A Receptor-binding Site as Revealed by the Crystal Structure of CfaE, the Colonization Factor Antigen I Fimbrial Adhesin of Enterotoxigenic Escherichia coli*

CfaE is the minor, tip-localized adhesive subunit of colonization factor antigen I fimbriae (CFA/I) of enterotoxigenic Escherichia coli and is thought to be essential for the attachment of enterotoxigenic E. coli to the human small intestine early in diarrhea pathogenesis. The crystal structure of an in cis donor strand complemented CfaE was determined, providing the first atomic view of a fimbrial subunit assembled by the alternate chaperone pathway. The in cis donor strand complemented variant of CfaE structure consists of an N-terminal adhesin domain and a C-terminal pilin domain of similar size, each featuring a variable immunoglobulin-like fold. Extensive interactions exist between the two domains and appear to rigidify the molecule. The upper surface of the adhesin domain distal to the pilin domain reveals a depression consisting of conserved residues including Arg181, previously shown to be necessary for erythrocyte adhesion. Mutational analysis revealed a cluster of conserved, positively charged residues that are required for CFA/I-mediated hemagglutination, implicating this as the receptor-binding pocket. Mutations in a few subclass-specific residues that surround the cluster displayed differential effects on the two red cell species used in hemagglutination, suggesting that these residues play a role in host or cell specificity. The C-terminal donor strand derived from the major subunit CfaB is folded as a β-strand and fits into a hydrophobic groove in the pilin domain to complete the immunoglobulin fold. The location of this well ordered donor strand suggests the positioning and orientation of the subjacent major fimbrial subunit CfaB in the native assembly of CFA/I fimbriae.

Microbial adherence to host surfaces represents a critical, early step in pathogenesis. Pili or fimbriae are one of several types of macromolecules that serve this function, projecting from the bacterial surface to dock with specific receptors on the host cell surface within preferred niches (1,2). Enterotoxigenic Escherichia coli (ETEC) 3 is a common bacterial cause of diarrhea of both humans and domesticated animals (3,4). As a human pathogen, it is second only to rotavirus as a cause of infant mortality from diarrhea in resource-limited countries and the leading cause of travelers' diarrhea (5,6). ETEC is a noninvasive pathogen that expresses fimbrial colonization factors to mediate small intestinal adherence and elaborates enterotoxins that induce fluid and electrolyte secretion (7)(8)(9)(10). Colonization factor antigen I (CFA/I) is one of the most prevalent adhesive fimbriae associated with human disease and is representative of the largest class (Class 5) of human-specific ETEC fimbriae (11)(12)(13)(14). Sequence analysis of the bioassembly components further divides this class of fimbriae into three subclasses, 5a, 5b, and 5c (15,16).
Studies of the biogenesis of CFA/I and related Class 5 fimbriae, referred to as the alternate chaperone (AC) pathway, have revealed four essential proteins (17)(18)(19). A periplasmic chaperone (CfaA) promotes subunit folding and transports subunits to an outer membrane usher (CfaC) where ordered assembly of the filamentous heteropolymer is achieved. The minor subunit (CfaE) nucleates fiber formation and localizes to the fimbrial tip, and the polymerized major subunit (CfaB) forms the fimbrial stalk. This structural model has been substantiated by immunoelectron microscopy showing the general architecture of CFA/I fimbriae (20).
Although adhesive phenotypes have been attributed to both the major and minor subunits of CFA/I (16,(21)(22)(23), neither the receptor-binding epitope nor the target intestinal receptor has been identified. Binding studies of CFA/I have largely exploited two in vitro adherence models. ETEC expressing CFA/I exhibits mannose-resistant hemagglutination (MRHA) of human, bovine, and chicken erythrocytes (16,24,25) and adheres to differentiated small intestinal Caco-2 cells in tissue culture (16,26). Earlier studies implicated the CfaB major subunit as the hemagglutinin (21), whereas more recent evidence has indicated that the CfaE minor subunit serves this role (16,20,23). Available evidence suggests that the erythrocyte receptor of CFA/I fimbriae is a sialylated protein. This is supported by the observations that sialic acid and related oligosaccharides inhibit MRHA, as does pretreatment of erythrocytes with neuraminidase and certain proteases (25,27,28).
Experimental structures for several minor adhesive subunits of Class I pilus and non-pilus systems have been determined, either in complex with their cognate periplasmic chaperone and/or as truncates bound to receptor analogs (29 -35). These structures all contain an immunoglobulin (Ig)-like fold with a missing strand and reveal features accounting for their divergent receptor specificities (36,37). A common feature of Class I pilus and non-pilus systems is the utilization of donor strand complementation and exchange in bioassembly (30,38). Recent evidence has implicated donor strand complementation and exchange in the assembly of CFA/I fimbriae and formed the basis for preparing a stable variant of the CfaE minor subunit, referred to as dscCfaE (20).
Exploiting the mechanism of donor strand complementation in engineering dscCfaE, the N-terminal ␤-strand of CfaB was added in cis to the C terminus of CfaE to produce a soluble recombinant protein. Biophysical profiling indicated that dsc-CfaE forms a stable monomer, and functional studies showed that it acts as the erythrocyte adhesin (20). Here, we present the crystal structure of dscCfaE, showing that it contains two Iglike domains. We provide both structural and mutational evidence that defines a surface site involved in erythrocyte binding. Furthermore, based on the observed fit of the donated ␤-strand, a specific mode of interaction is suggested for the articulation of minor and major subunits as well as inter-subunit interactions of the major subunits.

EXPERIMENTAL PROCEDURES
Strains and Chemicals-Plasmids, bacterial strains, and chemicals used in expression and purification of dscCfaE have been published elsewhere (20). Various chemicals for protein crystallization and crystal derivatization by heavy atoms were of highest grade possible and purchased commercially (39).
Crystal Structure Determination-Details on the expression, purification, and crystallization procedure of the recombinant dscCfaE were described previously (39). Diffraction experiments on dscCfaE crystal were carried out at the SER-CAT beamline of the Advanced Photon Source, Argonne National Lab. All data were collected at 100 K, processed, and scaled using the program HKL2000 (43). The phase problem for the diffraction data was solved based on a gold derivative by the single isomorphous replacement with anomalous scattering method using the program SOLVE (44).
Model Building and Structure Refinement-The initial C␣ tracing was carried out with the automated procedure in the programs SOLVE and RESOLVE (44) at 2.5-Å resolution, producing fragments covering several independent molecules in different asymmetric units. Extending resolution to 2.3 Å with a different native dataset yielded an electron density map of bet-ter quality, which was used for manual model building in the program O (45). Two anomalous difference Fourier maps were calculated: one based on the dataset of a selenomethionine variant crystal collected at the selenium atom absorption edge in crystal, and the other based on the dataset of a native crystal collected at a wavelength of 1.74 Å to enhance the anomalous signal of sulfur. Both were instrumental in assigning the amino acid sequence for the stretch of residues 237-240, MCFY, allowing subsequent sequence assignment without ambiguity (supplemental Fig. S1). Model refinement was carried out with the REFMAC program in the CCP4 package using two TLS groups for each monomer (46). Between REFMAC runs, manual adjustments to the model and addition of solvent and other molecules were performed in the program O. Stereochemistry of models at different refinement stages was examined with the program PROCHECK (47) ( Table 1).
Structure Analysis-Structures homologues to the CfaE, either N-or C-terminal domain, were obtained using the DALI server. Structure alignments were performed in the program O, in which comparisons were made pairwise at the domain level by severing the two domains at the residue Gly 202 . Calculations for the buried surface between two domains were performed with the program AREAIMOL (48) in the CCP4 package. The electrostatic potential surface was calculated with GRASP software (49).
Cloning, Expression, and Purification of the dscCfaE/R67A Mutant-Details of constructing the expression vector pET24dsc 19 cfaE(his) 6 used in the purification of dscCfaE[His] 6 are given elsewhere (20). A mutation was introduced into the gene encoding dscCfaE[His] 6 by site-directed mutagenesis, directing a change in residue 67 from Arg to Ala (dscCfaE[His] 6 /R67A), using the QuikChange site-directed mutagenesis kit (Stratagene). The modified vector was introduced into E. coli strain BL21(DE3) (Novagen) for expression. Purification was carried out using the same protocols as described for the dscCfaE (20).
E. coli BL21-SI(pMAM2) and derivative strains with CfaE point mutations were routinely grown on LB agar without NaCl (LBON), composed of (per liter) 10 g of Tryptone-peptone, 5 g of yeast extract, 15 g of agar, plus kanamycin (Kn, 50 g/ml) and grown at 30°C. To induce CFA/I expression, bacteria were grown in LBONϩKn agar at 30°C overnight. The culture were then passed onto LBONϩKn with 200 mM NaCl and grown at 30°C overnight. For phenotypic analyses, bacteria were harvested into phosphate-buffered saline with 0.5% D-mannose (PBS-M) and adjusted to an optical density (A 650 ) of 40, as the starting point for serial 2-fold dilutions. Positive controls included BL21-SI/pMAM2 grown in the same manner, and negative controls included BL21-SI (grown without Kn), and each test strain passed in parallel onto LBONϩKn agar without addition of NaCl.
MRHA and Inhibition Assay for Purified Mutant Protein-Human type A erythrocytes were obtained from a single volunteer, and bovine erythrocytes were obtained commercially (Lampire Biological Laboratories, Pipersville, PA). Erythrocytes were washed with PBS and stored at 4°C in Alsever's solution at a 10% concentration before usage within 2 weeks of blood drawing. Erythrocytes were washed three times in PBS and resuspended in PBS-M at a final concentration of 3% for use in the assay. MRHA was performed in 12-well porcelain tile plates with concave depressions. To test the effect of mutation R67A on adherence, the MRHA assay was performed with an admixture of erythrocytes and protein-coated beads as described here. Purified protein preparations were adsorbed to 3 M polystyrene beads (Polysciences, Inc.) using the manufacturer's suggested protocol with modifications. The beads were washed in boric acid buffer, pH 8.5, and protein adsorption was performed in 300-l volumes with the addition of 75 g of dscCfaE[His] 6 / R67A. Coated beads were then The sequence of a mature CfaE starts at residue 23 and ends at 360. In dscCfaE, the C terminus is extended (in order) by a short hairpin linker (DNKQ), the donor strand from CfaB, and a His 6 tag. The observed dscCfaE in crystal is from residue 23 to 378. B, stereoscopic view of the dscCfaE molecule in a ribbon representation. The CfaEad is red and the CfaEpd is green. The donor strand G is magenta, and the short linker between the two domains is shown as a blue coil. Three pairs of disulfide bonds and the residues Arg 67 and Arg 181 are shown as ball-and-stick models with carbon atoms in black, nitrogen in blue, and sulfur in yellow. The diagram was produced with the graphics programs MolScript (58), BobScript (59), and Povray interfaced with the GL-render graphics software (NCI, National Institutes of Health, convent.nci.nih.gov/glr/glrhome.html). C, the dscCfaE dimer present in the crystallographic asymmetric unit. Here, one monomer is shown in coral and the second in cyan. The donor strand is depicted in magenta. blocked in boric acid buffer with bovine serum albumin (0.05 mg/ml), pH 8.5. After blocking, the beads were pelleted and resuspended in PBS-M. Effective adsorption was confirmed by SDS-PAGE analysis of boiled beads. Beads coated with dscCfaE, CfaA, and CFA/I periplasmic chaperone were generated by the same method and used as positive and negative controls, respectively. The MRHA assay with human type A, and bovine erythrocytes was performed as previously described (16). Equal volumes (25 l) of 3% erythrocytes, bead suspension, and PBS-M were mixed, rocked on ice for 20 min, graded by visual inspection, and scored as follows: negative, no MRHA activity; 1ϩ, low, weak reaction; 2ϩ, moderate reaction; 3ϩ, strong reaction; and 4ϩ, nearly instantaneous and complete reaction involving all of the erythrocytes.
MRHA for E. coli with Mutant CFA/I-To each well was added 25 l of each erythrocyte suspension, bacterial suspension, and PBS-M (75 l, total volume), and the plates were incubated with rocking on ice for 20 min. For each bacterial preparation that gave a positive MRHA reaction at the starting concentration (i.e. A 650 ϭ 40), a 2-fold dilution series was performed using PBS-M as the diluent, and the dilution series was assayed for MRHA. The highest dilution yielding a positive MRHA reaction was recorded as the MRHA titer. All bacterial samples were tested in two separate experiments on different days, and each experiment was performed in duplicate. Positive and negative control bacteria included BL21-SI(pMAM2) and E. coli DH5␣, respectively, as well as the same recombinant bacterial sample grown in LBON (0 mM NaCl).
To confirm the expression of CFA/I fimbriae on the surface of each bacterial sample, an agglutination test was performed on glass slides. For each preparation, 8 l each of the bacterial suspension (diluted 1:8) and rabbit polyclonal anti-CFA/I serum (diluted 1:8) were mixed on the slide, and the presence or absence of agglutination was visually determined. BL21-SI(pMAM2) grown in LB with 200 mM NaCl was carried as the positive control. BL21-SI served as the negative control, as did BL21-SI (pMAM2) grown in LBON.

RESULTS AND DISCUSSION
Overview of the CfaE Structure-A variant of CfaE (Fig. 1A) was purified to homogeneity by nickel affinity, cation exchange, and size-exclusion chromatography (20). The first 22 residues of dscCfaE constitute the signal peptide, which is cleaved during export across the inner membrane. The dscCfaE protein was crystallized in a hexagonal form with the space group symmetry of P6 2 22. The structure was determined using a gold derivative and several native datasets and refined to 2.3-Å resolution; details of crystallization, x-ray diffraction data collection, and phasing were previously published (39). Statistics on the quality of diffraction data sets, single isomorphous replacement with anomalous scattering phasing, and the refined model are given in Table 1. Two dscCfaE molecules are present in one asymmetric unit. Both molecules are virtually identical with a root-mean-square deviation (r.m.s.d.) of 0.6 Å. One has 356 residues visible in the electron density from Ala 23 to Val 378 , whereas the other has 355 residues from Ala 23 to Pro 377 . Approximately 350 water molecules, in addition to a few malonate and polyethylene glycol moieties, were modeled in asso-ciation with each dscCfaE. Of the 19 residues that comprise the extension peptide at the C terminus (Fig. 1A), only 12 and 11 residues are modeled for the two molecules, respectively.
The dscCfaE molecule consists of two domains of roughly equal size (Fig. 1B). The N-terminal domain closely abuts the C-terminal domain, giving a cylindrical appearance to dscCfaE, with molecular dimensions of 110 ϫ 20 ϫ 20 Å. The two dsc-CfaE molecules in the asymmetric unit have a substantial number of contacts and pack against each other through the N-terminal domains, displaying an inverse Y shape (Fig. 1C).
Domain Structure of the dscCfaE-Previous work (16,22) led us to refer to the structurally defined N-terminal domain as the adhesin domain (CfaEad). This domain contains residues from Ala 23 to Asp 200 , which form one anti-parallel ␤-sheet (Sheet 1) and one mixed ␤-sheet (Sheet 2) (supplemental Fig. S2). The ␤-structure has a topology that resembles the v-type Ig fold (36) with nine ␤-strands. Sheet 1 consists of ␤-strands AЈ, B, BЈ, E, D, DЈ, CЈ, and CЉ, and Sheet 2 is composed of ␤-strands A, G, F, C, CЉ, and D (Figs. 1B and S2). Like many v-type Ig domains, ␤-strands are often first associated with one sheet and later reassociate with the second sheet, such as the strands A, CЉ, and D, which results in adoption of a ␤-barrel shape. There are two pairs of disulfide bonds in the CfaEad. The Cys 72 -Cys 83 pair fixes conformation of the loop between strands C and CЈ, and the Cys 130 -Cys 143 pair stabilizes the conformation of the loop between strands DЈ and E (Figs. 1B and S2). A cis-proline was found for residue Pro 38 .
The C-terminal domain immediately follows the short threeresidue linker (Lys 201 -Gly 202 -Asn 203 ). Based on its location and structural features, we infer that this domain mediates articulation of the adhesive subunit with the main body of the fimbria. It is therefore termed the pilin domain (CfaEpd). Sequence of the native CfaEpd starts with residue Ile 204 and ends at Leu 360 , but in dscCfaE it ends at residue Val 378 . The pilin domain folds into a ␤-sandwich with a topology reminiscent of the adhesin domain (Figs. 1B and S2), hence also conforming to the family   (1J8S, 1KLF, 1P5V, and 1USZ), FimH (1KLF) is the only one determined as a full-length two-domain adhesin like CfaE, and it is in complex with the FimC chaperone. The linkage between the two domains of FimH consists of four residues, including two consecutive glycines and the interaction between the two domains appears limited. Therefore, the two domains may be flexible with respect to each other especially in the absence of the chaperone FimC. Indeed, the largest r.m.s.d. between a pair of FimH (out of eight in an asymmetry unit of 1KLF) is Ͼ1.2 Å. The connection between CfaEad and CfaEpd is shorter, consisting of three residues that are highly conserved in the Class 5 pilus family (Fig. 3). As a result, the linker is entirely buried and the molecule takes the shape of a cylinder without an apparent demarcation between the two domains ( Fig. 2A).
The two dscCfaE molecules in a crystallographic asymmetric unit, although in different packing environments, are identical, with an r.m.s.d. of 0.6 Å when superimposed, commensurate with the notion of very rigid molecules. The short length of the linkage seems to contribute to the apparent rigidity of the molecule and may constrain relative motions between the two domains. In CfaE, the angle between the adhesin and pilin domains is close to 180°, as measured from the center of gravity of each domain to that of the connector. This observation is in contrast to the near 150°angle between the two domains of FimH observed in the crystal structure of the adhesin-chaperone complex, which has been suggested to be part of a mechanism for shear-force-enhanced receptor attachment (40).
We observed extensive interactions at the interface between the two domains of CfaE. A total of 17 hydrogen bonds, either direct or mediated by water molecules, were found (supple-  Fig. S1. The conserved sequence at the C terminus of minor subunits is indicated with red boldfaced letters below the alignment. Conserved residues that are part of the receptor-binding pocket are in boldface, and those residues subclass-specific that display differential responses to human type-A and bovine red cells are in boldface and colored in brown, blue, and magenta, respectively. mental Table S1). Additionally, a significant number of residues from either domain are within van der Waals contact distances, and the buried surface area between the two domains is nearly 700 Å 2 . The relative orientation and close association between the CfaE adhesin and pilin domains may be due to the crystal packing environment or may have a physiological basis whereby these interactions may place strong constraints on the relative movement between the adhesin and pilin domains of CfaE. Possible biological explanations include differences in receptor-binding mechanisms among different types of fimbriae, operating in the different environments in which the respective bacteria typically colonize, or in differential stability requirement for the respective fimbriae. It remains possible that the abutment of the adhesin and pilin domains observed here in the crystal structure of CfaE is peculiar to the apo form of the adhesin. Whether the association is artifactual or more dynamic and subject to alterations in orientation and proximity upon binding with its cognate chaperone or receptor awaits co-crystallization of the corresponding units.
Structural Alignment of CfaE with Class I Fimbrial Adhesins and Evolutionary Implications-The general shape of CfaE is similar to that of the type 1 pilus adhesin FimH. These molecules share a two-domain architecture, with each domain composed of a ␤ barrel and connected by a short linker. For both CfaE and FimH, the N-terminal domain mediates adhesion and the C-terminal domain appears to non-covalently fasten to the underlying filament via donor strand complementation (20,30,52). We compared the tertiary structure of the CfaE adhesin domain to that of the corresponding domain of FimH and related subunits assembled by the chaperone-usher (CU) pathway, including PapG of P pili, Caf1 of F1 capsular antigen, AfaE3 of Afa, DraE of Dr fimbriae, and GafD of F17 fimbriae. Except for DraE, which is not included in the alignment, all can be superimposed to CfaEad with r.m.s.d. values from 1.9 to 2.3 Å (see supplemental Table S2). Several are topologically similar to the jellyroll fold exemplified by the G subunit of bacteriophage X174 (Fig. 4) (53). FimH and DraE most closely resemble a typical jellyroll topology, whereas CfaE and PapG have the fewest topological equivalents. These tertiary structural similarities are intriguing when one considers the prevailing view that CFA/I and related fimbrial systems of the alternate chaperone pathway have developed along separate but convergent evolutionary paths from Class I pilus and non-pilus systems of the chaperone-usher pathway (18,19,54).
The argument for convergent evolution of the AC and CU pathway systems has centered on the lack of primary sequence similarities and absence of shared conserved sequence motifs between structural and assembly components of the two pathways (17,19). CfaE, however, shares 19 -21% primary sequence identity with FimH, PapGI, and GafD, which is not dissimilar to the range of identity (19 -25%) between FimH and the corresponding proteins as determined by pairwise alignments (data not shown). The ultrastructural appearance of the stalks of fimbriae formed by the CU and AC pathway are similar, but typically differ in the appearance of their distal segment. The former, exemplified by Type 1 and P pili, appear as a composite fiber with a tip fibrillar structure adjoined to the rigid, pilus stalk (55, 56), whereas the latter form a simple, rod-like struc-ture without a distinguishable tip substructure. 4 Given the ultrastructural diversity of macromolecules erected by the CU pathway (57), however, the difference between type I and CFA/I fimbrial tip morphology would not necessarily have any bearing on the evolutionary question. The similarities in three-dimensional folding between CfaE and fimbrial adhesins of the CU pathway described above and demonstration here and elsewhere (20) of the shared assembly mechanism of donor strand complementation argue that structures formed by the classic chaperone-usher pathway and the so-called alternate chaperone pathway have likely evolved from a common, distant progenitor system along divergent evolutionary paths.
Binding Site for Host Cell Receptor-In the crystal structures of four adhesin-ligand complexes, the binding pockets have been found in one of two general locations (Fig. 4). In FimH, the adhesin binds its receptor at the tip or surface loop region of the adhesin domain distal to the pilin domain (41). Adhesins can also attach to their receptor on the side of the adhesin domain along the surface of the ␤-sheet, as illustrated by PapGII binding to globoside (32), GafD to GlcNAc (35), and DraE to chloramphenicol (34). Comparison of the structures of adhesins with and without ligand receptors leads to two conclusions. First, although a wide variety of interactions are employed in the binding of adhesins to ligand receptors, largescale conformational changes on the part of adhesin are not observed. Second, despite structural similarities among adhesive molecules of known structures, binding specificities diverge widely due to sequence dissimilarity.
Prior demonstration that the residue Arg 181 of CfaE and the corresponding residue of the related adhesin CooD of CS1 fimbriae (Fig. 3) are required for fimbriae-mediated erythrocyte agglutination suggests that this conserved residue may lie within the erythrocyte receptor-binding pocket (23). Located at the upper surface of CfaEad distal to the CfaEpd, Arg 181 is found in a positively charged depression (Figs. 1B and 2A) and surrounded by a cluster of residues that are highly conserved in the Class 5 fimbrial adhesins, including residues from three different loops (i.e. BЈ-C, DЈ-E, and F-G loops) (Figs. 2C and 3). This pocket thus appears to be a suitable location to which a negatively charged sialylated receptor might bind (25,28). To confirm the role of this domain, Arg 67 , which is adjacent to Arg 181 , was mutated to Ala (dscCfaE/R67A), and the mutant protein was tested for MRHA. As shown in Fig. 5A, purified, bead-adsorbed dscCfaE/R67A failed to agglutinate human erythrocytes, similar to our previous findings for the dscCfaE/ R181A mutant (20). These results implicate the pocket anchored by these two residues as the putative receptor-binding domain.
Conserved residues in Class 5 fimbriae can be grouped into two types, those located on the surface of the molecule, and those that contribute to the core of the protein's hydrophobic interior. The conserved surface residues were mapped onto the structure (Fig. 2B). Two clusters were found: one cluster is located on the tip of CfaEad around the Arg 181 , including Tyr 58 , His 62 , Leu 64 , Asp 66 , and Arg 67 as indicated in Fig. 2B by a black circle. Another as marked by the black oval is found at the interface between two domains centered on the residue Trp 309 , consisting of two sequence motifs, one from residues Thr 105 to Arg 113 of CfaEad and the other between residues Asn 308 and Arg 311 of CfaEpd. As discussed above, the first conserved cluster represents the putative receptor-binding site, whereas the second one is yet to be associated with any particular functions of CfaE. , and PapG (1J8R). As a comparison, the ribbon and topology diagrams for the G subunit (2BPA) of bacteriophage X174 are provided, which has a typical jellyroll fold. Bound ligands, determined crystallographically, for GafD, FimH, DraE, and PapG are also shown and labeled. Using the topology of the X174 G subunit as a reference for the jellyroll fold, topologically equivalent ␤-strands are labeled for each adhesin domains, providing a basis for structural and evolutionary relationship among different bacterial adhesins.

Subclass-specific Residues Around the Receptor-binding Site
May Be Responsible for Host and Serotype Specificity-To determine the role in hemagglutination of individual residues in the neighborhood of Arg 181 , we introduced site-specific mutations into CfaE in the plasmid pMAM2, which encodes all components of the CFA/I and directs surface expression of mutant fimbriae with single site mutations of CfaE. Table 2 shows the MRHA phenotype (Fig. 5B) and titer of twelve such mutations involving residues that are either invariant (fully conserved) or are subclass-specific for Class 5 ETEC fimbrial adhesins. As mentioned earlier, the receptor-binding site is made of residues from loops BЈ-C, DЈ-E, and F-G. The residues from the BЈ-C loop are conserved in all Class 5 fimbriae, one of the two residues from the DЈ-E loop is conserved, whereas most residues from the F-G loops are subclass specific ( Table  2). All positively charged residues (Arg 181 , Arg 182 , and Arg 67 ) are absolutely required for receptor binding and cluster together to form a positively charged center. The conserved Ser 138 , which sits in a plane below the receptor-binding epitope, may be important for maintaining the structural integrity of the binding site rather than directly participating in binding.
The positively charged center of the binding pocket is surrounded by a band of subclass-specific residues mostly from the F-G loop with one (His 140 ) from the DЈ-E loop. Mutations of those residues display altered interactions with red cells, and several show discriminatory behavior to either human type-A or bovine red cell species. For example, the mutant T186A appears to enhance binding to both blood cells, whereas the two mutants (Y183A and D184A) exhibit discriminatory binding toward human type-A red cells ( Table 2). Mutations of H140A and Y187A reduce binding to both red cells, but to a different extent. The former lost binding completely to human blood cells and the latter to bovine red cells.
Unlike arginine residues found in most crystal structures, those found in the positively charged center of CfaEad were conformationally stable, as reflected from their low temperature factors obtained crystallographically ( Table 2). The stability of the positively charged cluster apparently stems from the support of surrounding residues, which are mostly subclassspecific. We speculate that the changes in the surrounding residues of different subclasses alter the packing arrangement of the central residues, leading to differential binding to various cell types.
Implication for the Assembly of CFA/I Fimbria-In a recent study on the assembly of the CS1 pilus (42), alanine-scanning mutagenesis was carried out on the major subunit CooA (CfaB homolog) for both the N-and C-terminal portions of the protein. The MRHA assay was used to identify residues essential  for assembly. In the C terminus of CooA, a conserved sequence motif AGXYXGX 6 T for the major subunits of Class 5 fimbriae was found, which is characterized by an alternating pattern of hydrophobic and hydrophilic residues in a zipper-like arrangement. A few non-conserved hydrophobic residues in this motif were found to be required for expression of a positive MRHA phenotype. This motif also exists in the pilin domains of the minor subunits of Class 5 fimbriae, displaying a greater number of conserved residues than for the corresponding major subunit domain. The adhesin-specific consensus shows the sequence motif AGQYXGX 4 TFT (Fig. 3). In the structure of dscCfaE, this motif constitutes the entire F strand of the pilin domain, hence the name "F zipper," providing essential interactions with the donor strand. The conserved residues in this motif are located at either end of the sequence motif. The residues AGQY form the last four main chain H-bonds with the donor strand peptide, whereas the residues TFT engage in interactions with the very beginning of the donor peptide (Fig. 6). In particular, the Thr 354 does not form main chain H-bonds with the donor peptide. Instead, the side-chain hydroxyl group of this residue could be in a position to form an H-bond with the N-terminal amine group of the donor peptide in a native fimbrial assembly. The hydrophobic residues that are critical for hemagglutination in the alanine-scanning mutagenesis all face inside, contributing to the hydrophobic core of the protein (42). Therefore, we believe that this conserved F zipper motif not only fixes the conformation of the donor strand by forming a large number of H-bonds, but also serves as a template that determines the orientation of the donor strand and, by extension, of the subjacent major fimbrial subunit.
Experiments with electron microscopy image reconstruction of the intact CFA/I fimbriae are underway, 4 and are expected to provide information on how major and minor subunits are assembled into a mature fimbria. However, detailed knowledge of interactions between the fimbrial subunits at atomic resolution will have to await structure solutions of binary complexes between CfaBs and between CfaE and CfaB. Despite obvious limitations, the donor strand complemented CfaE structure provides a basis for speculating how the fimbrial subunits are assembled in a quaternary structural organization. Secondary structure prediction of the major subunit CfaB indicates a very short connection (2-3 residues) between the donor strand and the first ␤-strand of the ␤-sandwich, suggesting a relatively rigid connection between the major and minor subunit.
In summary, the structure of dscCfaE provides the first atomic detail to a minor adhesive component of Class 5 fimbriae. The two-domain molecule appears to be rigid and features a positively charged surface depression at its upper surface, which likely serves as the receptor-binding pocket. Structure-based site-directed mutagenesis implicates a cluster of positively charged residues essential for receptor binding and suggests discriminatory roles in binding of surrounding subclass specific residues. The structure provides a molecular basis for the conservation of the F zipper, which defines the location FIGURE 6. Interaction of the conserved C-terminal F strand (AGQYXGX 4 TFT sequence) of the minor pilin with the major pilin donor strand. This conserved sequence forms the ␤-strand F (carbon atoms in black with overlay gray ribbon), guiding the alignment of the donor strand G (carbon atoms in yellow with overlay light green ribbon). The G and F strands are flanked by strands C, CЈ, and AЈ. Oxygen atoms are colored red, nitrogen blue, and sulfur yellow. and orientation of the major subunit in the native fimbrial assembly.