Structures of Sialylated O-Linked Oligosaccharides of Bovine Peripheral Nerve α-Dystroglycan THE ROLE OF A NOVEL O-MANNOSYL-TYPE OLIGOSACCHARIDE IN THE BINDING OF α-DYSTROGLYCAN WITH LAMININ

α-Dystroglycan is a heavily glycosylated protein, which is localized on the Schwann cell membrane as well as the sarcolemma, and links the transmembrane protein β-dystroglycan to laminin in the extracellular matrix. We have shown previously that sialidase treatment, but not N-glycanase treatment, of bovine peripheral nerve α-dystroglycan greatly reduces its binding activity to laminin, suggesting that the sialic acid of O-glycosidically-linked oligosaccharides may be essential for this binding. In this report, we analyzed the structures of the sialylated O-linked oligosaccharides of bovine peripheral nerve α-dystroglycan by two methods. O-Glycosidically-linked oligosaccharides were liberated by alkaline-borotritide treatment or by mild hydrazinolysis followed by 2-aminobenzamide-derivatization. Acidic fractions obtained by anion exchange column chromatography that eluted at a position corresponding to monosialylated oligosaccharides were converted to neutral oligosaccharides by exhaustive sialidase digestion. The sialidases from Arthrobacter ureafaciens and from Newcastle disease virus resulted in the same degree of hydrolysis. The neutral oligosaccharide fraction, thus obtained, gave a major peak with a mobility of 3.8-3.9 glucose units upon gel filtration, and its reducing terminus was identified as a mannose derivative. Based on the results of sequential exoglycosidase digestion, lectin column chromatography, and reversed-phase high-performance liquid chromatography, we concluded that the major sialylated O-glycosidically-linked oligosaccharide of the α-dystroglycan was a novel O-mannosyl-type oligosaccharide, the structure of which was Siaα2-3Galβ1-4GlcNAcβ1-2Man-Ser/Thr (where Sia is sialic acid). This oligosaccharide constituted at least 66% of the sialylated O-linked sugar chains. Furthermore, a laminin binding inhibition study suggested that the sialyl N-acetyllactosamine moiety of this sugar chain was involved in the interaction of the α-dystroglycan with laminin.

␣-Dystroglycan is a heavily glycosylated protein, which is localized on the Schwann cell membrane as well as the sarcolemma, and links the transmembrane protein ␤-dystroglycan to laminin in the extracellular matrix. We have shown previously that sialidase treatment, but not N-glycanase treatment, of bovine peripheral nerve ␣-dystroglycan greatly reduces its binding activity to laminin, suggesting that the sialic acid of O-glycosidically-linked oligosaccharides may be essential for this binding. In this report, we analyzed the structures of the sialylated O-linked oligosaccharides of bovine peripheral nerve ␣-dystroglycan by two methods. O-Glycosidically-linked oligosaccharides were liberated by alkaline-borotritide treatment or by mild hydrazinolysis followed by 2-aminobenzamide-derivatization. Acidic fractions obtained by anion exchange column chromatography that eluted at a position corresponding to monosialylated oligosaccharides were converted to neutral oligosaccharides by exhaustive sialidase digestion. The sialidases from Arthrobacter ureafaciens and from Newcastle disease virus resulted in the same degree of hydrolysis. The neutral oligosaccharide fraction, thus obtained, gave a major peak with a mobility of 3.8 -3.9 glucose units upon gel filtration, and its reducing terminus was identified as a mannose derivative. Based on the results of sequential exoglycosidase digestion, lectin column chromatography, and reversed-phase high-performance liquid chromatography, we concluded that the major sialylated O-glycosidically-linked oligosaccharide of the ␣-dystroglycan was a novel O-mannosyl-type oligosaccharide, the structure of which was Sia␣2-3Gal␤1-4GlcNAc␤1-2Man-Ser/Thr (where Sia is sialic acid). This oligosaccharide constituted at least 66% of the sialylated O-linked sugar chains. Furthermore, a laminin binding inhibition study suggested that the sialyl N-acetyllactosamine moiety of this sugar chain was involved in the interaction of the ␣-dystroglycan with laminin.
␣-Dystroglycan is an extracellular peripheral membrane glycoprotein anchored to the cell membrane by binding to a transmembrane glycoprotein, ␤-dystroglycan (1,2). These two dystroglycan subunits were originally identified as members of a sarcolemmal dystrophin-associated (glyco)protein complex. ␣and ␤-dystroglycans are encoded by a single gene and cleaved into two proteins by posttranslational processing (3). Based on recent amino-terminal sequencing of ␤-dystroglycan, it was proposed that the serine residue at position 654 of the precursor protein of dystroglycan is the cleavage site (4,5). The ␣-dystroglycan⅐␤-dystroglycan complex is widely expressed in many tissues (6) and thought to act as a transmembrane linker between the extracellular matrix and intracellular cytoskeleton (7), because ␣-dystroglycan binds extracellular matrix components, laminin-1 and -2 with high affinity (7)(8)(9)(10), and the intracellular domain of ␤-dystroglycan binds cytoskeletal proteins, dystrophin or its autosomal homologue, utrophin (2,11,12).
␣-Dystroglycan is heavily glycosylated. Whereas the deduced amino acid sequence predicts a ϳ74-kDa core peptide, ␣-dystroglycan is identified as a broad smear in SDS-polyacrylamide gel electrophoresis with an apparent molecular mass of 156 kDa in skeletal muscle (1), 120 kDa in brain and peripheral nerve (4,13,14), and 190 kDa in postsynaptic membrane of Torpedo electric organ (15). The differences in the molecular masses of ␣-dystroglycans obtained from different tissues seem to be due, not to differences in the primary structure, but to tissue-specific differential glycosylation of the core protein (6). Although the nature of the carbohydrate moiety of ␣-dystroglycan has not been fully clarified yet, chemical modification by treatment with periodic acid or trifluoromethanesulfonic acid resulted in the loss of laminin binding (7,16), suggesting that the sugar moiety is essential for this activity. The deduced amino acid sequence shows three potential N-glycosylation sites, and enzymatic cleavage of N-linked sugar chains reduced the molecular mass by approximately 4 kDa without affecting the laminin binding activity (1). There are also two conserved serine-glycine dipeptides (6), which are one type of potential glycosaminoglycan attachment site (17). However, nitrous acid treatment to degrade heparin or heparan sulfate or glycosaminoglycan lyase digestion affected neither the molecular mass nor laminin binding (7,10,13,15,16,18). Although ␣-dystroglycan from the C2 cell line (S27), which is defective in glycosaminoglycan synthesis (19), was found to bind laminin poorly (20), there is no direct evidence that N-linked oligosaccharides (or glycosaminoglycans, if present) are required for laminin binding.
The amino acid sequence of ␣-dystroglycan shows one feature of a mucin type O-glycosylation site (21) in the central region of the molecule. Threonine, serine, and proline are densely distributed between the 317th and the 488th amino acid residues, often clustering, and over one-half of the proline residues in this region are at positions Ϫ1 or ϩ3 relative to the threonine or serine residues. In addition, the susceptibility of ␣-dystroglycan to O-sialoglycopeptidase supports the hypothesis that ␣-dystroglycan is a sialylated mucin-type glycoprotein (4). Recently, we demonstrated that exhaustive sialidase treatment of ␣-dystroglycan or the addition of sialic acid to the incubation medium diminished the laminin binding activity of ␣-dystroglycan, suggesting that the sialic acid residues of ␣-dystroglycan, which are probably attached to O-linked oligosaccharides, were essential for this binding (10).
In this study, we have analyzed the structures of sialylated O-linked oligosaccharides of bovine peripheral nerve ␣-dystroglycan and demonstrate that a novel O-linked mannose-type oligosaccharide, Sia␣2-3Gal␤1-4GlcNAc␤1-2Man-Ser/Thr, 1 is the major component. The results of a binding-inhibition study suggest that this unique oligosaccharide contributes to the laminin binding activity of ␣-dystroglycan.
Purification of ␣-Dystroglycan from Crude Bovine Peripheral Nerve Membrane-␣-Dystroglycan was purified as described previously (25). Crude bovine peripheral nerve membranes were suspended at a protein concentration of 5 mg/ml in 50 mM Tris-HCl, pH 7.4, containing 0.5 M NaCl and a mixture of protease inhibitors: benzamidine (0.75 mM), phenylmethylsulfonyl fluoride (0.1 mM), pepstatin A (0.7 M), aprotinin (76.8 mM), and leupeptin (1.1 M). The suspension was titrated to pH 12 by slowly adding 1 N NaOH, extracted for 1 h, and then centrifuged at 140,000 ϫ g for 30 min at 25°C. The supernatant was titrated to pH 7.4 and centrifuged at 140,000 ϫ g for 30 min at 4°C. The supernatant was circulated over the wheat germ agglutinin-Sepharose overnight at 4°C. After extensive washing with buffer A (50 mM Tris-HCl, pH 7.4, 0.75 mM benzamidine, and 0.1 mM phenylmethylsulfonyl fluoride), the proteins bound to the wheat germ agglutinin-Sepharose were eluted with the buffer A containing 0.3 M N-acetylglucosamine. The eluate was then circulated over laminin-Sepharose (14) overnight at 4°C in the presence of 1 mM CaCl 2 and 1 mM MgCl 2 . After extensive washing with buffer A containing 1 mM CaCl 2 and 1 mM MgCl 2 , the proteins bound to the laminin-Sepharose were eluted with buffer A containing 10 mM EDTA. The eluate was extensively dialyzed against 50 mM NH 4 HCO 3 and used for the following analyses. The protein concentration was measured by using the bicinchoninic acid protein assay reagent. The purity of the sample was determined by SDS-polyacrylamide gel electrophoresis followed by Coomassie Brilliant Blue staining and by silver staining using the Silver Stain DAIICHI reagent kit. In the latter staining, to increase the sensitivity of glycoprotein detection, the periodic acid-silver stain method (26) was used with a modification; the gel was treated with 1% periodic acid for 10 min at room temperature prior to fixation instead of with 0.2% periodic acid treatment for 1 h at 4°C after fixation in the original method. With both staining methods, ␣-dystroglycan was detected as a prominent broad band approximately 116 kDa (data not shown).
Quantification of Sialic Acids-The DMB labeling method (27) was used with some modifications. Sialic acids were released from ␣-dystroglycan (155 ng of protein) in 60 l of solution by A. ureafaciens sialidase digestion (250 mU for 18 h at 37°C with 0.8% Triton X-100) or by acid hydrolysis (0.1 N HCl for 1 h at 80°C) (28). To this was added 240 l of DMB solution (4 mM DMB, 2 M acetic acid, 0.45 M ␤-mercaptoethanol, and 11 mM Na 2 S 2 O 4 ), and the mixture was incubated for 2.5 h at 50°C in the dark. A 60-l aliquot of the sample was applied to the Cosmosil 5C18-AR reversed-phase HPLC column (4.6 ϫ 250-mm), and the column was eluted with acetonitrile:methanol:water (9:7:84, v/v) at a flow rate of 0.75 ml/min at room temperature. The fluorescence was monitored at 448 nm (excitation, 373 nm). Authentic N-acetylneuraminic acid treated in the same way was used to make a standard curve.
Liberation of O-Glycosidically-linked Sugar Chains of ␣-Dystroglycan as Oligosaccharides-The sugar chains of ␣-dystroglycan were converted to oligosaccharide derivatives by two chemical methods. In the first procedure, an ␣-dystroglycan sample (31.0 g of protein) was thoroughly dried and subjected to hydrazinolysis for 5 h at 60°C according to the method of Patel et al. (23). The sample was subjected to N-acetylation followed by paper chromatography using 1-butanol:ethanol:water (4:1:1, v/v) for 18 h. The area of the paper from the origin to the position of authentic lactose was extracted with water, and the extracted oligosaccharides were labeled with 2-aminobenzamide (2AB) using the 2AB labeling kit (29). As for the 2AB-labeled oligosaccharides, fluorescence was monitored at 430 nm (excitation, 330 nm).
In the second procedure, a lyophilized ␣-dystroglycan sample (201.5 g of protein) was dissolved in 400 l of 0.05 N NaOH and 1 M NaBH 4 containing 25 mCi of NaB 3 H 4 and incubated for 16 h at 45°C (30,31). After adjusting the pH to 6 by adding acetic acid, the solution was passed through a column containing 1 ml of AG 50-X12 (H ϩ ), and the column was washed with 10 ml of water. The effluent and the wash volume were combined and evaporated to dryness. After the remaining borate was removed by repeated evaporation with methanol, the residue was subjected to paper chromatography using 1-butanol:ethanol: water (4:1:1, v/v) for 18 h. The area of the paper from the origin to the position of authentic GlcNAc␤1-2Man OT was extracted with water. Analytical Methods-The oligosaccharide samples dissolved in distilled water were applied to the Mono Q HR5/5 column. After elution of neutral oligosaccharides with 10 ml of water, acidic oligosaccharides were eluted with a 0 -1 M gradient of ammonium acetate, pH 4.0, at a flow rate of 1 ml/min at room temperature. Neutral oligosaccharides were applied to the Superdex Peptide HR10/30 gel filtration column (1 ϫ 60-cm) and eluted with distilled water at a flow rate of 0.4 ml/min at 60°C. PVL-Affi-Gel 10 column chromatography was performed as described previously (32). Reducing termini of tritium-labeled oligosaccharides were determined by using the Shodex SUGAR SP-1010 column (8 ϫ 300-mm) (33). Reversed-phase HPLC was carried out on the Cosmosil 5C18-AR column by eluting with a 50 mM acetate-acetonitrile gradient solvent system at a flow rate of 1.0 ml/min at 40°C. The 50 mM acetate-acetonitrile ratio was changed linearly from 100:0 to 95:5 (v/v) over 50 min from 5 min after injection for 2AB-labeled disaccharide analysis and from 100:0 to 98.5:1.5 (v/v) over 50 min from 10 min after injection and then maintained at that ratio for 2AB-labeled monosaccharide analysis.
Glycosidase Digestion-Oligosaccharides were incubated with one of the following mixtures for 18 h at 37°C: (i) A. ureafaciens sialidase (200 mU) in 80 l of 0.5 M sodium acetate or ammonium acetate buffer, pH 5.0; (ii) Newcastle disease virus sialidase (10 mU) in 40 l of 50 mM sodium acetate buffer, pH 5.5; (iii) diplococcal ␤-galactosidase (5 mU) in 40 l of 0.3 M citrate phosphate buffer, pH 6.0; and (iv) jack bean ␤-N-acetylhexosaminidase (0.5 U) in 55 l of 0.3 M citrate phosphate buffer, pH 5.0. One drop of toluene was added to all reaction mixtures to inhibit bacterial growth during incubation. Digestions were terminated by heating the reaction mixture in a boiling water bath for 3 min. Digested samples were desalted using ion exchange resin (300 l of AG3-X4(OH Ϫ ) and 300 l of AG50-X12(H ϩ )) for neutralized tritiumlabeled oligosaccharides or using the Sep-Pack C 18 (washing with 8 ml of distilled water and elution with 6 ml of acetonitrile:water (2:8, v/v)) for neutralized 2AB-labeled oligosaccharides. As for samples after sialidase digestions, ammonium acetate was removed by extensive evaporation, and sodium acetate was removed by passing the sample through a AG50-X12(H ϩ ) column (300 l) and evaporation.
Laminin Binding Inhibition Assay-Sixteen ng of ␣-dystroglycan were dispensed to each well of a polystyrene microtiter plate and dried under blowing cold air. After rinsing with LBB (10 mM triethanolamine, pH 7.6, 140 mM NaCl, 1 mM CaCl 2 , and 1 mM MgCl 2 ), the wells were blocked with LBB containing 3% bovine serum albumin (3% BSA-LBB). Fifty l of the biotinylated laminin-1 (2 nM in 3% BSA-LBB) containing various saccharides were added to the wells and incubated overnight at room temperature. After rinsing with LBB, the binding of laminin-1 was detected using the avidin-biotin complex (ABC) kit. o-Phenylenediamine dihydrochloride was used as a coloring substrate, and absorbance was measured at 492 nm. The experiments were done in triplicate for each added saccharide.

Quantification and Characterization of Sialic Acids-
The enzymatic cleavage and acid hydrolysis released about 14.9 and 13.8 mol of free sialic acids from 1 mol of ␣-dystroglycan, respectively. When the sialic acid residues thus obtained were analyzed by HPLC, N-acetylneuraminic acid and N-glycolylneuraminic acid were detected in a ratio of 4:1 by both methods. Although the DMB labeling method can distinguish Oacetyl derivatives of sialic acid from N-acetyl or N-glycolyl ones (27), no O-acetyl derivative was detected in ␣-dystroglycan.
Structural Analysis of Oligosaccharides Released by Mild Hydrazinolysis and Labeled with 2AB-O-Linked oligosaccharides of bovine peripheral nerve ␣-dystroglycan, released by mild hydrazinolysis and labeled with 2AB, were subjected to Mono Q column chromatography at pH 4.0 (Fig. 1A). The acidic fraction was exhaustively digested by A. ureafaciens sialidase, and the neutral oligosaccharides, thus obtained, were subjected to Superdex Peptide gel filtration column chromatography. Three major peaks eluting at 3.9 glucose units (GU) (F-I), 2.9 GU (F-II), and 0.9 -1.0 GU (F-III) were detected (Fig. 1B). Their percentage molar ratios calculated on the basis of their peak areas were 54, 30, and 16%, respectively.
The peak F-I released one galactose residue upon incubation with diplococcal ␤-galactosidase, which cleaves only the Gal␤1-4GlcNAc linkage (34) (Fig. 2A), and subsequently one N-acetylhexosamine residue was released upon jack bean ␤-N-acetylhexosaminidase digestion (Fig. 2B). The product eluted with 0.8 GU in Fig. 2B was identified to be Man-2AB by 2AB-derived monosaccharide analysis using reversed-phase HPLC (Fig.  3A). The component in Fig. 2A was eluted at the same retention time as that of authentic GlcNAc␤1-2Man-2AB standard in the reversed-phase HPLC (Fig. 4A), and more than 90% of it was retarded in the PVL-Affi-Gel 10 column, which specifically interacts with GlcNAc residues but not with GalNAc residues in the nonreducing terminus (32, 35) (Fig. 5).
Structural Analysis of Sialylated Oligosaccharides Released by ␤-Elimination Using Sodium Borotritide-The above results showed that the major sialylated O-linked oligosaccharide in bovine peripheral nerve ␣-dystroglycan was an O-mannosyltype oligosaccharide. However, the oligosaccharides in the fractions F-II and F-III, except for Glc-2AB, may be the peeling reaction products of the major component found in the fraction F-I. This raised a question whether the major component of the apparent O-mannosyl-type oligosaccharide also might be a peeled product of larger oligosaccharides. To solve this problem, we investigated whether oligosaccharides with the same composition could be obtained from the same material by the conventional ␤-elimination method, which releases O-linked oligosaccharides from the polypeptide backbone.
Tritium-labeled O-linked oligosaccharides released by alkaline ␤-elimination from the bovine peripheral nerve ␣-dystroglycan were separated by Mono Q column chromatography at pH 4.0 into neutral and acidic fractions (21 and 79%, respectively, based on the radioactivities). Three main peaks of the acidic fractions (T-A1, T-A2, and T-A3) were obtained (Fig. 6A). By exhaustive A. ureafaciens sialidase digestion, the peak T-A1 was completely converted to neutral oligosaccharides, whereas the peaks T-A2 and T-A3 were not digested (Fig. 6B), indicating that most of the sialylated O-linked oligosaccharides of the bovine peripheral nerve ␣-dystroglycan are monosialylated and do not contain any other anionic residues such as sulfated sugars and uronic acids. Sialidases from both A. ureafaciens and Newcastle disease virus gave the same results (data not shown), indicating that the sialic acid residues are linked at the C-3 position of the galactose (36). The predominance of ␣2-3linked sialic acid is compatible with the previous results of a lectin blot study in which bovine peripheral nerve ␣-dystroglycan was stained with Sia␣2-3Gal-specific Maackia amurensis agglutinin but not with Sia␣2-6Gal-specific Sambucus nigra   Fig.  4, B and C, respectively), and the fraction F-III (C) (the fluorescent component in Fig. 1B) were applied to the Cosmosil 5C18-AR column and eluted in a 50 mM acetate-acetonitrile gradient solvent system at a flow rate of 1.0 ml/min at 40°C. The 50 mM acetate:acetonitrile ratio was changed linearly from 100:0 to 98.5:1.5 (v/v) over 50 min from 10 min after sample injection and then maintained at that ratio. The elution positions of authentic monosaccharide-2AB derivatives are indicated by the arrowheads at the top of the figure. agglutinin (10).
The neutral fraction obtained by sialidase digestion was then subjected to Superdex Peptide gel filtration column chromatography. A peak eluting at 3.8 GU with a slight shoulder was detected (Fig. 7). It must be stressed here that no radioactive peak around 1 GU was detected. The shoulder portion was separated into two peaks of 3.8 and 2.7 GU by a second gel filtration. The combined fraction of the peaks of 3.8 GU was designated as fraction T-I, and the peak of 2.7 GU was designated as fraction T-II. The percentage molar ratios of T-I and T-II were 78 and 22%, respectively, on the basis of their radioactivities.
The fraction T-I released one galactose residue upon incubation with diplococcal ␤-galactosidase (data not shown). When the digested fraction was applied to the PVL-Affi-Gel 10 column, 84% of the radioactivity was recovered in the retarded fraction (data not shown). The retarded fraction was then heated in 4 N HCl for 3.5 h at 100°C followed by N-acetylation, and the tritium-labeled monosaccharide at the reducing terminus was determined by SP-1010 column chromatography. A radioactive peak appeared at the same retention time as that of authentic mannitol (data not shown). On the other hand, the fraction T-II was resistant to diplococcal ␤-galactosidase digestion. By SP-1010 column chromatography after acid hydrolysis of the fraction T-II, a peak with a retention time corresponding to N-acetylgalactosaminitol was detected (data not shown). Further structural studies of the oligosaccharides in fraction T-II could not be performed because of the limited amount of the sample.
Based on these results, the following structures were proposed for the major components of the fractions T-I and T-II: T-I, Gal␤1-4GlcNAc-Man OT ; and T-II, hexose-GalNAc OT . Based on the radioactivities of the fractions obtained by the Superdex Peptide and PVL-Affi-Gel 10 column chromatographies, the amount of the Gal␤1-4GlcNAc-Man OT was esti-mated to be at least 66% of the neutral oligosaccharides obtained by sialidase digestion.
Inhibition of Laminin Binding of ␣-Dystroglycan by Sialic Acid Compounds-We investigated the effect of sialic acid compounds, especially those having configurations similar to that of the major O-linked sialylated oligosaccharide of ␣-dystroglycan, on its laminin binding activity (Fig. 8). 3Ј-Sialyl N-acetyllactosamine, Neu5Ac␣2-3Gal␤1-4GlcNAc, which is the trisaccharide terminal portion of Neu5Ac␣2-3Gal␤1-4GlcNAc␤1-2Man, inhibited the binding of laminin, whereas both 3Ј-sialyllactose and 6Ј-sialyllactose did not. 6Ј-Sialyl Nacetyllactosamine, Neu5Ac␣2-6Gal␤1-4GlcNAc, which is an isomer of 3Ј-sialyl N-acetyllactosamine in the sialic acid linkage, also reduced the binding of laminin, but its effect was weaker than that of 3Ј-sialyl N-acetyllactosamine. These results indicate that the interaction between laminin and sialylated oligosaccharides of ␣-dystroglycan are not simply dependent on the anionic charge of sialic acid residues but also on the structure comprising the neutral sugar portion and the sialic acid linkage. DISCUSSION By two different analytical methods, we identified a sialidase-sensitive oligosaccharide with the same configuration as the major component. Based on these results, we conclude that the major sialylated O-linked oligosaccharide of the bovine peripheral nerve ␣-dystroglycan has the following structure: Neu5Ac (and Neu5Gc)␣2-3Gal␤1-4GlcNAc␤1-2Man-Ser/Thr (where Neu5Gc is N-glycolylneuraminic acid). Based on the radioactivity incorporated by the ␤-elimination method, these oligosaccharides constitute at least 66 molar percent of the total sialylated O-linked sugar chains. The components corresponding to fractions F-II and F-III (Fig. 1B), obtained by mild hydrazinolysis, were not detected by the ␤-elimination method. Therefore, Gal␤1-4GlcNAc-2AB and Gal-2AB in these fractions were considered to be the degradation products of the major oligosaccharide by a peeling reaction during hydrazinolysis. No reasonable explanation could be presented for the detection of Glc-2AB. Analysis of the fraction T-II was incomplete because of the limited amount of the radioactive sample. In view of the previous reports that lectin blots revealed the presence of a Gal␤1-3GalNAc group in ␣-dystroglycan (4, 7, 10), the hexose-GalNAc OT in the fraction T-II may represent this disaccharide. Mild hydrazinolysis has not been applied to various samples as frequently as conventional ␤-elimination to release their O-linked oligosaccharides. Although our results suggest the occurrence of a peeling reaction to some extent in the mild hydrazinolysis, this method has great advantages in that it requires a smaller amount of sample and allows a greater choice of labeling method. In the mild hydrazinolysis followed by 2AB derivatization, we could determine the major oligosaccharide structure by using less than one-sixth the amount of sample compared with the conventional alkaline borotritide treatment method. The fluorescent derivatives generally allow us to apply reversed-phase HPLC analysis, by which we could separate the disaccharide derivatives of the four possible GlcNAc␤1-Man isomers (Fig. 4A), and determine the exact linkage between the N-acetylglucosamine and mannose residues by using only about 2 pmol of the oligosaccharide derivative.
O-Mannosyl-type linkages, Man-Ser/Thr, have been found in the cell walls of microorganisms (37)(38)(39) and the skin collagen of the clam worm (40), but they are all neutral or glucuronosyl oligosaccharides. Thus far, a sialylated oligosaccharide with this type of linkage has been reported only in rat brain proteoglycans, which have a configuration of Neu5Ac␣2-3Gal␤1-4GlcNAc␤1-3Man1-Ser/Thr (31,41,42). The major sialylated O-linked oligosaccharide in ␣-dystroglycan identified in this study resembles this structure, but it is different in the linkage between the N-acetylglucosamine and the mannose residues. The type of linkage (GlcNAc␤1-2Man or GlcNAc␤1-3Man) would have a strong influence on the three-dimensional structure of the disaccharide. In the C1 conformer of D-mannose, the hydroxyl residue at C-2 takes an axial position, whereas that at C-3 is equatorial. Therefore, the carbohydrate chains attached to the nonreducing side of the mannose residue stretch out in completely different directions. Furthermore, the result of the disaccharide-2AB analysis by reversed-phase HPLC suggested that these different structures have different hydrophobicities (Fig. 4A).
Neu5Ac␣2-3Gal␤1-4GlcNAc␤1-2Man is commonly found as a side chain of complex type N-linked oligosaccharides. For example, human chorionic gonadotropin and human placental fibronectin contain such sugar chains (43,44), but laminin binding of such glycoproteins has not been reported. The de-duced amino acid sequence of ␣-dystroglycan predicts a mucin domain with a dense clustering of threonine, serine, and proline residues in the central region of the molecule (21). Consistent with this, a recent electron microscopic study of the cardiac muscle ␣-dystroglycan shows a rod-shaped segment with globular domains attached to both ends (45). Clustering of O-linked oligosaccharides in a mucin domain is thought to have important structural implications for the interaction between the oligosaccharides and their ligands (46). On the other hand, conformational analyses revealed that the side chains of complex type N-linked oligosaccharides do not always stretch out parallel to each other (47)(48)(49), suggesting that N-linked oligosaccharides are not proper to make sugar chains clustered toward a particular direction. Although the sialyl-N-acetyllactosaminyl-mannose structure is common, the functional significance may be different, depending on whether it links to the mannosyl-chitobiosyl core of an N-linked oligosaccharide or whether it links O-glycosidically to the core peptide directly.
We have recently demonstrated that both exhaustive sialidase treatment of ␣-dystroglycan and the addition of sialic acids to the incubation medium diminish the binding of laminin to the bovine peripheral nerve ␣-dystroglycan, suggesting that the sialic acid residues of ␣-dystroglycan are essential for this binding (10). In this study, we identified the structure of the major component of the sialylated O-linked oligosaccharides. The results of the inhibition study with sialylated oligosaccharides suggest that not only the anionic charge of sialic acid residues but also the sialyl N-acetyllactosamine configuration may be important in the interaction with laminin. To verify whether this unique type of O-linked oligosaccharide has a specific role in laminin binding, future studies need to clarify whether the ␣-dystroglycans of other species or tissues have the same structure of sialylated O-linked oligosaccharide as that found in the bovine peripheral nerve, because it is well known that glycosylation of proteins is both tissue-and speciesspecific (50).