Evaluation of Receptor Binding Specificity of Escherichia coli K88 (F4) Fimbrial Adhesin Variants Using Porcine Serum Transferrin and Glycosphingolipids as Model Receptors

Diarrheal disease caused by enterotoxigenic Escherichia coliexpressing the K88 (F4) fimbrial adhesin (K88 ETEC) is a significantsource of mortality and morbidity among newborn and weaned piglets.K88 fimbrial adhesins are filamentous surface appendages whoselectin (carbohydrate-binding) activity allows K88 ETEC to attachto specific glycoconjugates (receptors) on porcine intestinalepithelial cells. There are three variants of K88 adhesin (K88ab,K88ac, and K88ad), which possess different, yet related, carbohydrate-bindingspecificities. We used porcine serum transferrin (pSTf) andpurified glycosphingolipids (GSL) to begin to define the minimalrecognition sequence for K88 adhesin variants. We found thatK88ab adhesin binds with high affinity to pSTf (dissociationconstant, 75 µM), while neither K88ac nor K88ad adhesinrecognizes pSTf. Degradation of the N-glycan on pSTf by extensivemetaperiodate treatment abolished its interaction with the K88abadhesin, indicating that the K88ab adhesin binds to the singleN-glycan found on pSTf. Using exoglycosidase digestion of thepSTf glycan, we demonstrated that K88ab adhesin recognizes N-acetylglucosamine(GlcNAc) residues in the core of the N-glycan on pSTf. All threeK88 variants were found to bind preferentially to GSL containinga ß-linked N-acetylhexosamine (HexNAc), either GlcNAcor N-acetylgalactosamine, in the terminal position or, alternatively,in the penultimate position with galactose in the terminal position.Considering the results from pSTf and GSL binding studies together,we propose that the minimal recognition sequence for the K88adhesin variants contains a ß-linked HexNAc. In addition,the presence of a terminal galactose ß-linked to thisHexNAc residue enhances K88 adhesin binding.

INTRODUCTION

Top

Introduction

Three serologically distinguishable variants of K88 adhesin(K88ab, K88ac, and K88ad) have been identified (21, 40). Eachvariant consists of a conserved antigenic region shared by allthree variants, designated a, and variant-specific antigenicregions, designated b, c, and d for K88ab, K88ac, and K88ad,respectively (13, 29). The antigenic differences among the threeK88 variants can be ascribed exclusively to a small number ofnucleotide changes in the major fimbrial subunit gene, resultingin the amino acid substitutions that distinguish the three K88variants. Several researchers have investigated the molecularinteraction of K88 fimbrial adhesin variants with erythrocytes,intestinal mucus, and intestinal epithelial cells (4, 11, 12,17, 19, 37, 39, 51, 58). It has been clearly demonstrated thatall three K88 adhesin variants are lectins and that they recognizecarbohydrate structures expressed on host cell glycoconjugates.Each K88 variant has a different, but related, carbohydrate-bindingspecificity. The differences in carbohydrate specificity amongthe three K88 adhesin variants are reflected in the differenthemagglutination patterns of the K88 variants, determined byusing erythrocytes from various species (6), and in the presenceof multiple phenotypes of pigs whose intestinal epithelial cellscontain receptors that bind different combinations of K88 adhesinvariants (3, 5, 44).

Although it is clear that each K88 adhesin variant has a slightlydifferent receptor binding specificity, the carbohydrate structuresrecognized by each variant have not been clearly delineated.Data from monosaccharide blocking studies indicate that N-acetylglucosamine(GlcNAc) and N-acetylgalactosamine (GalNAc) may be part of thereceptor recognition site (2). Results from glycoprotein-blockingstudies indicate that terminal GlcNAc, GalNAc, and galactose(Gal) may play a role in the interaction of the K88 adhesinwith brush border receptors (2, 17). Also, Gal has been reportedto be an important residue in the recognition of putative intestinalmucus receptors and glycosphingolipids (GSL) by the K88ab adhesin(8, 42). Recently, we determined that ß-linked Galis an essential component in the recognition of intestinal mucin-typesialoglycoproteins (IMTGP) by K88ac adhesin (18) and that theK88ad adhesin recognizes the terminal Galß(1-4)GlcNAcon neolactotetraosylceramide (nLc₄; Galß1-4GlcNAcß1-3Galß1-4Glcß1-1Cer)(20). Despite all of this work, no minimal core recognitionsequence for the K88 adhesin variants has been identified. Theobjective of the present study was to more fully characterizethe carbohydrate specificities of the K88 adhesin variants inorder to define the core structure recognized by the K88 adhesinvariants. Knowledge of the receptor specificities of the K88fimbrial adhesins is essential for understanding the molecularmechanism of K88 ETEC adhesion and understanding the biochemicalbasis for genetic resistance of some piglets to K88 ETEC infections.

In the present paper, we used porcine serum transferrin (pSTf)and GSL standards as model receptors for the K88 fimbrial adhesinvariants. pSTf was demonstrated to be a specific model receptorfor the K88ab adhesin variant. Using exoglycosidase digestionstudies, we demonstrated that the K88ab adhesin recognizes GlcNAcresidues in the core of the N-glycan on pSTf. In addition, wecompared the carbohydrate specificities of the three variantsusing a group of known GSL as model receptors. All three K88variants were found to bind preferentially to GSL containinga ß-linked N-acetylhexosamine (HexNAc), either GlcNAcor GalNAc, in the terminal position or, alternatively, in thepenultimate position with Gal in the terminal position. Consideringthe results from pSTf and GSL binding studies together, we proposethat the minimal recognition sequence for the K88 adhesin variantscontains a ß-linked HexNAc. In addition, the presenceof a terminal Gal ß-linked to this HexNAc residueenhances K88 adhesin binding.

MATERIALS AND METHODS

Top

Materials and Methods

Materials. Organic solvents, GSL standards (lactocerebroside [Lc₂], globotriaosylceramide[Gb₃], globoside [Gb₄], globopentaosylceramide [Gb₅], asialogangliosides[GA₂ and GA₁], monosialogangliosides [GM₃, GM₂, and GM₁], disialogangliosides[GD_1a and GD_1b]), and orcinol were obtained from Sigma ChemicalCo. (St. Louis, Mo.). Precoated high-performance thin-layerchromatography (HPTLC) plates (silica gel 60) with aluminumbacking were obtained from E. Merck AG (Darmstadt, Federal Republicof Germany). Biotinylated plant lectins (concanavalin A [ConA]lectin, Griffonia simplicifolia lectin II, Ulex europaeus agglutinin[UEA], Ricinus communis agglutinin [RCA₁₂₀], Maackia amurensislectin I [MAL I], Erythrina cristagalli lectin [ECL], Solanumtuberosum lectin [STL], Lycopersicon esculentum lectin [LEL],and elderberry bark lectin [EBL]) were obtained from VectorLaboratories, Inc. (Burlingame, Calif.). Biotinylated K88 adhesinswere prepared as described by Erickson et al. (11).

Purified GSL were prepared as follows. nLc₄ and Lc₃ were obtainedsequentially by acid-catalyzed desialosylation and ß-galactosidasetreatment of bovine erythrocyte IV³NeuAc(Gc)nLc₄ as describedby Levery et al. (33). Lc₄ was obtained by partial conversionof Lc₃ with a recombinant human ß-1,3-galactosyltransferaseas described by Amado et al. (1). Le^x was obtained from humanadenocarcinoma (22, 23, 60). V³FucnLc₆ was obtained from humangranulocytes (16, 30, 54). VI²FucnLc₆ was obtained from humanO erythrocytes (32). Le^a was isolated from a mixture of Folchlower-phase GSL extracted from pooled liver and colonic adenocarcinoma(22) and was kindly provided by Mark R. Stroud (Department ofCell Surface Biochemistry, Northwest Hospital, and Divisionof Allergy and Infectious Diseases, Department of Medicine,University of Washington, Seattle, Wash.).

pSTf purification. pSTf was prepared by the procedure described previously by Grangeand Mouricout (19). The resulting preparation was further purifiedby loading the sample onto a ConA-Sepharose column (1.2 by 30cm) at room temperature (RT); the column was equilibrated withbuffer A (0.5 M NaCl, 1 mM CaCl₂, 1 mM MgCl₂, 0.1 M Tris [pH7.0]) at a flow rate of 9 ml/h. Components of the sample notrecognized by ConA were removed by washing the column with fivecolumn volumes of buffer A. Elution of the glycoproteins boundto the ConA column was accomplished by adding 30 ml of 0.3 Mmethyl- ${alpha}$ -D-glucopyranoside dissolved in buffer A at RT. The elutedfractions containing pSTf were pooled, desalted by extensivedialysis against water, and lyophilized. During the purificationprocedure, pSTf was detected by immunoblotting assays with rabbitanti-human transferrin polyclonal antibodies (Dako Corporation)or mouse anti-pSTf monoclonal antibodies as described by Grangeand Mouricout (19).

MLOA. For the microplate lectin overlay assay (MLOA), pSTf was dilutedinto 50 mM carbonate-bicarbonate, pH 9.6, and then 0.1-ml sampleswere immobilized to 96-well Immulon I polystyrene plates (Dynatech,Alexandria, Va.) at 37°C for 16 h. The wells were rinsedthree times with 0.2 ml of phosphate-buffered saline containing0.5% Tween 20 (PBS-Tween). Biotinylated lectins (0.1 ml), dilutedin PBS-Tween, were added to the wells and incubated at RT for30 min. The wells were rinsed three times with 0.2 ml of PBS-Tween.Horseradish peroxidase (HRP) conjugated to streptavidin (0.1ml of a 0.43-µg/ml solution in PBS-Tween; Pierce) wasadded and incubated for 1 h at RT. The wells were rinsed threetimes with 0.2 ml of PBS-Tween, and the bound HRP-streptavidinwas detected with chromogenic peroxidase substrate 2,2'-azino-bis(3-ethyl-benzthiazoline-6-sulfonicacid) as previously described (14).

WBLOA. Prior to electrophoresis, the protein concentrations of allsamples tested were determined using the modified Lowry assaydescribed by Peterson (43) with bovine serum albumin as thestandard. For the Western blot lectin overlay assay (WBLOA),pSTf-containing samples were separated by sodium dodecyl sulfate-polyacrylamidegel electrophoresis (SDS-PAGE) (31) and then transferred tonitrocellulose or polyvinylidene difluoride filters (55). Thefilters were washed three times (15 min, RT) with 20 ml of PBS-Tween.Binding activity was detected by incubating the filter with20 ml of PBS-Tween containing biotinylated lectin (0.5 µg/mlfor K88 adhesin variants and 10 µg/ml for plant lectins)for 30 min at RT, followed by three washings (15 min, RT) with20 ml of PBS-Tween. Bound biotinylated lectin was detected byincubating the filter with HRP-streptavidin (20 ml of a 0.43-µg/mlsolution in PBS-Tween) for 1 h at RT. The filter was washedthree times with 20 ml of PBS-Tween (10 min, RT), and the boundperoxidase activity was then detected with 3,3'-diaminobenzidinein the presence of CoCl₂ as previously described (24).

HPTLC lectin overlay assay. GSL were separated on HPTLC plates with chloroform-methanol-water(50:47:14 [vol/vol/vol]) containing 0.038% CaCl₂ as describedpreviously (45). Typically, two chromatograms were developedin parallel on the same HPTLC plate. One was sprayed with orcinol-sulfuricacid reagent to detect GSL (45), and the other was overlaidwith biotinylated K88 adhesin variants (20). After separation,the dried chromatogram was sprayed with buffer B (0.01 M Na₂HPO₄,0.14 M NaCl, 2% polyvinylpyrrolidone-40 [pH 7.2]). The plateswere then immersed in buffer B for 1.5 h at RT with gentle stirring(20 rpm). The plates were removed from the buffer solution,and the excess buffer was allowed to drip off the plate. Theplates were placed silica side down in buffered solutions containingbiotinylated K88 adhesin variants (10 µg/ml) and incubatedfor 2 h at RT with gentle stirring (20 rpm). Unbound moleculeswere then removed by washing the plate two times for 10 mineach in buffer B. Bound biotinylated K88 adhesins were detectedby incubating the plate with HRP-streptavidin (1 µg/mldiluted in buffer B) for 1 h at RT with stirring. After twowashes, the bound peroxidase activity was detected with 3,3'-diaminobenzidinein the presence of CoCl₂ as previously described (24).

Metaperiodate treatment of pSTf. Purified pSTf (2 µg) was separated by SDS-PAGE and transferredto nitrocellulose filters. Replicate filters (three at eachtime point) were then incubated in 10 mM sodium metaperiodatedissolved in 0.2 M sodium acetate, pH 4.1, for various periodsof time at 37°C on a rotator. Control digests were performedunder identical conditions with 10 mM sodium metaperiodate replacedwith 10 mM sodium iodate. The filters were then washed threetimes with MilliQ water for 5 min at RT. Replicate filters werestained with Coomassie blue R-250 and periodic acid-Schiff (PAS)stain (53) and were subjected to the WBLOA procedure describedabove.

Exoglycosidase treatments. Purified pSTf (1.8 mg) was digested with Vibrio cholerae neuraminidase(80 mU; Boehringer Mannheim Corp., Indianapolis, Ind.) in neuraminidasedigestion buffer (50 mM sodium acetate, 154 mM NaCl, 9 mM CaCl₂,pH 5.5) for 24 h at 37°C. The neuraminidase-digested pSTfwas dialyzed extensively against ß-galactosidase digestionbuffer (50 mM sodium phosphate, 1 mM dithiothreitol, 100 mMNaCl, pH 6.5), and then a portion (1.2 mg) of this digestedpSTf was treated with Diplococcus pneumoniae ß-galactosidase(30 mU; Boehringer Mannheim Corp.) for 24 h at 37°C. Theneuraminidase- and ß-galactosidase-digested pSTf wasdialyzed extensively against ß-N-acetylhexosaminidase(HexNAcase) digestion buffer (75 mM citrate-phosphate, pH 4.8),and then a portion (0.4 mg) of this digested pSTf was treatedwith jack bean meal HexNAcase (0.5 U; Oxford Glycosciences,Wakefield, Md.) for 24 h at 37°C. Control digestions withoutpSTf were performed under the identical conditions (same buffers,reactants, digestion times, and dialysis times) as those forthe exoglycosidase digestions of pSTf. All treated samples ofpSTf, along with control samples, were then tested for theirability to bind biotinylated lectins by the MLOA described above.The absorbance of the appropriate control digest was subtractedfrom the absorbance of the exoglycosidase-treated pSTf to determinethe final intensity of lectin binding.

RESULTS

Top

Results

Binding of K88 fimbrial adhesin variants to pSTf. To determine if pSTf is a suitable model receptor for the K88fimbrial adhesin variants, we first purified pSTf using theprocedure described by Grange and Mouricout (19). The purityof the pSTf preparation was assessed by SDS-PAGE analysis of2 µg of purified pSTf/lane followed by detection withCoomassie blue, antitransferrin antibodies, and PAS stain. Coomassieblue staining detected a 76-kDa band that represented over 98%of the total protein detected in that lane (Fig. 1BII, lane1). A polyclonal antibody to human serum transferrin and a monoclonalantibody to pSTf bound to this 76-kDa glycoprotein, indicatingthat this protein is pSTf (data not shown). PAS staining alsodetected the 76-kDa band, demonstrating that pSTf is the onlydetectable glycoprotein in the preparation (Fig. 1CII, lane1).

View larger version (31K):
[in this window]
[in a new window]
FIG. 1. Periodate inactivation of the binding of the K88ab fimbrial adhesin to pSTf. Purified pSTf was subjected to treatment with sodium metaperiodate (I) and sodium iodate (II) for 15 min (lane 1), 30 min (lane 2), 1 h (lane 3), 2 h (lane 4), 24 h (lane 5), and 48 h (lane 6) as described in Materials and Methods. The treated samples (2 µg of pSTf per lane) were run on SDS-PAGE (10% polyacrylamide) gels and then transferred onto nitrocellulose membranes as described in Materials and Methods. K88ab binding activity was detected with biotinylated K88ab adhesin (A), proteins were detected by Coomassie blue staining (B), and glycoproteins were detected by PAS staining (C). Arrowheads indicate the migration position of pSTF.

To determine if the K88 adhesin variants bind to pSTf, we immobilizedvarious amounts of purified pSTf to 96-well polystyrene plates,overlaid the pSTf with the three biotinylated K88 adhesin variants,and detected the amount of bound biotinylated K88 adhesin usingan HRP-streptavidin detection system. We found that K88ab, butnot K88ac or K88ad, adhesin bound strongly to pSTf (Fig. 2A).To assess the affinity of K88 adhesin variants for pSTf, weimmobilized pSTf to 96-well polystyrene plates, overlaid itwith various concentrations of the three biotinylated K88 adhesinvariants, and detected the amount of bound biotinylated K88adhesin (Fig. 2B). Only the K88ab adhesin variant showed strongbinding to pSTf (Fig. 2B). A double-reciprocal plot of thisbinding data was used to determine that the dissociation constantfor the binding of K88ab adhesin to pSTf is 75 µM.

View larger version (12K):
[in this window]
[in a new window]
FIG. 2. Binding of K88 adhesin variants to purified pSTF. (A) Purified pSTf (0.43 to 110 µg/ml) was immobilized to 96-well polystyrene plates and probed with biotinylated K88ab ( ${square}$ ), K88ac ( ${blacksquare}$ ), and K88ad (•) adhesin variants (2 µg/ml) for 1 h at RT. (B) Purified pSTf (60 µg per well) was immobilized overnight at 37°C onto a 96-well plate and probed with various concentrations of biotinylated K88 adhesin variants ranging from 0.01 to 2 µg/ml for 1 h at RT. Bound biotinylated K88 adhesins were detected with HRP-streptavidin as described in Materials and Methods.

Implication of the glycanic moiety of pSTf in recognition by the K88ab adhesin. The binding of K88ab ETEC to intestinal epithelial cells ismediated by the recognition of a specific carbohydrate on thesurface of intestinal epithelial cells by the K88ab adhesin(47). Consequently, for pSTf to be a useful model receptor,the K88ab adhesin must interact with the carbohydrate moietyof pSTf. To demonstrate that the recognition of pSTf by theK88ab adhesin was through the carbohydrate moiety, we treatedpurified pSTf with sodium metaperiodate to degrade the glycanon pSTf and then tested to see if the treated pSTf could bindK88ab adhesin (Fig. 1I). As a negative control, we treated pSTfwith sodium iodate, which does not degrade glycans (Fig. 1II).After 2 h of treatment, the metaperiodate-treated pSTf was nolonger recognized by the K88ab adhesin (Fig. 1AI, lane 4). Incontrast, iodate-treated pSTf was still recognized by the K88abadhesin even after 48 h of treatment (Fig. 1AII, lane 6). Theseresults indicate that the K88ab adhesin interacts with the glycanicmoiety of pSTf. In addition to testing for the K88ab adhesin-bindingability of the treated pSTf, we performed Coomassie blue (Fig.1B) and PAS (Fig. 1C) staining of each treated sample. Theseresults verified that periodate did not degrade the proteinmoiety of pSTf but slowly degraded the carbohydrate moiety,with complete destruction of the carbohydrate, as evidencedby the lack of PAS staining after 48 h of incubation with sodiummetaperiodate (Fig. 1CI, lane 6).

Characterization of the glycanic moiety of pSTf. Results from previous studies have shown that preparations ofpSTf contain a combination of glycoforms that each possess asingle biantennary complex N-glycan having the structures shownin Fig. 3 (49, 50, 56). To characterize the glycan present onthe pSTf that we isolated for this study, we used biotinylatedlectin binding studies and exoglycosidase digestion studies(Table 1). ConA lectin bound to pSTf, verifying the presenceof an N-glycan (Table 1). Purified pSTf also bound RCA₁₂₀ andEBL, indicating the presence of Gal and ${alpha}$ (2-6)-linked NeuAc,respectively, in terminal positions (Table 1). Also, ECL whichis specific for Galß(1-4)GlcNAc, bound to pSTf, whileclosely related lectin MAL I, which is specific for Galß(1-3)GlcNAc,did not react (Table 1). In addition, UEA, which is specificfor ${alpha}$ -linked fucose, weakly bound to pSTf, indicating that ourpSTf preparation contains at least some fucosylated pSTf glycoforms.

View larger version (30K):
[in this window]
[in a new window]
FIG. 3. Proposed structures of the N -glycan on pSTf before and after exoglycosidase digestions.

View this table:
[in this window]
[in a new window]
TABLE 1. Reactivities of biotinylated lectins with untreated and exoglyocosidase-treated pSTf

To further evaluate the structure of the glycan on pSTf andto determine the monosaccharides on pSTf recognized by the K88abadhesin, we treated pSTf with specific exoglycosidases to sequentiallyremove terminal monosaccharides and then verified the removalof the appropriate monosaccharides with biotinylated plant lectins.The effect of exoglycosidase treatment on the K88ab adhesin-bindingactivity of the receptors was also evaluated by K88ab adhesinoverlay assays. Treatment of pSTf with neuraminidase increasedK88ab adhesin-binding activity, inhibited EBL binding, and enhancedRCA₁₂₀ binding to pSTf (Table 1). These results indicate thatterminal NeuAc residues were removed to expose Gal residues,and removal enhanced recognition of pSTf by the K88ab adhesin(Fig. 3B). Since the reactivity of pSTf with RCA₁₂₀ was initiallylow and was then greatly enhanced when the NeuAc residues wereremoved, it is likely that at least some of pSTf in our preparationis the disialylated glycoforms (Fig. 3A). Sequential treatmentof pSTf with neuraminidase followed by ß-galactosidaseincreased the K88ab adhesin-binding activity in comparison tothat of the untreated control, completely abolished recognitionof pSTf by RCA₁₂₀ lectin, and enhanced recognition by G. simplicifoliaII lectin (Table 1). These results indicate that the Gal residuesthat are exposed by treatment of the pSTf with neuraminidaseare not required in the recognition of pSTf by the K88ab adhesin(Fig. 3C). Sequential treatment of pSTf with neuraminidase followedby ß-galactosidase and then HexNAcase increased theK88ab adhesin-binding activity of pSTf in comparison to thatof the untreated control, completely abolished recognition ofpSTf by G. simplicifolia II lectin, and enhanced binding byConA (Table 1). These results indicate that the pSTf containsNeuAc, ß-linked Gal, and ß-linked GlcNAc,which is consistent with the structures shown in Fig. 3A. Theseresults indicate that the terminal three carbohydrate residues[NeuAc ${alpha}$ (2,6)Galß(1,4)GlcNAcß] are not anessential part of the recognition site for K88ab adhesin onpSTf. In fact, removing these residues enhances binding of K88abadhesin to pSTF. Two possible explanations for this increasedbinding are that (i) the K88ab adhesin recognizes the newlyexposed Man (Fig. 3D) and (ii) removal of the terminal threeresidues on each antenna causes a conformational change in theremaining oligosaccharide, which makes it a better receptorfor K88ab adhesin. The second explanation seems more likelysince K88ab adhesin binding has been demonstrated to be mannoseresistant (48). Additional evidence for a conformational changein the oligosaccharide comes from experiments where we foundthat removal of the terminal three carbohydrate residues [NeuAc ${alpha}$ (2,6)Galß(1,4)GlcNAc-]from pSTf enhances binding of two lectins, STL and LEL, thatare known to react with GlcNAc oligomers (Table 1). From thesestudies, it seems likely that K88ab adhesin binds to one orboth of the GlcNAc residues that make up the core of the pSTfN-glycan and that the K88ab adhesin binds more intensely tohigh-mannose N-glycans (Fig. 3D) than to complex N-glycans (Fig.3A).

Determination of K88ab binding specificity with GSL. To further characterize the glycanic structures recognized bythe K88 adhesin variants, we tested 19 purified GSL standards,which can be divided into four groups (lactosyl, neolactosyl,globosyl, and gangliosyl) for their ability to bind the threeK88 adhesin variants (Fig. 4; Table 2). Within the lactosylgroup, all three K88 variants bound to Lc₃, K88ac bound to Lc₄,and K88ad bound to Lc₂. Within the neolactosyl group, all threeK88 variants bound to nLc₄ and nLc₆. Within the globosyl family,all three variants bound to Gb₃. Within the gangliosyl group,all three variants recognized GA₂ and GA₁ and K88ac bound toGM₃. Overall, the three K88 variants bind to many of the sameGSL, indicating that they have similar receptor binding specificities.However, there are also some differences in their receptor specificities,as evidenced by the variation in their recognition of some ofthe GSL.

View larger version (91K):
[in this window]
[in a new window]
FIG. 4. Binding of K88ac adhesin to GSL standards. Two micrograms of Lc₃ (lane 1), Lc₃ plus Lc₄ (lane 2), nLc₄ (lane 3), Le^a (lane 4), Le^x (lane 5), V³FucnLc₆ (lane 6), VI²FucnLc₆ (lane 7), and nLc₆ (lane 8) was separated on HPTLC plates. These GSL were incubated with biotinylated K88ac adhesin as described in Materials and Methods.

View this table:
[in this window]
[in a new window]
TABLE 2. Binding of K88 adhesin variants to GSL standards

DISCUSSION

Top

Discussion

With respect to receptor recognition, the P pilus adhesin foundon uropathogenic strains of E. coli is one of the best-characterizedfimbrial adhesins (26). As with the K88 adhesin, three differentvariants of P pili (class I, class II, and class III) have beenidentified (10, 35, 36). All three classes of P pili recognizegloboseries GSL that contain the Gal ${alpha}$ (1-4)Gal (galabiose) corestructure found on human P blood group antigens. However, thedifferent variants of P pili do not bind equally well to allgalabiose-containing receptors. Class I P pili preferentiallybind to Gb₃ [Gal ${alpha}$ (1-4)Galß(1-4)Glcß(1-1)Cer](35), while class II pili bind to Gb₄ [GalNAcß(1-3)Gal ${alpha}$ (1-4)Galß(1-4)Glcß(1-1)Cer](52) and class III pili prefer Forssmanantigen [Gb₅; GalNAcß(1-3)GalNAcß(1-3)Gal ${alpha}$ (1-4)Galß(1-4)Glcß(1-1)Cer](36). It seems logical that variants of a particular adhesinwould have related carbohydrate specificities, since it is notlikely that a whole new carbohydrate-binding pocket would becreated in each variant. It is much more likely that variantsof an adhesin have similar carbohydrate-binding pockets, whichhave been modified to produce slightly different binding specificities.The K88 adhesin variants may be similar to the P pilus variantsin having different, yet related, carbohydrate-binding specificities.Therefore, we hypothesized that the three K88 variants recognizea core mono-, di-, or trisaccharide structure and that eachvariant prefers different substitutions or modifications tothat core structure. The objective of the present study wasto define the core structure recognized by the three K88 adhesinvariants by comparing the binding specificities of the threeK88 variants by using pSTf and GSL as model receptors.

The cores of all N-glycans contain Manß(1,4)GlcNAcß(1,4)GlcNAcß(1,N)Asn(57). Consequently, the K88ab adhesin, which was shown in thepresent study to bind to GlcNAc residues in the core of theN-glycans, could potentially bind to all N-glycosylated proteins.This observation is consistent with previous reports that theK88ab adhesin binds to more glycoproteins in crude intestinalextracts than either of the other two K88 variants (7). Despitethe fact that all N-glycans possess the same core structure,it is not likely that K88ab adhesin binds equally well to allN-glycans, as the GlcNAc residues may be more accessible orin a more favorable conformation in certain types of N-glycans.For example in the present study, lectin binding to the N-glycanon pSTf was higher when the antennae of the N-glycans were shortenedby exoglycosidase treatment. This observation is supported bythe fact that STL and LEL, along with the K88ab adhesin, reactmuch more intensely with exoglycosidase-digested pSTf than withundigested pSTf. Consequently, GlcNAc-specific lectins, suchas LEL, STL, and K88ab adhesin, bind more intensely to high-mannoseN-glycans than to complex N-glycans. Modifications to the corestructure may also affect K88ab adhesin binding to N-glycans.One of the most common direct modifications to the core structureof N-glycans is the addition of Fuc ${alpha}$ (1,6)-linked to the GlcNAcresidue attached to the asparagine. This modification may inhibitthe binding of K88ab adhesin to N-glycans in a manner similarto what we observed with GSL, where fucosylation of the GlcNAcin nLc₄ to produce Le^x destroys its K88 adhesin-binding ability(Table 2). Further evidence that fucosylation may affect K88abbinding comes from a study by L'Hote et al. (34), where theyobserved that the K88ab adhesin does not interact with the A ${alpha}$ subunit of porcine fibrinogen, which contains the same fucosylatedN-glycan found on some glycoforms of pSTf (Fig. 3A).

Previously, Grange and Mouricout (19) reported that the K88abadhesin bound to porcine intestinal transferrin (pITf) on enterocytesfrom susceptible, but not resistant, pigs. In this same study,they also reported the binding of K88ab adhesin to the moreabundantly available form of transferrin found in the serum(pSTf) and found that pSTf from all pigs tested bound K88abadhesin. Thus, there is no correlation between K88ab bindingto pSTf and the animal's susceptibility to K88ab ETEC-induceddisease. It has been demonstrated that pSTf contains a singleglycan that is N linked to Asn-497, located near the C terminusof the molecule (49). This glycan has been shown to be a biantennarycomplex N-glycan with the structures shown in Fig. 3 (49, 50,56). The pSTf that we isolated in the present study is mostlikely a mixture of glycoforms containing fucosylated and unfucosylatedglycans that are either mono- or disialyated (Table 1; Fig.3A). The structure of the N-glycan on pITf has not been clearlyelucidated. However, Grange and Mouricout (19) reported thatthe carbohydrate compositions of the two forms pITf and pSTfare similar. It would be quite informative to compare the structuresof the N-glycans on pITf from susceptible versus resistant pigs.One possibility is that the pSTf from the resistant pigs mayconsist predominantly of fucosylated glycoforms that do notbind K88ab adhesin.

The core structure that is most consistently found in the GSLthat all three variants bind to is HexNAc ß-linked[either GalNAcß(1-4) or GlcNAc ß(1-3)] toGal. Addition of a Gal, either ß(1-4)-linked to aGlcNAc or ß(1-3)-linked to a GalNAc, does not disruptthe binding of the K88 variants to the core structure. In fact,addition of Gal ß(1-4)-linked to GlcNAc, as seen whencomparing binding to Lc₃ and binding to nLc₄, actually enhancesthe intensity of binding of all three variants. In contrast,modification of the HexNAc with Fuc destroys its K88 adhesin-bindingability, as demonstrated by the lack of binding to Le^x, whichis fucosylated nLc₄. In addition, modification of the Gal foundin the core structure with NeuAc, as seen when GA₂ and GA₁ areconverted to GM₂ and GM₁, respectively, abolishes the bindingof all three K88 variants. It appears that the core recognitionsequence for K88 adhesins, HexNAc ß-linked to Gal,must be positioned close to the nonreducing terminus. This isevidenced by the presence of a potential internal unsubstitutedcore recognition sequence for K88 adhesin in both VI²FucnLc₆and V³FucnLc₆ that does not bind any of the K88 adhesin variants.The binding of all three variants to Gb₃ is hard to reconcilewith the remainder of the GSL results, since Gb₃ does not containthe proposed core recognition sequence. One possibility is thatthe K88 variants recognize, weakly for K88ab and K88ac, theterminal Gal in Gb₃. It is difficult with the limited numberof GSL tested to define the structural basis of the differencesin the specificities of the three variants, but further studiesusing both model and natural receptors will allow these comparisonsto be made.

Many different types of studies, including studies of monosaccharideand glycoprotein blocking of hemagglutination and brush borderbinding, have been performed to determine the binding specificitiesof the three variants of K88 adhesin (2, 8, 17, 42). All ofthese studies conclude that Gal, GlcNAc, GalNAc, or some combinationof these monosaccharides has a role in K88 adhesin recognition.Recently, we determined that the K88ac adhesin interacts withß-linked Gal on IMTGP (18) and that the K88ad adhesinrecognizes the terminal Galß(1-4)GlcNAc on the neutralGSL, nLc₄ [Galß(1-4)GlcNAcß(1-3)Galß(1-4)Glcß(1-1)Cer](20). In addition, Seignole et al. (46) proposed that the K88acadhesin recognizes Galß(1-3)GalNAc and Fuc ${alpha}$ (1-2)Galß(1-3/4)GalNAcon IMTGP. We have been unable to verify that Fuc is involvedin K88 adhesin recognition of either IMTGP (P. A. Grange andA. K. Erickson, unpublished observation) or GSL (Table 2). Theresults in the present study are consistent with most of theearlier findings and extend these finding by directly comparingthe binding specificities of the three K88 adhesin variantsby using common model receptors. Based on the present findings,we propose that the minimal carbohydrate structure needed forrecognition by K88 adhesin variants contains a HexNAc ß-linkedto a Gal residue and that a terminal Gal ß-linkedto the HexNAc enhances K88 adhesin binding but is not essentialfor recognition of glycans by K88 adhesins.

One of our long-term goals is to determine the differences incarbohydrate specificity among the K88 adhesin variants. Theresults obtained in the present study shed some light on thesimilarities in binding specificities among the K88 variantsbut do not reveal much about the molecular basis of the differencesin carbohydrate specificity among the K88 variants. To addressthis problem, studies comparing the structures of the glycansrecognized by the K88 adhesin variants on the phenotype-specificreceptors (7) from intestinal epithelial cells (intestinal glycolipidK88ad receptor for K88ad [20], IMTGPs for K88ab and K88ac [7,11, 12, 15, 27, 46], and pITf for K88ab [19]) need to be performed.Determination of these structures will make it possible to beginto understand the changes in receptor specificity that occurduring evolutionary development of new adhesin variants andto synthesize structural analogues of the receptor recognitionsequence which can be used to prevent and treat K88 ETEC-induceddisease in pigs.

ACKNOWLEDGMENTS

We gratefully acknowledge USDA NRI-CGP grants 97-03708 and 2000-02200for providing financial assistance.

FOOTNOTES

* Corresponding author. Mailing address: Veterinary Science Department, P.O. Box 2175, South Dakota State University, Brookings, SD 57007. Phone: (605) 688-6544. Fax: (605) 688-6003. E-mail: alan_erickson{at}sdstate.edu.

Editor: V. J. DiRita

REFERENCES

Top