Purification and Characterization of a Glucuronyltransferase Involved in the Biosynthesis of the HNK-1 Epitope on Glycoproteins from Rat Brain*

The glucuronyltransferase involved in the biosynthesis of the HNK-1 epitope on glycoproteins was purified to an apparent homogeneity from the Nonidet P-40 extract of 2-week postnatal rat forebrain by sequential chromatographies on CM-Sepharose CL-6B, UDP-GlcA-Sepharose 4B, asialo-orosomucoid-Sepharose 4B, Matrex gel Blue A, Mono Q, HiTrap chelating, and HiTrap heparin columns. The purified enzyme migrated as a 45-kDa protein upon SDS-polyacrylamide gel electrophoresis under reducing conditions, but eluted as a 90-kDa protein upon Superose gel filtration in the presence of Nonidet P-40, suggesting that the enzyme forms homodimers under non-denatured conditions. The enzyme transferred glucuronic acid to various glycoprotein acceptors bearing terminal N-acetyllactosamine structure such as asialo-orosomucoid, asialo-fetuin, and asialo-neural cell adhesion molecule, whereas little activity was detected to paragloboside, a precursor glycolipid of the HNK-1 epitope on glycolipids. These results suggested that the enzyme is specifically associated with the biosynthesis of the HNK-1 epitope on glycoproteins. Sphingomyelin was specifically required for expression of the enzyme activity. Stearoyl-sphingomyelin (18:0) was the most effective, followed by palmitoyl-sphingomyelin (16:0) and lignoceroyl-sphingomyelin (24:0). Interestingly, activity was demonstrated only for sphingomyelin with a saturated fatty acid,i.e. not for that with an unsaturated fatty acid, regardless of the length of the acyl group .

glycoproteins (4), L1 (3), transiently expressed axonal glycoprotein-1 (5), and P0 (6), and some proteoglycans (7). Expression of the HNK-1 carbohydrate epitope is spatially and temporally regulated during development, and its highest expression is seen at the stages where the neural networks are constructed in the central and peripheral nervous systems (8 -10). The HNK-1 epitope is presumed to be involved in cell-to-cell interactions such as cell adhesion (11), migration (12), and neurite extension (13).
In our previous study, it was demonstrated that there are two types of glucuronyltransferases associated with the biosynthesis of the HNK-1 carbohydrate epitope in rat brain, one for glycolipid acceptors (18) and the other for glycoprotein acceptors (19). Similar glucuronyltransferase activities were found in chick (20) and rat (21) brains with paragloboside as an acceptor. In this study, a glucuronyltransferase specific for glycoprotein acceptors (GlcAT-P) was purified to apparent homogeneity from postnatal 2-week rat forebrains by means of various column chromatographies. The enzyme is a 45-kDa protein, which requires sphingomyelin (SM) for the expression of its transferase activity.
Glucuronyltransferase Assay-Glucuronyltransferase activity toward glycoprotein acceptors was measured essentially as described previously (19) with slight modification. Incubation was carried out 37°C for 3 h in an assay mixture comprising 20 g of ASOR, 100 M UDP-[ 14 C]GlcA (2ϫ10 5 dpm), 200 mM MES buffer, pH 6.5, 20 mM MnCl 2 , 0.5 mM ATP, 0.2% (v/v) Nonidet P-40, and 2 l of a 2% Nonidet P-40 extract of rat forebrain, which had been treated at 100°C for 3 min, in a final volume of 50 l. After incubation, the assay mixture was spotted onto a 2.5-cm Whatman No. 1 disc and the radioactivity of [ 14 C]GlcA-ASOR on the discs was counted with a liquid scintillation counter (Beckman LS-6000). Protein was quantitated with a Micro-BCA protein assay kit (Pierce) unless otherwise stated. Bovine serum albumin was used as a standard.
Extraction of Glucuronyltransferase-Preparation of the enzyme source and the following purification procedure were carried out at 4°C. Sixty frozen rat forebrains (60 g) were thawed and homogenized with five volumes of the homogenizing buffer. Each homogenate was centrifuged at 10,000 ϫ g for 10 min, and the resulting supernatant was centrifuged at 105,000 ϫ g for 1 h. The pellet was suspended in five volumes of the extracting buffer for 1 h, and then the suspension was centrifuged at 105,000 ϫ g for 1 h. The pellet was reextracted once more with three volumes of the extracting buffer. To the combined extracts glycerol was added to give a final concentration of 20%. The resulting solution (Nonidet P-40 extract) could be stored at Ϫ20°C for at least 2 months without loss of activity.
CM-Sepharose CL-6B Chromatography-CM-Sepharose CL-6B (50 ml) equilibrated with buffer A was added to the Nonidet P-40 extract prepared above. After shaking for 1 h, the suspension was filtered through a glass filter (CM-unbound fraction).
UDP-GlcA-Sepharose 4B Affinity Chromatography-The CM-unbound fraction was applied to a UDP-GlcA-Sepharose 4B column (100 ml; 5 ϫ 5.5 cm), which had been equilibrated with 500 ml of buffer B. After washing the column, the enzyme was eluted with buffer C. The flow-through fraction was applied to a UDP-GlcA-Sepharose 4B column once more, and the enzyme was eluted with buffer C. To the combined eluate, glycerol, MnCl 2 , and UDP were added to final concentrations of 20%, 20 mM, and 0.1 mM, respectively.
ASOR-Sepharose 4B Affinity Chromatography-The eluate from the UDP-GlcA-Sepharose 4B column was applied to an ASOR-Sepharose 4B column (30 ml; 5 ϫ 1.8 cm), which had been equilibrated with buffer D. After washing the column with buffer D and then buffer E, glucuronyltransferase activity was eluted with 200 ml of buffer E containing 10 mM N-acetyllactosamine.
Matrex Gel Blue A Chromatography-The eluate from the ASOR-Sepharose 4B column was dialyzed against buffer F. The dialysate was applied to a Matrex gel Blue A column (2 ml; 1.5 ϫ 1.2 cm), which had been equilibrated with buffer F. The enzyme was recovered by elution with buffer G.
Mono Q Anion Exchange Chromatography-The eluate from the Matrex gel Blue A column was dialyzed against buffer H, and then the dialysate was applied to a Mono Q column (1 ml; 0.6 ϫ 5.2 cm), which had been equilibrated with buffer H. Elution of the enzyme was carried out with buffer H containing a linear gradient of NaCl, from 10 mM to 1 M.
Metal Chelate Affinity Chromatography-The eluate from the Mono Q column was applied to a HiTrap chelating column (1 ml; 1.0 ϫ 2.4 cm), which had been chelated with copper ions and equilibrated with buffer I. The enzyme was eluted with buffer I containing a linear gradient of glycine, from 0 to 30 mM.
Heparin Affinity Chromatography-The eluate from the HiTrap chelating column was dialyzed against buffer J, and then applied to a HiTrap heparin column, which had been equilibrated with the same buffer. The enzyme was eluted with buffer J containing a gradient of NaCl, from 0.1 to 1.5 M.
SDS-Polyacrylamide Gel Electrophoresis-Enzyme fractions were treated with dithiothreitol at a final concentration of 100 mM prior to SDS-PAGE. SDS-PAGE was performed with a 10% polyacrylamide gel and the buffer system of Laemmli (25). The protein bands were stained with a silver stain kit (Wako, Osaka, Japan).
Superose 12 Gel Filtration of the Purified Glucuronyltransferase-Approximately 100 ng of the purified enzyme was applied to a Superose 12 column (25 ml; 1 ϫ 30 cm), which had been equilibrated with buffer K. The flow rate was 0.4 ml/min, and fractions of 0.4 ml were collected.
Preparation of a Neutral Glycolipid (Ceramide Mono-, Di-, and Trisaccharides) Fractions and SM from Rat Brain-One-fourth of the unbound lipid fraction recovered on DEAE-Sephadex A-25 chromatography was applied to a QAE-Sephadex A-25 (OH Ϫ form; Pharmacia LKB Biotechnology) column (2.5 ϫ 20 cm) equilibrated with chloroformmethanol-water, 30:60:8 (v/v). The unbound lipid fraction from the column was evaporated, acetylated, and then fractionated on a Florisil column (1 ϫ 20 cm, 60 -100 mesh; Floriden Co., New York, NY) by the method of Saito and Hakomori (27) with slight modification. The acetylated neutral glycolipid fraction and the acetylated SM fraction were deacetylated with 0.5 M KOH in methanol at 37°C for 6 h. The yield of the deacetylated neutral glycolipid fraction containing ceramide mono-, di-, and trisaccharides was 2.7 mg. The deacetylated SM fraction was applied to an Iatrobeads column (4.6 mm ϫ 25 cm, 6RS-8010), and the column was eluted with a solvent mixture of chloroform-methanolwater, 65:25:4 (v/v), at 40°C.
Partial Chemical Synthesis of SM with a Single Species of Fatty Acid-SMs were prepared by N-acylation of lyso-SM (sphingosylphosphorylcholine) with fatty acylchlorides (28). SM isolated from bovine brain was partially hydrolyzed with 6 M HCl-butanol, 1:1 (v/v), at 100°C for 1 h for preparation of lyso-SM (29,30). A mixture of 5 mg of lyso-SM, 0.4 ml of tetrahydrofuran, and 0.5 ml of 50% aqueous sodium acetate was added to about 5 mg of fatty acylchloride (stearoylchloride, oleoylchloride, lignoceroylchloride, or nervonoylchloride; Funacoshi Co., Tokyo, Japan), and the reaction mixture was vigorously stirred for 2 h at 20°C. The synthetic product was treated with 0.1 M methanolic NaOH and then purified by high performance liquid chromatography as described above. The purified SM was identified by TLC and infrared spectroscopy. The yields were 2.1 mg for N-stearoyl-, 2.5 mg for Noleoyl-, 3.0 mg for N-lignoceroyl-, and 3.1 mg for N-nervonoyl-SM, respectively.
Purification and Desialylation of NCAM-Frozen and thawed postnatal 2-week rat brains were homogenized with a Positron homogenizer in five volumes of 20 mM Tris buffer, pH 7.5, containing 0.15 M NaCl, 1 mM EDTA, 0.1 mg/ml PMSF, 10 M leupeptin, and 10 g/ml trypsin inhibitor. The homogenate was centrifuged at 105,000 ϫ g for 1 h, and the resulting pellet was suspended in five volumes of the same buffer containing 0.5% Nonidet P-40 to extract NCAM. After stirring for 1 h, the suspension was centrifuged at 105,000 ϫ g for 30 min. From the supernatant, NCAM was purified on an AF11-Sepharose 4B column (5 ml). NCAM bound to the column was eluted with 0.1 M diethylamine containing 0.1 M NaCl, 0.1% deoxycholic acid, and 1 mM EDTA. For desialization of the purified NCAM, 10 g of NCAM was treated with 50 milliunits of neuraminidase (Seikagaku Co., Tokyo, Japan) at 37°C for 4 h.

RESULTS AND DISCUSSION
Purification of a Glucuronyltransferase from Rat Brain-The results of purification of a glucuronyltransferase from rat brain are summarized in Table I. A preliminary study indicated that more than half of the glucuronyltransferase activity in the homogenate was recovered in the 105,000 ϫ g pellet, suggesting that the enzyme is mostly associated with the microsomal fraction. Treatment of the pellet with either 1 M NaCl or phosphatidylinositol-phospholipase C did not release the enzyme activity, suggesting that the enzyme is an intrinsic membrane protein. Extraction of the pellet with 0.5% Nonidet P-40 released the glucuronyltransferase activity into the soluble fraction.
As the first step of purification, the Nonidet P-40 extract was subjected to CM-Sepharose CL-6B cation exchange chromatography. Essentially all the glucuronyltransferase activity was recovered in the unbound fraction, with a specific activity increase of 1.2-fold.
The next step of the purification involved UDP-GlcA affinity chromatography. A preliminary experiment indicated that essentially all (97%) the glucuronyltransferase activity toward glycolipid acceptors, with neolactotetraose-phenyl-C 14 H 29 (nLc-PA 14 ) as a substrate (19), was recovered in the passthrough fraction, while a larger portion of the activity toward glycoprotein acceptors, with ASOR as a substrate, was recovered in the eluate fraction. Consistent with this preliminary result, approximately 93% of the glucuronyltransferase activity in the CM-Sepharose CL-6B-unbound fraction was recovered in the eluate fraction with a purification of about 2-fold.
The third step of the purification involved ASOR-conjugated Sepharose affinity chromatography. Among several buffers tested, a buffer containing 10 mM N-acetyllactosamine was effective for elution of the activity from the column. Because N-acetyllactosamine and glycerol inhibited the glucuronyltransferase activity, the apparent yield of the enzymatic activity in the eluate fraction was very low (9% that of the Nonidet P-40 extract; see Table I). However, after dialysis, the activity recovered to more than 30% that of the Nonidet P-40 extract. The purification achieved with this ASOR-Sepharose 4B affinity chromatography was over 1,000-fold.
The fourth step of the purification involved dye ligand affinity chromatography. A Matrex gel Blue A column was found to retain the glucuronyltransferase activity. 1 M NaCl was effective for dissociating the enzyme from the column, with a 10-fold increase in the specific activity.
The fifth step of the purification involved Mono Q anion exchange chromatography. The glucuronyltransferase activity was mainly eluted in fractions 43-47 ( Fig. 1), in which the concentration of NaCl was around 0.4 M. Because of the low amount of proteins in each fraction, protein quantification and SDS-PAGE were carried out with several (five to seven) fractions combined. Five-fold purification was achieved through this step. In the following purification steps (HiTrap chelating and HiTrap heparin columns), the protein concentrations were determined in the same way.
The sixth step of the purification involved HiTrap chelating metal chelate affinity chromatography. Mono Q eluate fractions 43-47 were pooled and applied to a HiTrap chelating column, which had been chelated with Cu 2ϩ . The glucuronyltransferase activity was eluted at a glycine concentration of 15 mM (fractions 22-28 in Fig. 2). Two-fold purification was achieved at this step.
The last step of the purification involved HiTrap heparin affinity chromatography, utilizing the inhibitory activity of heparin, as shown in Table II. The glucuronyltransferase bound to the column and was eluted in fractions 43-47 at a NaCl concentration of around 0.7 M (Fig. 3A). Upon SDS-PAGE, a major band of 45 kDa was observed, with a few minor bands. Since these minor bands were predominant components in the previous fractions, i.e. fractions 38 -42, the HiTrap heparin affinity chromatography was repeated. The glucuronyltransferase thus obtained gave a single band corresponding to 45 kDa upon SDS-PAGE (Fig. 3B).
The 45-kDa protein was shown to have a SH-group specifically protected by UDP-GlcA when the eluate fraction on Ma- trex gel Blue A column chromatography was treated with Nmaleimidopropionyl-biocytin (MPB) according to the procedure of Pukazhenthi et al. (31). Based on these results, the 45-kDa protein was tentatively concluded to be a glucuronyltransferase. This conclusion was finally confirmed by our recent successful cDNA cloning of a HNK-1-associated glucuronyl-transferase on the basis of the partial amino acid sequence of the purified protein (32).
Thus, the glucuronyltransferase involved in the biosynthesis of the HNK-1 epitope on glycoprotein acceptors (GlcAT-P) was purified to apparent homogeneity from postnatal 2-week rat forebrains, with 1,200,000-fold purification and a 4.0% overall recovery (Table I).
Superose 12 gel filtration chromatography of the purified glucuronyltransferase indicated that the molecular mass of the enzyme was approximately 90 kDa, as shown in Fig. 4. This value is almost 2 times larger than that determined by SDS-PAGE under reducing conditions (45 kDa, Fig. 3B), suggesting that the enzyme occurs as a homodimer of 45-kDa polypeptides under non-denaturing conditions.
Enzymatic Properties of GalAT-P: Substrate Specificity of GlcAT-P-In order to study the substrate specificity of purified enzyme, we analyzed the effects of various compounds on the glucuronyltransferase activity toward ASOR (Table II). N-Acetyllactosamine at the concentration of 5 mM had a very potent (95%) inhibitory effect. In contrast, lacto-N-biose (Gal␤1-3GlcNAc) and lactose (Gal␤1-4Glc) had no or little effect on the enzymatic activity (0% and 23%, respectively), indicating that the enzyme recognizes not only the terminal sugars on the acceptor molecules but also the penultimate sugars and their linkage positions. Hyaluronic acid and chondroitin had little or no inhibitory effect (9.5% and 0%, respectively), whereas heparin and heparan sulfate decreased the The combined eluate fraction obtained on Mono Q column chromatography was loaded onto a HiTrap chelating column, which had been pretreated with CuCl 2 , and was then eluted as described under "Experimental Procedures." The eluate was collected as 1-ml fractions, and then the glucuronyltransferase activity was measured. The fractions indicated by the bar were combined.

TABLE II
Substrate specificity of the purified enzyme, as measured by inhibition assaying The glucuronyltransferase activity was measured as described under "Experimental Procedures" in the presence of various inhibitors. The concentration of ASOR in the assay mixture was changed from 20 g to 5 g. The data are the mean values for three determinations in a representative experiment.

Inhibitor
Concentration Inhibition enzymatic activity (90% and 36%, respectively). These results may indicate that heparin and heparan sulfate act as acceptors for the enzyme but that hyaluronic acid and chondroitin do not. However, the purified enzyme did not show any transferase activity toward any of these glycosaminoglycans (data not shown).
With regard to donor specificity, UDP and UDP-GlcA exhibited strong inhibitory effects (99% and 98%, respectively), followed by UDP-GlcNAc (67%). In contrast, GlcA and CMP-NeuAc had no or little effect on the activity (0.0% and 1.5%, respectively). These results suggest that the enzyme principally recognizes the terminal non-reducing N-acetyllactosamine structure in the sugar chains on glycoproteins and the nucleotide portion of UDP-GlcA.
In order to determine the effect of the polypeptide portion of the acceptor glycoconjugates, various asialo-glycoproteins and glycolipids were tested as acceptors, as described under "Experimental Procedures." The purified enzyme sufficiently transferred the glucuronic acid to asialo-NCAM and asialofetuin (72% and 87% of that in the case of ASOR, respectively). In contrast, asialo-thyroglobulin, which contains high mannose-type sugar chains, as well as complex-type sugar chains, was a poor acceptor of the enzyme. Interestingly, GlcAT-P did not show any activity toward paragloboside, a precursor glycolipid of the HNK-1 epitope.
The effects of various divalent cations on the glucuronyl-   transferase activity were determined. Among those tested, Mn 2ϩ activated the enzyme most effectively. Co 2ϩ and Mg 2ϩ showed 20% and 14% of the activity of Mn 2ϩ , respectively. Ca 2ϩ , Ba 2ϩ , Ni 2ϩ , Cu 2ϩ , and Zn 2ϩ had no effect on the enzyme at all (data not shown).
GlcAT-P Required SM as an Activator-To our surprise, during the process of purification, the enzyme activity disappeared almost completely at the step of Matrex gel Blue A affinity chromatography (with approximately 22,500-fold purification). However, the activity was recovered when an aliquot of an Nonidet P-40 extract was added to the assay mixture, suggesting that GlcAT-P requires some kinds of activator(s) for its catalytic activity. This presumed activator was stable on heating at 100°C for 3 min. Therefore, we carried out the enzyme assay in the presence of a saturating amount of the heat-treated Nonidet P-40 extract in the following purification steps, as described under "Experimental Procedures." We tried to identify the activator present in postnatal 14-day rat brains. First, we found that the chloroform-methanol extract of rat brains can substitute for the heat-inactivated Nonidet P-40 extract. Upon Folch partitioning (26), the organic solvent layer (Folch lower phase) activated the enzyme effectively in a saturable manner, but the upper phase did not (Fig.  5, A and B). These lines of evidence indicated that the activator is a kind of lipid. Then, the respective lipid components were prepared from the Folch's lower phase according to the procedure described under "Experimental Procedures," and their stimulatory activity was measured. Table III shows the amount of each lipid that gives 50% of the full activity (in the presence of heat-treated Nonidet P-40 extract). It is clear that SM caused recovery of the enzymatic activity most effectively (5.2 g for 50% recovery), followed by PC. In contrast, PE, PS, PI, and neutral glycolipids did not have a positive effect. It should be noted that almost all the stimulatory activity in the Folch lower phase was accounted for by SM.
In order to determine the effect of the fatty acid composition on the stimulatory activity, several SM molecules with a single species of acyl group were synthesized, as described under "Experimental Procedures," and their activities were compared. Interestingly, the stimulatory activity of SM was extremely affected by the fatty acid composition (Table IV). As for the length of the acyl group, stearoyl-SM (18:0) was the most effective activator, followed by palmitoyl-SM (16:0) and lignoceroyl-SM (24:0). More interestingly, SM with a saturated acyl group activated the enzyme remarkably, while that with an unsaturated acyl group did not show any stimulatory activity regardless of the length of the acyl group. The most abundant acyl group in rat brain SM is the stearoyl group (37), which is in fact the most effective acyl group as an activator. Phosphatidylcholine, which has phosphocholine as a hydrophilic group like SM, stimulated the enzyme activity, although the activity was 20 times less than that of SM. It should be noted, in this context, that more than 90% of the phosphatidylcholine in the rat brain has an unsaturated acyl group. Ceramide, on the other hand, had no stimulatory effect, even if it had a stearoyl acyl group.
Recently, some phospholipids were reported to stimulate the activities of glycosyltransferases, such as hepatic glucuronyltransferase (38), ␤1-4-galactosyltransferase (39), and ␣2-3sialyltransferase (40). GlcAT-P was activated dramatically in the presence of SM. Many glycosyltransferases that have so far been cloned are type II transmembrane proteins located in the Golgi apparatus (41). GlcAT-P is also believed to be a type II transmembrane protein in the Golgi or ER membrane, on the basis of its characteristic hydropathy profile (32) and also the requirement of detergents for its solubilization. Two phospholipids, which contain phosphocholine as a hydrophilic group, stimulated the enzymatic activity in common, suggesting that the phosphocholine group of these phospholipids interacts with the luminal portion of the enzyme through an electrostatic interaction. Interestingly, in the cellular membrane system, SM is localized predominantly in the outer leaflet of the plasma membrane and the Golgi lumen (42). This is in good agreement with the putative location of the catalytic domain of this glucuronyltransferase, the Golgi lumen. These lines of evidence suggest that expression of the HNK-1 epitope on glycoproteins can be regulated not only by the expression of the enzyme protein but also by the micro-circumstances around the enzyme, especially by SM.