Biosynthesis of HNK-1 glycans on O-linked oligosaccharides attached to the neural cell adhesion molecule (NCAM): the requirement for core 2 beta 1,6-N-acetylglucosaminyltransferase and the muscle-specific domain in NCAM.

The HNK-1 glycan, sulfo-->3GlcAbeta1-->3Galbeta1-->4GlcNAcbeta1-->R, is highly expressed in neuronal cells and apparently plays critical roles in neuronal cell migration and axonal extension. The HNK-1 glycan synthesis is initiated by the addition of beta1,3-linked GlcA to N-acetyllactosamine followed by sulfation of the C-3 position of GlcA. The cDNAs encoding beta1,3-glucuronyltransferase (GlcAT-P) and HNK-1 sulfotransferase (HNK-1ST) have been recently cloned. Among various adhesion molecules, the neural cell adhesion molecule (NCAM) was shown to contain HNK-1 glycan on N-glycans. In the present study, we first demonstrated that NCAM also bears HNK-1 glycan attached to O-glycans when NCAM contains the O-glycan attachment scaffold, muscle-specific domain, and is synthesized in the presence of core 2 beta1,6-N-acetylglucosaminyltransferase, GlcAT-P, and HNK-1ST. Structural analysis of the HNK-1 glycan revealed that the HNK-1 glycan is attached on core 2 branched O-glycans, sulfo-->3GlcAbeta1-->3Galbeta1-->4GlcNAcbeta1-->6(Galbeta1-->3)GalNAc. Using synthetic oligosaccharides as acceptors, we found that GlcAT-P and HNK-1ST almost equally act on oligosaccharides, mimicking N- and O-glycans. By contrast, HNK-1 glycan was much more efficiently added to N-glycans than O-glycans when NCAM was used as an acceptor. These results are consistent with our results showing that HNK-1 glycan is minimally attached to O-glycans of NCAM in fetal brain, heart, and the myoblast cell line, C2C12. These results combined together indicate that HNK-1 glycan can be synthesized on core 2 branched O-glycans but that the HNK-1 glycan is preferentially added on N-glycans over O-glycans of NCAM, probably because N-glycans are extended further than O-glycans attached to NCAM containing the muscle-specific domain.

Neural cell adhesion molecules undergo critical post-translational modifications that modulate their affinity and in some cases alter their specificity for their cognate ligands (1,2). In particular, many neural cell adhesion molecules are heavily glycosylated, and the extent and composition of attached carbohydrates modify their adhesive properties. Among those, polysialic acid and HNK-1 are particularly notable. Polysialic acid is mostly attached to NCAM, 1 and polysialylated NCAM is abundantly present in embryonic brain (1,2). In the adult brain, polysialylated NCAM is restricted to certain tissues, such as the hippocampus and olfactory bulb, where neuronal regeneration persists. Polysialic acid is a linear homopolymer of ␣-2,8-linked sialic acid residues formed on ␣-2,3or ␣-2,6linked sialic acid in N-acetyllactosamine of N-glycans (3)(4)(5)(6). Polysialic acid is attached to two N-glycosylation sites in the fifth immunoglobulin-like domain in NCAM and thought to attenuate the adhesive property of NCAM (7)(8)(9).
The HNK-1 carbohydrate epitope was originally defined by a monoclonal antibody raised against human natural killer (HNK) cells (10). In nervous tissues, the HNK-1 carbohydrate was first recognized as an autoantigen involved in peripheral demyelinative neuropathy (11,12). The structural analysis of glycolipids reacting with these autoantibodies demonstrated that the HNK-1 epitope is sulfo33GlcA␤133Gal␤134GlcNA-c␤13 R (11,12). HNK-1 glycan is widely distributed in glycoproteins, glycolipids, and proteoglycans (2,(11)(12)(13). The expression of HNK-1 glycan is spatially and developmentally regulated and found on migrating neuronal crest cells, cerebellum, and myelinating Schwann cells in motor neurons but not on those in the sensory neurons (14 -16). Although it has been reported that HNK-1 glycan is attached to the N-glycosylation site in the third immunoglobulin-like domain of NCAM, the recent studies showed that HNK-1 glycan is attached to the second, third, fifth, and sixth N-glycosylation sites (17).
HNK-1 apparently plays critical roles in neural cell migration and axonal extension. HNK-1 glycolipids coated on plates facilitated neurite outgrowth, whereas no facilitation of neurite outgrowth was observed on sulfatide, 3Ј-sulfogalactosyl ceramide. The effect of HNK-1 glycan was abolished once the sulfate group was removed (18,19). The HNK-1 glycan was identified in NCAM and N-glycans isolated from P0 glycoprotein (20,21). It was also found as rather unusual structures of O-linked oligosaccharides, sulfo33GlcA␤133Gal␤134GlcNAc␤132Man, which then are attached to serine or threonine (22). This glycan can be attached to dystroglycan, and its defect may cause muscular dystrophy (23).
NCAM has numerous isoforms due to alternative splicing of precursor mRNAs. Among them, NCAM(MSD) contains an additional domain, the so-called muscle-specific domain (MSD), consisting of 37 amino acids between two fibronectin type IIIlike domains (24). This NCAM(MSD) is mostly present in skeletal myotubes and often anchored to the plasma membrane through glycophosphatidylinositol (24). The MSD is highly enriched with serine, threonine and proline and O-linked oligosaccharides were shown to attach to this domain (25). It has been shown that NCAM(MSD) from C2C12 myoblast cell line contains mucin-type O-linked oligosaccharides ϮNeuNAc-␣233Gal␤133(ϮNeuNAc␣236)GalNAc␣3 Thr/Ser (25). However, no studies have addressed whether O-glycans in NCAM contain an HNK-1-capping structure.
In this report, we first present evidence that HNK-1 glycan can be formed in O-glycans attached to the MSD in NCAM. We found that HNK-1 glycan is attached to core 2 branched Oglycans but not to core 1 O-glycans present in NCAM. We determined the HNK-1 glycan structure in mucin-type Olinked oligosaccharides as sulfo33GlcA␤133Gal␤134-GlcNAc␤136(Gal␤133)GalNAc. Finally, we found that HNK-1 glycans can be formed on N-glycans more efficiently than O-glycans on NCAM, although GlcAT-P and HNK-1ST utilize O-glycan and N-glycan oligosaccharides almost equally as an acceptor.
Preparation of Lec1 Cells Expressing HNK-1 Precursors and HNK-1 Glycans-To avoid the formation of HNK-1 in N-glycans, the majority of the experiments were carried out using a mutant Chinese hamster ovary cell line, Lec1. Since Lec1 cells are deficient in GnT-I, all of the N-glycans synthesized in this cell line remain as high mannose oligosaccharides (36). Lec1 cells were found to lack core 2 ␤1,6-N-acetylglucosaminyltransferase as in other Chinese hamster ovary cells (30). Lec1 cells were stably transfected with pcDNA3.1(Zeo)-Core2GlcNAcT-I as described before (29). Lec1-core 2 cells were first selected in the presence of zeocin (Invitrogen) and then by immunofluorescent staining using tomato lectin, which reacts with N-acetyllactosamine and poly-N-acetyllactosamines (37). As shown previously, the formation of core 2 branches results in the formation of small amounts of poly-N-acetyllactosamine (38). To confirm the integration of human Core2GlcNAcT-I cDNA into Lec1 cell chromosome, reverse transcription-PCR was carried out on poly(A) ϩ RNA derived from Lec1-core 2 cells. The primers for this reverse transcription-PCR were designed in exon 1 and exon 2, avoiding the amplification of hamster genomic DNA harboring Core2GlcNAcT.
Transient Expression of NCAM and NCAM(MSD)-Lec1 cells expressing the HNK-1 glycan were transiently transfected with pIG-NCAM⅐IgG or pIG-NCAM(MSD)⅐IgG with or without pcDNAI-Core2GlcNAcT-I, using LipofectAMINE Plus™ as described previously (29). The medium was changed to serum-free medium 24 h after the transfection and cultured for an additional 48 h. NCAM⅐IgG in the cultured medium was adsorbed to protein A-agarose as described before (8). NCAM⅐IgG molecules eluted from the medium were subjected to SDS-polyacrylamide (5%) gel electrophoresis and transferred onto a polyvinylidene difluoride membrane. The blot was then incubated with anti-NCAM antibody, anti-HNK-1 antibody, or M6749 antibody, with a secondary antibody as described before (29).
Metabolic Labeling of Carbohydrate Moiety of NCAM-IgG-Lec1-core 2⅐HNK-1 cells were transiently transfected with pIG-NCAM(MSD)⅐IgG or pIG-NCAM⅐IgG. Twenty-four hours after the transfection, the medium was replaced with sulfate-free or glucose-free RPMI 1640 medium containing dialyzed 10% fetal calf serum in the presence of Na 2 (41). After a 48-h incubation, the spent medium was harvested, and NCAM⅐IgG was isolated as described above. After confirming the purity of the isolated samples by SDS-polyacrylamide gel electrophoresis and fluorography, the samples were subjected to oligosaccharide characterization.
Isolation of O-Linked Oligosaccharides Containing HNK-1 Glycans-Metabolically labeled NCAM(MSD)⅐IgG was digested with Pronase, and labeled glycoproteins were isolated by Sephadex G-25 gel filtration as described previously (41). O-Glycan-containing glycopeptides were separated from high mannose glycopeptides and other glycans using ConA-Sepharose affinity chromatography. The purified glycopeptides were subjected to alkaline reductive treatment to release O-linked oligosaccharides as described previously (41). Oligosaccharides were applied to a Poly-Prep column (Bio-Rad) containing 2 ml of anionexchange QAE-Sephadex A-25. The resin was equilibrated with 2 mM pyridine-acetate buffer, pH 5.5, and eluted in a stepwise manner with 6 ml each of 70 mM, 250 mM, and 1 M NaCl in 2 mM pyridine-acetate buffer, pH 5.5. Fractions (1 ml) were collected, and aliquots were taken for determination of radioactivity.
The oligosaccharides eluted with 1 M NaCl were digested with Arthrobacter sialidase (Sigma) to remove sialic acid and subjected to QAE-Sephadex A-25 column chromatography. Those eluted with 1 M NaCl after sialidase treatment were subjected to structural characterization.
Structural Characterization of O-Glycans Containing HNK-1 Glycan-The O-linked glycans were subjected to solvolysis to remove a sulfate group as described previously (41,42). The resultant oligosaccharides, purified by QAE-Sephadex chromatography, were digested with 125 milliunits of Escherichia coli ␤-glucuronidase (Sigma) in 50 l of the reaction mixture under the same conditions as described previously (11). The obtained oligosaccharides were analyzed on an AX-5 amino-bonded column using a Gilson 306 HPLC apparatus at a flow rate of 0.8 ml/min. The column was equilibrated with 90% solvent A (80% acetonitrile, 20% 15 mM KH 2 PO 4 ) and 10% of solvent B (40% acetonitrile, 60% 15 mM KH 2 PO 4 ). After 5 min, the elution was carried out by linear gradient from 10% of solvent B to 100% of solvent B in 60 min. The column was finally washed for 5 min with solvent B. Fractions were collected every minute, and aliquots were taken for determination of the radioactivity by scintillation counting.
Periodate Oxidation of Oligosaccharides-Oligosaccharides were oxidized with 50 mM NaIO 4 in 50 mM sodium acetate buffer, pH 4.5, at room temperature, and oxidized samples were reduced as described previously (43). The samples, after destroying NaBH 4 by acetic acid, were passed over a small column of Dowex 50 ϫ 8 and dried after the addition of methanol. The samples were then hydrolyzed in 0.01 M HCl at 80°C for 1 h, neutralized with 0.1 M Tris-HCl buffer, pH 8.5, and applied to a column (1 ϫ 120 cm) of Bio-Gel P-4 (200 -400-mesh) equilibrated with 0.1 M pyridine-acetate buffer, pH 5.5. Oligosaccharides were also separated by QAE-Sephadex A-25 column chromatography as described above.
Assays for GlcAT-P and HNK-1ST Activity-To assay the activities of GlcAT-P and HNK-1ST, soluble protein A-GlcAT-P and soluble protein A-HNK-1ST were prepared and adsorbed to IgG-Sepharose as described previously (33). To assay for GlcAT-P activity, the reaction mixture (100 l) consisted of 500 M acceptors, 0.1 M HEPES buffer, pH 6.5, 10 mM MnCl 2 , 10 mM of galactonic acid-␥-lactone, 50 l of the soluble chimeric enzyme attached to beads, and 50 M UDP-[ 14 C]GlcA (294 mCi/mmol, PerkinElmer Life Sciences). The reaction mixture was incubated for 3 h at 37°C, and the reaction was stopped by adding 50 mM (final concentration) EDTA.
To assay for HNK-1ST activity, the donor, 3Ј-phosphate 5Ј-phospho-[ 35 S]sulfate, and various amounts of an acceptor were incubated with a solution of the 50% suspension of the chimeric protein A-HNK-1ST attached to beads under the conditions described previously (29,33). The reaction was stopped by 250 mM ammonium formate, pH 4, and labeled products were obtained by C18 reverse-phase chromatography as described previously (29). The K m for GlcAT-P and HNK-1ST was determined using a Lineweaver-Burk plot at various concentrations of the acceptor (2.5-2000 M).

Expression of HNK-1 Glycan in Mucin-type O-Glycans
Attached to NCAM-To determine the biosynthetic pathway of HNK-1 glycan in mucin-type oligosaccharides attached to NCAM, Chinese hamster ovary mutant Lec1 cells were utilized. Lec1 cells contain the deficient N-acetylglucosaminyltransferase I, which is the first key enzyme to form complex N-glycans (36). Because this enzyme is absent in Lec1 cells, N-glycans remain as the high mannose type, and N-acetyllactosamine, which is a precursor for HNK-1 glycan, is not formed in N-glycans.
To determine whether HNK-1 glycans can be attached to O-glycans in sequences other than MSD, NCAM⅐IgG lacking the MSD sequence was also expressed in the same Lec1 cells expressing Core2GlcNAcT-I and HNK-1ST glycan on the cell surface. As shown in Fig. 2B, NCAM synthesized in these Lec1 cells did not bear the HNK-1 glycan or glucuronylated precursor regardless of whether Core2GlcNAcT was present or absent. Minor HNK-1-positive glycoproteins were detected when Lec1 cells expressed Core2GlcNAcT-I, GlcAT-P, and HNK-1ST (Fig. 2B, lane 4). They were judged to be contaminating glycoproteins other than NCAM based on their mobility in SDSpolyacrylamide gel electrophoresis. These results indicate that the MSD sequence is essential for acquiring the HNK-1 glycan in O-glycans attached to NCAM.  Fig. 3 demonstrated that 35 S-labeled material was not bound to ConA-Sepharose. The glycopeptides eluted with 10 mM methyl ␣-glucoside and 250 mM methyl ␣-mannoside represent biantennary and high mannose type N-glycans, respectively, and they did not contain sulfate. The glycopeptides unbound to ConA-Sepharose (shown in the horizontal bar in Fig. 3), representing O-glycans and multiantennary N-glycans, were then subjected to alkaline borohydride treatment to release O-linked oligosaccharides.
The treated sample was separated into two peaks after Seph-adex G-25 gel filtration. The first peak of higher molecular weight contained both 35 S and 3 H radioactivity, whereas the second peak contained only 3 H radioactivity (data not shown). Upon QAE-Sephadex column chromatography, all of the 35 Slabeled material was recovered in the fraction eluted with 1 M NaCl in the pyridine-acetate buffer (Fig. 4A, Q3). This 35 S-and 3 H-labeled material eluted again in this highly anionic fraction after sialidase treatment (Fig. 4B). Since no 3 H radioactivity was released, peak Q3 did not contain sialic acid, which can be labeled from [ 3 H]glucosamine precursor. In contrast, the glycopeptides eluted with 2-250 mM NaCl (Q0, Q1, and Q2 in Fig.  4A) were mostly eluted at the void volume after sialidase treatment (data not shown), consistent with the fact that they were devoid of [ 35 S]sulfate. Since no sulfatase isolated to date is known to remove a sulfate group from HNK-1 glycan, we opted to remove sulfate by solvolysis. The sample after solvolysis eluted with 70 mM NaCl (Fig. 4C, peak b), where material containing one acidic group elutes. There was still substantial radioactivity eluted with 250 mM and 1 M NaCl (Fig. 4C, peak c). The latter material was mostly free [ 35 S]sulfate (fractions 24 -32 in Fig. 4D), while a small amount of remaining HNK-1 glycan existed (peak d in Fig. 4D). Peak b in Fig. 4C, desulfated HNK-1 glycan, was subjected to Bio-Gel P-4 gel filtration before and after ␤-glucu- ronidase treatment. The isolated oligosaccharide, peak b in Fig.  5B, was converted by ␤-glucuronidase digestion to Gal␤13 4GlcNAc␤136(Gal␤134)GalNAcOH, peak c in Fig. 5C. These results established that HNK-1 glycan, shown as peak a in Fig.  4B, has the structure sulfate3 GlcA␤13 Gal␤134GlcNAc␤13 6(Gal␤133)GalNAcOH.
Periodate Oxidation of HNK-1 Glycan and Its Desulfated Form-The above results establish that the HNK-1 glycan is attached to core 2 branched O-glycans. The above experiments, however, did not determine the linkage and attachment sites of sulfate and glucuronic acid. In order to obtain this information, the HNK-1 glycan and its nonsulfated form (peaks a and b in Fig. 5, A and B, respectively) were subjected to periodate oxidation followed by reduction and mild acid hydrolysis (Smith degradation). Periodate oxidation takes place in a cis-glycol group.
As a control experiment, core 2 O-glycan, Gal␤134GlcNAc␤-136(Gal␤133)GalNAc (Fig. 5C, c) was subjected to Smith degradation. This treatment produced two peaks: f and g (Fig.  5F). Peaks f and g represent GlcNAc and glycerol derived from galactose, respectively. Peak c derived from HNK-1 glycans yielded the same results. When peak b in Fig. 5B, nonsulfated HNK-1 glycan, was subjected to the same treatment, peaks e and g were produced (Fig. 5E). Peak e corresponds to Gal␤134GlcNAc, which was confirmed by exoglycosidase digestion. These results show that glucuronic acid prevented galactose from oxidation, indicating that glucuronic acid residue was attached to C-3 of galactose. Otherwise, peak b in Fig.  5B and peak c in Fig. 5C would produce the same product. When peak a in Fig. 5A was subjected to the same treatment, peak d was obtained (Fig. 5D). Peak d was larger than peak e and was judged to be produced from sulfated glycans. Peak e in Fig. 5D was probably produced from nonsulfated glycans that were formed during various preparation steps. These results thus indicate that sulfate protected the GlcA residue from oxidation, and peak d corresponds to sulfo33GlcA␤13 3Gal␤134GlcNAc. The presence of sulfate and glucuronic acid in peak d was confirmed by the fact that peak d still had two anionic charges as assessed by QAE-Sephadex column chromatography (data not shown).

NCAM in Mouse Embryonic Brain and Heart and C2C12 Cells Contain Negligible Amounts of HNK-1 Glycans Attached
to O-Glycans-We then examined whether NCAM in mouse embryonic brain and heart contain HNK-1 glycans attached to O-glycans. NCAM present in brain at embryonic day 12 exhibit broad high molecular weights that were converted to bands of 200 and 140 kDa after N-glycanase treatment (Fig. 6A, lanes 1  and 2). When the same blot was incubated with anti-HNK-1 antibody, no band corresponding to N-glycanase-treated NCAM reacted with the antibody (Fig. 6A, lanes 2 and 6). The same experiment showed that small amounts of HNK-1 glycan were attached to N-glycans in NCAM, displaying two bands that are slightly larger than 220 kDa (Fig. 6A, lane 5). The remaining HNK-1 antigen was mostly associated with molecules larger than 220 kDa, suggesting that those HNK-1 glycans are associated with proteoglycans. HNK-1 was barely detected in heart tissue, although NCAM was detected (Fig.  6A, lanes 3, 4, 7, and 8).
By using similar methods, the myoblast cell line, C2C12, was subjected to analysis. C2C12 cells were derived from mouse cells and can be induced to form myotubes (25). Our preliminary results showed that HNK-1 glycan was barely detected by Western analysis of cell lysates. We thus use 35 NaCl (shown by a horizontal bar in A) were subjected to the same QAE-Sephadex column chromatography after sialidase treatment (B) or removal of sulfate by solvolysis (C). D, peak c in panel C was applied for a short column of Bio-Gel P-4. Peaks d and e denote the elution positions of intact sulfated HNK-1 glycan and free sulfate, respectively. Open and closed circles represent 35 S and 3 H radioactivity, respectively. in molecular sizes. However, none of this polydispersed radioactive band was converted by N-glycanase treatment to a sharp band (Fig. 6B, lanes 1-4), whereas NCAM in day 0 and 4 culture apparently contained sulfate groups in N-glycans (Fig.  6B, lanes 1-4). It is possible that NCAM in days 4 and 7 may contain polysialic acid attached to O-glycans, or these broad bands may represent contaminating proteoglycans such as heparan sulfate. These results indicate that NCAM in C2C12 cells bear minimum amounts of HNK-1 glycan.

HNK-1 Glycan Synthesis Takes Place Preferentially on N-Glycans, although Both GlcAT-P and HNK-1ST Almost Equally Act on N-and O-Glycan Oligosaccharide Acceptors-
The above results could be obtained if one of the glycosyltransferases responsible for HNK-1 glycans prefers N-glycans acceptors over O-glycan acceptors. To test this hypothesis, we first assayed ␤1,3-glucuronyltransferase (GlcAT-P) using various synthetic acceptors. The results shown in Fig. 7 illustrate that GlcAT-P utilizes synthetic acceptor oligosaccharides that mimic both N-and O-glycans. GlcAT-P utilizes Gal␤134GlcNAc␤136Man␣136Man␤13octyl slightly better than core 2 branched O-glycan, but core 2 branched O-glycan acceptor is a slightly better acceptor than Gal␤13 4GlcNAc␤132Man␣136Man␤13octyl. We then compared HNK-1ST activity in three different acceptors. As shown in Fig.  8, HNK-1ST acts slightly better on GlcA␤133Gal␤134 GlcNAc␤132Man␣136Man␤13octyl (K m ϭ 169 M) than Despite the fact that the enzymes responsible for HNK-1 synthesis can utilize N-glycan and O-glycan acceptors almost equally well, HNK-1 was barely detected in NCAM O-glycans in tissues examined. These results obtained as a whole thus suggest that HNK-1 may be more efficiently added to N-glycans than O-glycans when they are attached to NCAM. In order to test this hypothesis, Lec1 cells expressing GlcAT-P and HNK-1ST were transiently transfected with cDNA encoding GnT-I. This transfection allows Lec1 cells to gain complex N-glycans, which can then become substrates for both GlcAT-P and HNK-1ST. The amount of HNK-1 glycan was significantly increased in NCAM(MSD) derived from Lec1 cells expressing GnT-I, compared with NCAM(MSD) derived from the parent Lec1 cells (Fig. 9A, compare lanes 7 and 9). The large majority of HNK-1 glycans were removed after N-glycanase treatment (Fig. 9A, lane 10), indicating that the majority of HNK-1 glycan is associated with newly synthesized N-glycans.
It is possible that sialylation inhibits HNK-1 glycan formation in core 2 branched O-glycans, since sialylation and glucuronylation compete for the same N-acetyllactosamine. To test this possibility, Lec2 cells that are deficient in sialylation were used to express HNK-1 glycan. Transfected Lec2 cells bearing both N-glycans and O-glycans expressed significant amounts of HNK-1 glycans, but almost all HNK-1 glycans were removed after N-glycanase treatment (Fig. 9A, lanes 11 and 12). Almost identical results were obtained when NCAM(MSD) were labeled with [ 35 S]sulfate (Fig. 9B). These results as a whole indicate that HNK-1 synthesis is shifted from O-glycans to N-glycans once N-glycan acceptor substrates are available. DISCUSSION The present study demonstrated that HNK-1 can be synthesized in O-glycans when three enzymes, GlcAT-P, HNK-1ST, and Core2GlcNAcT, are transfected into Lec1 cells, which lack the ability to form the HNK-1 glycan attached to complex and hybrid N-glycans. Since Lec1 cells lack Core2GlcNAcT, a simultaneous expression of Core2GlcNAcT is essential to form HNK-1 glycan on O-glycans. Our structural analysis of HNK-1 glycan indeed demonstrated that HNK-1 glycan is synthesized on N-acetyllactosamine in core 2 branched O-glycans (Fig. 1). This requirement of Core2GlcNAcT was also demonstrated for sialyl Lewis x synthesis in mucin-type O-glycans, since sialyl Lewis x, Sia␣233Gal␤134(Fuc␣133)GlcNAc␤13 R, is also FIG. 6. HNK-1 glycan present in mouse fetal brain and heart and C2C12 myoblast cells. A, lysates derived from mouse brain and heart at embryonic day 12 were separated by SDS-polyacrylamide gel electrophoresis and subjected to Western analysis using anti-NCAM (lanes 1-4) or anti-HNK-1 (lanes 5-8) antibody. The lysates were subjected to analysis before (Ϫ) and after (ϩ) N-glycanase treatment. B, NCAM was immunoprecipitated from the cell lysates of [ 35 S]sulfatelabeled C2C12 cells and subjected to SDS-polyacrylamide gel electrophoresis and fluorography. Fig. 4B (A) and peak b in Fig. 4C before (B) and after ␤-glucuronidase treatment (C) were subjected to Bio-Gel P-4 gel filtration. D-F, the samples shown in A-C were subjected to Smith degradation and subjected to the same Bio-Gel P-4 gel filtration. Peaks a-c correspond to sulfo33GlcA␤-133Gal␤134GlcNAc␤136(Gal␤133)-GalNAcOH, GlcA␤133Gal␤134GlcNAc-␤136(Gal␤133)GalNAcOH, and Gal␤13 4GlcNAc␤136(Gal␤133)GalNAcOH, respectively. Peaks d-g are explained under "Results." Gal␤134GlcNAc␤136-(Gal␤133)GalNAcOH was also prepared from HL-60 cells as described previously (43). synthesized on N-acetyllactosamine backbone, which is formed by Core2GlcNAcT (41,48).

FIG. 5. Bio-Gel P-4 gel filtration of HNK-1 glycans before and after various treatments. A-C, peak a in
The present study also demonstrated that mucin-type Oglycans bearing HNK-1 can be synthesized only when NCAM contains an MSD. MSD is enriched with threonine, serine, and proline, which is characteristic for amino acid sequences that attach mucin-type O-glycans (23,24). A short amino acid sequence that consists of these amino acids is also present in the other part of NCAM, but it apparently does not serve for O-glycosylation attachment sites. MSD is encoded by four different exons, and differential splicing of NCAM precursor mRNA leads to the formation or absence of MSD. The expression of MSD-containing NCAM is temporally and spatially restricted, and myotubes express NCAM containing MSD (24).
Since the amino acid sequence of MSD is enriched with proline, threonine, and serine, MSD most likely does not have a unique conformation, such as an ␣-helical or ␤-sheet structure (49). Since O-glycosylation takes place in the Golgi apparatus, O-glycosylation sites need to be exposed to the environment after a protein folds. Non-␣-helical and non-␤-sheet structure satisfies such a requirement. In addition, it is most likely that such an amino acid sequence does not take a particular conformation, and its structure is flexible (49). It is thus possible that MSD may function as a hinge region where Oglycans are attached. This situation is similar to the hinge region of IgA, where multiple O-glycans are attached (50,51). Lysosomal membrane glycoproteins (LAMP-1 and LAMP-2) also contain a hingelike structure, and the hingelike structure of human LAMP-1 has homology to the IgA hinge region (52,53). Future studies are of importance to determine whether MSD in NCAM also functions as a hingelike structure. Despite the fact that HNK-1 glycan can be added to Oglycans to NCAM(MSD) upon transfection with GlcAT-A and HNK-1, NCAM isolated from embryonic brain, heart, and C2C12 cells did not bear detectable amounts of HNK-1 epitope in O-glycans. This is most likely not due to the absence of Core2GlcNAcT, since Core2GlcNAcT-I is expressed in the brain and heart (54). One possible explanation for this discrepancy is that the expression levels of GlcAT-P and HNK-1ST are lower in these tissues compared with Lec1 cells or HeLa cells (data not shown) transfected with GlcAT-P and HNK-1ST. On the other hand, the same experiment clearly detected HNK-1 glycan in molecules with high molecular weights. Since those HNK-1-positive molecules still showed a polydispersity after N-glycanase treatment, it is likely that these molecules represent proteoglycans. These results indicate that the method used was sensitive enough to detect HNK-1 glycan. As a whole, it can be concluded that HNK-1 glycan is only barely, if at all, added on O-glycans attached to NCAM in the tissues examined.
The above results could be obtained if GlcAT-P and HNK-1 utilize N-glycans as acceptors better than O-glycans. However, GlcAT-P and HNK-1ST did not utilize N-glycan acceptors better than O-glycans when synthetic oligosaccharides that mimic a portion of N-glycans or O-glycans were used as acceptors. Moreover, the synthesis of HNK-1 glycans on NCAM was substantially increased when Lec1 cells were converted to synthesize complex-type N-glycan acceptors for HNK-1 glycan by transfecting with GnT-I. The amount of HNK-1 glycan on Oglycans is much less than that on N-glycans (for Lec1 cells bearing complex type N-glycans). These results indicate that GlcAT-P and HNK-1ST are more accessible to N-glycan acceptors than to O-glycan acceptors in glycoproteins such as NCAM. This is probably because N-glycans are more extended from the polypeptide backbone than O-glycans. On the other hand, PSGL-1 glycoprotein containing numerous O-glycans has an extended rodlike structure and has been shown to have sulfated N-acetyllactosamines, presumably on the O-glycans (55,56). These combined results strongly suggest that MSD of NCAM may not display as preferable a conformation for enzymes that synthesize O-glycans as the conformation of PSGL-1.
HNK-1 glycan has been shown to bind to laminin (2), whereas sialylated core 1 O-glycan, NeuNAc␣233Gal␤133-(NeuNAc␣236)GalNAc, was shown to bind Siglec such as myelin-associated glycoprotein (57). These results suggest that the addition of a HNK-1-capping structure on NCAM O-glycans may switch the interaction of NCAM with myelin-associated glycoprotein to that with laminin. Such switching probably plays a critical role in neural cell development.
Our studies demonstrated that HNK-1 synthesis may be synthesized in a complex manner and depends on how much N-glycan acceptors versus O-glycan acceptors are available.
Similarly, it has been demonstrated that GlcNAc-6-O-sulfotransferase preferentially acts on N-glycans when glycoprotein acceptors such as CD34 bear both N-and O-glycans (41). By contrast, the same enzyme can act efficiently on O-glycans when the enzyme acts on GlyCAM-1, which almost exclusively contains O-glycans (58). In addition, it has been demonstrated that N-acetyllactosamine formation in core 2 branched O-glycans is much less efficient than in N-glycans (59). These results suggest that the synthesis of N-acetyllactosamine in core 2 glycans can be a rate-limiting step for the addition of HNK-1 epitope structure on O-glycans. These results combined indicate that the synthesis of terminal, functional oligosaccharide groups in O-glycans is regulated in a complex manner and that the availability of acceptors in N-or O-glycans and their con-formation play a critical role in displaying such functional groups in different glycans.