INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Escherichia coli is a major cause of bacterial diarrhea and urinary tract infection (UTI) in humans. It is also a major pathogen in agriculture, causing enteric and fatal septicemic colibacillosis in neonatal and young animals. Pathogenic E. coli strains have several virulence properties. One such property, adhesin production, plays an essential role in the colonization of the mucosal epithelium (14). The adhesins of pathogenic E. coli causing infections in humans are specific for a particular disease. The only known exception is that of afimbrial adhesins, which are encoded by the afa gene clusters and are produced by both uropathogenic and diarrhea-associated E. coli strains (14, 16, 17, 34). The afa family of gene clusters also includes the daa and dra operons, which encode the fimbrial F1845 and Dr adhesins, respectively (4, 47). Epidemiological studies have shown that afa-positive strains are enteric pathogens in children (17, 18, 35) and that the same afa-positive isolate may cause both diarrhea and cystitis in an individual child (16). Several studies have suggested that afa-positive strains play an important role in UTI pathogenesis. Such strains are common in E. coli isolates from pregnant woman (46) and from children (2, 3) with UTI and are associated with recurring UTI (10). An animal experimental chronic pyelonephritis model has also provided evidence that the expression of the dra gene cluster is important in persistent colonization of the urinary tract (19).

The afa gene clusters control the formation of an afimbrial sheath composed of the AfaE and AfaD proteins, an adhesin and an invasin, respectively (12, 26). The adhesin-encoding afaE gene is highly heterogeneous, producing antigenically different adhesins (31). The AfaE-I and AfaE-III adhesin subtypes mediate the mannose-resistant hemagglutination (MRHA) of human erythrocytes and specific attachment to epithelial cells via recognition of the decay-accelerating factor (DAF) molecule as a receptor (45).

afa-related sequences have been detected in isolates from diseased piglets and calves (21, 37). Hybridization and PCR analyses have suggested that the afa operons of animal-pathogenic E. coli are diverse and structurally different from the operons of human E. coli isolates (37, 38). In this study, we characterized the genes carried by the animal-pathogenic strains. We report here the cloning of two afa operons, designated afa-7 and afa-8, from bovine isolates and the analysis of similarity between these related gene clusters and the afa family of gene clusters from human-pathogenic E. coli. The binding properties of the AfaE-VII and AfaE-VIII adhesins were different from those of the AfaE adhesins produced by human-pathogenic strains. The afa-8 gene cluster appears to be widespread in bovine-pathogenic E. coli strains.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains, plasmids, tissue culture cells, and culture conditions. Nineteen E. coli strains were isolated from intestinal or extraintestinal sites in calves with signs and/or lesions typical of septicemia or enteritis in European countries (Belgium, France, and Spain) and Canada. All 19 strains have been characterized (37). All carried sequences that hybridized with the afa family of gene clusters in colony hybridization experiments but test negative for afa gene clusters by PCR; the afa genes were located on the chromosome in 11 strains and on Vir plasmids in 8 strains (37). Four of the 19 strains produced the CNF1 (cytotoxic necrotizing factor 1) toxin, and 12 produced the CNF2 toxin. Sequences coding for P and F17 adhesins have been detected in six and four strains, respectively (36a, 37). Six F165-positive E. coli isolates carrying afa-related sequences were kindly provided by J. M. Fairbrother and J. Harel (Faculté de Medecine Vétérinaire, Université de Montréal, Saint-Hyacinthe, Québec, Canada). They were originally isolated from calves (five strains) or piglets (one strain) with diarrhea. Strain HB101(pSSS1) (4) was kindly provided by S. Moseley (University of Washington, Seattle). E. coli J96 (28) and A30 (31) were used as positive controls for the presence of pap (coding for P adhesins) and afa gene clusters, respectively. E. coli HB101 (6) and MC1061 (39) were used as recipients for the recombinant plasmids listed in Table 1.

Hemagglutination and adhesion assays. Hemagglutination of washed human or animal (cattle, cat, dog, horse, piglet, and sheep) erythrocytes, in phosphate-buffered saline (pH 7.4) containing 2% (wt/vol) -methyl mannoside, was tested as previously described (2, 9). The DAF specificity of agglutination was tested with human erythrocytes previously pretreated with the DAF-specific monoclonal antibodies (MAbs) BRIC 230, BRIC 110, 1H4, and 8D11 (directed against the SCR-1, SCR-2, SCR-3, and SCR-4 domains of DAF, respectively) as previously described (50). The DAF-specific MAbs 1H4 and 8D11 were generously provided by D. M. Lublin (Washington University, St. Louis, Mo.). BRIC 230 and BRIC 110 were purchased from the International Blood Group Reference Laboratory (Bristol, Great Britain). Bacterial agglutination of P latex particles (P1 disaccharide latex; Chembiomed Ltd., Edmonton, Alberta, Canada) was used to determine P-adhesin-specific adhesion.

DNA analysis and genetic techniques. Plasmids were routinely isolated by alkaline lysis (39). Total plasmid DNA was extracted by the Kado method (27). Whole-cell DNA was prepared by cesium chloride gradient (31). Standard procedures were used for restriction endonuclease digestion and other common DNA manipulations (39). The afa and pap PCRs were as described previously (33). Four sets of primers were chosen, based on the sequences of the afaD and afaE genes from the afa-7 and afa-8 gene clusters. For afaD7 and afaE7 PCR assays, two sets of primers flanked segments of 377 and 618 bp, respectively. The upstream and downstream afaD primers were 5'-GCAGAGCTGAGTCTTGATGTCCGT-3' and 5'-CGTAATTATTCCTGTGAACGGCTGTCCA-3', respectively, and the upstream and downstream afaE primers were 5'-GCTAAATCAACTGTTGATGTT-3' and 5'-GGACAATCCAAATGGCGAATTA-3', respectively. For afaD8 and afaE8 PCR assays, two pairs of oligonucleotides flanking internal fragments of 354 and 302 bp of the afaD and afaE genes, respectively, were used; their sequences were 5'-GTTGAACTGAGTCTTAATACCAGTG-3' and 5'-TGAGCATTCTCCGCTAACTGATAAT-3' for afaD8 and 5'-CTAACTTGCCATGCTGTGACAGTA-3' and 5'-TTATCCCCTGCGTAGTTGTGAATC-3' for afaE8. PCR was carried out as previously described (33).

Cosmid and plasmid libraries. Genomic DNA was isolated from E. coli 239 KH 89 and 262 KH 89 and was partially digested with the restriction endonuclease Sau3A. For the cosmid library, restriction fragments from 239 KH 89, 35 to 50 kb in size as determined by sucrose gradient (10 to 40%), were ligated into the BamHI-digested and alkaline phosphatase-treated cosmid vector pHC79 as previously described (39). Cosmids were packaged in vitro into phage lambda particles with the  DNA in vitro packaging module (Amersham International). These particules were then used to infect E. coli HB101. For the plasmid library, restriction fragments from E. coli 262 KH 89, 6 to 12 kb in size, were fractionated by preparative agarose gel electrophoresis and ligated into pILL570 which had been BamHI digested and alkaline phosphatase treated. Recombinant plasmids were used to transform E. coli MC1061. Carbenicillin-resistant HB101 transductants and spectinomycin-resistant MC1061 transformants were screened by colony hybridization, using the 4.6-kb SphI fragment of pILL1101 containing the conserved region of the afa operons as a probe (Fig. 1A) (31), in low-stringency conditions.

View larger version (18K): [in this window] [in a new window]   FIG. 1.   Genetic organization of the afa-7 and afa-8 gene clusters. (A) Physical and genetic maps of pILL1101, the plasmid carrying the afa-3 gene cluster (13). Locations of the six afa genes are shown by boxes and single-headed arrows indicating the direction of the transcription. The conserved region of the afa gene clusters from the human-pathogenic isolates, corresponding to the 4.6-kb SphI fragment, is represented by dashed lines (32). Only the restriction sites of interest are indicated. (B) Restriction maps of pILL1191 and pLLL1194 resulting from the cloning of the afa-7 gene cluster from total DNA of E. coli 262 KH 89. ORFs deduced from the partial nucleotide sequence are indicated by boxes. Sequenced regions are highlighted. (C) Restriction maps of pILL1226, pILL1222, pILL1235, and pILL1244 resulting from the cloning of the afa-8 gene cluster from total DNA of E. coli 239 KH 89. ORFs deduced from the partial nucleotide sequence are indicated by boxes. Sequenced regions are highlighted. Thin lines represent the inserts; thick lines represent pILL570, pBR322, and pUC18 vector DNAs. Abbreviations: A, AccI; B, BamHI; E, EcoRI; H, HincII; P, PstI; Pv, PvuI; S, SmaI; Sp, SphI.

Hybridization. Bacteria grown for 3 h on nitrocellulose filters were used for colony hybridization as described by Grunstein and Hogness (20). For Southern blot hybridization, DNA restriction fragments were fractionated by agarose gel (0.7%) electrophoresis and transferred to nitrocellulose sheets (0.45-µm pore size; Schleicher & Schuell, Inc.) as previously described (56). Hybridization was performed under stringent (68°C) or nonstringent (50°C) conditions with probes labeled with ³²P by using the Megaprime DNA labeling system (Amersham). The bound probes were detected by autoradiography with Amersham Hyperfilm-MP.

DNA sequencing. Double-stranded DNA was sequenced by dideoxynucleotide chain termination with a Thermo Sequenase radiolabeled terminator cycle sequencing kit (Amersham). If required, oligonucleotide primers were synthesized on the basis of the sequence acquired. Nucleotide and amino acid database searches were performed with BLAST, and sequences were aligned with the GAP and PILEUP programs of the Genetics Computer Group sequence analysis software package, version 7-UNIX.

Electron microscopy. Negatively stained preparations of bacteria were examined for the presence of fimbriae on the bacterial surface as previously described (34). For transmission electron microscopy, infected HeLa cells were embedded as previously described (26), and ultrathin sections stained with uranyl acetate and lead citrate were examined with a JEOL 1010 transmission electron microscope operating at 80 kV. For scanning electron microscopy, glass coverslips coated with infected HeLa cells were processed as previously described (41) and examined with a JEOL JSM 35CF electron microscope.

Data imaging. Photographic negatives and raw picture data were scanned into Adobe Photoshop with an AGFA Studio scanner, and the scanned images were printed with a Hewlett Packard Deskjet 870Cxi color printer.

Nucleotide sequence accession numbers. The GenBank accession numbers for the sequences reported herein are AF072900, AF072901, and AF076152.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Characterization of the afa-positive strains 239 KH 89 and 262 KH 89. Nineteen of the bovine clinical E. coli isolates in our collection were afa positive. Of these, strains 239 KH 89 and 262 KH 89 were selected for cloning experiments, based on the intestinal or extraintestinal site of isolation and the virulence factors carried. Strain 239 KH 89 was isolated from internal organs and contained sequences coding for the P adhesin and the CNF1 toxin (37). The gene coding for the CS31 adhesin was also detected in this strain by colony hybridization (data not shown). E. coli 239 KH 89 causes the MRHA of human and sheep erythrocytes but does not adhere to HeLa cells. In contrast, strain 262 KH 89 was isolated from feces and carried no sequences encoding P adhesin or CNF toxins (37). This strain causes the MRHA of human, bovine, and porcine erythrocytes and adheres to HeLa cells with a diffuse adhesion pattern.

Molecular cloning of afa-related sequences from strain 262 KH 89. Recombinant plasmids containing portions (6 to 12 kb) of 262 KH 89 DNA were used to transform E. coli MC1061. Screening of the 322 clones of the library with the afa probe identified one clone which contained a hybrid plasmid, designated pILL1191 (Table 1). This plasmid resulted from the insertion of a 7.7-kb fragment (Fig. 1B). The insert of pILL1191 conferred a hemagglutination phenotype on the host bacteria and a HeLa cell adhesion pattern similar to that of strain 262 KH 89 (Fig. 2). Electron microscopy of negatively stained preparations of both 262 KH 89 and recombinant HB101(pILL1191) revealed no characteristic fimbriae on the bacterial surface (Fig. 3).

View larger version (68K): [in this window] [in a new window]   FIG. 2.   Photomicrographs of HeLa cells infected with E. coli 262 KH 89 (A) and HB101(pILL1191) (B). Bacteria adhered to the cell surface in a diffuse pattern.

View larger version (50K): [in this window] [in a new window]   FIG. 3.   Electron micrographs of E. coli strain preparations stained with ammonium molybdate. (A) HB101(pSSS1), used as a positive control for the detection of fimbriae; (B) HB101(pILL1191). Magnification, ×100,000; bar, 100 nm.

Molecular cloning of afa-related sequences from strain 239 KH 89. Bacteriophage lambda particles carrying recombinant cosmid molecules with portions (35 to 50 kb) of 239 KH 89 DNA were prepared and used to infect E. coli HB101. Screening of 500 recombinant HB101 with the afa probe identified 10 positive clones. Eight of these clones agglutinated human erythrocytes. This MRHA phenotype was not due to the pap operon of strain 239 KH 89 being present on the recombinant cosmids because, unlike the parental strain, none of the clones gave positive amplification in the pap PCR assay or agglutinated P latex particles. The recombinant cosmids carried by these eight MRHA clones were analyzed by digestion with three restriction enzymes (BamHI, EcoRI, and HindIII). Comparison of the restriction profiles suggested that all of the cosmids carried the same region of 239 KH 89 genomic DNA. Hybridization with the afa probe showed that in three of the cosmids, including pILL1211 (Table 1), the sequences binding to the afa probe were located on a single 5-kb EcoRI fragment. pILL1222, which results from the insertion of this 5-kb EcoRI fragment from pILL1211 into pUC18 (Table 1; Fig. 1C), did not confer the property of MRHA on HB101. The MRHA-expressing DNA fragment was subcloned by partially cleaving pILL1211 with Sau3A such that 6- to 12-kb fragments were generated and ligated into BamHI-digested and alkaline phosphatase-treated pBR322. Recombinant plasmids were used to transform E. coli HB101. The carbenicillin-resistant transformants were tested for MRHA and were screened for afa-related sequences by using the 5-kb EcoRI fragment of pILL1222 as a probe. One positive clone carried the recombinant plasmid pILL1226, which has a 9.5-kb insert (Table 1; Fig. 1C). This clone was chosen for further characterization of the afa-related sequences. Unlike the parental strain, which agglutinates both human and sheep erythrocytes, HB101(pILL1126) causes the MRHA only of human erythrocytes. Electron microscopy of negatively stained preparations of HB101(pILL1226) showed that no fimbriae were present (data not shown).

Characterization of the afa-7 and afa-8 gene clusters. Restriction mapping and Southern analyses were performed with the cloned afa-related sequences from the two bovine strains. Comparison of the sequences of the insert from pILL1191 and total DNA from 262 KH 89 digested by PstI and SmaI (data not shown) and of the insert from pILL1226 and total DNA from 239 KH 89 digested by EcoRI and PstI (Fig. 4A and B, lanes 1 and 2) showed that the afa-related sequences were not rearranged during the molecular cloning processes. The restriction maps of pILL1191 and pILL1226 were compared to each other and to that of pILL1101, which carries the well-characterized and sequenced afa-3 gene cluster (Fig. 1). The restriction maps of pILL1191 and pILL1226 differed, and neither plasmid contained the three PstI fragments of 2.6, 1.1, and 0.4 kb characteristic of the conserved region of the afa gene clusters of human-pathogenic strains. Reciprocal hybridizations between restriction fragments from the inserts of the three plasmids indicated that (i) the inserts of both pILL1191 and pILL1226 contained, upstream from the conserved sequences of afa gene clusters, sequences related to the 1.9-kb PstI fragment of pILL1101, corresponding to the promoter region of the afa-3 operon, and (ii) the afa-related sequences in pILL1191 from strain 262 KH 89 were only slightly similar to those from strain 239 KH 89, which were inserted into pILL1226. These results suggest that the afa-related sequences inserted into pILL1191 and pILL1226 correspond to two different afa-related operons, designated afa-7 and afa-8, respectively.

View larger version (64K): [in this window] [in a new window]   FIG. 4.   Search for homology between the insert of pILL1226 and total and plasmid DNA from strain 239 KH 89 and between the insert of pILL1226 and plasmid DNA from strain 89.201.2/3. Plasmid or total DNA was digested with EcoRI (A) or PstI (B). Lanes: 1, pILL1226; 2 and 3, total and plasmid DNA from strain 239 KH 89, respectively; 4, plasmid DNA from strain 89.201.2/3. The resulting fragments were separated by electrophoresis in a 0.7% agarose gel, transferred to nitrocellulose filters, and probed with either pBR322 or pILL1226. The sizes indicated are those of the internal EcoRI and PstI fragments of the pILL1226 insert. The EcoRI and PstI fragments hybridizing specifically with the insert of pILL1226 are indicated by arrows.

Genetic organization of the afa-7 gene cluster. Determination of the sequence of a 4.7-kb region of the pILL1191 insert confirmed that the genetic organization of the afa-7 gene cluster of strain 262 KH 89 was very similar to that of the afa operons of human-pathogenic strains (Fig. 1A and B). Computer analyses identified four ORFs transcribed in the same orientation (Fig. 1B). Two of the ORFs mapped to the same loci as the afaC and afaD genes in the afa-3 operon. The ORF designated afaC encoded a peptide of 838 amino acids (aa) similar (77% identity and 81% similarity) to AfaC, an outer membrane usher protein (data not shown). Following this ORF was a short intergenic region of 17 bp (rather than 15 bp as in the afa-3 gene cluster) and the afaD ORF, which encoded a peptide with 46% identity and 56% similarity to AfaD. The first ATG codon of this ORF was preceded by a Shine-Dalgarno sequence (data not shown) and followed by a probable signal peptide sequence with a classical cleavage site (Ser-Gln-Ala) that was recognized by signal peptidase I (Fig. 5). The predicted mature AfaD molecule contains 123 aa and has a deduced molecular mass of 13.6 kDa and a pI of 5.27. 

View larger version (26K): [in this window] [in a new window]   FIG. 5.   Sequence alignment of AfaD proteins from afa-3, afa-7, and afa-8 gene clusters. Dashes mark identical residues; gaps indicated by dots have been inserted to optimize the alignment. Numbers correspond to amino acid positions in the protein of the afa-3 gene cluster. The predicted translational start sites of AfaD-III, AfaD-VII, and AfaD-VIII proteins are indicated by +1.

View larger version (38K): [in this window] [in a new window]   FIG. 6.   Protein sequence alignments. Dashes mark identical residues. Numbers correspond to amino acid positions in the AfaE-VII and AfaE-VIII adhesins. The predicted translational start sites of AfaE-VII and AfaE-VIII adhesins are indicated by +1. (A) Protein sequence of AfaE-VII aligned with that of the AAF/I fimbrial adhesin. The two cysteine residues conserved among the afimbrial and fimbrial AfaE-related adhesins are underlined; gaps indicated by dots have been inserted to optimize the alignment. (B) Sequence alignment of AfaE-VIII and BmaE proteins.

Genetic organization of the afa-8 gene cluster. Based on hybridization experiments, a 4.2-kb region of the insert of pILL1226 was presumed to contain the 3' end of the afa-8 gene cluster. Computer analyses of the sequence of this region showed three ORFs transcribed in the same orientation. They were homologous to the afaC and afaD genes and to an adhesin-encoding gene (Fig. 1C), respectively, confirming that the genetic organizations of the afa-3, afa-7, and afa-8 gene clusters were similar. This similarity was confirmed by sequencing of the 5' terminus of the insert of pILL1222 (Table 1; Fig. 1C), which detected a partial ORF (afaA') with 75% identity (over 164 bp) to the afaA gene of the afa-3 gene cluster. The AfaA product has been shown to be similar to the PapB regulatory protein (13).

Comparison of the adhesion to various cell lines of E. coli strains expressing the afa-3, afa-7, and afa-8 gene clusters. The production of AfaE adhesins by E. coli strains pathogenic for humans conferred on the bacteria the capacity to bind to HeLa cells, uroepithelial cells (32), and Caco-2 cells (29). The expression of the afa-7 gene cluster by HB101 strains conferred on the bacteria the capacity to bind to human HeLa, Caco-2, and UROtsa cell lines (Table 2; Fig. 2B and 7A). HB101 strains producing the AfaE-VII adhesin, but not those producing AfaE-III, adhered to bovine kidney-derived MDBK cells (Table 2; Fig. 7A). Scanning electron microscopy of cultured HeLa cells infected with HB101(pILL1191) showed bacteria adhering in a diffuse pattern and interacting with microvillar extensions of HeLa cells, with some bacteria enclosed by elongations (Fig. 8). These observations were similar to those previously reported (26) for interactions between HB101(pILL1101) bacteria expressing the afa-3 gene cluster and HeLa cells.

View larger version (75K): [in this window] [in a new window]   FIG. 7.   Photomicrographs of Caco-2, UROtsa, MDBK, and MDCK cells infected with HB101(pILL1191) (A) and HB101(pILL1244) (B). The infection time was 3 h except for UROtsa cells, which were infected for 5 h with HB101(pILL1244).

View larger version (158K): [in this window] [in a new window]   FIG. 8.   Scanning electron micrograph showing adhesion to HeLa cells of strain HB101(pILL1191) after 3 h of infection. Arrows indicated bacteria totally surrounded by microvillar extensions. Magnification, ×12,000; bar, 1 µm.

Comparison of the invasion properties of E. coli strains expressing the afa-3, afa-7, and afa-8 gene clusters. Examination of HeLa cells infected with HB101(pILL1191) showed intracellular bacteria within membrane-bound vacuoles (Fig. 9A). The AfaD-VII and AfaD-VIII products were also similar to the AfaD protein, which is encoded by the afa-3 gene cluster and has been reported to be an invasin (26). We investigated the role of AfaD from the afa-7 and afa-8 gene clusters in bacterial internalization by complementing the invasion deficiency of an afa-3-expressing HB101 with a mutated afaD gene by producing the AfaD-VII or AfaD-VIII protein. HeLa cell monolayers were infected with HB101 harboring two plasmids: one (pILL1189) containing the afa-3 gene cluster with a mutation in the afaD gene, and the other (pILL1194 or pILL1235) containing the afa-7 or afa-8 cluster deleted of the afaE gene. Electron microscopy showed that both HB101(pILL1189, pILL1194) (Fig. 9B) and HB101(pILL1189, pILL1235) (data not shown) adhered to HeLa cells because the AfaE-III adhesin was present, and it invaded HeLa cells due to the presence of AfaD-VII or AfaD-VIII proteins. As for internalized afa-3-expressing bacteria, internalized HB101(pILL1189, pILL1194) and HB101(pILL1189, pILL1235) were found within membrane-bound vesicles.

View larger version (173K): [in this window] [in a new window]   FIG. 9.   Transmission electron micrographs of recombinant strains interacting with HeLa cells 3 h postinfection. (A) HB101(pILL1191), expressing the afa-7 gene cluster. Intracellular bacteria were observed within membrane-bound vacuoles, as showed in the inset. (B) HB101(pILL1189, pILL1194), producing both AfaD-VII and AfaE-III. AfaE-III mediated adhesion of the bacteria to HeLa cells; AfaD-VII mediated internalization of the bacteria into the cells. Magnifications, ×15,000 (A), ×31,000 (inset), and ×14,000 (B); bar, 1 µm.

Distribution of afa-7 and afa-8 operons in bovine and porcine E. coli isolates reported to harbor afa-related sequences. In addition to strains 262 KH 89 and 239 KH 89, 17 other bovine strains containing afa-related sequences (37) were screened for afa-7 and afa-8 sequences by colony hybridization in high-stringency conditions. The probes used were afaD and afaE PCR products obtained from strains 262 KH 89 and 239 KH 89 (see Materials and Methods for DNA amplification). None of the 17 strains hybridized with the afaD and afaE probes from the afa-7 gene cluster. In contrast, all 17 strains contained sequences that hybridized with both the afaD and afaE genes from the afa-8 gene cluster. We evaluated a PCR approach for the detection of afa-8 sequences, using the two sets of primers derived from the sequences of the afaD and afaE genes in amplification reactions to detect the presence of the corresponding sequences. All 17 strains positive by hybridization tested positive by PCR for both afaD8 and afaE8. The specificity of the PCR assays was validated by direct sequencing of the products obtained from nine isolates. The afaD amplification products differed by no more than 1 bp between any two sequences. Similar results were obtained for comparison of the sequences of the nine afaE amplification products.

The afa-8 gene cluster is plasmid or chromosome borne. Mainil et al. reported that the 19 isolates from cattle that he provided for this study carry afa-related sequences either on the chromosome or on Vir plasmids (37). The afa-related sequences carried by E. coli 239 KH 89 were described as chromosome borne (37). We confirmed that this was the case by DNA hybridizations of plasmid or total DNA from 239 KH 89 digested with EcoRI (Fig. 4A) and PstI (Fig. 4B), with pBR322 and pILL1226 used as probes. None of the plasmid-generated fragments had sequences similar to that of the pILL1226 insert (Fig. 4A and B, lanes 3). In contrast, two EcoRI (Fig. 4A, lane 2) and three PstI (Fig. 4B, lane 2) fragments of total DNA from 239 KH 89 specifically hybridized with the pILL1126 insert. Of these, the two EcoRI fragments and one PstI fragment had the same electrophoretic mobility as the EcoRI (5 and 2.3-kb) and PstI (5.7-kb) internal fragments carrying the afa-8 operon in the pILL1226 insert (Fig. 1C). Of the E. coli isolates testing positive for afa-8 by PCR that have been studied, E. coli 89.201.2/3 has been reported to carry afa-related sequences on the Vir plasmid (37). We investigated the location of the afa-8 operon in this strain by hybridization between each of the two probes described above and digested plasmid DNA from strain 89.201.2/3. Sequences similar to that of the pILL1226 insert were clearly detected in three EcoRI (Fig. 4A, lane 4) and two PstI (Fig. 4B, lane 2) plasmid fragments, indicating that the afa-8 determinants are plasmid borne in strain 89.201.2/3. However, although one hybridizing EcoRI fragment (5 kb) was similar to the internal EcoRI fragment of the pILL1226 insert that carried most of the afa-8 genes, we detected no hybridizing fragment that was the same size as the second internal EcoRI fragment (2.3 kb) of the pILL1226 insert that contained the afaE gene and flanking sequences. In the same way, one PstI fragment was similar in size to the internal PstI fragment (5.7 kb) of the pILL1226 insert that contained the afa-8 genes, but the second hybridizing fragment differed in size from the second hybridizing PstI fragment detected in total DNA from 239 KH 89. Thus, the plasmid and chromosomal sequences flanking the afa-8 gene cluster differ, suggesting that the presence of afa-8 sequences in the chromosome did not result from the insertion of a Vir plasmid.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Neonatal septicemia and diarrhea due to E. coli are common problems in farm animals, especially calves (15). Fimbrial adhesins, virulence factors produced by these pathogenic E. coli strains, have been extensively detected and studied. Recently, sequences related to the afa gene clusters that encode afimbrial adhesins in human-pathogenic E. coli associated with intestinal and extraintestinal infections have been detected in strains isolated from calves and piglets with diarrhea and septicemia (21, 37). We describe here two afa-related gene clusters from bovine E. coli isolates. We found that these two systems also code for afimbrial adhesins and demonstrated that the genetic organization of these two gene clusters, designated afa-7 and afa-8, is very similar to that of the afa gene clusters from human E. coli isolates. Hybridization and sequencing experiments indicated the presence in both gene clusters of regulatory sequences followed by an afaC gene encoding an outer membrane anchor protein, an afaD gene encoding an invasin, and an afaE gene encoding an adhesin. The afaC genes were very similar, the afaD genes were less similar, and the afaE genes were completely divergent. The afa-8 operon is very prevalent in a collection of bovine E. coli strains associated with intestinal and extraintestinal infections. These preliminary epidemiological results show a high prevalence of afa-8 genes in E. coli isolates from diseased animals. Like the afa-3 and afa-5 gene clusters from human strains (60), the afa-8 operon has been found on both large plasmids and the chromosome. We demonstrated that the chromosomal location of the afa-8 genes was not the result of chromosomal insertion of a virulence plasmid. The afa-3 gene cluster has been shown to translocate from a plasmid to the E. coli chromosome via IS1-specific recombination mediated by a recA-independent mechanism (13). Thus, all the available evidence strongly suggest that the gene clusters of the afa family may be carried by mobile genetic elements, facilitating their dissemination among E. coli strains.

Sequence analysis of the afa-7 gene cluster showed that directly upstream from the afaE gene is an ORF (afaG) with no homolog in the other afa gene clusters, which maps to the same locus as the newly described daaP gene in the daa operon (36). The DaaP polypeptide is involved in the regulation of expression of the daa operon via endoribonucleolytic processing of the polycistronic mRNA (36). Interestingly, afaG encodes a peptide with some similarity to the group II intron maturases involved in intron RNA processing, and the mRNA cleavage sequence described by Loomis et al. is also present in the afa-7 gene cluster. It is therefore possible that AfaG is involved in the regulation of expression of the afa-7 gene cluster.

The afa gene clusters characterized from human-pathogenic strains have the unique feature of encoding both an adhesin and an invasin, which seem to be involved in a two-stage process of interaction with epithelial cells (26). The results of this study demonstrate that the afa-related gene clusters expressed by animal-pathogenic strains are also involved in the adhesion to and invasion of epithelial cells. The AfaD proteins encoded by both the afa-7 and afa-8 gene clusters mediated internalization of bacteria into HeLa cells although they were only 46 and 42% identical, respectively to the AfaD invasin encoded by the afa-3 gene cluster. Thus, these three AfaD proteins may indeed be the first members of a new family of invasins.

Previous studies have demonstrated the great heterogeneity of the AfaE-related adhesins of E. coli strains isolated from humans. The AfaE-VII protein is genetically different from the AfaE adhesins produced by human-pathogenic strains and also has different binding specificities. Our data indicate that the receptor of the AfaE-VII adhesin is not human DAF, a species-specific molecule previously thought to be the receptor for ll of the AfaE adhesins produced by human-pathogenic strains (11, 45, 50). AfaE-VII mediates MRHA of human, bovine, and porcine erythrocytes and the adhesion of bacteria to human (HeLa, Caco-2, and UROtsa cells) and bovine (MDBK cells) epithelial cells. It therefore appears that its receptor is widely distributed in various tissues, at least in humans and cattle.

AfaE-VII is slightly similar to the fimbrial adhesin AAF/I produced by EAggEC isolates implicated in pediatric diarrhea in the developing world (43). Not only does AAF/I belong to a putative subfamily of AfaE adhesins, but the organization of the AAF/I biogenesis genes is typical of the afa family of operons (54), suggesting that the agg and afa operons may belong to the same phylogenetic family of adhesin-encoding gene clusters. The gene encoding EAST1 of E. coli is located approximately 1 kb downstream from the AAF/I (53)- and AfaE-VII-encoding structural genes, indicating that there may be a genetic association between such adhesin-encoding genes and this toxin-encoding gene. The EAST1-encoding gene has been also detected in enteropathogenic, enterohemorrhagic, and diffusely adherent E. coli strains associated with diseases in humans (53, 55, 57) and in enterotoxigenic E. coli isolates from humans, cattle, and pigs (57, 58). This widespread distribution among pathogenic E. coli led Yamamoto and Escheverria to suggest that the EAST1-encoding gene may be located on a transposon (57). Our results showing identity between the regions upstream (59 nucleotides) and downstream (147 nucleotides) from the EAST1-encoding gene on the chromosome of the afa-7-expressing strain, 262 KH 89, and on pCS1 in enterotoxigenic E. coli strains strongly support this notion.

The gene encoding AfaE-VIII, the second AfaE adhesin from animal-pathogenic strains characterized in this study, is also different from those of known AfaE adhesins produced by human-pathogenic strains, and AfaE-VIII binds to a different receptor, yet to be identified. AfaE-VIII is very similar (97% identity) to the M agglutinin, an afimbrial adhesin encoded by the bma gene cluster in human-uropathogenic strains (52). The bma gene cluster consists of five genes, the sequences of which, with the exception of that of the adhesin-encoding gene, are not available in databases. It is therefore not possible to compare the sequences of the bma and afa-8 gene clusters. As the prevalence of M-agglutinin-producing strains is estimated to be very low among human isolates (25), the biological function and significance of this adhesin in human diseases are unknown. In this study we have shown that the AfaE-VIII adhesin is frequent and highly conserved among E. coli strains producing either CNF1 or CNF2, factors that are pathogenic for calves. AfaE-VIII-producing strains have also been detected among non-CNF-producing strains isolated from calves and piglets. All of these afa-8-expressing bacteria are associated with diarrhea, septicemia, and internal organ infections in which the site of entry is the intestine. Bloodstream infections in which the bacteria are derived from the intestinal flora by bacterial translocation are common in patients with cancer (1). Preliminary results obtained with the afa-8-specific PCR assay in our laboratory (32a) show that afa-8 gene clusters are present in CNF1-producing E. coli strains recently associated with this type of bacteremia (22). It therefore seems likely that afa-8 is involved in the development of extraintestinal infections associated with a primary colonization of the intestine.

In conclusion, we have demonstrated that animal-pathogenic E. coli isolates express different afa gene clusters, all encoding products involved in adhesion to and internalization into epithelial cells. The role of the adhesins and invasins produced by afa-expressing strains in the pathophysiology of diarrheal and extraintestinal infections in young animals is unclear. We have described specific PCR assays that may be useful tools for the epidemiological studies of animals necessary to confirm the association of the afa-8 gene cluster with CNF-producing strains. The high degree of conservation of the AfaE-VIII adhesins between strains suggests that this antigen could be used as a component of vaccines in animal husbandry.