Interaction of Human Macrophage C-type Lectin withO-Linked N-Acetylgalactosamine Residues on Mucin Glycopeptides*

A fluorescein-labeled synthetic peptide, PTTTPITTTTK, was converted into O-glycosylated glycopeptides with various numbers of attachedN-acetyl-d-galactosamines (GalNAcs) by in vitro glycosylation with UDP-GalNAc and a microsomal fraction of LS174T human colon carcinoma cells. Glycopeptides with 1, 3, 5, and 6 GalNAc residues (G1, G3, G5, and G6) were obtained, and their sizes were confirmed by matrix-assisted laser desorption ionization time-of-flight mass spectrometry. Their sequences were determined by a peptide sequencer to be PTTTGalNAcPITTTTK for G1, PTGalNAcTTPITGalNAcTGalNAcTTK for G3, PTTGalNAcTGalNAcPITGalNAcTGalNAcTGalNAcTK for G5, and PTGalNAcTGalNAcTGalNAcPITGalNAcTGalNAcTGalNAcTK for G6. A calcium-type human macrophage lectin (HML) was prepared in a recombinant form, and its interaction with these glycopeptides was investigated by surface plasmon resonance (SPR) spectroscopy and fluorescence polarization. The affinity of recombinant HML (rHML) for immobilized glycopeptides increased, as revealed by SPR, in parallel with the number of GalNAc. The highest affinity was obtained when the G6-peptide was immobilized at high density. Fluorescence polarization equilibrium-binding assays also revealed that the affinity of rHML for soluble gly-copeptides increased, depending on the number of attached GalNAcs. Carbohydrate recognition domain (CRD) fragments of HML were prepared, and their affinity for these four glycopeptides was also determined, this affinity was apparently lower than that of rHML. Affinity constants of rHML for the G3- and G5-peptides were 11- and 38-fold higher, respectively, than for the G1-peptide, whereas those of CRD fragments were only 2- and 6-fold higher, respectively. A chemical cross-linking study revealed that rHML but not recombinant CRD forms trimers in an aqueous solution. Thus, preferential binding of densely glycosylatedO-linked glycopeptides should be due to the trimer formation of rHML.

prise a large family of recognition molecules in the immune system. Their specificity and potential regulatory functions demand a perspective far beyond the traditional concept of lectins (1,2). A C-type lectin specific for galactose (Gal) and N-acetyl-D-galactosamine (GalNAc) as a monosaccharide has been found in histiocytic macrophages in mice (3). A very similar lectin is also expressed in human macrophages as preferentially bound glycopeptides containing consecutive serine (Ser) and threonine (Thr) residues with attached GalNAc, which is well known as a carcinoma-associated Tn antigen (4). While these C-type lectins are assumed to play roles in the recognition of malignant cells, their immunological roles are actually more diverse (5)(6)(7). In rats, a lectin homologous to these molecules was shown to be up-regulated in the area of chronic rejection of heart transplants in rats (8). It has long been known that other C-type lectins produced by hepatocytes are involved in asialoglycoprotein uptake from the circulation; macrophage C-type lectins from rodents seem to function in the same fashion (5,9).
Because the interactions of some lectins with simple monosaccharides are sometimes weak, the direct binding of these lectins with a monosaccharide may be difficult to determine. If the cooperative effects of polyvalent carbohydrate ligands are present, however, and the lectin has multivalency, a dramatic increase in the affinity could prevail (2,10,11). Members of the C-type lectin family often form oligomeric structures through the stalk regions containing the ␣-helical domain; this provides an advantage in polyvalent binding (12,13). A similar observation has been made in the binding of carbohydrate-specific antibodies (14). For example, synthetic neoglycoproteins and saccharide-derivatized polyacrylamide polymers have been used to represent a multivalent ligand that has a high affinity for lectins (15,16). In other studies, the clustering effects of carbohydrate chains were evaluated in detail through the use of trivalent glycosides (17)(18)(19). Although the multivalency of carbohydrate ligands is seen in a variety of native glycoproteins, particularly O-linked glycoproteins, its functions in carbohydrate-recognition pathogenic processes are not well known. Furthermore, differences in the kinetics of proteincarbohydrate interactions between monosaccharide and multivalent ligands were unknown until recently. The first step toward determining the significance of HML in the pathogen-esis of a variety of diseases was the cDNA cloning and sequencing of human macrophage lectin (HML), described in a previous report (4). The next step should be to elucidate the nature of natural ligand for this lectin. In the previous study, HML was demonstrated to preferentially bind clusters of N-acetylgalactosamine linked to Ser or Thr in a carcinoma-associated mucinlike configuration (4).
Epithelial mucins are characterized by their tandem repeat domains, which contain many Ser and Thr residues (20,21). The tandem repeat portion of the MUC2 mucin consists of 23 amino acids, including 14 Thr residues; these Thr residues are reported to be glycosylated up to 78% in the LS174T colon carcinoma cell line (22). Although the tandem repeat domain of MUC2 mucin in other cells is also thought to be highly Oglycosylated, whether the glycosylation patterns in different cells and tissues are unique is not yet understood. Mucin architecture represents an ideal framework for a variety of multiple ligand arrangements that should serve as a recognition unit. In the present study, we applied the surface plasmon resonance (SPR) biosensor and fluorescence polarization (FP) spectroscopy to analyze ligand-density dependence in carbohydrate-protein interactions.

EXPERIMENTAL PROCEDURES
Preparation of rHML and Its CRD Fragments-The gene encoding HML lacking the transmembrane domain was prepared by polymerase chain reaction and constructed into an expression vector pET-8c to yield the plasmid pET-8c-HML as described previously (4). The constructed plasmid was introduced into Escherichia coli strain BL21(DE3) cells and expressed. Soluble recombinant HML (rHML) was purified by galactose-Sepharose 4B affinity chromatography. For the preparation of its CRD portion, rHML (2 mg) was digested with 100 g/ml of N-tosylphenylchloromethyl ketone-treated trypsin (Sigma) in 10 mM HEPES buffer (pH 7.5) containing 2 mM CaCl 2 and 0.15 M NaCl (Ca-HBS) at 37°C for 3 h. CRD fragments were separated by dialysis from digested proteins. The purity of both the rHML and the CRD fragments was confirmed by polyacrylamide gel electrophoresis (PAGE) on 12.5% gels, as single bands of 29 and 19 kDa, respectively.
Determination of Oligomeric Structures of rHML through Chemical Cross-linking-rHML in aqueous solutions was cross-linked with homobifunctional cross-linkers. Disuccinimidyl suberate (DSS; Pierce), which potentially cross-links two amino groups located 1.14 nm apart, was used at a concentration of 1 mM in 100 l of Ca-HBS at room temperature for 3 h. Ethylene glycolbis(succinimidylsuccinate) (EGS; Pierce), which potentially cross-links two amino groups located 1.61 nm apart, was used at a concentration of 1 mM under the same conditions. Aliquots of the incubation mixtures were analyzed by 10% SDS-PAGE and visualized by Coomassie Blue staining. Their molecular weights were calculated according to the migration distance of molecular weight markers (Marker II, Daiichi Pure Chemicals Co., Ltd. Tokyo, Japan). Products of cross-linking reactions were also analyzed by matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF MS) on Voyager Elite (Nihon PerSeptive Ltd., Tokyo, Japan) using ␣-cyano-4-hydroxycinnamic acid as a matrix.
Preparation of Glycosylated Peptides-The fluorescein-labeled peptide (FITC-PTTTPITTTTK) corresponding to the MUC2 mucin tandem repeat domain (23) was prepared. The peptide was purified by reversedphase high performance liquid chromatography (HPLC) (Jasco, Tokyo, Japan) on 6 ϫ 150-mm column of Puresil (C18, Nihon Waters, Tokyo, Japan). The fidelity and molecular mass of the peptide was confirmed on an amino acid sequencer (PSQ-1, Shimadzu, Kyoto, Japan) and MALDI-TOF MS. These peptides were labeled with fluorescein isothiocyanate at its N-terminal amino acid at pH 7.5 adjusted with 100 mM HEPES buffer. The lysine residue was not substituted with FITC as revealed by carboxypeptidase digestions. The peptide was glycosylated in vitro with crude UDP-GalNAc:polypeptide ␣-GalNAc-transferases (EC 2.4.1.41) derived from LS174T colon carcinoma cell lines, as described in detail elsewhere. 2 The glycosylated peptides were fractionated by reversed-phase HPLC equipped with a 10 ϫ 250-mm column of Cosmosil (5C 18 -AR, Nacarai Tesque, Kyoto, Japan) or a Puresil column as above. The columns were eluted with gradients of solvent A (0.05% trifluoroacetic acid in water) and solvent B (0.05% trifluoroacetic acid and 70% propanol in acetonitrile). A linear gradient of 0 -50% solvent B in solvent A was used for Cosmosil, and a linear gradient of 10 -30% solvent B in solvent A was used for Puresil, at a flow rate of 2 ml/min for 30 min. Elutions were monitored by fluorescence intensity at an emission wavelength of 520 nm, using excitation at 492 nm. The number of GalNAc attached to these peptides was determined by the mass number according to MALDI-TOF MS using ␣-cyano-4-hydroxycinnamic acid as a matrix, as described elsewhere. 2 The linear mode with delayed extraction setting was applied. For the biotinylation of glycopeptides, the glycopeptides were incubated with 1 mg of either sulfosuccinimide-biotin (SNHS-biotin) or sulfosuccinimidyl-6-(biotinamido)hexanoate (SNHS-LC-biotin) (Pierce) in 100 mM HEPES buffer (pH 8.0) at room temperature for 1 h. The biotinylated glycopeptides were purified by reversed-phase HPLC equipped with C18 columns; structures were confirmed by MALDI-TOF MS analysis.
Amino Acid Sequencing-Pulsed liquid Edman degradation amino acid sequencing of glycopeptides was performed on an Applied Biosystems 490 Procise protein sequencing system (Perkin-Elmer). With this system, a phenylthiohydantoin (PTH) derivative of GalNAc-Thr was identified as a pair of peaks eluted near the positions of PTH-Ser and PTH-Thr (24). Amino acid sequencing of the fully glycosylated peptide (PT GalNAc T GalNAc T GalNAc PIT GalNAc T GalNAc T GalNAc T GalNAc K) confirmed these eluting positions (data not shown).
Interaction of rHML with Immobilized Glycopeptides-The interaction of rHML with immobilized glycopeptides was determined with BIAcore (Pharmacia Biosensor AB, Uppsala, Sweden), in which the measurement is based on the SPR (25). Each biotinylated glycopeptide was introduced onto the surface of a streptavidin-coated sensor chip (SA5) at a flow rate of 10 l/min. Based on the saturated response of anti-FITC monoclonal antibody (mAb) (clone FL-D6, Sigma) at 2500 -3000 response units, the amount of each glycopeptide immobilized onto different sensor chips was confirmed to be almost identical. rHML at various concentrations (0.1-10 M) dissolved in 30 l of Ca-HBS was applied over each glycopeptide-immobilized surface at 25°C. Anti-trimeric Tn mAb (clone B231.5, kindly donated by Dr. Mark Reddish, Biomira Inc., Alberta, Canada) (26) was also used. To minimize an effect of the mass transport limit, flow rate was maintained at 10 l/min (27). The surface was regenerated with HEPES-buffered saline containing 10 mM EDTA for the binding of rHML, and with 200 mM glycine-HCl buffer (pH 2.7) for the binding of anti-trimeric Tn mAb. Association and dissociation rate constants (k a and k d ) were calculated using BIAevaluation 2.0 software (Pharmacia Biosensor AB).
Determination of Interaction of rHML and Its CRD Fragment with Soluble Glycopeptides-To determine the interaction of rHML with soluble glycopeptides, FP equilibrium-binding assays were performed. The principal advantage of this method is that the assay is done in fluid phase. In addition, no separation of bound glycopeptide from free glycopeptide is required (28). To reach an equilibrium, fluorescence-labeled glycopeptides (0.1 pmol; total intensity of 200) were incubated with the indicated concentration of rHML or CRD fragments in 100 l of Ca-HBS at room temperature for 90 min. FP values were measured by Beacon 2000 (Takara Shuzo, Shiga, Japan) at 25°C.

RESULTS
Properties of rHML and CRD Fragments-rHML cDNA lacking its transmembrane domain was expressed in E. coli. Soluble rHML protein was purified by affinity chromatography, as described previously (4). Its CRD fragments were obtained by digestion with N-tosylphenylchloromethyl ketone-treated trypsin. As shown in Fig. 1, the purity of rHML and its CRD fragments was confirmed by SDS-PAGE on 12.5% gels and the apparent relative molecular mass values were 29 and 19 kDa, respectively. The oligomeric status of these preparations in aqueous solution was examined by chemical cross-linking. Following treatment with either DSS or EGS, the resultant proteins were analyzed by SDS-PAGE (Fig. 1). rHML was crosslinked with DSS and with EGS, as shown by bands migrating at positions corresponding to a molecular mass of 58 and 87 kDa, respectively. These sizes suggested that the proteins are dimeric and trimeric forms of rHML. Such cross-linked products were not observed, however, when DSS-or EGS-treated CRD fragments were analyzed by SDS-PAGE, indicating that CRD fragments were present predominantly in monomeric form (Fig. 1). The major cross-linked product was an 87-kDa molecule, even after a prolonged incubation for 20 h and incubation with high concentrations of cross-linking agents (data not shown).
MALDI-TOF MS analysis indicated that the average molecular masses of rHML and CRD fragments were 25,334 and 17,780, respectively; these findings are consistent with predicted masses of 25,322 and 17,732, respectively. Following the cross-linking of rHML, signals of 26,673, 53,163, and 79,845 were obtained (data not shown). For the evaluation of the binding kinetics of HML and CRD, the predicted mass figures of M r 25,322 and 17,732 were used.
Carbohydrate Specificity of rHML toward Polyvalent Oligosaccharide Derivatives-As subjects in an initial screening of the specificity of rHML, as shown in Fig. 2, we chose polyvalent oligosaccharide derivatives constructed on a polyacrylamide backbone (15). Results of ELISA indicated that GalNAc␣-and GalNAc␤-BPPs strongly bind immobilized rHML in a calciumdependent manner; as no binding of other monosaccharide BPPs, including Gal␤-, GlcNAc␤-, and Man␣-BPPs, were observed at concentrations up to 30 nM ( Fig. 2A). Two BPPs closely related to GalNAc␣-BPP, a Forssman disaccharide de- rivative (GalNAc␣1-3GalNAc␤-BPP) and a blood group A-type trisaccharide derivative (GalNAc␣1-3[Fuc␣1-2]Gal␤-BPP), showed no affinity for rHML. Competitive inhibition experiments with various monosaccharides demonstrated that Gal-NAc inhibits the interaction of GalNAc␣-BPP with rHML and that Gal does not. When 2-5 mM GalNAc were added to the binding assay in which 1 nM rHML was used, the binding of GalNAc␣-BPP dropped to half of the maximum level (Fig. 2B). Even though rHML was purified by affinity chromatography on Gal-Sepharose, neither the binding of Gal-BPP to rHML nor the inhibition of GalNAc-BPP binding by Gal was detected under the conditions used.
Characterization of Glycopeptide Prepared by Cell-free Glycosylation of MUC2 Mucin Peptides-To determine the effects of the density of GalNAc residues in glycopeptides on their affinity with rHML, undecapeptides corresponding to the tandem repeat portion of the MUC2 mucin with various numbers of attached GalNAc residues were generated in vitro as described elsewhere. 2 Fluorescence-labeled peptides with one, three, five, or six GalNAc residues (G1-, G3-, G5-, and G6peptides) were immobilized by biotinylation on streptavidincoated sensor chips (see below). The fidelity of the products was confirmed by MALDI-TOF MS analysis as shown in Fig. 3. Mass data revealed that all of the peptides contained 1 mol of FITC and biotin (or LC-biotin) and the predicted number of GalNAc residues. Additional peaks observed in the areas of lower mass numbers seemed to represent degradation products generate during the mass analysis, because each glycopeptide eluted as a single peak on the reversed phase HPLC under various conditions. Furthermore, as shown in Fig. 4, a significant portion of the material with a mass number of 2094.4 was identified as PTTT GalNAc PITTTTK by peptide sequence analysis. When an analytical C18 column was applied in place of a preparative column and eluted with less steep gradient, material corresponding to the peak with three attached GalNAc residues was separated into two peaks. The second peak was found to be composed primarily of PT GalNAc TTPIT GalNAc T GalNAc TTK. It was this glycopeptide, with three attached GalNAc residues, that was examined to determine its interaction with HML. Sequence analysis of the glycopeptide with five attached GalNAc residues was shown to be PTT GalNAc T GalNAc PIT GalNAc T GalNAc T GalNAc TK and that with six attached residues was shown to be PT GalNAc T GalNAc T GalNAc PIT GalNAc T GalNAc T GalNAc TK.
Interaction of rHML with Immobilized Glycopeptides as Determined by SPR Spectroscopy-Glycopeptides (0.1 pmol in 100 l) were introduced onto each flow cell. Approximately equal amounts of glycopeptides were immobilized, as confirmed by a similar resonance response following anti-FITC mAb binding (ϳ3000 response units). For the kinetic analysis, rHML was applied at concentrations of 0.15-10 M to each flow cell; association and dissociation were monitored in real time.
As can be seen in the representative sensorgraphs of the  interactions between rHML and the immobilized glycopeptides shown in Fig. 5, rHML bound G1-, G3-, G5-, and G6-peptides in a dose-dependent manner (Fig. 5, A-C and E), while no binding to the GalNAc-free peptides was not observed. As shown in Table I, k a , k d , and K D constants were calculated according to a single class receptor-ligand binding equation. The affinity of rHML for glycopeptides, as well as increases in k a and decreases in k d , were dependent upon the number of bound Gal-NAc residues resulting in apparent K D constants of 1.0 ϫ 10 Ϫ6 , 4.4 ϫ 10 Ϫ7 , 1.9 ϫ 10 Ϫ7 , and 1.5 ϫ 10 Ϫ7 M for G1-, G3-, G5-, and G6-peptide, respectively.
Using the same flow cells, with the peptides incorporating different residues, anti-trimeric Tn mAb (ϳ30 g/ml) was tested (Fig. 6). It did not react with the GalNAc-free and the G1-and G3-peptides, but it did react with the G5-and G6peptides (over 1000 response units). These results confirmed that G5-and G6-peptides contain trimeric-Tn within three and four consecutive Thr residues.
It should be noted that when greater than 10 pmol of peptides were immobilized, anti-trimeric Tn mAb also reacted with G1-peptides (Fig. 6). rHML had very high affinity in flow cells for densely immobilized G6-peptides, as shown in the bottom rows in Table I (K D : 5.3 ϫ 10 Ϫ8 M). However, because k d decreased drastically, the affinity of rHML for G6-peptides immobilized in a more sparse density was much lower (K D : 1.5 ϫ 10 Ϫ7 M). Binding of anti-trimeric Tn mAb was not intense to G1-peptides immobilized at 0.1-10 pmol (100 l, 10 l/min) (Fig. 6). When rHML was tested for it binding to immobilized G1-peptides in an even more sparse density, its affinity constant was calculated to be lower (K D : 1.3 ϫ 10 Ϫ6 M; : 0.2 ϫ 10 Ϫ6 M) (data not shown). Specific binding of rHML to immo-bilized glycopeptides was completely inhibited by the presence of EDTA, as shown in Fig. 5.
Interaction of rHML with Soluble Glycopeptide-In the measurement of the affinity constant, one of several ways to avoid the effect of distance between each immobilized glycopeptide is to determine the interaction of rHML and glycopeptides in solutions. This can be achieved with the FP equilibriumbinding assay system (28). Before assays were conducted, it was determined that the time required to reach equilibrium was less than 5 min (data not shown). With the assay system, as shown in Fig. 7, fluorescence-labeled glycopeptides were incubated with the indicated concentration of rHML at room temperature for 90 min. The binding of rHML to glycopeptides, as indicated by changes in FP intensity, depended upon its concentration (Fig. 7A) and calcium ion (data not shown); no binding to GalNAc-free peptides was observed. Affinity constants of rHML were obtained by Scatchard analysis (Fig. 7B), yielding apparent K D of 6.8 ϫ 10 Ϫ6 , 6.4 ϫ 10 Ϫ7 , and 1.8 ϫ 10 Ϫ7 for G1-, G3-, and G5-peptides, respectively. The affinity constants for G3and G5-peptides were 11-and 38-fold higher, respectively, than the constant for G1-peptides. These findings were in agreement with the results from the interaction of rHML with immobilized glycopeptides except for the G1-peptide.
Interaction of CRD Fragments with Soluble Glycopeptides-Interactions of CRD fragments from rHML with glycopeptides were primarily evaluated with the FP equilibrium-binding assay system (Fig. 8). The signals revealed by SPR spectroscopy with CRD fragments were not high enough to permit the calculation of affinity constants. Interactions were detected, however, when relatively high doses of CRD fragments were analyzed; the maximum response was observed in the interaction FIG. 4. Profiles of amino acid sequencing chromatograms of the oligopeptide and its derivatives with incorporated GalNAc residues (G1-, G3-, G5-, and G6-peptides). Following separation by reversed-phase HPLC, these glycopeptides were analyzed on the Applied Biosystems 490 Procise protein sequencing system. A, untreated FM2-1. B, FM2-1 with one GalNAc residue. C, FM2-1 with three GalNAc residues. D, FM2-1 with five GalNAc residues. E, FM2-1 with six GalNAc residues. Single asterisks and double asterisks indicate putative peaks of PTH derivatives of ␣-GalNAc-Thr (T GalNAc ).
with G5-peptides (ϳ14 M). Scatchard analysis (Fig. 8B) revealed that affinity of CRD is lower than that of rHML. They were 8.3, 3.9, and 1.4 ϫ 10 Ϫ6 M for G1, G3, and G5-peptide (Table II), respectively. The affinity of CRD fragments was 2-fold greater for G3-peptides and 6-fold greater for G5-peptides than for G1-peptide. As was found regarding the specific binding of rHML to immobilized glycopeptides, the binding of CRD fragments to each of the these four soluble glycopeptides was inhibited by the presence of EDTA (data not shown). DISCUSSION To examine the interaction of HML with GalNAc residues arranged on peptides, we prepared short synthetic glycopeptides corresponding to a portion of the tandem repeat domain of MUC2 mucin with different numbers of attached GalNAc residues. Molecular sizes were determined by MALDI-TOF MS, and the locations of the incorporated residues were determined by peptide sequencing. Kinetics of the binding were examined in both the fluid-solid phase with SPR and the fluid-fluid phase with FP; both equilibrium-binding assays of the interaction of rHML with immobilized glycopeptides revealed that the number of GalNAc residues does affect K D . Affinity of the G3peptide (PT GalNAc TTPIT GalNAc T GalNAc TT[K]) for HML was 11fold greater than that of the G1-peptide (PTTT GalNAc PITTTTK) and that of the G5-peptide was 38-fold greater; when K D values were normalized according to the number of GalNAc residues, the increases were 3.5-and 7.6-fold, respectively.
Differences in affinity for CRD fragments of G3 and G5 relative to G1, based on the molar amount of glycopeptide, were 2-and 6-fold, respectively; differences based on the number of GalNAc residues were 0.7-and 1.2-fold, respectively. Therefore, rHML fragments but not CRD fragments revealed a drastic increase in affinity that depended on the number of GalNAc residues present. The differences among G1-, G3-, and G5peptides should be due to the increased avidity of polyvalent ligands. This effect was more prominent with rHML than with CRD fragments due to the oligomeric nature of rHML, whereas CRD fragments apparently remained in monomeric in aqueous solutions (Fig. 1).
The G5-and G6-peptides were shown to contain three consecutive Thr-linked GalNAc residues, i.e. Tn antigen. Antitrimeric Tn mAb (26) was shown to bind these glycopeptides. It has already been shown that rat hepatic lectin, a type of trimeric C-type lectin, also bound preferentially to clustered Gal-NAc residues, as shown by using synthetic cluster glycosides based on Tris (17,18). Therefore, in addition to the impact of increased number of GalNAc residues, three consecutive Thr attached with GalNAc residues might also contribute to the high binding affinity for HML.
Dense immobilization of all of the synthetic glycopeptides leads to high affinity association. The difference between densely immobilized and sparsely immobilized glycopeptides seemed to be due to the dramatic decrease in the dissociation rates (Table I). Although the organization of the densely immobilized glycopeptides on the membranes was unknown, the binding of anti-trimetric Tn mAb strongly suggested that some GalNAc residues from separate glycopeptides were located very close to each other; thus, the clustering of GalNAc residues may well be one of the more important factors in the high affinity of HML (Fig. 6). Because the immobilization was done via biotinstreptavidin, it may be unconceivable that the distance of Gal-NAc would be as close as to mimic a single amino acid distance between them. Thus, the spacing among the carbohydrate chains may contribute to macromolecular ligand interactions with HML, as observed in amebic lectin binding (16). There may be another level at which clusters in macromolecular carbohydrate ligands contribute to high affinity (18). Simulation by x-ray crystallography (29) has revealed three binding sites of trimeric mannose binding protein spaced approximately 4.5 nm apart. If HML also accommodates such a structure, GalNAc residues spaced more sparsely than trimeric Tn might be required for viable ligand-carbohydrate interactions.
The results of FP equilibrium-binding assays, with one exception, were consistent with the findings from the experi-ments using the SPR biosensor, K D for G1-peptides from SPR (1.0 M for rHML) was higher than K D based on FP binding data (6.8 M for rHML and 8.3 M for CRD). The difference might be due to an increase in conformational stability on solid surfaces. It is also probable that the solid phase configuration was more appropriate for the formation of the multiple attachment sites preferred by oligomeric lectin than are glycopeptides in solutions. Collectively, slightly lower constants from FP assays in fluid-fluid phases than those from SPR measurements seemed to be reasonable. From FP assays conducted under ideally diluted conditions, K D values between monovalent lectin and GalNAc monosaccharide can be obtained by the Scatchard plot (Fig. 6). All K D values reported here were based on the monovalent ligand-monovalent receptor model. Such analysis was used because the binding of multivalent lectins to multivalent glycopeptides is inherently more complex than are antigen-antibody interactions, and thus it is difficult to simulate the factors involved in a simple equation (14). The results, however, fit well with the single-site binding model. It is possible, then, that, regardless of the valency, binding of rHML might be an "all-or-nothing" event with no intermediate forms, as discussed by Adler and co-workers (16). In microtiter assays, rHML bound only GalNAc ␣or ␤-monosaccharide-BPPs linked directly to the spacer arm on the polymer backbone. A GalNAc residue located away from the polymer backbone, such as those in the Forrsman disaccharide and the blood group A-type trisaccharide (Fig. 2), did not have strong affinity with HML. On the contrary, a low albeit significant affinity for HML by these oligosaccharides was detected through measurement on the SPR biosensor (data not shown). Such a discrepancy might be due to different steric hindrances in the access of this lectin to the binding sites using these two methods. The preferred ligand structure for HML binding was multiple GalNAc residues attached to linear backbones, regardless of the specific synthetic polymers or oligopeptides. As noted above, the Gal␤ residue attached to the same polymer neither bound directly to the lectin nor inhibited the binding of GalNAc to rHML. These results were consistent with our previous data, which showed that rHML bound neither N-linked carbohydrate chains with terminal Gal residues nor the Nterminal octapeptide of human glycophorin A with three consecutive desialized O-linked carbohydrate chains (4). Greater specificity for GalNAc than Gal was also observed with a human hepatic lectin (30), in contrast to a mouse macrophage lectin that preferentially bound Gal instead of GalNAc residues (31,32). Using chimeric and mutagenized CRD fragments of hepatic and macrophage lectin, amino acid residues likely to be involved in the selective binding of GalNAc in the human hepatic lectin were predicted (30) (His-256 and Asn-208). These amino acids apparently correspond to His-260 and Tyr-212 in HML, which may be the most important residue in the interaction between HML and GalNAc. HML could serve as an useful model to further investigate the nature of GalNAc-protein interactions.
When solutions of either rHML or CRD fragments were incubated in the presence of cross-linking agents, a homotrimer of rHML was the only major oligomeric lectin protein (Fig. 1). The CRD fragments lacking the HML stalk region did not appear to form oligomers. Other members of the C-type lectin family, including CD23, CD72, rat asialoglycoprotein receptor, chicken hepatic receptor, Kupffer cell receptor, and mannosebinding protein, are also known to form oligomers (12,29). The primary amino acid sequence of the stalk region of rHML has a structure known to be involved in oligomer formation. When the amino acid sequence of the stalk region (from Ile-87 to Glu-153) deduced from the nucleotide sequence was arranged in order to accommodate two heptads (a, b, c, d, e, f, and g), hydrophobic residues were arranged at positions a and d in a way that should result in an ␣-helix coiled-coil. Thus, HML forms a trimeric configuration and would have high affinity for ligands with GalNAc residues such as in the case of CD23 with IgE (13).
We have determined affinity constants between HML and peptides with a single GalNAc residue using FP equilibriumbinding assays. A carbohydrate density-dependent increase of affinity for the lectin was revealed, and the increase was more prominent with intact rHML than with CRD fragments. In the present study, variations in the distance between multiple GalNAc residues on a single peptide were not systematically investigated. More dispersed arrangements of GalNAc residues than those used in this study may generate more preferable configurations as functional ligands for HML, and therefore this will be an important subject for future studies.