the three gene cluster sequences. This probe is effective for identifying strains carrying AAF-encoding operons (14).

The AggB, AafB, and Agg-3B products are similar to AfaD invasins, suggesting that they also may be invasins. They all contain the highly conserved 8-amino-acid region (ATGRXXCR) described for all proteins of the family (28). Agg-3B is 58, 59, and 54% identical with AggB, AafB, and AfaD-III, respectively. AafB is 56% identical with AggB and 59% identical with AfaD-III. AggB is 51% identical with AfaD-III. Like Afa/Dr DAEC isolates, some EAEC strains invade HeLa cells at low frequency (6, 14) (Fig. 5E). It is still unknown, however, whether this bacterial invasion is specifically promoted by AAF products related to AfaD invasins. This is plausible because, like AfaD invasins, both AggB and Agg-3B proteins can promote internalization of polystyrene beads into epithelial cells (C. Bernier and C. Le Bouguénec, Abstr. 102^nd Gen. Meet. Am. Soc. Microbiol., 2002, and reference 28) and the AggB product complements an AfaD-negative mutant (28). AAF-producing EAEC, like Afa/Dr DAEC strains, are enteric pathogens associated with persistent diarrhea (72). The ability of these isolates to invade epithelial cells may reflect an evolutionary strategy involving establishing bacterial reservoirs and persistence within the host. However, only two studies report the potential of EAEC strains to be internalized in HeLa cells (6, 14). The relative lack of clinical evidence of invasiveness may be because of poorly efficient internalization into epithelial cells both in vivo and in vitro.

Adhesin-Encoding Genes.

The adhesin structural genes are highly divergent. They display no significant similarity to any protein sequence in the database. However, alignments of the AggA, AafA, and Agg-3A sequences highlighted the conservation of residues described as typical of the family of adhesins encoded by the afa operons. The conserved residues include 2 cysteines separated by 31 to 33 amino acids in the N-terminal moiety of the mature product and a glycine near the C terminus (14, 22). The distribution of the agg, aaf, and agg-3 gene clusters among EAEC strains is highly variable and depends on the geographical area and the tools used for the typing (89). Moreover, a high degree of heterogeneity of the fimbrial subunits encoded by the three operons has been suggested (14, 88). It would be very informative to type the fimbrial subunits encoded by the various AAF-encoding operons, as a mean of identifying the subunits associated with pathogenicity. No receptor has yet been identified for these bacterial adhesins.

Transcriptional Regulation.

The aggR genes in the three gene clusters are highly conserved (99 to 100% identity), suggesting a similar mechanism of action for this regulator highly related to DNA-binding proteins of the AraC family (73). In the agg gene cluster, AggR functions as an activator of aggA expression under a wide variety of conditions, including variations in growth medium, temperature, and oxygen tension, but it is inactive at extremes of temperature and in minimal medium (73). In the aaf gene cluster, AggR regulates transcription of aafA and aafD in region 1. AggR apparently acts directly to regulate the aafD promoter; however, the role of AggR in aafA transcription is not clear (24). The mechanism of regulation described for the afa operons implicates mRNA processing and enhanced stability of the resulting adhesin-encoding gene transcript due to the presence of a downstream hairpin structure. Consistent with these data, a candidate hairpin has been identified downstream from aafA (24).

Biofilm and Persistence

A recently described protein called dispersin forms a loosely associated layer on the surface of EAEC strains. Dispersin seems to counter the strong aggregating effects of the AAF adhesin, perhaps facilitating spread across the mucosal surface or penetration of the mucus layer. This protein is encoded by the aap (antiaggregative protein) gene immediately upstream from aggR in EAEC strain 042 (94). Export of dispersin from the bacteria requires a specific ATP-binding cassette transporter system encoded by a genetic locus on the pAA plasmid. This aat locus comprises a cluster of five genes recognized by the EAEC probe (76). Transcription of both the dispersin gene and the aat genes depends on AggR (76).

MINOR DAEC AND EAEC ADHESINS

Adhesin involved in diffuse adherence (AIDA) has been isolated and characterized from the E. coli diarrhea isolate 2787 (0126:H27) (8). It belongs to the family of autotransporter proteins that are used by a range of pathogenic gram-negative bacteria to transport virulence factors to the extracellular environment (42). Two genes (aah and aidA) located on a ~100-kb plasmid are involved in AIDA activity: aidA codes for the AIDA autotransporter system, which is synthesized as a pre-pro-protein of 132 kDa and which undergoes N- and C-terminal processing to give both the adhesin AIDA-I (~100 kDa, α domain) and the outer membrane-integrated β-domain (AIDA^C) of 47.5 kDa; the latter domain serves as the translocator for AIDA-I (7, 10, 97). Experiments with the electron microscope reveal that AIDA-I does not form filamentous fimbriae-like structures on the cell surface, so it is considered to be an afimbrial adhesin. To be fully active, the adhesin needs to be posttranscriptionally modified by heptose residues at several sites. This modification is mediated by the "autotransporter adhesin heptosyl-transferase" (45 kDa), which is the cytoplasmic product of the aah gene (9). AIDA-I binds to an N-glycosylated membrane protein of 119 kDa (called AIDAR) in HeLa cells. This AIDA-I receptor seems to be expressed on various cell lines (52). Although the AIDA system was first characterized in a clinical E. coli isolate responsible for infant diarrhea, AIDA genes have been detected only in about 2 to 4% of human isolates (32, 46). The existence of an animal reservoir for AIDA genes has been confirmed by the description of a much higher frequency in porcine E. coli isolates causing edema disease and postweaning diarrhea than in human and various mammalian E. coli isolates (75).

In DAEC strains, whether or not they carry sequences related to the afa family of operons, a major surface protein of 16 kDa (the CF16K adhesive factor) is involved in the diffuse adherence phenotype in vitro to Caco-2 and HEp-2 cells (46). However, no further characterization of CF16K has been performed. Additionally, an outer membrane protein of 57 kDa associated with a diffuse adherence pattern to both HeLa cells and glass coverslips and adherence to mucus and colonic epithelium was first characterized in strains that did not hybridize with the afa/daa or AIDA probes. This adherence factor is plasmid encoded (105). More recent data strongly suggest that a plasmid-encoded outer membrane protein of 58 kDa, called the aggregative protein 58 or Ap58, is associated with the adherence properties of some EAEC strains. The ap58 gene is widespread among EAEC strains of O111:H12 serotype (67). It has not been determined whether Ap58 is fimbrial or afimbrial. The N-terminal amino acid sequence of Ap58 is similar to the 57-kDa adherence factor described by Yamamoto (105) in DAEC strains, suggesting that these two adherence factors belong to the same family of molecules. Other fimbrial and afimbrial adhesins have recently been implicated in the AA pattern of EAEC strains. Some of these studies suggested a role in EAEC adherence for type-IV pili (E. Dudley, C. Abe, and J. P. Nataro, Abstr. 103^rd Gen. Meet. Am. Soc. Microbiol., 2003), and for outer membrane protein homologs of the heat-resistant agglutinin encoded by hra1 (K. Unger, T. Harris, and I. N. Okeke, Abstr. 103^rd Gen. Meet. Am. Soc. Microbiol., 2003).

The type 1 fimbria is the most common adhesin found in Enterobacteriaceae. This fimbria has been reported to be an important virulence factor in uropathogenic E. coli (21). The involvement of type 1 fimbriae in the establishment of an AA pattern and in EAEC biofilm formation has also been suggested recently (68). It was hypothesized that these phenotypes are multifactorial; type 1 fimbria may be responsible for the first step of attachment, and this step may be followed by the expression of other fimbrial or afimbrial adhesins responsible for the establishment of the definitive AA pattern.