Purification and characterization of a Neu5Acalpha2-6Galbeta1-4Glc/GlcNAc-specific lectin from the fruiting body of the polypore mushroom Polyporus squamosus.

A lectin has been purified from the carpophores of the mushroom Polyporus squamosus by a combination of affinity chromatography on beta-D-galactosyl-Synsorb and ion-exchange chromatography on DEAE-Sephacel. Gel filtration chromatography, SDS-polyacrylamide gel electrophoresis, and N-terminal amino acid sequencing indicated that the native lectin, designated P. squamosus agglutinin, is composed of two identical 28-kDa subunits associated by noncovalent bonds. P. squamosus agglutinin agglutinated human A, B, and O and rabbit red blood cells but precipitated only with human alpha(2)-macroglobulin, of many glycoproteins and polysaccharides tested. The detailed carbohydrate binding properties of the purified lectin were elucidated using three different approaches, i.e. precipitation inhibition assay (in solution binding assay), fluorescence quenching studies, and glycolipid binding by lectin staining on high-performance thin layer chromatography (solid-phase binding assay). Based on the results obtained by these assays, we conclude that although the P. squamosus lectin binds beta-D-galactosides, it has an extended carbohydrate-combining site that exhibits highest specificity and affinity toward nonreducing terminal Neu5Acalpha2, 6Galbeta1,4Glc/GlcNAc (6'-sialylated type II chain) of N-glycans (2000-fold stronger than toward galactose). The strict specificity of the lectin for alpha2,6-linked sialic acid renders this lectin a valuable tool for glycobiological studies in biomedical and cancer research.

Lectins are proteins (or glycoproteins), other than antibodies and enzymes, that bind specifically and reversibly to carbohydrates, resulting in cell agglutination or precipitation of glycoconjugates (1). They are ubiquitous in the biosphere, having been found in viruses, bacteria, fungi, plants, and animals (2).
Lectins of known specificity recognizing sialic acid serve as valuable reagents in glycobiological research. They can be employed for the detection and preliminary characterization of sialic acid-containing glycoconjugates on the surface of cells and for assaying the incorporation of sialic acid into complex carbohydrates in biosynthetic studies. In their immobilized form, these lectins can be used for the resolution and isolation of sialic acid-containing glycoconjugates. Lectins are found in greatest quantity and are most readily purified from plant sources, especially higher plants, although relatively few sialic acid-binding lectins have been identified in the plant world (including fungi), which lacks sialic acid.
During the last decade, there has been a growing interest in fungal lectins, largely due to the discovery that some of these lectins exhibit antitumor activities, e.g. Volvariella volvacea lectin shows antitumor activity against sarcoma S-180 cells (3), Grifola frondosa lectin is cytotoxic to Hela cells (4), Agaricus bisporus lectin possesses antiproliferation activities against human colon cancer cell lines HT29, breast cancer cell lines MCF-7 (5), and Tricholoma mongolicum lectin inhibits mouse mastocytoma P815 cells in vitro and sarcoma S-180 cells in vivo (6). Fungal lectins have recently been reviewed (7)(8)(9). However, apart from the lectin from A. bisporus, which binds to Gal␤1,3GalNAc␣-Ser/Thr (T-disaccharide) (10), the detailed carbohydrate specificities of these fungal lectins have not been investigated in depth.
We report herein the purification and characterization of Polyporus squamosus lectin (designated PSA), a Neu5Ac␣2-6Gal␤1-4Glc/GlcNAc-specific lectin present in the carpophores (fruiting bodies) of this member of the Polyporaceae family.

MATERIALS AND METHODS
Carpophores of P. squamosus (Huds.) Fr. were collected in late summer 1998 from a decaying Ulmus stump in Ann Arbor, Michigan. A voucher specimen (Goldstein (MICH) 27953) was deposited in the University of Michigan herbarium.
Unless stated otherwise, saccharides, their derivatives, and glycoproteins (including fetuin, asialofetuin, transferrin, thyroglobulin, ␣ 2macroglobulin, ␣ 1 -acid glycoprotein, bovine mucin, etc.) were purchased from Sigma. Ovine submaxillary mucin was a gift of Dr. R. N. Knibbs (University of Michigan). Except for asialofetuin, asialoglycoproteins were prepared by heating the corresponding native glycoproteins in 0.1 M hydrochloric acid at 80°C for 1 h, followed by dialysis and lyophilization; the removal of sialic acid was confirmed by the thiobarbituric acid assay (11).
Purification of the Lectin-All procedures were conducted at 4°C. The fruiting bodies from P. squamosus (fresh weight, 80 g) were homogenized and extracted overnight with 400 ml of extraction buffer (10 mM sodium phosphate, 0.15 M NaCl, 0.135 mM CaCl 2 , 0.04% sodium azide, pH 7.2 (PBS), containing 10 mM thiourea, 0.25 mM phenylmethylsulfonyl fluoride, and 1 g/liter ascorbic acid). The homogenate was squeezed through two layers of cheesecloth and centrifuged at 10,000 ϫ g for 30 min. To the supernatant solution was added solid ammonium sulfate to 25% saturation. After standing overnight, the precipitate was removed by centrifugation, and the supernatant solution was applied directly onto a ␤-D-galactosyl-Synsorb 100 column (18 ϫ 1.2 cm; bed volume, 80 ml), which had been equilibrated with 10 mM PBS (pH 7.2) containing 1 M ammonium sulfate. The column was washed with the same solution until the absorbance at 280 nm of the effluent had fallen below 0.01. The affinity-adsorbed lectin was desorbed with 0.2 M lactose in 10 mM PBS, collected, dialyzed extensively against distilled water, and lyophilized (designated crude PSA). Approximately 18 mg of crude lectin was obtained from 80 g of fresh fruiting body.
Crude lectin was reconstituted in 50 mM phosphate buffer, pH 7.8, and percolated slowly through a DEAE-Sephacel column (17 ϫ 1.2 cm, bed volume 75 ml) preequilibrated and eluted with the same buffer (50 mM phosphate buffer, pH 7.8). The elution was monitored by absorbance at 280 nm until it become negligible, whereupon the adsorbed protein was eluted with 1 M sodium chloride. Both the DEAE-unbound (pass through) and DEAE-bound peak fractions were collected, dialyzed against distilled water, lyophilized, reconstituted in 10 mM PBS, pH 7.2, and tested for electrophoretic homogeneity and agglutination activity.
Protein Estimation-Protein concentration was determined by the method of Lowry et al. (12), using bovine serum albumin as a standard.
PAGE and SDS-PAGE-Native gel electrophoresis using a 12.5% slab gel was carried out in alkaline buffer system (Tris/glycine, pH 8.3) (13). SDS-PAGE in the presence and absence of ␤-mercaptoethanol using a 12.5% slab gel was conducted in Tris/tricine buffer system as described by Schagger and von Jagow (14).
Hemagglutination Assay-The hemagglutinating activity of the lectin was determined by a 2-fold serial dilution procedure using formaldehyde-treated (15) human and rabbit erythrocytes as described previously (16). The hemagglutination titer was defined as the reciprocal of the highest dilution still exhibiting hemagglutination.
Molecular Mass and Molecular Structure-The molecular mass and molecular structure of the purified PSA was determined by gel filtration and SDS-PAGE performed in the presence and absence of ␤-mercaptoethanol.
Gel filtration chromatography of the lectin was carried out on a Bio-Gel P-150 column (1.45 ϫ 120 cm; bed volume, 198 ml) operating at room temperature in PBS, pH 7.2, with or without 0.2 M lactose, at a flow rate of 10 ml/h. Fractions of 10 min/tube (approximately 1.7 ml/ tube) were collected and monitored at A 280. The column was calibrated with the following standard proteins: bovine ␥-globulin (158 kDa), bovine serum albumin (67 kDa), chicken ovalbumin (45 kDa), equine myoglobin (17 kDa), and vitamin B-12 (1.35 kDa). Blue dextran 2000 was used for determination of the void volume of the column.
Amino Acid Composition Analysis and N-terminal Sequence Analysis-The amino acid composition and the N-terminal sequence of the purified lectin were analyzed by the Protein and Carbohydrate Structure Core facility on this campus as described previously (17,18). Tryptophan was estimated spectrophotometrically in 6 M guanidine hydrochloride (19). Sulfhydryl groups were estimated by release of 2-nitro-5-mercaptobenzoic acid (⑀ 0 ϭ 13, 600 M Ϫ1 cm Ϫ1 at 410 nm) from 5,5Ј-dithiobis(2-nitrobenzoic acid).
Quantitative Precipitation and Hapten Inhibition Assays-Quantitative precipitation assays were performed by a microprecipitation technique as described previously (17). Briefly, varying amounts of glycoproteins or polysaccharides, ranging from 0 to 100 g, were added to 10 g of purified PSA in a total volume of 160 l of PBS, pH 7.2. After incubation at 37°C for 1 h, the reaction mixtures were stored at 4°C for 48 h. The precipitates formed were centrifuged, washed three times with 150 l of ice-cold PBS, dissolved in 0.05 M NaOH, and determined for protein content by Lowry's method using bovine serum albumin as standard.
For hapten inhibition assays, increasing amounts of various haptenic saccharides were added to the reaction mixture consisting of 10 g of the purified lectin and 5 g of ␣ 2 -macroglobulin in a final volume of 160 l of PBS, pH 7.2. After incubation at 37°C for 1 h and storage at 4°C for 48 h, the precipitated proteins were centrifuged, washed, and determined. The percentage of inhibition was calculated, and inhibition curves were constructed. The minimum concentration of each haptenic sugar required for 50% inhibition was obtained from corresponding complete inhibition curves. Preparation of Biotin-Lectin Conjugate-The biotinylation of the purified lectin was achieved using EZ-Link NHS-LC-Biotin according to the manufacturer's instructions, except that 0.2 M lactose was added to the reaction mixture to protect the carbohydrate-binding sites. After coupling, the lectin activity was ascertained by hemagglutination assay.
HPTLC of Glycolipids and Lectin Staining-Two identical sets of glycolipids were separated chromatographically in parallel on the same aluminum-backed silica gel 60 HPTLC plate (Merck, Darmstadt, Germany) using chloroform/methanol/water (65:25:4, by volume) for neutral glycolipids or chloroform/methanol/aqueous 0.25% KCl (50:40:10, by volume) for gangliosides as developing solvent. The reference chromatogram was chemically visualized by spraying the plate with orcinol reagent.
The lectin staining was performed as follows: after drying, the plates were blocked by overlaying with 1% gelatin in PBS containing 0.1% NaN 3 and incubating overnight at room temperature. The plates were then overlaid with biotin-labeled lectin diluted in PBS containing 1% bovine serum albumin and 0.05% Tween 20, and incubated at 37°C for 2 h. After rinsing the plates five times with 10 mM Tris/HCl buffer, pH 9.5, containing 0.05% Tween 20, they were overlaid with alkaline phosphatase-streptavidin diluted with 0.1 M Tris/HCl, pH 9.5, containing 100 mM NaCl and 5 mM MgCl 2 , incubated at 37°C for 1 h, washed five times with the same buffer, and finally visualized with 5-bromo-4chloro-3-indolylphosphate-nitroblue tetrazolium chloride.
␣2,6-Sialylation of Asialofetuin-Asialofetuin, 2.5 mg (40 nmol) in 1.0 ml of 50 mM MES buffer, pH 6.0, containing 10 mM MnCl 2, 0.15M NaCl, 10% glycerol, 2 mol of CMP-sialic acid, and 10 milliunits of rat liver ␣2,6-sialyltransferase (Roche Molecular Biochemicals, 8 units/mg of protein) was incubated 24 h at 37 o , followed by addition of another 1 mol of CMP-sialic acid and 5 milliunits of sialyltransferase. After an additional 24 h of incubation, the reaction mixture was passed through a column (1 ϫ 15 cm) of BioGel P-10. Fractions containing protein were pooled, dialyzed against distilled water, and lyophilized. The resialylated fetuin contained 3.1 mol of sialic acid/mol of protein, compared with a negligible amount (Ͻ0.1 mol of sialic acid/mol of protein) in the asialofetuin and approximately 10 mol of sialic acid/mol of protein in native fetuin.
Absorbance and Fluorescence Spectra-Ultraviolet absorbance spectra were recorded on a Shimadzu model UV160U spectrophotometer; fluorescence spectra were recorded on an ISA JodinYvon-Spex Fluoromax-2 spectrofluorometer.

RESULTS
Hemagglutinating Activity-In a survey of mushrooms for hemagglutination activity by Pemberton (20), it was found that the extract of P. squamosus exhibited strong hemolytic activity. In the present study, by using hemolysis-resistant, formaldehyde-treated erythrocytes, we observed that the crude extract of P. squamosus contained a lectin(s) that agglutinated rabbit and human erythrocytes, irrespective of blood group type (type B erythrocytes were agglutinated slightly better than those of types A and O). The hemagglutinating activity was inhibited by D-galactose and D-galactose-related carbohydrates, such as Dfucose, L-arabinose, melibiose, and lactose.
Purification of the Lectin-Because the hemagglutination activity of the crude extract of P. squamosus was inhibited by D-galactose and lactose, ␤-D-galactosyl-Synsorb was used as an affinity absorbent for isolation of the lectin. As shown in Fig. 1, upon nondenaturing PAGE at pH 8.3, the lectin preparation obtained from affinity chromatography on ␤-D-galactosyl-Synsorb showed two protein bands, which we later found could be separated from each other by ion exchange chromatography on DEAE-Sephacel in 0.05 M phosphate buffer, pH 7.8 ( Fig. 1). Under these conditions, the major portion of the lectin activity was not retained on the DEAE column. This DEAE-unbound fraction was electrophoretically pure (designated purified PSA). The minimal concentration of the purified lectin required for the agglutination of formaldehyde-treated human type B erythrocytes was 0.6 g/ml. Removal of divalent metal ions by extensive dialysis of this fraction against buffer containing 1.25 mM EDTA had little effect on its agglutination activity. The DEAE-bound fraction exhibited about 30 times lower agglutinating activity, 19 g/ml being required for agglutination; therefore, in the present study, characterization of only the DEAE-unbound fraction was pursued.
Molecular Mass and Molecular Structure-Upon gel filtration chromatography on Bio-Gel P-150, the purified lectin eluted as a single, symmetric peak, irrespective of the presence of 0.2 M lactose, at an elution volume corresponding to an apparent molecular mass of 52 kDa (not shown). On the other hand, upon SDS-PAGE, with or without ␤-mercaptoethanol, purified PSA gave a single band with an apparent mass of 28 kDa (Fig. 2). Taken together, these data suggest that at neutral pH, the lectin exists as a homodimer of 28-kDa subunits associated by noncovalent bonds. No neutral carbohydrate was detected using the phenol-sulfuric acid assay.
Amino Acid Composition and N-terminal Amino Acid Sequence-As shown in Table I, purified PSA contains an extremely high proportion of hydrophobic amino acids (Ala, Ile, Leu, Val, and Phe) that account for one-third of the total amino acids, high contents of acidic and hydroxyl amino acids (Asx and Glx account for 22%; Ser and Thr, 14%), and relatively high amounts (11.5%) of aromatic amino acids, accounting for the high absorbency at 280 nm of the lectin (A 1% ϭ 29.0). The lectin also contains two residues each of methionine and cysteine. No free sulfhydryl groups were detected by reaction with 5,5Јdithiobis(2-nitrobenzoic acid); together with the observation that ␤-mercaptoethanol has no effect on SDS-PAGE migration, the presence of an intrachain disulfide linkage is indicated. A single N-terminal amino acid sequence, H 2 N ϩ -PFEGHGIY-HIPSVNTANVRI, was determined. A search of the protein data base revealed no significant homology of this N-terminal sequence to any sequence in the data base.
Quantitative Precipitation and Precipitation Inhibition-Inasmuch as the hemagglutinating activity of the crude extract of P. squamosus was specifically inhibited by galactose and the lectin was initially isolated by affinity chromatography on ␤-Dgalactosyl-Synsorb, various glycoproteins were chemically desialylated to expose their penultimate D-galactosyl residues and tested for their ability to precipitate the lectin. However, none of these asialoglycoproteins formed a detectable precipi-tate with the lectin. The galactomannan from Cassia alata, which contains multiple terminal ␣-D-galactosyl residues, also failed to precipitate with the lectin. To our great surprise, of the many native glycoproteins tested, including fetuin, transferrin, thyroglobulin, ␣ 1 -acid glycoprotein, bovine mucin, and ovine submaxillary mucin, only human ␣ 2 -macroglobulin, but not its desialylated form, gave a pronounced precipitation reaction with the lectin. Therefore, human ␣ 2 -macroglobulin was employed as a precipitant in the inhibition assays.  The results of sugar hapten inhibition are shown in Table II. Among the monosaccharides tested, only D-galactose, its derivatives, and D-galactose-related carbohydrates (i.e. D-fucose and L-arabinose) were inhibitory, whereas epimers of D-galactose (i.e. D-talose (C-2 epimer), D-gulose (C-3 epimer), and D-glucose (C-4 epimer)) were all noninhibitory up to 100 mM. However, the most striking observation was that both Neu5Ac␣2, 6Gal␤1,4Glc (6Ј-sialyllactose) and Neu5Ac␣2,6Gal␤1,4GlcNAc (6Ј-sialylLacNAc), but not their ␣2,3 isomers, were very strong inhibitors, being 2000-fold more inhibitory than D-galactose and 250 -300 times stronger than lactose and LacNAc. Neither free N-acetylneuraminic acid, its p-nitrophenyl glycoside, nor its ␣2,8-linked polymer (colominic acid) was inhibitory.
HPTLC of Glycolipids and Lectin Staining-A panel of neutral glycosphingolipids and gangliosides with well defined carbohydrate structures (Table III) was also used to investigate the binding specificity of purified PSA. The chromatograms are shown in Fig. 3. Of the neutral glycosphingolipids tested, only lactosylceramide was bound by the lectin. This result is in good agreement with the results obtained by precipitation inhibition assays, in which lactose is a fairly good inhibitor of the PSA/ ␣ 2 -macroglobulin precipitation reaction, but oligosaccharides similar to those found in the other neutral glycolipids were not inhibitory. On the other hand, all gangliosides tested failed to react with the lectin. It is especially noteworthy that GM3, which contains the lactosylceramide moiety but is substituted by ␣2,3-linked Neu5Ac at the ␤-linked galactosyl residue, also failed to react with the lectin, further confirming that this lectin is exclusively specific for ␣2,6-linked Neu5Ac.
Spectral and Fluorescence Studies-Changes in absorbance or fluorescence spectra were examined as probes for binding studies. It had been observed that p-nitrophenyl ␣-D-mannoside, upon binding to the mannose/glucose-specific lectin concanavalin A, undergoes a small but definite spectral change with a maximum decrease in absorbance at 317 nm, apparently due to interaction of an aromatic residue with the chromophore (21). Accordingly, we looked for such a spectral change with both pNP␤Gal and pNP␤Lac upon reaction with PSA, but none was observed (data not shown). Another indication of the interaction of an aromatic aglycon with residues in the lectin is the complete quenching of fluorescence of methylumbelliferyl ␣-D-mannoside (excitation at 320 nm, emission peak at 375 nm) upon binding to concanavalin A (22). Again, however, fluorescence of methylumbelliferyl ␤-D-galactoside was unaffected by titration with PSA, suggesting that an aromatic residue of the protein is not in a position near the binding site to interact with the chromophore or fluorophore of these ␤-galactosides.
We also examined the effects of ligands on the intrinsic fluorescence of tryptophanyl residues of the protein, as measured by excitation at 280 nm and emission at 300 -400 nm. Free monosaccharides, free oligosaccharides, or methyl glycosides had no effect on the intrinsic fluorescence peak at 340 nm. As expected because of its strong UV absorbance, all p-nitrophenyl glycosides quenched intrinsic fluorescence and caused a slight shift in the emission peak toward longer wavelengths. However, we observed that quenching by pNP␤Gal or pNP␤Lac was significantly greater than that caused by a nonreactive p-nitrophenyl glycoside, such as p-nitrophenyl ␣-D-mannoside, at the same concentration. Furthermore, addition of competing oligosaccharides, such as lactose or LacNAc, reversed this additional quenching in a saturable manner, indicating that it is caused by the p-nitrophenyl ␤-galactosides binding in the vicinity of a fluorophoric group on the lectin. We thus refer to this phenomenon as "specific quenching." Thus, intrinsic fluorescence titration of oligosaccharide ligands in the presence of a fixed amount of pNP␤Gal yielded apparent inhibition constants (IC 50 values) for the titrant from a plot of 1/⌬F versus 1/[L], where ⌬F is the increase in peak fluorescence caused by the titrant at a concentration of [L], in a manner analogous to Lineweaver-Burk plots. The ordinate intercept gives the reciprocal of the maximum fluorescence change ⌬F max (in arbitrary fluorescence units), which is a function of concentration and binding affinity of the quenching probe. Table IV summarizes data for several oligosaccharides and chromophoric probes. At a lectin concentration of 0.09 mg/ml (3.2 M in monomers) and pNP␤Gal of 100 M, the maximum fluorescence change was approximately 120,000 units, as compared with a total quenching of about 800,000, or about 15% of the total quenching. The use of pNP␣Gal at the same concentration also caused a specific quenching signal, but the ⌬F max was only about 40,000 units, consistent with the weaker interaction of the lectin with ␣-galactosides.

DISCUSSION
By definition, a lectin is a sugar-binding protein or glycoprotein of nonimmune origin that agglutinates cells and/or precipitates glycoconjugates (1). In order to form detectable precipitate, however, both the lectin and the glycoconjugate must be multivalent as well as in an appropriate stoichiometric ratio. a N-Acetyl-glucosamine, D-arabinose, 2-deoxy-ribose, L-fucose, D-mannose, L-and D-rhamnose, L-and D-ribose, L-and D-xylose, cellobiose, chitobiose, gentiobiose, maltose, isomaltose, sucrose and trehalose were all noninhibitory up to 200 mM (cellobiose to saturation).
b Minimum concentration required for 50% inhibition of the PSA/␣ 2macroglobulin precipitation reaction, unless otherwise noted.
c Maximum concentration tested (percentage of inhibition observed).
Of all the sialoglycoproteins we assayed for their ability to form a precipitate with the P. squamosus agglutinin, only human ␣ 2 -macroglobulin precipitated the lectin. Human ␣ 2 -macroglobulin is composed of four identical subunits, each of which contains 1451 amino acid residues and eight N-linked sugar chains (23) that are 70% sialylated, exclusively in Neu5Ac␣2,6 linkage (24). On the other hand, most native glycoproteins, including ␣ 1 -acid glycoprotein, fetuin, transferrin, and thyroglobulin, contain both ␣2,3and ␣2,6-linked N-acetylneuraminic acids. In addition to the prevalence of ␣2,3-linked sialic acids at the nonreducing termini of these glycoproteins, many ␣2,6-linked Neu5Ac residues occur not at the nonreducing termini, but rather attached to the penultimate GlcNAc residues of the tri-and tetraantennary structures, thus precluding lectin binding. It is noteworthy that these ␣2,6-linked sialic acid groups are also resistant to cleavage by neuraminidase from Clostridium perfringens (25). To further confirm the specificity of PSA for terminal ␣2,6-linked Neu5Ac, we resialylated asialofetuin using rat liver ␣2,6-sialyltransferase, which sialylates specifically at the 6-position of nonreducing terminal ␤1,4linked galactose (26). Despite incorporating only 3 Neu5Ac groups per molecule, this resialylated fetuin precipitated PSA much more strongly than either asialofetuin or fully sialylated native fetuin.
Taking these facts into consideration, it is understandable why the native glycoproteins tested, except for ␣ 2 -macroglobulin, failed to precipitate with PSA. In fact, we observed that these native glycoproteins inhibited the PSA/␣ 2 -macroglobulin precipitation reaction (data not shown), indicating that they do contain a few nonreducing terminal Neu5Ac␣2,6Gal␤1,4Glc/ GlcNAc structures, but not a sufficient number to act as a multivalent precipitating ligand for the lectin.
To elucidate the detailed carbohydrate binding specificity of purified PSA, precipitation inhibition assays were carried out using ␣ 2 -macroglobulin as the precipitant. The data in Table II permit some conclusions to be drawn regarding the specificity of the lectin. The 2-, 3-, and 4-hydroxyl groups in the galactopyranose ring are evidently important in binding, because the epimers in these positions (D-talose, D-gulose, and D-glucose, respectively) are unreactive. The reactivity of ␤-D-galactopyranosides, compared with free D-galactose, indicates a preference for the ␤-D-pyranose form, which is predominantly in the 4 C 1 chair conformation (27). Because D-gulose and D-glucose in solution equilibrium are 60 -70% in the same conformation and anomeric configuration, their lack of reactivity indicates that the 3-and 4-hydroxyl groups in the equatorial and axial epimer, respectively, are essential for binding. The conclusion regarding the 2-epimer, D-talose is less clear; it exists in the ␤-D-pyranose form only to the extent of 29% (27), and the two chair conformations are approximately energetically equivalent. However, its complete lack of reactivity at 100 mM, together with the poor reactivity of 2-deoxy-D-galactose and Gal-NAc, indicate that the 2-hydroxyl group is also important in binding. The fact that D-fucose (i.e. 6-deoxy galactose) and 6-O-methyl-D-galactose are nearly equal to D-galactose in inhibitory potentcy suggests that the 6-hydroxyl group is not involved, either as a hydrogen bond donor or acceptor. However, the poor reactivities of L-arabinose and methyl ␤-L-arabinopyranoside are enigmatic. The free pentose in solution is approximately 61% in the ␣-L-pyranose form (27), which is homomorphous with ␤-D-galactose except for the lack of the hydroxymethyl group on C-5. This suggests that the methylene or methyl carbon (C-6) of D-galactose or D-fucose, respectively, contributes significantly to the binding. However, the effect of the hydroxylmethyl group on C-5 may be indirect, e.g. by its stabilizing effect of the 4 C 1 chair conformation of the pyranose ring.
Both methyl ␤-D-galactopyranoside and p-nitrophenyl ␤-Dgalactopyranoside are three times more potent inhibitors than the corresponding ␣-glycosides, as indicated both by precipitin inhibition and extent of specific fluorescence quenching, suggesting that PSA has an anomeric preference for the ␤-configuration. This ␤-anomeric preference is further confirmed by the oligosaccharides tested, of which most with nonreducing terminal ␤-galactosyl residues (e.g. lactose and N-acetyllactosamine) are 3-8-fold more potent inhibitors than D-galactose and 10 -20-fold better than ␣-galactosyl-terminated disaccharides (e.g. melibiose and raffinose). A major exception is Gal␤1,4Gal, which is a relatively poor inhibitor. This could be explained by the fact that in Gal␤1,4Gal, the nonreducing terminal galactosyl residue is ␤-linked to an axial C-4 oxygen atom, causing the two hexose residues to lie at a considerably sharper planar angle to each other than in equatorial oxygen-linked disaccharides, such as lactose, LacNAc, and Gal␤1,4Man, increasing the possibility of steric hindrance of the hydroxyl groups involved in binding. The corresponding ␣-digalactoside is also a very poor inhibitor, probably because of the bulk of the substituent in the ␣-position. In neither case could the reducing galactose moiety occupy the binding site, because of the necessity of a free C-4 hydroxyl group.
The substitution of the C-6 hydroxyl group of the nonreducing terminal ␤-galactosyl residue with ␣ 2,6-linked N-acetylneuraminic acid increased the inhibitory potency by 3 orders of magnitude (2000-fold) as compared with galactose and 250 - 300 times compared with the parent sugars (i.e. lactose and LacNAc). However, free N-acetylneuraminic acid, p-nitrophenyl ␣-sialoside, and the Neu5Ac␣ 2,8 polymer colominic acid did not react, whereas galactose derivatives having an acidic function near the C-6 position (D-galacturonic acid, D-galactose 6-sulfate, and D-galactose 6-phosphate) are moderate to very weak inhibitors. Furthermore, the addition of sialic acid to the C-3 position of the nonreducing terminal galactose abolished the lectin binding almost totally. This is in contrast to S. nigra agglutinin, a plant lectin isolated from elderberry (Sambucus nigra) bark, which recognizes not only the terminal Neu5Ac␣2,6Gal/GalNAc sequence but also the ␣2,3-linked isomer, to a 100-fold smaller extent (28). The exclusive specificity of PSA is consistent with the C-3 equatorial hydroxyl group of the terminal galactosyl group being a critical locus, so that its substitution abrogates the lectin binding, probably due to steric effects. However, the C-6 hydroxyl group of the D-pyranose ring is not involved in lectin binding, so that substitution at that position of the galactose is tolerated. Furthermore, sialylation at the C-6 position evidently creates additional interactions via hydrogen bonds and/or charge interactions between the carboxylate group and positively charged amino acids in the vicinity of the carbohydrate-binding site of the lectin. However, because free sialic acid, or sialosides, its polymer, or its 2,3-substituted form does not react, the lectin does not appear to have an independent binding site for sialic acid and cannot be considered as an N-acetylneuraminic acid-binding lectin. A lectin with similar affinity was isolated from tubers of Trichosanthes japonica (Cucurbitaceae) (29). That lectin also recognized ␤-galactosyl residues and was greatly enhanced by ␣2,6-sialylation but blocked by ␣2,3 sialylation. However, that lectin did not react with the Gal␤1,3GlcNAc group, showed tolerance for C-2 epimerization of the galactose (i.e. D-talose), and had little or no preference for ␤-galactosides versus ␣-galactosides. Those workers also observed that 6-sulfated lactose and 6Ј-sialyllacto-N-neotetraose strongly interacted with the lectin. We have tested neither of these latter compounds, but would expect them to react at least as well as lactose and 6Ј-sialyllactose, respectively. We did, however, observe that galactose 6-sulfate is equivalent in its inhibitory potency to galactose. On the other hand, a lectin isolated from a related polypore mushroom, Laetiporus sulfureus, appears to be completely different in its molecular structure, amino acid composition, and carbohydrate binding specificity (30).
FIG. 3. HPTLC of glycolipids and lectin staining. Two identical sets of glycolipids were separated in parallel on the same aluminumbacked silica gel 60 HPTLC plate. Solvents were chloroform/methanol/ water (65:25:4, by volume) (A) and chloroform/methanol/0.25% KCl (50:40:10, by volume) (B). The reference chromatograms were visualized chemically by orcinol reagent; lectin staining was conducted as described under "Materials and Methods." A, neutral glycosphingolipids (for structures, see Table III). Lane 1, lactosyl ceramide; lane 2, globotriosyl ceramide (note: the two bands are due to heterogeneity of the lipid moiety; the upper band containes nonhydroxy fatty acid side chains, and the lower band contains mostly hydroxylated fatty acid side chains); lane 3, globotetraosyl ceramide; lane 4, globopentaosyl ceramide; lanes 5 and 6, mixture of the four glycolipids. Lanes 1-5 were detected by orcinol reagent, and lane 6 was stained with biotinylated PSA. Among neutral glycolipids tested, only lactosyl ceramide reacted with the lectin. B, acidic glycosphingolipids (gangliosides; see Table III for structures). Lanes 1 and 3, monosialoganglioside mixture including GM3, GM2, and GM1; lanes 2 and 4, disialoganglioside mixture containing GD3, GD1a, and GD1b. Lanes 1 and 2 were visualized by orcinol spray, and lanes 3 and 4 were stained with biotinylated PSA. None of the gangliosides tested (all have ␣2,3-linked Neu5Ac, but lack ␣2,6linked Neu5Ac) were bound by the lectin, further confirming the results obtained by precipitation inhibition assays that PSA is exclusively specific for ␣2,6-linked Neu5Ac. Of equal importance is the finding that ovine submaxillary mucin, which contains a prodigious number of Neu5Ac␣2, 6GalNAc␣1-Ser/Thr moieties per molecule, is neither a precipitant nor an inhibitor of PSA, in contrast to S. nigra agglutinin, which requires only a sialyl␣2,6Gal/GalNAc moiety (28). This observation is consistent with the preference of PSA for three structural features lacking in the mucin O-linked glycans: a ␤-galactosidic linkage (cf. Gal␤OMe versus Gal␣OMe), the necessity of a free equatorial 2-OH group on the galactosyl residue (cf. galactose versus GalNAc, 2-deoxy galactose, or D-talose), and the preference for an additional sugar at the ␤-galactoside linkage (cf. lactose or LacNAc versus Gal␤OMe).
The relative inhibitory constants of oligosaccharides tested in the specific fluorescence quenching assay (Table IV) are consistent with those observed by precipitin inhibition (Table  II), although absolute values are lower in the former case. This difference is understandable, because the reporter ligand in the fluorescence assay is monovalent, whereas in the precipitin assay, it is a polyvalent ligand leading to precipitation of an extensive network of lectin and glycoprotein molecules, requiring higher concentrations of a given competitive monovalent ligand to cause inhibition. Preliminary measurements of ligand binding using isothermal titration calorimetry showed PSA to bind methyl-␤-galactoside, lactose, and Neu5Ac␣2,6Lac with K d values of 393, 92, and 0.59 M, respectively (data not shown), values that are also 30 -40-fold lower than the IC 50 values for precipitation inhibition (Table II) but comparable to K i values for specific fluorescence quenching (Table IV). Calorimetric studies with PSA, as well as further studies of this specific fluorescence quenching phenomenon, which we have observed with several lectins, are in progress and will be published subsequently.