INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Enteropathogenic Escherichia coli (EPEC) is one of several pathovars of E. coli capable of causing diarrhea (45). While human EPEC infections, which are manifest primarily in infants, were once commonplace in industrialized nations (56), they are now identified primarily in developing countries (10). EPEC strains possess distinct virulence factors not found in most E. coli strains (45). Typical EPEC strains produce bundle-forming pili (BFP), long, flexible, rope-like structures composed of intertwining fibers (19). Based on protein sequence analysis and morphology, BFP belong to the type IV group of fimbriae or pili found on a variety of gram-negative bacteria, many of which are human, animal, or plant pathogens (67, 70, 77). BFP have recently been shown to elicit an antibody response in natural infections (37, 38, 53) and a modest response in experimentally infected adults (13). BFP are required for two phenotypes of EPEC that can be studied in vitro and probably play a role in colonization in vivo. The first is autoaggregation, which is the ability of EPEC to reversibly form multicellular clusters in liquid culture (5, 76). The second is localized adherence (LA), which is the capacity of EPEC to form defined multicellular clusters (microcolonies) upon epithelial cells (7, 59) and upon human intestinal tissue (26). Both of these phenomena seem to be manifestations of the ability of BFP to intertwine, creating a fibrous network connecting individual bacteria. BFP also appear to adhere directly to epithelial cells (19, 31), although no BFP receptor on epithelial cells has been identified. During the course of tissue culture infection, BFP bundles undergo a transformation to a longer and thicker form that is correlated with dispersal of bacteria from microcolonies (31). In a recent clinical study, volunteers fed mutant EPEC strains lacking BFP developed diarrhea at a much lower frequency and of less severity than did volunteers receiving equivalent doses of an isogenic strain producing BFP (5). Thus, BFP are dynamic fimbriae that contribute to EPEC pathogenesis.

BFP biogenesis is specified by a cluster of 14 bfp genes (65, 66) located on large (~50- to 70-MDa) EPEC adherence factor (EAF) plasmids (3, 47, 48). The first gene of this cluster is bfpA, which encodes bundlin, a pilin protein that is the only known structural component of BFP. Like other type IV pilins, bundlin is synthesized with a leader peptide removed by a prepilin peptidase, has a hydrophobic region at its mature N terminus, and has two cysteine residues forming a disulfide bond near the C terminus. Bundlin is membrane associated (2, 76). It seems likely that the ultimate function of most of the bfp gene products is to remove bundlin subunits from the bacterial membrane and assemble them into pili (2).

Little is known about the structure of bundlin monomers or how they assemble into BFP. However, some information on these topics can be inferred from studies of other type IV pili. The crystal structures of type IV pilin monomers from Neisseria gonorrhoeae and Pseudomonas aeruginosa have been solved (15, 25, 52). These pilins are ladle shaped with a long N-terminal -helical handle and a globular head. Models of the way in which pilin monomers pack together into the pilus fiber have been proposed for the gonococcal and Pseudomonas pili (16, 25, 52), as well as the toxin-coregulated pili (TCP) of Vibrio cholerae (R. Chattopadhyaya and A. C. Ghose, unpublished data; Protein Data Bank accession number 1QQZ).

Extensive sequence variability has been observed in the type IV pilins of Dichelobacter nodosus, Eikenella corrodens, Moraxella bovis, P. aeruginosa (reviewed in reference 70), N. gonorrhoeae (52), N. meningitidis (1), and Vibrio cholerae (27, 43, 50, 55). In some of these species, pilin variability is clearly linked to changes in antigenicity. The host immune response may have played a major role in selecting for pilin variants with altered sequences. In the N. gonorrhoeae genome, extra copies of partial pilin genes are also present. These normally silent copies can be exchanged with the active copy via recombination, leading to antigenic variation that can aid the gonococcus in evading the host immune response (reviewed in references 41 and 62). A comparison of many gonococcal pilin gene products shows that extensive sequence variation occurs in amino acid residues of the "head" region but hardly any occurs in residues of the "handle" region (52). A hypervariable region of gonococcal pilin occurs between the disulfide-bonded cysteine residues. This region, as well as a disaccharide moiety and the extreme C terminus of the pilin, is predicted to be exposed on the outside of the pilus fiber (16, 52). In P. aeruginosa pilins, the disulfide region is also variable in sequence and serves as a binding domain for cell surface receptors (reviewed in reference 23).

In contrast to the considerable variation noted in other type IV pilins, the bundlins encoded by the two known bfpA alleles contain only a single amino acid difference (11, 64). Bundlin proteins from diverse EPEC strains might be expected to exhibit more significant sequence variation than has previously been noted. The current study was undertaken to assess the level of sequence diversity in bundlin proteins among EPEC strains and to gain insight into structural and functional constraints on different regions of the molecule.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

EPEC strains. The origins of the EPEC strains examined in this study are listed in Table 1. To analyze the motility of each strain, a loopful of culture was stabbed into a motility agar plate (1% Bacto-Tryptone, 0.5% NaCl, 0.35% agar) and examined after overnight incubation at 37°C. Based on this analysis, strains lacking H serotype information were designated H+ (motile) or NM (nonmotile) (Table 1).

Isolation and sequencing of bfpA genes. Partial and complete bfpA genes were amplified by PCR using DeepVent DNA polymerase in accordance with the manufacturer's (New England Biolabs) instructions. EPEC colonies were picked from Luria agar plates and boiled for 10 min in ThermoPol buffer (New England Biolabs) containing 2 to 6 mM MgSO₄. DeepVent DNA polymerase (2 U), oligonucleotide primers (0.35 to 1.2 µM), and deoxynucleoside triphosphates (400 µM each) were added to bring the total reaction volume to 100 µl before beginning the PCR. Primer pairs used for PCR were Donne-28 and -29, -362 and -363, -423 and -363, -6 and -382, and -423 and -382 (Table 2). The precise reaction conditions varied with the strain and primers used. The PCR products were examined by agarose gel electrophoresis. If the PCR mixture contained a single DNA species of the expected size, it was purified directly using the Wizard PCR Preps DNA Purification System (Promega). If additional DNA species were present, the desired PCR product was isolated from a preparative agarose gel. Automated DNA sequencing of PCR products was carried out by the staff of The Biopolymer Laboratory (University Of Maryland School of Medicine) using two or more of the primers listed in Table 2. For some strains, the entire double-stranded sequence of the bfpA gene could be obtained using only primers Donne-423 and Donne-363.

Sequence analysis. DNASIS (Hitachi Software) was used to align bfpA nucleotide sequences from each strain and determine a consensus sequence. Multiple sequence alignment was performed on the bfpA gene sequences and the deduced bundlin amino acid sequences using DNASIS and CLUSTAL W (71). Gene trees were constructed with the computer program MEGA (Molecular Evolutionary Genetics Analysis, version 1.0; S. Kumar, K. Tamura, and M. Nei, Institute of Molecular Evolutionary Genetics, The Pennsylvania State University, 1993 [http://www.bio.psu.edu/People/Faculty/Nei/Lab/megaform.txt]). The proportions of polymorphic synonymous (p_S) and nonsynonymous (p_N) sites were calculated by the method of Nei and Gojobori (49). To examine variation in the functional constraints of different parts of the molecule, these statistics were tabulated in a sliding-window analysis of 30 codons along the gene by the program PSWIN. The theoretical three-dimensional structure of bundlin was analyzed and colored using RASMOL (58) (http://www.umass.edu/microbio/rasmol/).

MLEE. Every strain was characterized by its profile of electromorphs for 20 enzymes by multilocus enzyme electrophoresis (MLEE) (63, 78, 79). Genetic relationships between strains were determined based on a matrix of genetic distances between all pairs constructed by comparison of the allelic arrays. A dendrogram was constructed with MEGA using the neighbor-joining algorithm.

Immunoblotting. Overnight cultures of EPEC in LB were diluted 1:100 into 20 ml of Dulbecco's modified Eagle medium-nutrient mixture F-12 (DMEM-F-12) containing 15 mM HEPES buffer and lacking phenol red (Gibco-BRL Life Technologies catalog no. 11039-021). After 6 h of growth at 37°C with shaking, the bacteria were pelleted by centrifugation at 2,500 × g for 5 min and then resuspended in 1.2 ml of cell lysis buffer (20 mM Tris-HCl [pH 8.0], 500 mM NaCl, 0.1 mM EDTA, 0.1% Triton X-100). The optical density at 600 nm of a portion of these samples was measured on a spectrophotometer. Aliquots (75 µl or less) of these samples were mixed with 25 µl of 4× sodium dodecyl sulfate-polyacrylamide gel electrophoresis loading buffer (2) and then adjusted to a total volume of 100 µl with additional cell lysis buffer such that the resulting gel samples were derived from roughly equivalent densities of bacteria. The gel samples were boiled for 10 min, and then 5 µl of each was loaded per lane of a sodium dodecyl sulfate-15% polyacrylamide gel. After electrophoretic separation, the protein in the gel was electrotransferred to a polyvinylidene difluoride membrane. The membrane was blocked with phosphate-buffered saline containing 0.1% Tween 20 plus 5% nonfat dry milk and then incubated sequentially with a polyclonal antibody (1:5,000 dilution) directed against six-His-tagged prebundlin (81), an anti-rabbit horseradish peroxidase conjugate (1:30,000 dilution), and ECL Western blotting detection reagents (Amersham Pharmacia Biotech).

Transmission electron microscopy. To prepare samples for electron microscopy, 1-ml aliquots were removed from EPEC cultures grown in DMEM-F-12 for 5 h and the bacteria were pelleted by brief centrifugation. Most of the medium was decanted, and the bacterial pellet was gently resuspended in the remainder. Aliquots (10 µl) were spotted onto electron microscopy grids, which were dried for 10 min and then washed, stained, and examined for BFP as described previously (66).

Autoaggregation assay. EPEC strains were cultured overnight at 37°C in LB. The resulting stationary-phase cultures were diluted 1:250 (making appropriate adjustments for optical density at 600 nm) into 20 ml of DMEM-F-12. These cultures were shaken at 250 rpm for 5 h in 50-ml conical polypropylene tubes at 37°C. Autoaggregation was initially gauged by visually inspecting the cultures for bacterial aggregates and sedimentation. For some strains, it was necessary to use light microscopy in order to detect aggregates. Culture aliquots (5 µl) were examined in hanging-drop slides at ×40 magnification.

LA assay. LA to HeLa epithelial cells was assayed as described previously (12) using the eight-well chamber slide modification.

Nucleotide sequence accession numbers. The nucleotide sequences of the bfpA genes isolated from all of the strains listed in Table 1 (except prototype strain E2348/69) have been deposited in GenBank under accession numbers AF304468 through AF304486.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Identification of novel bfpA alleles. The goal of this study was to assess the amount of sequence variation among bfpA genes found in EPEC strains. To this end, we initially analyzed 13 strains from our laboratory collection (a subset of those listed in Table 1) representing a variety of serotypes and well-documented outbreaks (or sporadic cases) of human diarrhea. These strains were originally isolated from diverse geographical locations between 1947 and 1986. They belong to classical EPEC O serogroups O55, O86, O111, O119, O127, O128, and O142. Three of them (C771, E990, and Stoke W) are type strains for the preparation of O antisera. Canine EPEC strain CA89-4221 and avian EPEC strain Z188-93, which are both probe positive for the bfpA gene, were also examined. PCR was used to amplify the bfpA gene from each EPEC strain. The DNA sequences of both strands of the PCR products were determined, and the resulting bfpA coding sequences were compared to each other and to the known bfpA sequences from prototypic strains E2348/69 and B171 (GenBank accession numbers Z12295 and U27184). Eight different bfpA alleles were identified among the 15 strains. Six of these alleles have not been previously described. Each bfpA allele specifies a prebundlin protein carrying a variant amino acid at one or more positions (Fig. 1).

View larger version (46K): [in this window] [in a new window]   FIG. 1.   Variable amino acids are clustered in prebundlin. An alignment of the prebundlin amino acid sequences encoded by the eight bfpA alleles described in this study is shown. Note that amino acids 1 to 13 comprise the cleaved leader peptide. The 1 (ALPHA1.AMI) prebundlin prototype sequence is listed on the top line. In the remaining seven sequences (ALPHA2.AMI through ALPHA3.AMI and BETA1.AMI through BETA5.AMI), invariant amino acids are indicated by dots. Variant amino acids are boxed and indicated by single-letter abbreviations.

Genetic relationships among bfpA alleles and correspondence to O serogroups. Sequence comparisons of the bfpA alleles show that they fall into two distinct clonal groups, which we term bfpA alpha () and bfpA beta () (Fig. 2 and Table 3). The three  alleles show polymorphism at only 7 (1.2%) of 579 nucleotides. The  prebundlins differ at only 6 (3.1%) of 193 amino acid residues. The previously reported 1 allele is found in prototype O127 strain E2348/69 (11), as well as in the single O86 and O55 strains initially examined in the current study. The previously reported 2 allele is carried by prototype O111 strain B171 (64) and each of the other O111 strains examined. The 1 allele varies from 2 at a single nucleotide, resulting in an amino acid substitution. The 3 allele of bfpA is found in both of the O119 strains but in only one of the two O128 strains we initially examined. This allele varies from 1 and 2 at six nucleotides located in the 3' half of the gene. Five of these six nucleotide changes specify altered amino acids relative to 2 prebundlin. Three of the sequence changes are identical to those found throughout the  group, suggesting a closer relationship to a common / ancestor.

View larger version (25K): [in this window] [in a new window]   FIG. 2.   Eight bfpA alleles cluster into two major groups. To the left is a phylogenetic tree for the bfpA gene from 17 EPEC strains constructed by the neighbor-joining algorithm based on the gamma distance with  = 2. Branch lengths in terms of nucleotide substitutions per 100 sites are given above the major branches. Allele designations are listed for each strain. To the right is a graph of the locations of polymorphic nucleotide sites, which are marked as vertical lines that indicate nucleotides differing from those of the 2 allele at each position. Gaps indicate codons absent from specific alleles.

Evidence of selection for sequence diversification near the 3' end of the bfpA gene. To determine how the level of selective constraint varies along the length of the bfpA gene, the proportions of nonsynonymous and synonymous codon changes (p_N and p_S, respectively) were calculated within a 30-codon sliding window (Fig. 3). Nonsynonymous codon changes result in amino acid substitution; synonymous codon changes do not. The difference, p_N  p_S, is a measure of the degree of selective constraint. The zero-difference line indicates selectively neutral variation, where p_S and p_N are equal. The more positive the value, the less the contribution of synonymous substitutions and the greater the contribution of replacement substitutions. A positive difference in a particular region suggests that some selective pressure is driving diversification in that region, whereas a negative difference indicates selection against mutations resulting in amino acid variation. The results shown in Fig. 3 suggest that most bundlin is under selective constraint. However, a single region of bfpA, located between codons 143 and 174, exhibits a positive value of p_N  p_S. This region corresponds to amino acids 130 through 161 of mature bundlin, which are located between the disulfide-bonded cysteines. An attractive possibility is that this region of bundlin is located on the surface of the BFP filament and that the host immune system has selected for antigenic diversity at this site.

View larger version (23K): [in this window] [in a new window]   FIG. 3.   Evidence for diversifying selection near the 3' end of the bfpA gene. A sliding-window plot of the proportions of synonymous and nonsynonymous sites (pS and pN, respectively) that have mutated among eight bfpA alleles is shown. The difference (pN  pS) is a measure of the level of selective constraint on various parts of the molecule. Note that pN  pS > 0 for codons 143 to 174 (indicated by the arrow), a pattern consistent with the effect of diversifying selection.

Genetic relationships between EPEC strains. MLEE was carried out to define the overall phylogenetic relationships between selected EPEC strains. A dendrogram resulting from this analysis (Fig. 4) shows that all of the 2 alleles are found in a closely related cluster of O111 strains. Among the remaining EPEC strains, however, there is a curious lack of correspondence between the deduced clonal pattern of the strains and the pattern of bfpA alleles they possess. Alleles of both the  and  types are found among strains in either of the two major clonal EPEC groups, EPEC 1 and EPEC 2 (54, 78). In the EPEC 1 group, the 1, 3, 1, 3, and 5 alleles are interspersed among each other in a way that obviously does not correlate with the overall strain lineage. These results strongly suggest that the bfpA gene has been recently transferred between EPEC strains through multiple horizontal transfer events.

View larger version (30K): [in this window] [in a new window]   FIG. 4.   The alleles of bfpA identified in EPEC strains do not correlate precisely with the overall clonal lineage. Shown is a dendrogram of genetic relationships of EPEC strains from this study and representative DEC strains (79). The genetic distance was estimated in terms of electrophoretically detectable codon differences per enzyme locus for 20 enzymes. Previously identified DEC clusters (EPEC 1 and 2 and EHEC 1 and 2) are indicated (54, 78). Strain serotypes are in parentheses. Alleles of bfpA identified in specific EPEC strains are displayed in large, shaded type.

Examination of bfpA alleles in recent EPEC isolates. To determine whether bfpA  alleles have remained stable over recent decades, we examined a set of four EAF⁺ EPEC strains of very recent Brazilian origin (Table 1). The bfpA alleles derived from these strains were isolated by PCR and sequenced. As expected from their serogroups, the O86, O111, and O119 strains (HSP19/4, RN191/1, and RN410/1) possessed bfpA genes of the 1, 2, and 3 varieties, respectively. These results indicate that bfpA  sequences have been retained unchanged in current EPEC strains. The O55 strain (MA551/1), however, did not carry the 1 allele found in the previously examined O55 EPEC (strain ). Rather, it carried a 5 allele identical to that carried by O142 strains C771 and E851/71. This unexpected finding provides a further suggestion of the horizontal spread of a particular bfpA allele.

Expression and function of BFP variants. Selected EPEC strains were examined to determine whether they produced detectable bundlin and BFP and whether they exhibited the BFP-dependent phenotypes of autoaggregation and LA. Some of these characteristics have been previously reported for a subset of these strains (see references in Table 1). Prototype EPEC strain E2348/69 served as a positive control for all of our studies. To identify bundlin, a polyclonal antiserum was used to probe whole-cell extracts from each of the EPEC strains (Fig. 5 and Table 3). Bundlin protein was readily detected in extracts from 10 out of 12 strains tested carrying  alleles of bfpA and 5 of 7 strains carrying  alleles. Bundlin was not identified in strain 009-271082 or 2309-77 carrying the 2 allele or in strain E56/54 or C771 carrying the 4 or 5 allele. Particular  bundlins migrated more slowly (1 and 2) or more quickly (5) than did  bundlins (Fig. 5 and data not shown). The relative migration correlates well with the lengths (180 amino acids for 1 and 5 and 182 amino acids for 1 and 2) and molecular masses (19,986 Da for 5, 20,269 Da for 1, 20,306 Da for 2, and 20,328 Da for 1) predicted for these bundlin types. Such a migration anomaly has been noted for bundlin from other strains in previous studies (18, 20, 38).

View larger version (37K): [in this window] [in a new window]   FIG. 5.   Immunoblot of bundlin protein from EPEC whole-cell extracts. The arrow indicates the position of bundlin. The extracts are from the following strains: lane 1, E2348/69; lane 2, 0659-79; lane 3, E851/71; lane 4, 2309-77; lane 5, 10; lane 6, 012-050982; lane 7, 009-271082; lane 8, 010-311082; lane 9, DIF043256; lane 10, ; lane 11, Stoke W; lane 12, E56/54; lane 13, C771; lane 14, E990. The remaining strains exhibited detectable bundlin but are not shown here.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

In this study, we have identified eight alleles of the bfpA gene by surveying 19 EPEC strains of diverse origins. Previous studies using prototype EPEC strains E2348/69 and B171 had identified two of these alleles, 1 and 2, that differ by only a single nucleotide (11, 64). The eight bfpA alleles fall into two distinct groups based upon sequence analysis. By analogy to the intimins, we call these groups  and . The  group, isolated from 12 of the EPEC strains examined here plus prototype strains B171 and E2348/69, contains three different alleles showing little sequence variation. The  group, represented by seven strains, exhibits more significant sequence variation, both internally and when compared to  alleles. We note that additional alleles of bfpA have recently been deposited in sequence databases. Bortolini and colleages have described a nonfunctional bfpA allele (GenBank accession number AF119170) that is prematurely truncated by an insertion sequence and is found in certain clonally related O128ab:H2 and O119:H2 EPEC strains (6). We did not identify this allele in our study; however, the residual 5' bfpA sequence of this allele is identical to that of the 3 allele, with the exception of a single nucleotide. Incomplete bfpA sequences can be found in GenBank records with accession numbers AF233895 through AF233899 (S. Subpasu, M. Yamasaki, J. Yatsuyanagi, O.-A. Ratchatrachenchai, and K. Ito, unpublished data). One of these sequences appears to be from an 1 allele, but the remaining four differ from our 3, 1, and 4/5 alleles at 1 to 10 nucleotide positions. These sequences support our division of bfpA alleles into two groups and attest that there is further sequence diversity to be identified in the bfpA gene. The existence of multiple bfpA alleles has important implications for the study of BFP structure and function, for our understanding of the evolution of EPEC as a pathogen, and for EPEC diagnosis and vaccine development as detailed below.

Implications for the structure of bundlin and BFP. To better understand the structural significance of amino acid polymorphism in bundlin, we carried out a graphical analysis of variability in the sequence of the protein (Fig. 6A). We also mapped the variable residues onto a theoretical three-dimensional (3D) structure model of 1 bundlin (Fig. 6B and C) that has been recently deposited in the Protein Data Bank (http://www.rcsb.org/pdb) under accession number 1QT2 (R. Chattopadhyaya and A. C. Ghose, unpublished data). The variable amino acids of bundlin are generally clustered, within both the linear sequence and the theoretical 3D structure. Variable residues which group together in the 3D structure are sometimes derived from different segments of the linear bundlin sequence (Fig. 6). Most of the highly variable amino acids are located in the C-terminal third of bundlin. Many of the variable residues are located on the pilin surface in the theoretical 3D model. Variable residues 58, 61, 63 and 64, 79, 81, and 92 (numbered in accordance with the 180-residue mature 1 bundlin prototype) occur along one side of the pilin head (shown in green in Fig. 6B and C). Variable residues 33, 95, 97 and 98, 100 and 101, 136 to 142, and 163 (shown in red) are located in a cleft on the lower convex face of the pilin head. Residues 110, 149 to 151, 154 and 155, 165, and 167 to 169 (shown in blue) are found on one side of the upper convex face of the pilin head. Residues 145 and 146 and residue 174 (shown in violet) are located on the flattened reverse face of the pilin. One possibility is that these variable positions define regions of molecular interactions either between pilin monomers or between BFP filaments. A more appealing possibility is that these variable positions define regions on the surface of BFP that are directly exposed to the host immune system. Interestingly, other regions of bundlin show little or no variation. These include the cleaved leader peptide, the -helical handle of the pilin, and five beta strands that probably serve to create the structure of the pilin head. Such regions may lack variation due to constraints inherent in critical structural domains.

View larger version (39K): [in this window] [in a new window]   FIG. 6.   Variable amino acids may be clustered on the surface of bundlin. (A) Graphical analysis of bundlin variation. The bars represent amino acid residues that vary among the eight bundlin types identified in this study. The heights of the bars are proportional to the numbers of alternate residues per site. The positions of conserved cysteine residues are indicated below the x axis. (B and C) Theoretical 3D structure of bundlin (Chattopadhyaya and Ghose, unpublished). Colored amino acid residues are those that vary between bundlin types. These residues have been assigned to four color groups (green, red, blue, and violet) in accordance with their relative locations in the molecule. These colors correspond to those displayed in the graph (A). The disulfide-bonded cysteines are yellow. Panel B shows the convex face of bundlin that is likely to be facing outward on the surface of the BPF filament. Panel C shows the flattened face on the opposite side of the molecule that is likely to be facing toward the center of the filament, based on existing models of gonococcal and V. cholerae pili.

Implications for the role of BFP in the evolution of EPEC as a pathogen. EPEC strains are defined as diarrheagenic E. coli (DEC) strains producing attaching and effacing histopathology but lacking Shiga-like toxins (28). Typical EPEC strains also carry the EAF plasmid and adhere in a BFP-dependent localized fashion to cultured epithelial cells. The importance of the EAF plasmid and plasmid-encoded BFP in pathogenicity has been demonstrated in volunteer studies (5, 35). Further support is provided by some case-control studies, where EAF⁺, but not EAF, strains have been significantly associated with diarrhea (reviewed in reference 28). As bundlin is a virulence factor and a potentially important antigen of EPEC, it is useful to understand whether bfpA alleles are capable of being mobilized between different clonal groups of EPEC strains and how this might take place.

Implications for EPEC diagnosis. The presence of multiple bfpA alleles in nature requires a re-evaluation of the methods used to identify typical EPEC strains. Molecular diagnosis of bfpA⁺ EPEC has previously been carried out with hybridization probes consisting of restriction digest fragments, PCR products, or labeled oligonucleotides (4, 18, 42, 45, 64). PCR amplification has also been described for the identification of bfpA from E. coli (22, 74, 75, 80). The current results indicate that oligonucleotides for bfpA identification should be designed such that the annealing regions represent sequences conserved in all known bfpA alleles. Use of a poorly conserved oligonucleotide sequence might result in the incorrect designation of bonafide bfpA⁺ strains as BFP negative. Results from the functional studies we conducted, along with the findings of others (38), show the limitations of using probes or PCR for diagnosissome strains carrying an intact bfpA gene may not actually produce readily detectable levels of bundlin or BFP. Therefore, accurate designation of an E. coli strain as BFP positive ultimately requires one or more tests of BFP expression or function. Our results have shown for the first time that diverse EPEC strains exhibit autoaggregation, a phenotype that is BFP dependent. While the autoaggregation assay is not yet in common use, it appears to be an acceptable and facile substitute for LA in those laboratories lacking tissue culture facilities but having a simple microscope.

Antigenic variation of bundlin and implications for vaccine development. BFP has been proposed as a potential component of an EPEC vaccine (33, 61). This concept is supported by the detection of antibundlin antibodies in children with natural EPEC infections (38) and in healthy children and mothers living in areas where EPEC is endemic (37, 53). Further support is provided by the finding that antisera raised against BFP blocks LA (19). Studies of EPEC infections in adult volunteers have demonstrated a correlation between the level of antibundlin antibodies and the propensity for loss of the plasmid encoding BFP (13). Since the plasmid-cured strain is less pathogenic than the plasmid-containing strain (35), this association establishes a link between antibundlin antibodies and virulence. The results of the current study have a direct bearing on considerations for vaccine formulations. Strains of serogroups O55, O111, and O119 have the highest prevalence in areas where EPEC is endemic (8, 21, 36, 73). Our results, which demonstrate that O55 strains, at least, may possess either  or  bfpA alleles, suggest that successful vaccines may need to include both classes of antigen. Further studies are needed to test the hypotheses that immune responses discriminate between these classes of bundlin proteins and that the immune response is a significant factor driving the evolution of bfpA allelic variants of EPEC.