INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

One of the indispensable steps in several infections caused by Escherichia coli and other gram-negative bacteria is the specific adherence to host cell surface carbohydrates linked to glycoproteins or to glycolipids. This phenomenon is known to be mediated either by nonfimbrial adhesins or by hair-like heteropolymeric bacterial surface appendages called fimbriae or pili (4, 8-10, 19, 20, 28).

Accumulating data argue that the glycolipids present on eukaryotic cell surfaces play a significant role in cell-cell and cell-ligand interactions (3, 43). Several different types of fimbrial adhesins have been described to mediate attachment of bacteria to the carbohydrate sequence of glycolipids present on host cell surfaces (7, 16, 18, 26, 30, 35, 39, 48, 49).

Urinary tract infections in humans are strongly associated with E. coli producing P, type 1, S, and F1C fimbriae. P fimbriae are defined by their ability to mediate binding to the Gal1-4Gal saccharide in glycolipids of the globoseries (13, 14, 26, 47). Type 1 fimbriae bind specifically to mannose residues (9, 31). The S fimbria super family has been described to consist of S fimbriae, F1C fimbriae, and S- and F1C-related fimbriae (34, 40, 42). S fimbriae bind to receptors containing sialic acid sugar moieties. They have the capacity to agglutinate human and bovine erythrocytes (24, 33, 52). Their binding to brain glycolipids has also been demonstrated (39).

Another member of this family, namely F1C fimbria, has been described as a nonhemagglutinating adherence factor and is expressed by approximately 14% of the E. coli known to cause urinary tract infections and 7% of E. coli fecal isolates (11, 32, 37, 38, 45). A cluster of eight genes (foc) is necessary for the biogenesis of F1C fimbriae (21, 22, 40, 50, 51). The F1C fimbrial complex is composed of the major subunit protein FocA (16 kDa) and the minor subunits FocF (17 kDa), FocG (15 kDa), and FocH (30 kDa) (40). These fimbriae share high sequence homology to the major and minor subunits of the S fimbriae (11). E. coli harboring F1C fimbriae have been reported to bind to epithelial cells in the distal tubules and collecting ducts as well as to endothelial cells of human kidney and bladder (53). Their binding specificity has also been suggested to be similar to that of S fimbriae (27). However, their exact receptor specificity has not yet been identified.

In the present study, we identify glycolipid receptors for purified F1C fimbriae as well as for F1C fimbriated recombinant E. coli HB101(pPIL110-54) by using thin-layer chromatography (TLC) overlay and solid-phase binding assays.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains and growth conditions. The E. coli K-12 strain HB101 used in this study was transformed with the plasmid pPIL110-54 (pACYC184 plasmid vector containing the complete foc gene cluster cloned from the uropathogenic E. coli strain AD110) (40, 51). Recombinant bacteria were grown overnight (with shaking) in liquid broth supplemented with the appropriate antibiotic (50 µg of chloramphenicol per ml).

Reagents and chemicals. Purified glycolipids, saccharides, and enzymes used were obtained from Sigma, Deisenhofen, Germany. Recombinant -N-acetylhexosaminidase was from New England Biolabs, Beverly, Mass. Carbohydrates conjugated to bovine serum albumin (BSA) and disaccharides used as inhibitors were purchased from Dextra Laboratories (Reading, United Kingdom); Toronto Research Chemicals, North York, Canada; and Glycorex, Lund, Sweden. All chemicals used were of analytical grade.

Antisera. Polyclonal antibodies against purified F1C fimbriae were raised in chicken (Nano-Tools, Freiburg, Germany). Antiserum was adsorbed with HB101 host strain before use.

Seroagglutination. The capacity of bacteria to express F1C fimbriae was assayed by agglutinating bacteria by using chicken polyclonal anti-F1C antibody as described earlier (20). Briefly, diluted (1,000 times) antiserum (10 µl) was mixed with the same volume of liquid bacterial culture grown for 16 h (with shaking), and agglutination was observed. Nonfimbriated strain HB101 carrying plasmid vector pACYC184 was used as a negative control. Fimbriated bacteria did not show agglutination with phosphate-buffered saline (PBS).

Purification of F1C fimbriae. F1C fimbriae were harvested from recombinant strain HB101 (pPIL110-54) by using commercial blendor (Omnimixer; Waring) and were purified essentially as described by Khan and Schifferli (20). Moreover, like other fimbriae (20), F1C fimbriae were heat extractable. Fimbriated bacteria from broth cultures were pelleted by centrifugation, were suspended in 0.5 mM Tris-HCl (pH 7.4) containing 75 mM NaCl, and were treated at 60°C for 30 min in a shaking water bath. Bacteria were removed by centrifugation, and F1C fimbrial preparations were concentrated from the supernatant by using an ultrafiltration pressure cell (YM100 membranes; Amicon, Beverly, Mass.).

Biotinylation of bacteria. Overnight cultures of bacteria were pelleted at 1,940 × g for 15 min. After being washed three times with PBS bacteria were resuspended in PBS containing 0.1 mM CaCl₂ and 1 mM MgCl₂ (pH 7.6) to a final concentration of 5 × 10⁸ bacteria/ml. To this, an equal volume of PBS-CaCl₂-MgCl₂ containing 0.2 to 0.3 mg of sulfo-NHS-LC-biotin (sulfosuccinimidyl-6-[biotinamido]-hexanoate) (Pierce Chemical Company) per ml was added. Bacterial surface proteins were allowed to biotinylate for 2 h by incubating the mixture at room temperature with gentle shaking. The bacteria were collected, washed four times, resuspended to a final concentration of 2 × 10⁹ bacteria/ml in PBS (pH 7.4) containing 1% BSA, and stored at 4°C. This preparation was used for 3 days.

TLC fimbriae overlay assay. TLC was performed as described earlier (12, 17, 18) with a solvent system involving chloroform-methanol-water (60:35:8). Briefly, pure glycolipids (Sigma) were separated for analytical purpose on high-performance TLC (HPTLC) aluminium-backed silica gel 60 plates (Merck, Darmstadt, Germany). Glycolipids were stained with orcinol spray (Sigma). For the overlay assay, the plates were sequentially treated for 2 min with polyisobutylmethacrylate (0.1% in hexane), were blocked with 2% BSA in Tris-buffered saline (TBS; 10 mM Tris-HCl [pH 7.4], 154 mM NaCl) for 90 min at room temperature, were washed once with TBS, were overlaid with pure F1C fimbriae (20 µg/ml in 1% BSA-TBS) for 2 h, and were washed three times with TBS-0.05% Tween 20 and once with TBS. Fimbrial antigen was detected sequentially with anti-F1C fimbrial chicken polyclonal antibody and horseradish peroxidase-conjugated anti-chicken antibody (Sigma) by conventional Western blot technique by using the DAB peroxidase substrate tablet set (Sigma) as described earlier (20).

Neuraminidase treatment of disialogangliosides separated by HPTLC. The disialoganglioside fraction containing O-acetylated derivatives of the gangliosides (GD3, disialosylparagloboside [DSPG], disialosyllacto-N-nor-hexaosylceramide [DSnHC], and disialosyllacto-N-nor-octaosylceramide [DSnOC]) purified from human leucocytes (23) were separated on five glass-backed HPTLC plates (Merck) in parallel and were fixed with polyisobutylmethacrylate as described above. One plate was used for fimbrial overlay assay and the other one was immunostained as described earlier (23) by using CDw60 monoclonal antibody M-T21, which is specific for GD3, 9-O-acetylated GD3, and terminally disialyated gangliosides of the neolacto series such as DSPG and DSnHC. In order to remove sialosyl residues linked to gangliosides, three plates were incubated overnight at 37°C with neuraminidase (10 mU) per ml of Arthrobacter ureafaciens (Boehringer GmbH, Mannheim, Germany) in 0.1% CaCl₂, were washed three times with PBS, and were blocked with 2% BSA-PBS. After the washing of blocked plates, one plate was overlaid with the monoclonal antibody 1B2 specific for the terminal Gal1-4GlcNAc-carbohydrate structure found in glycolipids of the neolacto series and was further processed as described earlier (23). The second plate was overlaid with purified fimbriae. The third plate was used as a control where neither fimbriae nor monoclonal antibody 1B2 was overlaid. They were further processed as described above.

Periodate oxidation. Microtiter wells containing either asialo-GM₁ (GgO₄Cer) or asialo-GM₂ (GgO₃Cer) were washed twice with PBS, incubated with 10 mM sodium metaperiodate in PBS (pH 7.1) at 37°C for 90 min (31), washed four times with PBS, blocked with 2% BSA-PBS at 37°C for 2 h, and then overlaid with twofold serially diluted bacteria as described below. Bacterial binding was determined as mentioned below.

Solid-phase binding and binding inhibition assay. Polyvinyl chloride plates (Falcon; Becton Dickinson) were coated with glycolipids per well, (2 µg of glucosylceramide, 2 µg of galactosylceramide 1 and 2-[hydroxylated and nonhydroxylated fatty acids], 2 µg of sulfatide, 2 µg of lactosylceramide [LacCer], 2 µg of Gb₃Cer, 2 µg of GM1, 2 µg of asialo-GM₁ (GgO₄Cer), 1 µg of asialo-GM₂ (GgO₃Cer), 2 µg of paragloboside) in chloroform-methanol (1:9 [vol/vol]). The solvent was allowed to evaporate overnight at room temperature, and the wells were blocked with 2% BSA-PBS (pH 7.37) for 2 h at 37°C and were washed three times with PBS. For each glycolipid, a serial dilution of F1C fimbriated E. coli HB101(pPIL110-54), control strain HB101(pACYC184), or purified F1C fimbriae was first tested to determine the numbers of bacteria and the quantity of purified fimbriae required to obtain 50% binding. The number of fimbriated HB101(pPIL110-54) and nonfimbriated control strain HB101(pACYC184) was determined by measuring absorbance at a wavelength of 550 nm, with a standardized chart correlating absorbance with viable counts. After removal of the blocking solution and washing, 100 µl of serially diluted bacterial suspensions (2 × 10⁸ bacteria in the first well) or purified fimbrial solution (10 µg of fimbriae in the first well) in PBS-0.5% BSA was added to each well, and plates were incubated for 2 h at 37°C. After wells were washed as previously described, rabbit anti-E. coli polyclonal antibody (Biodesign International, Kennebunk, Maine) in PBS-0.5% BSA (1:2,000) (or for wells with fimbriae, anti-F1C chicken antibody [polyclonal; 1:400]) was added for 1 h at 37°C (100 µl/well). Following another washing step, peroxidase-conjugated goat anti-rabbit immunoglobulin G (Dako, Hamburg, Germany) or anti-chicken immunoglobulin G whole molecule (Sigma) in PBS-1% BSA (1:2,000) was added for 1 h at 37°C (100 µl/well). Following a final wash, the bound enzyme was detected by the addition of 100 µl of substrate (Pierce ImmunoPure TMB Substrate Kit) per well for 5 to 30 min. The reaction was stopped by adding 100 µl of 2 M H₂SO₄ to each well. Absorbance was measured at a wavelength of 450 nm with an enzyme-linked immunosorbent assay (ELISA) reader. Similar experiments were also performed by using biotinylated bacteria, and binding was detected with horseradish peroxidase-conjugated streptavidine (Pierce Chemical Company) and its substrate as above. The control wells were treated in the same manner except that blank control wells had no bacteria or fimbriae.

Enzymatic treatment of asialo-GM₁ (GgO₄Cer). Purified asialo-GM₁ (GgO₄Cer) (Sigma) was immobilized on ELISA plates and was sequentially treated with bovine -galactosidase (-Gal) (Sigma) and then with recombinant -N-acetylhexosaminidase (New England Biolabs). Asialo-GM₁ (GgO₄Cer) was suspended in methanol (10 µg/ml), and 100 µl of the suspension was added to wells of microtiter plates. After evaporation of methanol, immobilized glycolipids were incubated with 0.4 U of -Gal in Tris buffer [10 mM Tris, 1.7 M (NH₄)₂SO₄, 10 mM MgCl₂, pH 7.3] for 2 h at 37°C. The wells were washed with PBS, and the glycolipids were further treated with 10 U of -N-acetylhexosaminidase in citrate buffer (50 mM sodium citrate, pH 4.5) per well for 6 h at 37°C in a shaking water bath. After performing washing and blocking steps, bacterial or fimbrial binding was determined as described above.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Analysis of F1C fimbrial binding to purified glycolipids. The efficiency of F1C fimbrial binding to purified glycolipids was studied by using the HPTLC overlay assay (Table 1). Purified fimbriae bound to glucosylceramide (GlcCer), galactosylceramide 2 (nonhydroxylated fatty acids), lactosylceramide (LacCer), paragloboside (nLc₄Cer), lactotriaosylceramide (Lc₃Cer), asialo-GM₁ (GgO₄Cer), asialo-GM₂ (GgO₃Cer), and globotriaosylceramide (Gb₃Cer) (Table 1) but did not bind to sulfatide (SFT), globotetraosylceramide (Gb₄Cer), the Forssman antigen (Table 1), GM₁ (Fig. 1), GM₂, GM₃, GD₂, GD₃ (Table 1), DSPG, DSnHC, DSnOC (see Fig. 3), or ceramide (Cer) (Table 1). A weak binding was observed with GlcCer and GalCer2 (Fig. 1). F1C fimbriae bound better to LacCer, Gb₃Cer, Lc₃Cer (data not shown), and nLc₄Cer (Fig. 1). Binding to GalCer1 containing hydroxylated ceramide portion was even weaker than GlcCer or GalCer2 (Table 1). Optimal binding of purified F1C fimbriae or F1C fimbriated recombinant bacteria to asialo-GM₂ (GgO₃Cer) and asialo-GM₁ (GgO₄Cer) on HPTLC plates was observed (Fig. 1 and 2).

View larger version (52K): [in this window] [in a new window]   FIG. 1.   HPTLC-separated glycolipids and overlay analysis of the binding of purified F1C fimbriae. Chromatograms of glycolipids separated on HPTLC silica gel 60 and stained with orcinol (A and C), probed with purified F1C fimbriae, and immunostained with anti-fimbrial antibodies and horseradish peroxidase-conjugated anti-chicken antibodies (B and D) are shown. (A) Lane 1, (from top to bottom) glucosylceramide (4 µg, GlcCer), sulfatide (4 µg, SFT), asialo-GM2 (GgO3Cer) (2 µg), and sialylated GM1 (4 µg, GM1); lane 2, (from top to bottom), galactosylceramide 2 (4 µg, GalCer2), lactosylceramide (3 µg, LacCer), paragloboside (1 µg, nLc4Cer), and asialo-GM1 (GgO4Cer) (2 µg). (B) Lane 1, fimbrial binding to GlcCer (upper double band) and to asialo-GM2 (GgO3Cer) (lower double band); lane 2, fimbrial binding from top to bottom, GalCer2, LacCer, nLc4Cer, and asialo-GM1 (GgO4Cer). (C) Lane contains (from top to bottom) globotriaosylceramide (4 µg, Gb3Cer) and globotetraosylceramide (4 µg, Gb4Cer). (D) Lane contains fimbriae bound to globotriaosylceramide (Gb3Cer).

View larger version (17K): [in this window] [in a new window]   FIG. 2.   Dose-dependent binding of biotinylated F1C fimbriated recombinant E. coli (3 × 107/well) to purified glycolipids immobilized in microtiter wells. The twofold serially diluted glycolipids in 50 µl of chloroform-methanol (1:9), asialo-GM1 (GgO4Cer), asialo-GM2 (GgO3Cer), LacCer, and GalCer2 were added to microtiter wells. Bound bacteria were detected with horseradish peroxidase-conjugated streptavidine as described in Materials and Methods. Data represent the average of triplicate determinations.

Glycolipid binding specificity of F1C fimbriae and fimbriated bacteria. The ability of purified fimbriae and fimbriated bacteria with selected glycolipids was studied in microtiter plate assay. Both fimbriae and bacteria bound to glucosylceramide (GlcCer), galactosylceramide (GalCer1 and -2), lactosylceramide (LacCer), paragloboside (nLc₄Cer), globotriaosylceramide (Gb₃Cer), asialo-GM₁ (GgO₃Cer), and asialo-GM₂ (GgO₄Cer) in a dose-dependent manner (data not shown). Neither purified fimbriae nor fimbriated bacteria showed any binding to glycolipids containing sialic acid residues (Table 1). Even sialic acid residue linked to internal galactose of GM₁ was able to block the binding, probably due to steric hinderance (Fig. 1). Pretreatment of the glycolipids with sodium metaperiodate almost completely eliminated the binding (data not shown).

Treatment of glycolipids with neuraminidase or glycosidases. Purified fimbriae did not bind to disialoganglioside fraction containing O-acetylated derivatives of the gangliosides (GD₃, disialosylparagloboside, disialosyllacto-N-nor-hexaosylceramide, and disialosyllacto-N-nor-octaosylceramide) purified from human leucocytes and separated on HPTLC plates (Fig. 3B). After the treatment of the separated disialogangliosides with neuraminidase, the binding of fimbriae could be observed (Fig. 3D). The removal of terminal sialic acid residues was confirmed by using specific monoclonal antibody 1B2 to Gal1-4GlcNAc carbohydrate sequences (Fig. 3C).

View larger version (62K): [in this window] [in a new window]   FIG. 3.   Binding of purified F1C fimbriae prior to and after A. ureafaciens neuraminidase treatment to the disialoganglioside fraction of unseparated human leucocytes. An aliquot of the disialoganglioside fraction was manually loaded and separated simultaneously on five small HPTLC plates (glass backed) and was fixed with polyisobutylmethacrylate as described in Materials and Methods. The glycolipids were immunostained with the CDW60 monoclonal antibody M-T21 specific for O-acetylated disialogangliosides (A) showing doublets of the 9-O-acetylated forms of GD3, DSPG, DSnHC, and DSnOC. The same glycolipids as mentioned in the legend of Fig. 1 were probed with purified F1C fimbriae (B). After neuraminidase treatment, the resulting asialogangliosides were immunostained with the monoclonal antibody IB2 specific for glycolipids containing an external Gal1-4GlcNAc structure (C) and were probed with purified F1C fimbriae in a way similar to that shown in panel B (D). The strongest bands visible in panel D are supposed to originate from 7-O-acetylated GD3 and GD3, respectively, which are known to migrate between 9-O-acetyl GD3 and 9-O-acetyl DSPG. These two gangliosides were not visualized by M-T21 (A), but their presence was deduced by IB2 stain (C). A control plate was neuraminidase treated and probed with PBS containing no fimbriae but 1% BSA (E).

View larger version (15K): [in this window] [in a new window]   FIG. 4.   Effect of enzymatic sequential treatment of asialo-GM1 (GgO4Cer). Asialo-GM1 (GgO4Cer) (1 µg/well) was coated in microtiter wells and treated with bovine -Gal. The ability of purified F1C fimbriae (1.5 µg/well) and biotinylated F1C fimbriated recombinant E. coli (3 × 107/well) to bind to asialo-GM1 (GgO4Cer) and to -Gal-treated asialo-GM1 (GgO4Cer), i.e., to asialo-GM2 (GgO3Cer) was determined by ELISA. The binding of fimbriae or bacteria was determined as described in the legends of Fig. 1 and 2. Results were compared with the binding of the purified F1C fimbriae and F1C fimbriated bacteria to asialo-GM1 (GgO4Cer). Significant increase (two- to threefold) in binding due to the generation of asialo-GM2 (GgO3Cer) with -Gal was noted. Data represent the average of quadruplicate determinations.

Carbohydrate inhibition studies. In order to obtain information about the carbohydrate receptor structure for F1C adhesin, the ability of saccharides corresponding to -Gal, lactose, and GalNAc1-4Gal recognition sites to inhibit the F1C ligand binding to asialo-GM₁ (GgO₄Cer) and asialo-GM₂ (GgO₃Cer) was studied (Table 2). Attachment of HB101(pPIL110-54) to either glycolipid was inhibited most effectively (>60%) with 1 mg of neoglycoprotein (GalNAc1-4Gal-spacer-BSA) per ml, representing 0.26 mM of GalNAc1-4Gal disaccharide. The other mono- and disaccharides used as inhibitors reduced the bacterial adhesion to about 50% at 10 mM concentration (Table 2). At 1 mM, none of these sugars inhibited adhesion to a significant degree. In contrast, the GalNAc1-4Gal containing neoglycoprotein inhibited the adhesion even at a concentration of only 0.026 mM to a similar extent as the other sugars at more than 380-fold higher concentration (Table 2). These results confirm that disaccharide GalNAc1-4Gal specifically binds to F1C fimbriae. The fact that some saccharides were not efficient inhibitors reflects a multivalency effect in the binding of F1C fimbriae to asialo-GM₂ (GgO₃Cer) or asialo-GM₁ (GgO₄Cer).

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Binding of bacteria to glycolipids either separated by TLC (16, 49) or immobilized on microtiter plates appears to be a convenient way of detecting and characterizing carbohydrate receptor(s). In this study, we identified for the first time ligand-receptor interactions between the F1C fimbrial adhesin complex of uropathogenic E. coli and reference glycolipids of different carbohydrate compositions (Table 1). This was accomplished by performing binding assays with purified F1C fimbriae or F1C fimbriated recombinant E. coli strain HB101(pPIL110-54) and a panel of naturally occurring purified glycolipids resolved by TLC or immobilized in microtiter wells (Table 1 and Fig. 1, 2, and 3).

The binding affinity expressed as the amount of glycolipid required to achieve 50% maximal binding in the microtiter plates was the same for both F1C fimbriae and fimbriated E. coli being asialo-GM₂ (GgO₃Cer) (0.18 µg) > asialo-GM₁ (GgO₄Cer) (0.35 µg) > LacCer (1.50 µg) > GalCer2 (3.40 µg) (Fig. 2). The twofold difference in binding affinity of the F1C fimbriae to LacCer compared to GalCer2 suggested that a disaccharide sequence is needed for more efficient binding of F1C fimbriae. The binding of F1C adhesin increased by a factor of 8 after the addition of a GalNAc1-4 residue to LacCer as in asialo-GM₂ (GgO₃Cer) (Fig. 2). However, the addition of Gal1-3 residue to GalNAc1-4 in asialo-GM₂ (GgO₃Cer) approximately doubles the quantity of glycolipid (asialo-GM₁ [GgO₄Cer]) to reach 50% binding of F1C fimbriae (Fig. 2). This must be due to the presence of the -linked galactose residue at the nonreducing end of asialo-GM₁ (GgO₄Cer), which may partially inhibit the accessibility of the rest of the glycolipid containing the "real" receptor structure. Alternatively, F1C fimbriae might also bind to some extent Gal residues terminally linked to asialo-GM₂ (GgO₃Cer) which would interfere with binding to the internal sequence of asialo-GM₁ (GgO₄Cer) (GalNAc1-4Gal).

This notion is supported by the finding that preincubation of F1C fimbriated recombinant E. coli strain HB101(pPIL110-54) with 10 mM free disaccharide (Gal1-3GalNAc) corresponding to the nonreducing end of asialo-GM₁ (GgO₄Cer) or with the same concentration of free D-galactose reduced the binding to asialo-GM₁ (GgO₄Cer) by approximately 50% (Table 2). Moreover, maximum inhibition was only achieved with a very low amount (1 mg/ml and 0.1 mg/ml of GalNAc1-4Gal containing neoglycoprotein equivalent to 0.26 mM and 0.026 mM of disaccharide GalNAc1-4Gal (Table 2). The inhibitory effect of neoglycoprotein containing disaccharide GalNAc1-4Gal was 38- to 380-fold more potent than that of the mono- and disaccharides tested.

This indicates that terminally linked -Gal alone, as in asialo-GM₁ (GgO₄Cer) and internally positioned -Gal alone or terminally linked -N-acetylgalactosamine alone, as in asialo-GM₂ (GgO₃Cer) are not sufficient for effective binding of purified F1C fimbriae or F1C fimbriated bacteria. This is supported by the fact that bacteria did not bind to galabiosylceramide, which contains a single internally linked Gal residue, or to globotetraosylceramide (Gb₄Cer), which contains a terminal GalNAc residue (Table 1, Fig. 1). They also did not bind to Forssman glycosphingolipid, which contains the GalNAc1-3GalNAc sequence (Table 1). Hence, it appears that for the high-affinity binding of F1C fimbriae, the disaccharide sequence GalNAc1-4Gal is required, and the minimum carbohydrate moiety of glycolipids needed for F1C fimbrial binding is the 1-linked galactose or glucose, provided they are presented in the correct conformation and configuration.

However, as has been noted for other adhesin systems, the presence of the minimum receptor structure in glycolipids does not necessarily correlate with binding (16). For example, despite containing -Gal in their oligosaccharide sequence, GM₁, GM₂, GM₃, GD₂, GD₃, DSPG, DSnHC, and DSnOC do not show any affinity towards F1C fimbrial adhesin (Table 1 and Fig. 3B). This might be due to steric hinderance from sialyl residues linked to those glycolipids, preventing the access of fimbriae to the binding epitope on the glycolipids. Furthermore, purified F1C fimbriae or fimbriated bacteria did not bind to GM₁ but did bind very efficiently to asialo-GM₁ (GgO₄Cer) or to asialylated GD₃, asialylated DSPG, asialylated DSnHC, and asialylated DSnOC (Fig. 1B and Fig. 3C).

The findings that sialyl residues abolish binding corroborate a previously reported study which showed the inability of several pulmonary pathogenic bacteria to bind to gangliosides (25), although they bind to the receptor structure GalNAc1-4Gal of asialo-GM₁ (GgO₄Cer) and asialo-GM₂ (GgO₃Cer). In addition, the inability of bacteria to recognize glycolipids such as globotetraosylceramide (Gb₄Cer) and Forssman antigen, which do not carry sialic acid residues, could be attributed to the conformational restrictions on the presentation of specific carbohydrate receptors due to the saccharide sequence and the configuration of the oligosaccharide moieties of those glycolipids.

In addition, the findings of glycosidase treatment showing that the removal of terminal Gal1-3 residue from asialo-GM₁ (GgO₄Cer) increased the binding by approximately two- to threefold demonstrated again the importance of the internal GalNAc1-4Gal disaccharide sequence of asialo-GM₁ (GgO₄Cer) or the terminal GalNAc1-4Gal disaccharide sequence of asialo-GM₂ (GgO₃Cer) for high-affinity binding of F1C fimbriae. The specificity for internal sequences is similar to that reported for the P fimbrial adhesin, which recognizes terminal or internal Gal1-4Gal sequences in glycosphingolipids (5) but is different from K88 fimbriae of enterotoxigenic E. coli and pH6 antigen of Yersinia pestis (35, 36).

These observations support a role for the F1C fimbrial adhesin as a mediator of bacterial adherence. They also suggest that GalCer- or GlcCer-sensitive adherence may be the mechanism of low-affinity transient binding of bacteria. For low-affinity binding, these data further suggest the requirement of the terminal Gal at the disaccharide level in the case of LacCer binding. Part of the internal Glc apparently is also involved in the binding epitope since this residue is described to be quite exposed in the conformation responsible for a correct presentation of the epitope (1). Furthermore, the inhibitory effect of neoglycoprotein containing disaccharide GalNAc1-4Gal together with the binding affinity and specificity for glycolipids (Table 1) suggests the involvement of disaccharide GalNAc1-4Gal sequence as a high-affinity receptor for F1C fimbriae.

The poor binding of the F1C fimbrial adhesin to GalCer1 containing hydroxylated fatty acids and comparatively better binding to GalCer2 which contains nonhydroxylated fatty acids may contribute to the nonoptimal presentation of the relevant epitope. In support of this, it was reported that the Gal headgroup of GalCer with hydroxylated fatty acid is in an L-like conformation in relation to ceramide. In contrast, the Glc headgroup of GlcCer with nonhydroxylated fatty acid projects straight up (29, 46). However, this could not be resolved using TLC overlay assay or solid-phase binding assay. Moreover, an increased affinity of ligands for glycoceramides with higher levels of hydroxylation has been described in other systems (1, 2, 16, 18, 35, 36, 39, 41, 49). This dependence on the ceramide structure for bacterial binding has been reported to be limited at the two sugar level, which disappeared upon elongation of the saccharide chain (48).

Several other fimbrial adhesins of bacteria have also been shown to be able to bind to glycolipids (18, 35, 39, 44, 47). For instance, the Pseudomonas aeruginosa fimbrial receptor was determined to be the carbohydrate sequence GalNAc1-4Gal of asialo-GM₁ (GgO₄Cer) or asialo-GM₂ (GgO₃Cer) (6, 44). Moreover, a protein distinct from fimbriae present on the surface of Neisseria gonorrhoeae has also been shown to mediate the binding of gonococci to LacCer, Gb₃Cer, asialo-GM₂ (GgO₃Cer) and asialo-GM₁ (GgO₄Cer) (48).

The data presented in this paper demonstrate the high-affinity binding of F1C fimbriae to the GalNAc1-4Gal sequence of glycolipids, i.e., asialo-GM₁ (GgO₄Cer) and asialo-GM₂ (GgO₃Cer). An additional binding to carbohydrate structures GlcNAc1-3Gal, Gal1-4Glc, Gal, and Glc of glycolipids may indicate functional low-affinity receptor sites. Studies are in progress to demonstrate the relevance of this binding for uropathogenecity of F1C fimbriated E. coli.