The Distinct Binding Specificities Exhibited by Enterobacterial Type 1 Fimbriae Are Determined by Their Fimbrial Shafts*

Type 1 fimbriae of enterobacteria are heteropolymeric organelles of adhesion composed of FimH, a mannose-binding lectin, and a shaft composed primarily of FimA. We compared the binding activities of recombinant clones expressing type 1 fimbriae from Escherichia coli, Klebsiella pneumoniae, and Salmonella typhimurium for gut and uroepithelial cells and for various soluble mannosylated proteins. Each fimbria was characterized by its capacity to bind particular epithelial cells and to aggregate mannoproteins. However, when each respective FimH subunit was cloned and expressed in the absence of its shaft as a fusion protein with MalE, each FimH bound a wide range of mannose-containing compounds. In addition, we found that expression of FimH on a heterologous fimbrial shaft, e.g. K. pneumoniae FimH on the E. coli fimbrial shaft or vice versa, altered the binding specificity of FimH such that it closely resembled that of the native heterologous type 1 fimbriae. Furthermore, attachment to and invasion of bladder epithelial cells, which were mediated much better by native E. coli type 1 fimbriae compared with native K. pneumoniae type 1 fimbriae, were found to be dependent on the background of the fimbrial shaft (E. coli versus K. pneumoniae) rather than the background of the FimH expressed. Thus, the distinct binding specificities of different enterobacterial type 1 fimbriae cannot be ascribed solely to the primary structure of their respective FimH subunits, but are also modulated by the fimbrial shaft on which each FimH subunit is presented, possibly through conformational constraints imposed on FimH by the fimbrial shaft. The capacity of type 1 fimbrial shafts to modulate the tissue tropism of different enterobacterial species represents a novel function for these highly organized structures.

Fimbrial lectins, which are filamentous organelles of bacterial adhesion, are generally classified according to the structure of the carbohydrates they recognize (1). Their specificity is usually defined as the simplest carbohydrate structure, typically a monosaccharide, that best inhibits lectin-mediated adhesion. This is referred to as the primary sugar specificity of the lectin (2). Within lectins possessing the same primary sugar specificity, differences in the binding of different oligo-saccharides is often observed. This is referred to as the fine sugar specificity (2). Type 1 fimbriae, the primary sugar specificity of which is for D-mannose, are expressed in Escherichia coli as well as almost in all enterobacteria. Based primarily on studies of E. coli type 1 fimbriae, it is known that these organelles are heteropolymers composed of a major subunit and at least three minor subunits (3)(4)(5)(6)(7). The minor subunit FimH is responsible for the sugar specificity of type 1 fimbriae because inactivation of the fimH gene abolishes the binding activity of the bacteria without any apparent effect on fimbrial expression (3,8). Moreover, isolated FimH mimics many of the mannose-specific binding reactions of type 1 fimbriae (9,10). On the other hand, FimA is the major subunit that makes up Ͼ95% of the fimbrial shaft and is structurally and antigenically heterogeneous among different species (11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22). FimH has been crystallized, and the sugar-binding region was mapped to the N-terminal half (residues 1-156) of the molecule, whereas the region that associates with the fimbrial shaft was mapped to the C-terminal half (residues 160 -277) of the FimH molecule (23). Because fimbrial proteins are not typically soluble in solution, FimH was crystallizable only when bound to its periplasmic chaperone, FimC (23).
One of the earliest indications of heterogeneity in the fine sugar specificity among type 1 fimbriae from different strains was the finding that the binding reactions mediated by various enteric type 1 fimbriae exhibit differing sensitivities to competition with defined mannosecontaining compounds (24,25). These reports proposed that the receptor-binding pocket on Salmonella type 1 fimbriae is different from that found on the fimbriae of E. coli or other enteric bacteria. Subsequently, heterogeneity in fine sugar specificity was discovered even within the same species (26). Significantly, it was found that different E. coli type 1 fimbriae can be classified into those that bind either trimannose or monomannose and those that recognize trimannose only (26). Two mechanisms have been suggested in the literature to explain these interspecies and intraspecies differences in the fine sugar specificity of the type 1 fimbrial lectin. Madison et al. (27) reported that the interspecies (E. coli and K. pneumoniae) differences in fine sugar specificity is influenced by the fimbrial shaft, which may induce distinct conformational changes in its FimH subunit. Sokurenko et al. (26) found that the intraspecies heterogeneity in the fine sugar specificity of type 1 fimbriae from different E. coli isolates can be attributed to the allelic variation in the primary amino acid structure of FimH. The heterogeneity between different enterobacterial species and within the same species has been implicated in contributing to different infectious processes. For example, it has been argued that uropathogenic E. coli clones emerge in a process called pathoadaptation to express type 1 fimbriae with specificity for monomannose and trimannose, whereas fecal isolates are specific for trimannose only (28). However, the biological significance of the interspecies differences among enterobacteria discussed above has not been studied.
Here, we investigate whether the fimbrial shaft-driven modulation of the fine sugar specificity of various enterobacterial species influences tissue tropism. We show that the apparent heterogeneity in binding among type 1 fimbriae is modified by the fimbrial shaft and is responsible for variations in tissue tropism.

EXPERIMENTAL PROCEDURES
Bacterial Strains, Culture, and Plasmids-The bacterial strains and plasmids used in this work are described in TABLE ONE. All strains were cultured statically in Luria broth (Difco) with the appropriate antibiotics (80 g/ml chloramphenicol for strains containing pACYC184based plasmids, 100 g/ml ampicillin for strains containing pUC18based plasmids, and both antibiotics for strains containing both plasmids).
3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium Bromide (MTT) Bacterial Adherence Assay-The adherence of various fimbriated bacteria to the BECs and GECs was examined as described previously (30). Briefly, cells were plated onto 96-well plates, incubated for 24 -48 h, and then fixed overnight with 2.5% paraformaldehyde. The monolayers were washed three times with sterile phosphate-buffered saline (PBS) and pretreated for 1 h at room temperature with blocking buffer (3% bovine serum albumin (BSA; Sigma) in PBS). 100 l of each bacterial strain to be tested (A 600 nm ϳ 1.0) in PBS was incubated with the monolayers for 30 -60 min at room temperature on a rotating shaker. Nonadherent bacteria were removed by washing the cell monolayers three times with PBS. 50 l of Luria broth was applied to each monolayer and incubated for 15 min at 37°C. MTT (Sigma) was dissolved in PBS at 2 mg/ml, filter-sterilized, and stored at 4°C until used. 50 l of 2 mg/ml MTT was added to all wells, and the plates were incubated for 15 min at 37°C to allow reduction of MTT to formazan by live bacteria. 150 l of isopropyl alcohol and hydrochloric acid at a ratio of 24:1 was added to each well to solubilize the formazan, and the absorbance was measured at 620 nm using a Tecan Sunrise remote microplate reader.
Bacterial Invasion Assay-Bacterial invasion of BECs was determined as described previously (31). Briefly, 5637 cells were seeded into 96-well plates and grown to confluence. Bladder cells were infected at a multiplicity of infection of 50 -100 bacteria/host cell by the addition of 100 l of each bacteria to be tested diluted in serum-free cell culture medium containing 10 mg/ml BSA (A 600 nm ϳ 0.05). Plates were then incubated at 37°C for 1 h, and the medium was replaced with fresh culture medium containing the membrane-impermeable antibiotic gentamycin (100 g/ml; Invitrogen) to kill extracellular bacteria. Incubation was continued for an additional 1 h, after which each well was washed three times with PBS, and the cells were lysed by the addition of 100 l of 0.1% Triton X-100 in PBS and plated onto Luria broth-agar plates with the appropriate antibiotics.
Aggregation of Mannose-containing Glycoproteins-The aggregation of various mannosylated glycoproteins by various type 1 fimbriated bacteria was undertaken as described previously (30,32). Briefly, cultures of bacterial cells were washed twice with sterile PBS before being resuspended in PBS. 980 l of each bacterial suspension (A 600 nm ϳ 1.75) was transferred to a cuvette. 20 l of the glycoprotein being tested (10 mg/ml in PBS) was then added to each cuvette. The cuvettes were covered with Parafilm and shaken gently for 1 min, at which point they were placed into a spectrophotometer, and a reading at 600 nm was taken. As a control, 980 l of bacterial suspension (A 600 nm ϳ 1.75) was mixed with 20 l of PBS. Subsequent readings were taken at 600 nm after a total elapsed time of 5 and 15 min and every 15 min thereafter for 1 h. The A 600 nm reading dropped as the bacteria aggregated and settled at the bottom of the cuvettes. Aggregation was expressed as absorbance as a function of time. To measure agglutination of Saccharomyces cerevisiae, 950 l of each bacterial suspension (A 600 nm ϳ 1.75) was transferred to a cuvette. 50 l of a 0.5% (w/v) suspension of S. cerevisiae was added to each cuvette, and the absorbance was measured as described above. The type 1 fimbriae-deficient E. coli strain ORN103 was used as a control. All glycoproteins employed in this assay were purchased from Sigma and are listed in TABLE TWO. DNA Manipulations-Restriction enzymes and DNA-modifying enzymes (New England Biolabs Inc., Beverly, MA) were used as recommended by the supplier. Oligonucleotides were synthesized and purified commercially (Integrated DNA Technologies, Inc., Coralville, IA). A complete list of the plasmids that were generated as well as their gene products are described in TABLE ONE.
Determination of the DNA Sequences of the fimF, fimG, and fimH Genes of E. coli and K. pneumoniae-The sequences of the E. coli strain J96 and K. pneumoniae strain IA551 fimF and fimG genes carried on the pSH2 and pBP7 plasmids, respectively, were determined using primers FimF1 (5Ј-TCG CCA ATT ATC AAC TGC CAC-3Ј) and FimF2 (5Ј-CCT GTA TCG TCG CAC TTG C-3Ј) for fimF and primers FimG1 (5Ј-TAA TGG CGA CAC AGG TGC-3Ј) and FimG2 (5Ј-ATG ACC AGG CAT TTA CCG AC-3Ј) for fimG. The sequence of the E. coli strain J96 fimH gene on the pSH2 plasmid was determined using primers KP1 (5Ј-TGG TCG GTA AAT GCC TGG TCA TTC-3Ј) and KP2 (5Ј-CAT TAG CAA TGT CCT GTG ATT TCT-3Ј), derived from the published fimH sequence of E. coli K12 strain PC31 (5). To obtain the sequence of the K. pneumoniae strain IA551 fimH gene, a 3.5-kb PvuI-SalI fragment containing fimHK was released from the pBP7 plasmid, gel-purified, blunt-ended, and cloned into the SmaI site of the pUC19 cloning vehicle. This construct was named pKT201, and the insert was sequenced completely using an automated fluorescence sequencer (PerkinElmer Life Sciences).
Construction of Bacterial Strains Expressing Hybrid Type 1 Fimbriae-These strains were produced as described previously (27). Briefly, a 5.0-kb BamHI fragment from plasmid pSH2 or a 5.5-kb BamHI-SalI fragment from plasmid pBP7 was inserted into a similarly cut site in a pUC18 cloning vector to make plasmids pBM20 and pBM10, respectively. These fragments contained E. coli or K. pneumoniae fimH as well as the minor fimbrial subunit genes fimF and fimG, which are thought to allow optimal functional activity of FimH on a heterologous fimbrial filament (27). The plasmids were then transformed into E. coli strain ORN103 that had been previously transformed with either pUT2002 (E. coli FimH Ϫ type 1 fimbriae) or pMD7 (K. pneumoniae FimH Ϫ type 1 fimbriae). Transformant designations such as FimHE-ShaftE and FimHK/ShaftE show the presence of FimH and the fimbrial shaft, the source of each (E for E. coli and K for K. pneumoniae), and the presence of one or two plasmids (hyphen or slash, respectively). Plasmid pMD7 (in this work, used interchangeably with pBP799 (27)) was produced by replacing the BglII-SalI fragment of pBP7 (containing the majority of K. pneumoniae fimH) with a short linker made from oligonucleotides MD1 (5Ј-GAT CTT TGA TAA GAG CTC TGA TAG-3Ј) and MD2 (5Ј-TCG ACT ATC AGA GCT CTT ATC AAA-3Ј). A depiction of the plasmids created and a description of the fimbrial phenotype of each strain produced are given in supplemental Fig. 2 and Fig. 4, respectively. To rule out the possible effects of heterologous FimF and FimG, two additional plasmids were created by PCR to clone fimH alone and to insert it into SacI-SphI-cut pUC18. The following primers were used to clone E. coli fimH: forward, 5Ј-GAG AGC TCA ACC CGA AGA GAT GAT TGT A-3Ј; and reverse, 5Ј-GCC GCA TGC TTA TTG ATA AAC AAA AGT CAC-3Ј. The forward primer 5Ј-GAG AGC TCA ACC CGA AGA GAT AAT TGT C-3Ј was used to clone K. pneumoniae fimH in conjunction with the reverse primer used to clone E. coli fimH. Sequencing of each fimH clone was performed to verify that the amino acid sequence of each FimH protein was correct. The plasmids carrying K. pneumoniae or E. coli fimH (designated pMD108 and pMD109, respectively) were then transformed into E. coli strain ORN103 that had been previously transformed with either pUT2002 or pMD7 to create hybrid type 1 fimbriae in the absence of heterologous FimF and FimG.
Protein Sequence Analysis-Software programs used to analyze the DNA-derived protein sequence data have been described (30).
Assays for Fimbriation-Bacterial cell expression of type 1 fimbriae was confirmed by mannose-sensitive yeast agglutination.
Purification of MalE/FimH Fusion Proteins, SDS-PAGE, and Western Blotting-The fusion proteins were expressed and purified as described previously (10). SDS-PAGE analysis and immunostaining of specific proteins were performed as described previously (10) using antisera raised against FimHE-(1-100).
Verification of FimA and FimH Expression-To examine the relative amounts of the FimA shaft subunit and the FimH adhesin subunit expressed by each bacterial strain, each strain was grown statically for 48 h and washed with PBS by centrifugation. Bacteria (ϳ1 ϫ 10 10 ) of each strain were boiled in acid for 5 min to dissociate the fimbrial subunits, neutralized by the addition of NaOH, brought to 1ϫ SDS-PAGE sample buffer, and boiled for an additional 2 min (33). The protein concentration of each sample was determined using the Bio-Rad protein assay, and 100 g of total protein from each sample was subjected to SDS-PAGE and transferred to a polyvinylidene difluoride membrane. Western blotting was performed using rabbit antisera raised against E. coli FimA or FimHE-(1-100).
Mouse BEC and GEC Overlay Assay-This assay was performed as described previously (10). Briefly, purified fusion proteins (MalE/ FimHE, MalE/FimHK, and MalE/FimHS) and MalE were subjected to SDS-PAGE and transferred to nitrocellulose membranes. After block- ing in 3% BSA in PBS for 1 h, the nitrocellulose blot was overlaid with 5 ϫ 10 6 biotinylated mouse BECs or GECs in the presence and absence of 100 mM methyl ␣-D-mannopyranoside. BECs and GECs were biotinylated as described previously (12). After 1 h of incubation at room temperature, the blot was rinsed several times with PBS, and the bound cells were probed with alkaline phosphatase-conjugated avidin, followed by the substrate 5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium.
Glycoprotein Overlay Assay-This assay was performed as described previously (30). Briefly, purified fusion proteins (MalE/FimHE, MalE/ FimHK, and MalE/FimHS) and MalE were subjected to SDS-PAGE and transferred to nitrocellulose membranes. After blocking the nitrocellulose blot in 3% BSA in PBS for 1 h, the membrane was overlaid with a glycoprotein-containing solution at a concentration of 100 g/ml in PBS in the presence and absence of 100 mM methyl ␣-D-mannopyranoside. After 1 h of incubation at room temperature, the blot was rinsed several times with PBS. Bound glycoproteins were probed with alkaline phosphatase-conjugated concanavalin A (Sigma), followed by 5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium.

Bacteria Expressing E. coli, K. pneumoniae, or S. typhimurium Type 1 Fimbriae Exhibit Different Binding Specificities for Mouse BECs and
GECs-We found that ORN103(pSH2) bacteria expressing E. coli strain J96 type 1 fimbriae bound well to BECs and moderately to GECs, whereas ORN103(pBP7) bacteria expressing K. pneumoniae strain IA551 type 1 fimbriae bound moderately to BECs but exhibited minimal binding to GECs. In contrast, ORN103(pISF101) bacteria expressing S. typhimurium strain 6704 type 1 fimbriae exhibited minimal binding to BECs and bound very well to GECs (Fig. 1, A and B). Thus, there exists remarkable variation among enterobacterial type 1 fimbriae in their capacity to bind mucosal cells of gut and urinary tract origin. The binding of the fimbriated bacteria belonging to the three species was minimal in strains containing plasmids pUT2002, pBP799, and pKT303, which encode E. coli, K. pneumoniae, and S. typhimurium fimbriae, respectively, which are deficient in FimH. In addition, binding was inhibited by methyl ␣-D-mannopyranoside (data not shown). Taken together, the data suggest that FimH is the decisive determinant on the fimbriae responsible for epithelial cell binding.

Enterobacterial Type 1 Fimbriae Exhibit Considerable Diversity in Their Ability to Aggregate Soluble Mannose-containing Glycoproteins-
We examined the ability of each of the strains to bind and aggregate 10 soluble glycoproteins, most of which are known to be mannosylated. The capacity of type 1 fimbriated bacteria to aggregate mannosylated proteins is dependent not only on the availability of mannose residues in these molecules, but also on their number and correct spatial arrangement within the molecule. Only mannan and p-aminophenyl ␣-D-mannopyranoside/BSA were aggregated by all three type 1 fimbriated bacteria, albeit at different rates (Fig. 2, A and B; and TABLE TWO). The two glycoproteins were aggregated the fastest by bacteria expressing Salmonella fimbriae, followed by bacteria expressing E. coli fimbriae, with bacteria expressing Klebsiella fimbriae exhibiting the slowest and the lowest level of aggregation (Fig. 2, A and B). In addition, horseradish peroxidase (HRP) and porcine thyroglobulin were readily aggregated by bacteria expressing E. coli fimbriae, but not by either of the other type 1 fimbriated bacteria ( Fig. 2C and TABLE TWO), whereas RNase B was not aggregated by any of the strains (Fig. 2D). Thus, E. coli, S. typhimurium, and K. pneumoniae type 1 fimbriae were highly selective in the nature of the mannoproteins that they aggregated. Moreover, there appeared to be distinct differences between fimbriae as to the level of aggregation in each case. Because none of the bacteria expressing FimH Ϫ fimbriae exhibited any aggregating ability, FimH is a critical determinant of glycoprotein aggregation by type 1 fimbriated bacteria.
The Distinct Binding Traits of Enterobacterial Fimbriae Cannot be Readily Ascribed to Variation in the Primary Structures of Their Respective FimH Adhesins-The DNA sequences of the fimH genes from plasmid pSH2 containing the E. coli strain J96 fim cluster and plasmid pBP7 containing the K. pneumoniae strain IA551 fim cluster (GenBank TM / EBI accession numbers AY914173 and AY914172, respectively) were determined, and the amino acid sequences were deduced (supplemental Fig. 1). Earlier work had determined the fimH sequences of E. coli K12 strain PC31 (5) and K. pneumoniae strain IA565 (15); however, our use of the fim operon from two different strains (E. coli J96 and K. pneumoniae IA551) required us to determine their respective fimH sequences. The primary structure of S. typhimurium strain 6704 FimH (FimHS) encoded by pISF101 has already been reported (GenBank TM accession number L19338). A comparison of the predicted amino acid sequences of E. coli strain J96 FimH (FimHE) and K. pneumoniae strain IA551 FimH (FimHK) revealed a high level of identity (98.6%). Both   located in the C-terminal pilin domain (amino acids 160 -277), which mediates the interactions of FimH with the fimbrial shaft (23). Mature Salmonella FimH is composed of 313 amino acids and is therefore considerably larger than either FimHE or FimHK. Moreover, even after multiple sequence alignments of mature FimHS and FimHE or FimHK, very limited homology can be seen (supplemental Fig. 1). Therefore, few clues to the molecular basis for the diversity in binding among enteric type 1 fimbriae are apparent from determining the primary sequences of their respective FimH lectins. Enterobacterial FimH Subunits in the Absence of Their Respective Fimbrial Shafts Exhibit Similar Binding Traits-We investigated whether the differential binding among enteric type 1 fimbriae could be attributed to the intrinsic properties of their respective FimH lectins. We subcloned each of the fimH genes and expressed them as fusion proteins with MalE. MalE stabilizes FimH and also allows the fusion protein to be readily isolated from the recombinant clones (10). The periplasmic fraction from each of the isopropyl-␤-D-thiogalactopyranoside-induced clones was isolated, and the MalE/FimH fusion proteins were affinity-purified on an amylose column and analyzed by SDS-PAGE. As shown in Each of the fusion proteins was subjected to SDS-PAGE and then electrophoretically transferred to nitrocellulose strips. The immobilized proteins were then exposed to biotinylated BECs or GECs. In contrast to the selective binding traits that are characteristic of their respective fimbriae, all three FimH lectins bound readily to both BECs (Fig. 3B) and GECs (Fig. 3C). No epithelial cell binding was seen with MalE alone, indicating that MalE was functionally inert. All of the epithelial cell binding reactions exhibited by the MalE/FimH proteins were inhibitable by methyl ␣-D-mannopyranoside (data not shown), confirming the mannose specificity of the binding reaction.
We next examined the ability of the three MalE/FimH fusion proteins to bind the battery of glycoproteins listed in TABLE TWO. MalE/FimH fusion proteins were electrophoretically transferred to nitrocellulose and incubated with each glycoprotein. To detect any bound mannoproteins, we probed the nitrocellulose blots with alkaline phosphataselabeled concanavalin A (a well known probe for the detection of mannosylated compounds), followed by an appropriate substrate. Interestingly, not only did the three fusion proteins exhibit comparable binding activity, but the binding range of each of the FimH proteins included most of the mannose-containing glycoproteins tested. The binding of each of the fusion proteins to HRP and RNase B is shown in Fig. 3 (D and  E), and this is typical of other mannoproteins tested. Moreover, each of these binding reactions was inhibitable by 100 mM methyl ␣-D-mannopyranoside (data not shown). Thus, all three isolated FimH subunits bound comparably to a wide range of mannosylated substrates.

Bacteria Expressing Hybrid Fimbriae Composed of FimH on a Heterologous Fimbrial Shaft
Have an Altered Binding Specificity-In previous studies, Madison et al. (27) found that the fimbrial shaft can modify the fine sugar specificity of FimH. In these studies, the FimH subunits of E. coli and K. pneumoniae were swapped on the fimbrial shafts of the two species. To see whether such a presentation also affects tissue tropism, we expressed FimHE on the K. pneumoniae fimbrial shaft (FimHE/ShaftK) and vice versa (FimHK/ShaftE). Because the hybrid fimbriae are derived from genes encoded on two separate replicons (diagrammed in supplemental Fig. 2), the stoichiometry of the fimbrial proteins produced may differ from that of proteins encoded by fimbrial genes present on a single replicon, such as pSH2 and pBP7. Therefore, as a control, we created reconstituted wild-type fimbriae with the plasmids used to create hybrid fimbriae by complementing fimH-deficient fimbrial gene clusters with homologous fimH. The strain designations and the type of fimbriae produced are shown in Fig. 4A. Attempts to con- struct hybrid fimbriae expressing FimHS on either the E. coli or K. pneumoniae type 1 fimbrial shaft did not lead to the production of functional fimbriae based on their inability to agglutinate yeast (S. cerevisiae) (data not shown). Whole cell lysates of each strain were examined for both FimA and FimH production by Western blotting (Fig. 4B). Although some differences in fimbrial subunit production were observed, such as overproduction of FimHK by both the hybrid (FimHK/ShaftE) and reconstituted wild-type (FimHK/ShaftK) fimbria-producing strains, these differences did not affect the yeast agglutination profiles of the reconstituted wild-type fimbriae (FimHE/ShaftE and FimHK/ShaftK) compared with those of the wild-type fimbriae (FimHE-ShaftE and FimHK-ShaftK) (Fig. 5A). Thus, the type 1 fimbriae encoded by genes present on two replicons were functional, and the effects of any differences in the stoichiometry of fimbrial subunit production were minimal. The yeast agglutination profiles of the two strains producing hybrid type 1 fimbriae (FimHK/ShaftE and FimHE/ShaftK) were found to be very similar to the profiles of the native fimbriae from which their shaft was derived (e.g. FimHK/ShaftE similar to FimHE-ShaftE and FimHE/ ShaftE; FimHE/ShaftK similar to FimHK-ShaftK and FimHK/ShaftK) (Fig. 5A). The importance of the fimbrial shaft in determining the binding specificity of FimH was further demonstrated by measuring aggregation of HRP and porcine thyroglobulin (Fig. 5, B and C), two glycoproteins that were aggregated by wild-type E. coli type 1 fimbriae, but not by wild-type K. pneumoniae fimbriae (Fig. 2 and TABLE TWO). The ability of the hybrid fimbriae to recognize and aggregate HRP and porcine thyroglobulin was dependent on the background of the fimbrial shaft rather than the background of FimH such that FimHK/ShaftEexpressing bacteria aggregated both HRP and porcine thyroglobulin, similar to wild-type E. coli fimbriae, whereas FimHE/ShaftK fimbriae resembled wild-type K. pneumoniae fimbriae by not aggregating either of the glycoproteins (Fig. 5, B and C). Thus, FimH expressed on a heterologous shaft acquired the glycoprotein specificity of its heterologous FimH (e.g. FimHK on the E. coli type 1 fimbrial shaft acts like FimHE on the E. coli type 1 fimbrial shaft). Neither of the hybrid fimbriae aggregated RNase B, indicating no dramatic increase in the breadth of mannoproteins recognized by FimH expressed on a heterologous type 1 fimbrial shaft (Fig. 5D). To rule out the possible effects of heterologous FimF and FimG, we performed the previous aggregation experiments using strains expressing heterologous FimH in the absence of heterologous FimF and FimG and found the same results (supplemental Fig. 3).
We next tested the ability of wild-type E. coli, wild-type K. pneumoniae, and hybrid type 1 fimbriae to mediate the previously described tropism for the bladder epithelium (Fig. 1A). Both wild-type fimbriae mediated adherence to BECs, but E. coli type 1 fimbriae (FimHE-ShaftE and FimHE/ShaftE) demonstrated a much greater capacity to mediate both adherence and the previously described bacterial invasion of BECs compared with wild-type K. pneumoniae type 1 fimbriae (FimHK-ShaftK and FimHK/ShaftK) (Fig. 6, A and B) (31,36). FimHK/ShaftE hybrid fimbriae mediated adherence to and invasion of BECs at levels equal to those of wild-type E. coli type 1 fimbriae, whereas FimHE/ ShaftK hybrid fimbriae mediated BEC interactions that were much more similar to those of wild-type K. pneumoniae type 1 fimbriae, thus further demonstrating the potential importance of the role the type 1 fimbrial shaft plays in determining FimH binding specificity and tissue tropism.

DISCUSSION
Because type 1 fimbriae are expressed by virtually all enterobacterial species, including nonpathogenic strains, there is controversy regarding their contribution to the infectious process. This controversy was fueled, at least in part, by the fact that, for many years, the only information regarding the binding properties of type 1 fimbriae was their uniform binding specificity for D-mannose (37,38). However, once it was realized that type 1 fimbriae from various strains exhibit differences in their fine sugar specificity, studies were initiated examining the relevance of such differences to bacterial virulence. For example, uropathogenic E. coli strains that avidly bind uroepithelial cells and efficiently colonize the mouse bladder express type 1 fimbriae that possess a fine sugar specificity for monomannose and trimannose residues (26,28). In contrast, fecal E. coli isolates that bind weakly to uroepithelial cells and poorly colonize the mouse bladder expressed type 1 fimbriae with specificity for only trimannose residues (26,28). These studies revealed that the intraspecies variation in E. coli type 1 fimbrial fine binding specificity correlates with the capacity of distinct E. coli strains to colonize a particular tissue.
In this study, we have shown that interspecies variation in tissue tropism can be ascribed, in part, to differences in the fine sugar specificity driven by the association of FimH with the fimbrial shaft rather than the primary structure of FimH. This notion is based on the following findings. (i) Different enterobacterial species differentially adhere to BECs and GECs. S. typhimurium bound avidly to GECs; E. coli bound well to BECs and moderately to GECs; and K. pneumoniae bound moderately to BECs only. (ii) When K. pneumoniae FimH was expressed on the shaft of E. coli, the tissue tropism of the strain expressing the hybrid fimbriae was similar to that of the strain expressing the native E. coli fimbriae and vice versa. (iii) Once FimH was disengaged from the fimbrial shaft and expressed as a fusion protein with MalE, the differences in binding to epithelial cells were abolished irrespective of the origin of FimH. MalE/FimHE, MalE/FimHK, and MalE/FimHS also exhibited comparable binding to various soluble mannose-containing proteins as  NOVEMBER 11, 2005 • VOLUME 280 • NUMBER 45 well. Therefore, the distinct binding property exhibited by each enterobacterial fimbria is not solely an intrinsic property of their respective FimH adhesins. We determined that, although the primary structures of FimHE and FimHK are 98.6% identical, the FimHS structure exhibits very limited homology to either of the other two FimH proteins. The E. coli FimH mannose-binding pocket is created by discontinuous regions of the molecule and seems to be conformation-dependent, so the similar basic mannose specificity of both E. coli and S. typhimurium FimH could arise from similar tertiary conformations, even though their primary structures are divergent (23,39,40). Because the MalE/FimH fusion proteins mediated mannose-binding properties on the overlay assays following SDS-PAGE, these proteins must have renatured and presumably regained their functionally competent conformation. Many proteins are known to renature and regain their conformation during electrophoretic transfer from SDS-polyacrylamide gels to nitrocellulose membrane (41)(42)(43).

Determinants of Type 1 Fimbrial Binding Specificity
Presumably, the mannose-binding pocket in each FimH is sufficiently flexible to interact with many mannose configurations. The role of the fimbrial shaft in modulating the binding spectrum of FimH could be by restricting the flexibility of the mannose-binding pocket or by limiting its accessibility to particular mannose configurations. The heterogeneity in the binding traits of FimHE, FimHK, and FimHS after they are incorporated into their native fimbriae suggests that the restrictive impact of each shaft is distinct. Because FimA, which constitutes Ͼ95% of the fimbrial shaft, is structurally and antigenically heterogeneous among various enterobacteria with sizes ranging from 14 to 22 kDa (11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22), the conformational change undergone by FimH is likely to be different in each fimbrial structure.
FimH is not only located at the tips of the fimbriae, but is also intercalated at intervals along the fimbrial shaft (35,44). One explanation for our results is that the location (tip versus shaft) of FimH influences its fine sugar specificity; and therefore, if FimH is intercalated into the shaft of one fimbria but is present only at the tip of another, they would differ in their fine sugar specificity. However, for this to be true, FimH presented at the fimbrial tip as well as FimH intercalated along the shaft would both have to be able to mediate adhesion. Much previous work has indicated that only FimH present at the tips of fimbriae is functional in mediating adhesion (35,(45)(46)(47)(48). For example, type 1 fimbriae have been observed to bind to erythrocyte membranes as well as the uroepithelium by the tips of the fimbriae (46,49). In addition, non-aggregated cell-free type 1 fimbriae do not agglutinate erythrocytes, whereas aggregated type 1 fimbriae and whole bacteria both cause hemagglutination (45,46,48), strongly indicating that, with respect to adhesion, type 1 fimbriae are monovalent. The work of Ponniah et al. (47) supports this idea by showing that fragmentation of type 1 fimbriae greatly increases their mannose-binding activity by exposing FimH that was previously intercalated into the fimbrial shaft, indicating that shaft-located FimH is not functional in adhesion. In addition, the expression of FimH on a heterologous fimbrial shaft leads to the acquisition of a specificity that appears to be dependent on the background of the fimbrial shaft, not the background of FimH, indicating that heterologous FimH and homologous FimH are incorporated into the shaft in a very similar manner. For example, the specificity of FimHK expressed on the E. coli type 1 fimbrial shaft closely mimics the specificity of FimHE expressed on the E. coli type 1 fimbrial shaft, indicating that FimHK and FimHE are incorporated into the E. coli type 1 fimbrial shaft in a similar manner, viz. at the fimbrial tip and intercalated along the length of the shaft. We therefore believe that the differences in fine sugar specificity are influenced by the manner in which FimH present at the fimbrial tip interacts with the fimbrial shaft.
Although FimA is the main component of the fimbrial shaft, the fimbrial tip structure is made up not only of FimH but also the minor fimbrial subunits FimF and FimG (50,51). The potential for these sub-units to modulate FimH specificity has not been ruled out by these studies; however, we demonstrated that strains expressing heterologous FimH in the absence or presence of heterologous FimF and FimG did not differ in their ability to aggregate yeast, HRP, or porcine thyroglobulin (supplemental Fig. 3). In addition, sequencing fimF and fimG of E. coli strain J96 (GenBank TM accession number DQ090770) and K. pneumoniae strain IA551 (GenBank TM accession number DQ090769) revealed 99.4% (one dissimilarity at amino acid residue 41) and 100% homology between mature E. coli and K. pneumoniae FimF and mature E. coli and K. pneumoniae FimG, respectively (supplemental Fig. 3), further minimizing the likelihood of their impact on our studies of type 1 fimbrial specificity.
Most work to date has directly related differential binding among E. coli type 1 fimbriae to alterations in the primary sequence of FimH (52)(53)(54)(55). Sequence variations that diminish the mannose-binding ability of FimH were found to be located within or close to the sequences that make up the FimH binding pocket as defined by the FimH crystal struc- ture (23,40,54), although the location of the FimH binding pocket in intact fimbriae may vary somewhat from that determined by the crystal structure of FimH complexed with only its chaperone, FimC (23). However, sequence alterations that increase the monomannose-binding ability of FimH, which is thought to play a role in the tropism of uropathogenic E. coli for the bladder epithelium, are not located near the FimH binding pocket, but are instead located in the lower part of the FimH lectin domain defined by the tertiary structure of FimH (26,28,39,40,54). Therefore, rather than altering the FimH binding pocket directly, these sequence variations may instead alter the conformational stability of the binding pocket, such as we hypothesize would occur by altering the FimH interaction with the fimbrial shaft. In fact, replacement of the pilin domain of FimH (amino acids 160 -279), which directs incorporation of FimH into the type 1 fimbrial shaft, with the corresponding pilin segment of the E. coli type 1C fimbrial adhesin FocH significantly alters the binding phenotype of the FimH/FocH hybrid type 1 fimbriae without altering the primary sequence of the FimH lectin domain (amino acids 1-156) (23,39). Presumably, this is because of an altered conformation of the FimH binding site when incorporated into the type 1 fimbrial shaft via the FocH pilin domain (39). The importance of quaternary association in polymeric adhesins in determining binding specificity was recently revealed in studies of the plant snowdrop and garlic lectins (56). Although subunits of both lectins are structurally similar in many respects, their carbohydrate specificity varies considerably, and this is directly linked to differences in their oligomerization (56).
In summary, we have determined that enterobacterial FimH intrinsically has the ability to exhibit a broad range of mannose-specific binding interactions, but that this trait is modulated by the fimbrial shaft on which it is presented. We believe that the influence of the shaft on FimH forms part of the molecular basis for the tissue tropism exhibited by different type 1 fimbriae. This supplemental binding mechanism (e.g. modulation of the basal specificity of FimH for D-mannose by the fimbrial shaft) may have evolved as different enterobacteria adapted to colonize a particular niche. This capacity of each fimbrial shaft to modulate the binding activities of its FimH subunit could be critical to microbial pathogenesis because it enables the pathogen to selectively colonize sites in the host that are presumably supportive to its growth. Our findings reveal a novel and physiologically relevant function for these highly organized cell-surface organelles.