Cloning and Expression of a Novel Galactoside β1,3-Glucuronyltransferase Involved in the Biosynthesis of HNK-1 Epitope*

We isolated a cDNA encoding a novel glucuronyltransferase, designated GlcAT-D, involved in the biosynthesis of the HNK-1 carbohydrate epitope from rat embryo cDNA by the degenerate polymerase chain reaction method. The new cDNA sequence revealed an open reading frame coding for a protein of 324 amino acids with type II transmembrane protein topology. The amino acid sequence of GlcAT-D displayed 50.0% identity to rat GlcAT-P, which is involved in the biosynthesis of the HNK-1 epitope on glycoproteins. Expression of GlcAT-D in COS-7 cells resulted in the formation of the HNK-1 epitope on the cell surface. The enzyme expressed in COS-7 cells transferred a glucuronic acid (GlcA) not only to asialo-orosomucoid, a glycoprotein bearing terminal N-acetyllactosamine structure, but also to paragloboside (lacto-N-neotetraosylceramide), a precursor of the HNK-1 epitope on glycolipids. Furthermore, substrate specificity analysis using a soluble chimeric form of GlcAT-D revealed that GlcAT-D transfers a GlcA not only to Galβ1–4GlcNAcβ1–3Galβ1–4Glc-pyridylamine but also to Galβ1–3GlcNAcβ1–3Galβ1–4Glc-pyridylamine. Enzymatic hydrolysis and Smith degradation of the reaction product indicated that GlcAT-D transfers a GlcA through a β1,3-linkage to a terminal galactose. The GlcAT-D transcripts were detected in embryonic, postnatal, and adult rat brain. In situ hybridization analysis revealed that the expression pattern of GlcAT-D transcript in embryo is similar to that of GlcAT-P, but distinct expression of GlcAT-D was observed in the embryonic pallidum and retina. Regions that expressed GlcAT-D and/or GlcAT-P were always HNK-1-positive, indicating that both GlcATs are involved in the synthesis of the HNK-1 epitope in vivo.

Structures of HNK-1 epitope determined to date almost invariably carry HSO 3 -3GlcA␤1-3Gal␤1-4GlcNAc at nonreducing termini (16 -19). The precursor of the epitope, Gal␤1-4GlcNAc sequence, is commonly found on glycoproteins and glycolipids, but expression of HNK-1 epitope is spatially and temporally restricted. The key enzymes in the biosynthesis of HNK-1 are a ␤1,3-glucuronyltransferase, which transfers a GlcA to a terminal galactose, and a sulfotransferase, which adds a sulfate group to the GlcA. Recently, GlcAT-P, a glucuronyltransferase involved in the biosynthesis of HNK-1 on glycoprotein, was purified, and its cDNA was cloned (20,21). Subsequently, a sulfotransferase that directs a final step of the biosynthesis of HNK-1 was cloned by an expression cloning strategy that involved cotransfection of GlcAT-P cDNA (22,23). On the other hand, GlcAT-I, another glucuronyltransferase involved in the biosynthesis of the linkage region of proteoglycans (EC 2.4.1.135), was cloned by PCR 1 strategy based on motifs conserved in GlcAT-P with putative proteins in Caenorhabditis elegans and Schistosoma mansoni (24).
We examined the enzymatic features of glycosyltransferases, including ␤1,3-glucuronyltransferase, expressed in rat. We found ␤1,3-glucuronyltransferase activity to glycolipids as well as glycoproteins in rat embryonic brain, suggesting that this enzyme is a novel glucuronyltransferase. To clone cDNA of the novel enzyme, we used the RNA of rat embryonic day 13 (E13) brain as a template of reverse transcription-PCR and designed degenerate primers to the highly conserved regions found in * This work was supported in part by the Yamada Science Foundation and by Grants-in-Aid for Scientific Research from the Ministry of Education, Science, Culture, and Sports of Japan. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AF106624.
ʈ To whom correspondence should be addressed: the alignment of amino acid sequence of GlcAT-P with GlcAT-I. In this study, we describe the cDNA cloning of a new member of the glucuronyltransferase family, GlcAT-D. The expression of cDNA revealed that GlcAT-D is a glucuronyltransferase involved in the biosynthesis of HNK-1 epitope on both glycolipid and glycoprotein. We also demonstrate that GlcAT-D transfers a GlcA to a Gal residue of Gal␤1-3/4GlcNAc by ␤1,3-linkage.
PCR-based Cloning of a New Glucuronyltransferase-Based on the amino acid sequence alignment of rat GlcAT-P (21) and human GlcAT-I (24), degenerate oligonucleotides to elements conserved in motifs II and IV were synthesized. The sequences of the 5Ј-and 3Ј-primers were 5Ј-TSGTSTAYTTYGCYGAYGAYGA-3Ј and 5Ј-TTYTCNGTNCKNGT- and Y ϭ C ϩ T), respectively. PCR was performed using first strand cDNA synthesized with total RNA of rat E13 brain. Thirty cycles (94°C for 45 s, 50°C for 45 s, and 72°C for 90 s) were run using AmpliTaq Gold polymerase (PE Applied Biosystems). The PCR product of around 380 base pairs in length was subcloned into the pGEM-T Easy vector (Promega), and 10 clones were sequenced. Two of the 10 clones contained a novel putative glucuronyltransferase (GlcAT-D) fragment that was used as a probe for screening a rat E15 brain cDNA library (25). One positive clone, 58-2-2, was isolated and sequenced.
To obtain the entire coding sequence of GlcAT-D, rapid amplification of cDNA 5Ј-end was employed with 5Ј-AmpliFINDER RACE Kit (CLONTECH) using rat E15 brain total RNA as a template. The amplified cDNA of about 1.6 kilobase pairs in length was subcloned into pGEM-T Easy and sequenced. Several clones were sequenced to compensate for misreading by AmpliTaq Gold polymerase.
Northern Blot Analysis-Total RNA was extracted from embryo, newborn, and adult rat brain using ISOGEN reagent (Nippon gene). Equal amounts of total RNA (20 g in each lane) were run on formaldehyde 1.2% agarose gel and transferred to a nitrocellulose membrane. Multiple tissue Northern blot of rat poly(A) ϩ RNA was purchased from CLONTECH. The blots were hybridized overnight with 32 P-labeled EcoRI fragment of 58-2-2 at 42°C in hybridization buffer. Then the filters were washed in 0.5ϫ SSC plus 0.1% SDS for 1 h at 65°C.
In Situ Hybridization-For the construction of riboprobes for in situ hybridization analysis, we obtained the cDNA fragment of GlcAT-D (nucleotide positions 285-1005) by digestion of 58-2-2 plasmid with EcoRI and subcloned into pBluescript II SK(Ϫ) (Stratagene). The GlcAT-P cDNA was obtained by reverse transcription-PCR in which the cDNA template was synthesized from the total RNA taken from E13 rat embryo brain. Oligonucleotides used to amplify the GlcAT-P cDNA (21) are 5Ј-AAACCTGCTGCCACAATGGGTAA-3Ј (nucleotides Ϫ15 to 8) and 5Ј-ATGGGAGAGGATGGAAGCCAAGAT-3Ј (nucleotides 1297 to 1274). The identity of the PCR product was confirmed by sequencing the subcloned fragment. The cDNA fragment of GlcAT-P was cloned into pBluescript II SK(Ϫ) (Stratagene). The antisense RNA probes were generated with T3 or T7 RNA polymerase. Corresponding sense probes were used to check the specificity of hybridization signals.
In situ hybridization with probes of GlcAT-D and GlcAT-P was performed as described previously (26) with slight modifications. Briefly, the head region of E13.5 Sprague-Dawley rat embryos was fixed in 4% paraformaldehyde in phosphate-buffered saline overnight at 4°C. Cryosections, 14 m in thickness, were thaw-mounted onto VECTABOND® (Vector Laboratories)-coated glass slides. After rehydration in phosphate-buffered saline with 0.1% Tween 20, the sections were postfixed in 4% paraformaldehyde for 20 min, treated with proteinase K (1 g/ml) at 37°C for 4 min, postfixed again in 4% paraformaldehyde for 20 min, and hybridized with digoxigenin-labeled probes (1 g/ml) overnight at 65°C. Then the sections were washed in 50% formamide, 5ϫ SSC, and 1% SDS for 30 min at 65°C, subsequently twice in 50% formamide and 2ϫ SSC for 45 min at 65°C, and finally in Tris-buffered saline containing 0.1% Tween 20. The slides were then immersed in blocking solution and incubated overnight at 4°C with anti-digoxigenin Fab-alkaline phosphatase conjugate diluted to 1:2000 by blocking solution. The hybrids were visualized by the alkaline phosphatase reaction with nitro blue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate substrate.
Immunohistochemistry-Immunostaining patterns of HNK-1 antibody were examined basically as described previously (27,28). Frozen sections were prepared as described above, rehydrated in Tris-buffered saline containing 0.1% Tween 20, and boiled in 0.01 M sodium citrate for antigen enhancement. After preincubation with the blocking solution for 30 min, the sections were incubated with Leu7 monoclonal antibody (Becton Dickinson) diluted to 1:25 by the blocking solution overnight at 4°C. Biotinylated anti-mouse IgM (Zymed Laboratories Inc.) was used as a secondary antibody, and the sections were quenched with hydrogen peroxide. Immunoreactivity was detected using the ABC kit (Vector Laboratories) and the metal enhanced diaminobenzidine kit (Pierce).
Construction of Expression Vector Containing a Full-length GlcAT-D cDNA-A DNA fragment, containing the open reading frame of GlcAT-D, from nucleotide positions Ϫ34 to 1004 was amplified by PCR with cDNA reverse-transcribed from total RNA of rat E13 brain using a 5Ј-primer (5Ј-CCCAAGCTTGAGGGTGGTGTCCGAGACGCT-3Ј) containing a HindIII site and a 3Ј-primer (5Ј-GGAATTCCCTTCTCTCCT-CAGCGGCTGCTC-3) containing an EcoRI site. After restriction enzyme digestion, the PCR fragment was subcloned into HindIII and EcoRI sites of pBluescript II SK(Ϫ) (Stratagene), yielding pBS-GlcAT-D. pBS-GlcAT-D was digested with HindIII and NotI and then cloned into HindIII and NotI sites of pCDM8, yielding pCDM8-GlcAT-D.
Construction of Soluble Form of GlcAT-D-A truncated form of GlcAT-D, lacking the first 35 amino acids of GlcAT-D, was amplified by PCR with pBS-GlcAT-D as a template using a 5Ј-primer (5Ј-GGAATTC CCGACCCTACTTCTCTCCGCATA-3Ј) containing an EcoRI site and the same 3Ј-primer as used in the amplification of the fragment containing the open reading frame of GlcAT-D. The PCR product was digested with EcoRI and cloned into the EcoRI site of pPROTA vector (29). A recombinant plasmid with the correct orientation, pPROTA-GlcAT-D, was used for expression.
Assays of Glucuronyltransferase Activity to Glycoproteins and Glycolipids-COS-7 cells (1 ϫ 10 6 cells) were transfected with one of the expression plasmids, pCDM8-GlcAT-D or pCDM8 (10 g each) using LipofectAMINE (Life Technologies, Inc.) according to the manufacturer's instructions. After 3 days, the cells were harvested, washed with phosphate-buffered saline, homogenized in 100 l of 0.3% Triton X-100, and used as an enzyme source.
When a glycolipid was used as the acceptor, glucuronyltransferase activity was measured in 0.1 M MES, pH 6.0, 6 g of LNnT-Cer, 100 M UDP-[ 14 C]GlcA, 5 mM CDP-choline, 10 mM GlcA, 10 mM MnCl 2 , and 10 l of the enzyme source (total volume, 30 l). After incubation at 37°C for 2 h, the reaction mixtures were applied to a 1cc Sep-Pak Vac Cartridge (Millipore) preactivated sequentially with methanol, chloroform/methanol (2:1) and then methanol again and subsequently equilibrated with water. The column was washed with 50 ml of water and eluted with 0.9 ml of methanol and 1.0 ml of chloroform/methanol (2:1). The eluant was evaporated by SpeedVac and resuspended with 30 l of chloroform/methanol (2:1). The product was applied to a silica gel 60 high performance thin-layer chromatography plate (Merck) and devel-oped with chloroform/methanol/0.5% CaCl 2 (55:45:10). The glycolipids were visualized with orcinol reagent, and the mobility was compared with that of GlcA-LNnT-Cer purchased from Wako Pure Chemicals. The position and radioactivity of the product were estimated with a Fujix BAS 2000 Bioimage Analyzer (Fuji Photo Film).
Assays of Glucuronyltransferase Activity to Pyridylaminated Glycans-For the expression of the soluble form of GlcAT-D, 10 g of pPROTA-GlcAT-D or pPROTA plasmid was transfected to COS-7 as described above. After 3 days, 20 l of the 50% suspension of IgG-Sepharose (Pharmacia) was added to 1 ml of culture medium and rotated overnight at 4°C. The beads were washed twice with reaction buffer and used as an enzyme source.
Glucuronyltransferase activity to pyridylaminated glycans was measured in 0.1 M sodium cacodylate, pH 6.5, 0.1 mM pyridylaminated glycan, 1 mM UDP-GlcA, 5 mM CDP-choline, 10 mM GlcA, 20 mM MnCl 2 , and 10 l of the enzyme source (total volume, 30 l). After incubation at 37°C for 4 -16 h, the reactions were stopped by boiling for 5 min. After centrifugation 10,000 ϫ g for 5 min, 10 l of the supernatant was subjected to HPLC analysis using PALPAK Type N column (Takara), equilibrated, and eluted with acetonitrile/200 mM acetic acid-triethylamine, pH 7.3 (55:45) at a flow rate of 1.0 ml/min at 40°C. The elution profile was monitored with a fluorescence spectrophotometer (Shimadzu). The amounts of the products were determined from their fluorescence intensities using PA sugar chain 041 (Takara).
␤-Glucuronidase Digestion of the Reaction Products-The products from the glucuronyltransferase reaction using Gal␤1-4GlcNAc␤1-3Gal␤1-4Glc-PA or Gal␤1-3GlcNAc␤1-3Gal␤1-4Glc-PA were isolated by HPLC using PALPAK Type N column in the conditions described above. The isolated products were digested with 1 milliunit of ␤-glucuronidase from bovine liver overnight at 37°C or with 100 units of ␤-glucuronidase from H. pomatia for 2 days at 37°C in a total volume of 50 l of 0.1 M sodium acetate buffer, pH 5.0. The digests were analyzed by HPLC using the same PALPAK type N column as that noted above.
Smith Degradation of the Reaction Products-The products from the glucuronyltransferase reaction using Gal␤1-4GlcNAc␤1-3Gal-PA or Gal␤1-3GlcNAc␤1-3Gal-PA were isolated by HPLC using PALPAK Type N column eluted with acetonitrile/200 mM acetic acid-triethylamine, pH 7.3 (66:34). The isolated products (0.5 nmol each) and intact Gal␤1-4GlcNAc␤1-3Gal-PA and Gal␤1-3GlcNAc␤1-3Gal-PA (2.5 nmol each) were subjected to Smith degradation according to the method of Hase et al. (31) with slight modification. A pyridylaminated glycan was dissolved in 20 l of 80 mM NaIO 4 in 50 mM sodium acetate buffer, pH 4.0, and left to stand at 4°C for 48 h in the dark. The reaction was terminated by adding 5 l of 20% ethylene glycol and kept at room temperature for 1 h. Aldehyde groups produced were reduced with 1 mg of sodium borohydride in 100 l of 0.1 M borate buffer, pH 8.0, overnight at room temperature. After neutralizing with acetate, the product was desalted on a Superdex Peptide HR 10/30 column and concentrated. Mild acid hydrolysis of the products was carried out in 100 l of 0.05 M H 2 SO 4 at 80°C for 1 h. After adding 10 l of 1 M NaOH, the product was desalted on a Superdex Peptide HR 10/30 column and concentrated. Half an aliquot of the Smith degradation products thus obtained was digested with 25 milliunits of ␤-galactosidase (bovine testes) overnight at 37°C in a total volume of 50 l of 0.1 M sodium citrate phosphate buffer, pH 4.3, containing 10% glycerol and 1% bovine serum albumin. These products were analyzed by HPLC using PALPAK Type N column eluted with acetonitrile/200 mM acetic acid-triethylamine, pH 7.3 (75: 25).

RESULTS
Cloning of a New Glucuronyltransferase-Our preliminary experiment of the enzymatic characterization of glycosyltransferases from rat embryo suggested that a glucuronyltransferase other than GlcAT-P, which is involved in the biosynthesis of HNK-1 epitope, may be expressed in the rat embryonic nervous system. To clone new members of the glucuronyltransferase gene family, we designed two degenerate primers to the sequences conserved in motifs II and IV of rat GlcAT-P and human GlcAT-I. The PCR using these primers with rat E13 brain cDNA as a template resulted in the amplification of a product around 380 base pairs in length. After subcloning the PCR product, we sequenced and characterized 10 individual clones: one clone was GlcAT-P; seven clones were putative rat GlcAT-I, a sequence of which had 93% identity with human GlcAT-I 2 ; and the other two clones had sequences that were similar to but distinct from those of GlcAT-P and GlcAT-I, suggesting that these clones contained a novel glucuronyltransferase (GlcAT-D) fragment. Using the PCR fragment as a probe, we screened a rat E15 brain cDNA library and obtained one clone, 58-2-2. The sequence of 58-2-2 contained a putative stop codon but no in-frame ATG. To obtain the entire coding sequence of GlcAT-D, rapid amplification of cDNA 5Ј-end was employed. The nucleotide sequence of the overlapping cDNA fragments revealed that GlcAT-D has a single open reading frame consisting of 324 amino acids, with a molecular mass of 37,177 Da, and two potential N-glycosylation sites (Fig. 1A). The predicted translation initiation site conformed to Kozak's consensus sequence (32), and the upstream region contained an in-frame stop codon. Hydropathy analysis (Fig. 1B) indicated the presence of a potential transmembrane domain at the Nterminal region (from Ala-4 to Val-25), suggesting that the protein has type II transmembrane topology, which has been found in almost all glycosyltransferases cloned to date. The domain (from Asp-26 to Pro-81) next to the transmembrane region was characterized by its high proline content (20%; 11 of 56 amino acids), as seen in several other glycosyltransferases including rat GlcAT-P and human GlcAT-I. The putative catalytic domain (from Leu-80 to Val-324) showed a very high sequence identity with those of both rat GlcAT-P and human GlcAT-I (Fig. 2). A data base search indicated that a hypothetical protein of Drosophila melanogaster (accession number AL033125) has high homology with these three GlcATs in addition to those of C. elegans and S. mansoni, as pointed out by Terayama et al. (21).
Northern Blot Analysis-Of the adult rat tissues examined, GlcAT-D was expressed only in the brain (Fig. 3A). Three transcripts of 1.1 kb (major), 2.3 kb (minor), and 4.0 kb (faint) were detected in the adult whole brain (Fig. 3A), cerebral cortex, and cerebellum (Fig. 3B). In the postnatal (P1 and P7) cerebral cortex, the 1.1-kb transcript was more intense than the 2.3-kb one, whereas these two transcripts in the corresponding cerebellum showed almost equal intensity. In the E18 brain, the 1.1-and 2.3-kb transcripts showed similar intensity. A faint single 2.3-kb transcript was detected in the E13 brain.
Localization of HNK-1 Epitope and Gene Expression of GlcATs in Embryonic Brain-HNK-1 immunohistochemistry was performed on E13.5 rat brain sections. Intense staining was observed in the outer layer (preplate) of the cerebral cortex, the lateral and medial ganglionic eminences, and the broad subventricular area of the basal ganglia. Specific staining was also detected in the retina and slightly in the lens (Fig. 4A). Adjacent brain sections were used for in situ hybridization to detect the in vivo expression patterns of GlcAT-D and GlcAT-P. Expression of GlcAT-D was observed in the preplate of the cerebral cortex and ventral regions of the basal ganglia (Fig.  4B). A focal staining spot was also noted in the pallidal subventricular zone (arrow in Fig. 4B). GlcAT-P basically showed similar expression in the telencephalon (arrowheads in Fig.  4C), although no signal was observed in the pallidum (compare with arrow in Fig. 4B). In the eye primordium, GlcAT-D mRNA was detected in the retina, whereas GlcAT-P was very weakly expressed in the lens. The regions that expressed GlcAT-D and/or GlcAT-P were always HNK-1-positive, although transcripts of both GlcATs were below the detection level in the putative piriform cortex (white arrowheads in Fig. 4C).

Expression of GlcAT-D in COS-7 and Characterization as Glucuronyltransferase Involved in Biosynthesis of HNK-1
Epitope-To prove the enzymatic activity of the cDNA product, expression plasmid containing GlcAT-D cDNA was transfected into COS-7 cells. The cell homogenate was assayed for glucuronyltransferase activity using a glycoprotein and a glycolipid, both of which contained Gal␤1-4GlcNAc sequences at nonreducing termini, as acceptor substrates. As shown in Table I, significant activity was detected using both glycoprotein and glycolipid acceptors. Cells transfected with pCDM8 vector showed much less activity. As shown in Fig. 5, the reaction product using a glycolipid, LNnT-Cer, as an acceptor was detected on high performance thin-layer chromatography at a position corresponding to GlcA-LNnT-Cer. Moreover, a signif-  icant portion of COS-7 cells transfected with GlcAT-D cDNA were stained with anti-CD57, showing that these cells expressed HNK-1 epitope (Fig. 6). Cells transfected with pCDM8 vector did not express the epitope. These results are consistent with those of Terayama et al. (21), who showed that COS-1 cells transfected with GlcAT-P cDNA expressed HNK-1 epitope. The results of expression in COS-7 indicate that GlcAT-D is a glucuronyltransferase involved in the biosynthesis of HNK-1 epitope.
Expression of a Soluble Form of GlcAT-D-To clarify the function of the cDNA product, a soluble form of GlcAT-D was generated by fusing the putative stem and catalytic domain of the protein to a secreted form of the protein A IgG-binding domain. The fused protein was expressed in COS-7 cells and absorbed on IgG-Sepharose beads from culture medium, and then the enzyme-bound beads were used as an enzyme source.
Characterization as a Galactoside ␤1,3-Glucuronyltransferase-Reactions using pyridylaminated glycans as acceptors were performed on a large scale to identify the products, and those isolated by HPLC were subjected to further analyses.
To prove whether a GlcA was transferred to the C-3 position of Gal of Gal␤1-4GlcNAc and Gal␤1-3GlcNAc sequence by Smith degradation, the glucuronyltransferase reactions using Gal␤1-4GlcNAc␤1-3Gal-PA and Gal␤1-3GlcNAc␤1-3Gal-PA as acceptors were performed on a large scale. The products were detected in the reaction mixture using both trisaccharide-PAs as acceptors (data not shown), and isolated by HPLC for further analyses. Matrix-assisted laser desorption/ionization time-of-flight mass spectrometry of both products showed a molecular ion [M-Hϩ2Na] ϩ at m/z 843. Smith degradation (SD) of the intact acceptors and isolated products was carried out as described under "Experimental Procedures," and the predicted products are shown in Fig. 9. The terminal galactoses of Gal␤1-4GlcNAc␤1-3Gal-PA and Gal␤1-3GlcNAc␤1-3Gal-PA are cleaved by periodate, and thus the SD products are both Glc-NAc␤1-3-L-threosyl-PA (Fig. 9A). When the GlcA was transferred to the C-3 position of the nonreducing terminal galactose by GlcAT-D, as O-3-substituted Gal was not cleaved, the SD products of the enzyme products using Gal␤1-4GlcNAc␤1-3Gal-PA and Gal␤1-3GlcNAc␤1-3Gal-PA as acceptors were Gal␤1-4GlcNAc␤1-3-L-threosyl-PA (Fig. 9B) and Gal␤1-3Glc-NAc␤1-3-L-threosyl-PA (Fig. 9C), respectively. When the GlcA was linked at any position other than the C-3 position of galactose, the galactose was cleaved, and thus the SD products were both GlcNAc␤1-3-L-threosyl-PA (Fig. 9A). Fig. 10 shows the HPLC analysis of the SD products. The SD products of the two acceptors were eluted at the same retention time (6.1 min, Fig. 10, A and D), whereas two species of the SD products of the enzyme reaction products were eluted at different times from those of the acceptors (Fig. 10, B and E) and eluted at the same time (6.1 min) after ␤-galactosidase digestion (Fig. 10, C and  F). These results indicate that the SD products of the glucuronyltransferase reaction products have a ␤-linked Gal at their nonreducing termini and therefore that the C-3 position of Gal of the enzyme reaction products is substituted with a GlcA. In conclusion, the cloned GlcAT-D is characterized as a galactoside ␤1,3-glucuronyltransferase.
a ND, not detected (less than 0.1 pmol/h/ml medium).

DISCUSSION
We cloned a new member of the glucuronyltransferase gene family that transfers a GlcA to the glycan chain terminated by a Gal residue. We named the cloned enzyme GlcAT-D, meaning a glucuronyltransferase with dual specificity for both glycolipid and glycoprotein acceptors as mentioned below. The GlcAT-D showed high homology with the two glucuronyltransferases cloned to date, rat GlcAT-P (21) and human GlcAT-I (24) (overall amino acid identity of 50.0 and 45.7%, respectively). These three enzymes may constitute a family of galactoside ␤1,3glucuronyltransferases, which has high homology with hypothetical proteins of D. melanogaster, C. elegans, and S. mansoni (Fig. 2). In addition, a data base search suggested that more than five additional isoforms have been identified in C. elegans and that plant Arabidopsis thaliana and rice might have glucuronyltransferases homologous to this family (data not shown). The galactoside ␤1,3-glucuronyltransferase appears to be commonly conserved from plant, nematode, and insect to mammal.
The ␤-glucuronidase digestion and Smith degradation of the glucuronyltransferase reaction products demonstrated that the GlcAT-D catalyzed the formation of GlcA␤1-3Gal sequence, which is consistent with the determined structure of glycans containing the HNK-1 epitope, sulfated GlcA␤1-3Gal (16)(17)(18)(19). For the first time in this cloned glucuronyltransferase family, the enzyme was confirmed to be a galactoside ␤1,3-glucuronyltransferase.
The GlcAT-D catalyzed the transfer of a GlcA to both glycoprotein and glycolipid containing Gal␤1-4GlcNAc sequence at nonreducing termini. The transfer is assumed to form the precursor of HNK-1 epitope, which is used by HNK-1-sulfotransferase as an acceptor substrate. We observed that the COS-7 cells transfected with the cloned cDNA expressed HNK-1 epitope and that the sites in situ detected with the transcript were also stained with HNK-1 antibody. These results support the idea that this enzyme is a glucuronyltransferase involved in the biosynthesis of HNK-1 epitope.
The glucuronyltransferase activity involved in the biosynthesis of HNK-1 epitope has been studied using brain extracts from chick embryo (33), rat embryo (34), and postnatal rat (35,36) as well as purified enzyme from postnatal rat forebrain (20). The present study showed that GlcAT-D acts on both glycoprotein and glycolipid acceptors in vitro, whereas GlcAT-P has been reported to be specific to glycoprotein acceptors (20,35). It remains to be determined what these glucuronyltransferases use as acceptor substrates in the nervous system. Our study examined the substrate specificity as to the carbohydrate sequence using pyridylaminated glycans as acceptor substrates. The GlcAT-D transferred a GlcA to not only type 2 (Gal␤1-4GlcNAc) but also type 1 (Gal␤1-3GlcNAc) glycan chains, suggesting that the HNK-1 epitope expressed on type 1 glycan chain may occur, although the existence of such glycans has not previously been reported. This result is in conflict with those of previous studies, which showed that glucuronyltransferase from rat brain extract has no or little activity transferring a GlcA to a glycolipid containing type 1 glycan (34,36). Because type 3 (Gal␤1-3GalNAc) glycan did not serve as an acceptor of GlcAT-D, it remains to be examined whether the GlcAT-D can transfer a GlcA to Gal␤1-3Gal, which is a good acceptor for GlcAT-I (24). In addition, the GlcAT-D did not transfer a GlcA to Fuc␣1-2Gal␤1-4GlcNAc or Gal␤1-4(Fuc␣1-3)GlcNAc, the product formed by ␣1,2or ␣1,3-fucosyltransferase acting on type 2 glycan chain, suggesting that the glucuronyltransferase may also compete with these fucosyltransferases for acceptor substrates in vivo.
The Northern blot analysis showed that the ratio of the two major transcripts (1.1 and 2.3 kb) changes according to the developmental stage. Combined with the structure of the cloned cDNA, we speculated that the major initiation site of the transcription might differ between the early embryo and adult brain, although the structure of transcripts is yet to be analyzed. Smith degradation was carried out as described under "Experimental Procedures." The Smith degradation products of Gal␤1-4GlcNAc␤1-3Gal-PA (A) and Gal␤1-3GlcNAc␤1-3Gal-PA (D) and the glucuronyltransferase reaction products of Gal␤1-4GlcNAc␤1-3Gal-PA (B) and Gal␤1-3Glc-NAc␤1-3Gal-PA (E) were subjected to HPLC analysis using PALPAK type N column. After ␤-galactosidase digestion, the Smith degradation products of glucuronyltransferase reaction products of Gal␤1-4Glc-NAc␤1-3Gal-PA (C) and Gal␤1-3GlcNAc␤1-3Gal-PA (F) were analyzed with the same HPLC column.
In situ hybridization and immunohistochemistry on the developing brain revealed that regions that expressed GlcAT-D and/or GlcAT-P were always HNK-1-positive. This fact, along with the above-mentioned in vitro data, indicates that both GlcATs are involved in the synthesis of the HNK-1 epitope in vivo. The GlcAT-D expression pattern was slightly different from that of GlcAT-P. GlcAT-D transcripts were positive in the pallidum and retina where GlcAT-P expression was not specifically observed, suggesting different in vivo functions between the two GlcATs. GlcAT-D had catalytic activity on both glycoprotein and glycolipid acceptors in our in vitro assay, whereas GlcAT-P was reported to be active on only glycoprotein acceptors (20). This may reflect different localization patterns of the GlcATs, where the GlcAT-D is expressed in a wider region than GlcAT-P. It should be noticed that there were some HNK-1positive areas where both GlcATs were negative. This observation suggests several possible mechanisms: 1) transcripts below the detection level by in situ hybridization, 2) existence of unknown GlcAT genes, or 3) proliferation or migration of HNK-1-positive neurons into transcript-negative areas. Further studies will reveal the mechanism of spatiotemporally restricted expression of HNK-1 epitope.