Formation of HNK-1 Determinants and the Glycosaminoglycan Tetrasaccharide Linkage Region by UDP-GlcUA:Galactose β1,3-Glucuronosyltransferases*

While expression-cloning enzymes involved in heparan sulfate biosynthesis, we isolated a cDNA that encodes a protein 65% identical to the UDP-GlcUA:glycoprotein β1,3-glucuronosyltransferase (GlcUAT-P) involved in forming HNK-1 carbohydrate epitopes (3OSO3GlcUAβ1,3Gal-) on glycoproteins. The cDNA contains an open reading frame coding for a protein of 335 amino acids with a predicted type II transmembrane protein orientation. Cotransfection of the cDNA with HNK-1 3-O-sulfotransferase produced HNK-1 carbohydrate epitopes in Chinese hamster ovary (CHO) cells and COS-7 cells. In vitro, a soluble recombinant form of the enzyme transferred GlcUA in β-linkage to Galβ1,3/4GlcNAcβ-O-naphthalenemethanol, which resembles the core oligosaccharide on which the HNK-1 epitope is assembled. However, the enzyme greatly preferred Galβ1,3Galβ-O-naphthalenemethanol, a disaccharide component found in the linkage region tetrasaccharide in chondroitin sulfate and heparan sulfate. During the course of this study, a human cDNA clone was described that was thought to encode UDP-GlcUA:Galβ1,3Gal-R glucuronosyltransferase (GlcUAT-I), involved in the formation of the linkage region of glycosaminoglycans (Kitagawa, H., Tone, Y., Tamura, J., Neumann, K. W., Ogawa, T., Oka, S., Kawasaki, T., and Sugahara, K. (1998) J. Biol. Chem.273, 6615–6618). The deduced amino acid sequences of the CHO and human cDNAs are 95% identical, suggesting that they are in fact homologues of the same gene. Transfection of a CHO cell mutant defective in GlcUAT-I with the hamster cDNA restored glycosaminoglycan assembly in vivo, confirming its identity. Interestingly, transfection of the mutant with GlcUAT-P also restored glycosaminoglycan synthesis. Thus, both GlcUAT-P and GlcUAT-I have overlapping substrate specificities. However, the expression of the two genes was entirely different, with GlcUAT-I expressed in all tissues tested and GlcUAT-P expressed only in brain. These findings suggest that, in neural tissues, GlcUAT-P may participate in both HNK-1 and glycosaminoglycan production.

Glucuronic acid has been found in several types of complex carbohydrates expressed by vertebrate and invertebrate cells, including the sulfated glycosaminoglycan chains of proteoglycans and HNK-1 carbohydrate epitopes (3OSO 3 GlcUA␤1,3Gal-R). HNK-1 (human natural killer cell carbohydrate antigen-1) was originally described on human natural killer cells (1), but later studies showed that it was present in greatest abundance in the nervous system on subsets of glycolipids (2,3), glycoproteins (4), and proteoglycans (5). In contrast, glycosaminoglycans are ubiquitously distributed among tissues, usually in covalent linkage to proteoglycan core proteins (6). Both HNK-1 and glycosaminoglycans can bind a variety of proteins that participate in cell-cell, cell-extracellular matrix, and cell signaling during development (4,6). The presence of GlcUA in both types of glycans and the partial overlap in their ligand binding properties suggest the possibility that their synthesis may be coordinated as well.
Much information is available about the enzymes involved in the addition of GlcUA to these glycans. Assembly of HNK-1 occurs by the transfer of GlcUA from the high energy donor UDP-GlcUA to a terminal galactose residue linked ␤1,4 to GlcNAc, followed by sulfation of the GlcUA residue at C-3. The glucuronosyltransferase associated with HNK-1 was first demonstrated in embryonic chick brain extracts using neolactotetraosylceramide as acceptor (7). The same activity was found later in rat brain using both neolactotetraosylceramide and asialoorosomucoid as substrates (7)(8)(9). By partially purifying the enzymes and noting differences in phospholipid activation and pH dependence, Oka et al. (9) concluded that the glucuronosyltransferase involved in the synthesis of HNK-1 epitopes on glycoproteins (GlcUAT-P) 1 differs from the one that acts on glycolipids (GlcUAT-L). This hypothesis was confirmed recently in studies of recombinant GlcUAT-P, which selectively adds GlcUA to glycoprotein substrates (10). The gene and cDNA encoding GlcUAT-L have not yet been identified.
Glycosaminoglycan biosynthesis begins by the formation of the tetrasaccharide linkage intermediate -GlcUA␤1,3Gal␤1, 3Gal␤1,4Xyl␤-O-Ser. This intermediate serves as the primer for heparan sulfate and chondroitin sulfate assembly, which arises from the alternating addition of ␤-GlcNAc and ␤-GlcUA or ␤-GalNAc and ␤-GlcUA residues, respectively, to the linkage tetrasaccharide. Three GlcUA-transferases are thought to catalyze the addition of GlcUA: one involved in the formation of the linkage region tetrasaccharide (GlcUAT-I) (11,12) and two that polymerize the different chains (13). The latter activities may be part of bifunctional enzymes in which the same protein catalyzes the alternating addition of a HexNAc residue and GlcUA (14,15). GlcUAT-I, in contrast, is much like GlcUAT-P in that it transfers GlcUA from UDP-GlcUA to a ␤-linked Gal residue. The enzyme was first described in embryonic chick cartilage (12) and partially purified from embryonic chick brain (11) and a mouse mastocytoma (16). Interestingly, these early studies showed that crude enzyme preparations transferred GlcUA not only to substrates derived from the linkage region, such as Gal␤1,3Gal and Gal␤1,3Gal␤1,4Xyl, but also to lactose (Gal␤1,4Glc) and N-acetyllactosamine (Gal␤1,4GlcNAc), the precursor of HNK-1.
These findings raised the question of whether formation of the linkage region and HNK-1 determinants is catalyzed by the same enzyme. Curenton et al. (17) provided evidence that GlcUAT-I is distinct from the enzyme involved in HNK-1 formation based on partial separation of the activities and substrate competition studies. The cloning of a cDNA for GlcUAT-P confirmed that at east two enzymes exist, but detailed analysis of substrate specificity was not done. In the present report, we isolated a cDNA encoding a hamster glucuronosyltransferase that is 65% identical to GlcUAT-P (10) and 95% identical to human GlcUAT-I (18), which was cloned while these experiments were under way. Analysis of the recombinant enzymes showed significant overlap in substrate specificity, and transfection experiments revealed that both enzymes will produce HNK-1 carbohydrate epitopes and facilitate glycosaminoglycan biosynthesis.

EXPERIMENTAL PROCEDURES
Cell Cultures-Chinese hamster ovary (CHO-K1, ATCC CCL-61), COS-7 (ATCC CRL-1651), and Lec2 (ATCC CRL-1736) cells were obtained from the American Type Culture Collection (Manassas, VA). Mutants pgsG-110, -114, and -224 were isolated by direct selection of glycosaminoglycan-deficient CHO-K1 cells and will be described in greater detail elsewhere. 2 Lec2-GlcUAT-P is a subclone of Lec2 stably expressing HNK-1 GlcUAT-P and was kindly provided by E. Ong and M. Fukuda (Burnham Institute, La Jolla, CA). All of the cell lines were grown under an atmosphere of 5% CO 2 in air and 100% relative humidity. CHO cells and the various transfectants were maintained in Ham's F-12 growth medium (Hyclone Laboratories) supplemented with 7.5% (v/v) fetal bovine serum (Hyclone Laboratories), 100 g/ml streptomycin sulfate, and 100 units/ml penicillin G. Sulfate-free medium was prepared from individual components (19), substituting chloride salts for sulfate and fetal bovine serum that had been dialyzed exhaustively against phosphate-buffered saline (20). COS-7 cells were maintained in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 10% fetal bovine serum and antibiotics. Lec2 cells were maintained in ␣-minimal essential medium (Life Technologies, Inc.) supplemented with 10% fetal bovine serum and antibiotics. Lec2-GlcUAT-P cells were maintained in complete medium with 0.2 mg/ml (active) G418 (Life Technologies, Inc.).
Cloning of a Novel Glucuronosyltransferase from Chinese Hamster Ovary Cells-pgsD-H661, a CHO mutant defective in heparan sulfate biosynthesis (14), was stably transfected with a CHO-K1 cDNA library in pcDNA1 (Invitrogen) and screened for restoration of heparan sulfate biosynthesis. 3 A PCR fragment was prepared using the genomic DNA from the correctant as a template and the SP6 and T7 flanking sequences of the integrated vector as primers. PCRs were carried out with Taq DNA polymerase (Life Technologies, Inc.) in a Perkin-Elmer Model 2400 thermal cycler (35 cycles at 94°C for 1 min, 55°C for 2 min, and 72°C for 3 min, with a final incubation at 72°C for 10 min). The PCR fragment was cloned into pGEM-T (Promega), and the sequence was determined on both strands by the dideoxy chain termination method using Taq polymerase (dye terminator cycle sequencing, Perkin-Elmer) with a DNA automatic sequencer (ABI PRISM genetic analyzer).
cDNA Library Screening and Hybridization-Approximately 6 ϫ 10 5 colonies from a CHO-K1 cDNA library in pcDNA1 were transferred to Duralon-UV TM membranes (Stratagene) and fixed by alkaline lysis as recommended by the manufacturer. The filters were prehybridized in a solution containing 0.02 M PIPES, 0.8 M NaCl, 50% (v/v) deionized formamide, 0.5% SDS, and 100 g/ml denatured salmon sperm DNA for 2 h at 42°C. The PCR product described above was labeled with [ 32 P]dCTP by random oligonucleotide priming (Prime-IT II labeling kit, Stratagene) and purified with an Elute-tip (Schleicher & Schü ll). Hybridization was carried out for ϳ16 h at 42°C in the same buffer containing ϳ1 ϫ 10 6 cpm/ml 32 P-labeled probe, and positive clones were detected by conventional autoradiography. One of the plasmids obtained in this way (1B-2) contained the full-length sequence and was named pcDNA1-GlcUAT-X in initial experiments. This was later shown to encode a section of GlcUAT-I and therefore was renamed GenBank™/ EBI accession number AF113703. Mouse multiple-tissue poly(A) ϩ RNA (CLONTECH) was hybridized using gel-purified full-length cDNA for GlcUAT-I as a probe essentially according to the manufacturer's recommendation. Briefly, the solution was prewarmed to 68°C, and the blot was prehybridized for 30 min. The Expresshyb solution was replaced with fresh solution containing 1-2 ϫ 10 6 cpm/ml 32 P-labeled probe and hybridized at 68°C for 1 h. The blot was rinsed and washed for 30 -40 min at room temperature in 2ϫ SSC containing 0.05% SDS with several changes of buffer and then for 40 min at 50°C in 0.1ϫ SSC containing 0.1% SDS, with one change of solution. Hybridization was detected with a PhosphorImager (Storm 860, Molecular Dynamics, Inc.).

Expression of the Protein A-GlcUAT-I and Protein
A-GlcUAT-P Fusion Proteins-The cDNA fragment encoding amino acids 30 -335 of GlcUAT-I (the putative stem region and catalytic domain) was prepared by PCR using pcDNA1-GlcUAT-I as a template. The fragment was fused in frame to the C terminus of protein A in pRK5-F10-PROTA (21). The 5Ј-primer for PCR was GGCGAATTCACCATGTGACTGCCTC-CTCC, and the 3Ј-primer was GGCGAATTCCAGTCCCACAAGGTAT-GTGCC (the EcoRI site is shown in boldface letters, and the coding sequence of GlcUAT-I is underlined). PCR was carried out with Pfu polymerase (CLONTECH; 25 cycles at 94°C for 1 min, 55°C for 1 min, and 72°C for 2 min, followed by a final incubation at 72°C for 7 min). The PCR products were cloned into pCR-Script Amp SK(ϩ) (Stratagene), and the sequences were determined. The clone with the correct sequence was digested by EcoRI and ligated into the EcoRI site of pRK5-F10-PROTA to yield pPROTA-GlcUAT-I.
The cDNA fragment encoding amino acids 39 -347 of rat GlcUAT-P (GenBank TM /EBI accession number D88035) (10) was prepared by PCR using pcDNA3-GlcUAT-P as a template (kindly provided by M. Fukuda). The 5Ј-primer was TCCGGAATTCCCAGAGCAGCCTCGCA-CCT, and the 3Ј-primer was GCCCTCGAGTGTGTAGTTTCAGATCT-CCAC (the EcoRI and XhoI sites are shown in boldface letters). Expression of soluble recombinant enzyme was measured after transfection of COS-7 cells using LipofectAMINE (Life Technologies, Inc.) according to the manufacturer's instructions. The supernatant was centrifuged for 5 min at top speed in an IEC clinical centrifuge at 4°C to sediment cell debris. The supernatant was collected and incubated with rabbit IgGagarose beads (10 l of beads/ml of sample; Sigma) with end-over-end mixing at 4°C for 24 -48 h. The samples were centrifuged for 5 min, and the supernatant was aspirated. The beads were washed twice with 10 ml of 20% (v/v) glycerol and 50 mM Tris-Cl, pH 7.4, and resuspended in the same buffer containing protease inhibitors (10 mM phenylmethylsulfonyl fluoride, 1 g/ml leupeptin, and 1 g/ml pepstatin) to achieve an ϳ50% (v/v) slurry. The immobilized enzyme was stable at 4°C for at least 4 months.

Isolation of Cell Lines Stably Expressing GlcUAT-I and GlcUAT-P-
The cDNA insert of pcDNA1-GlcUAT-I was digested by HindIII and XhoI and cloned into pcDNA3 (Invitrogen), yielding pcDNA3-GlcUAT-I. Lec2, wild-type CHO-K1, and pgsG mutant cells were transfected with pcDNA3-GlcUAT-I or pcDNA3-GlcUAT-P using LipofectAMINE, and stable transfectants were selected using 0.4 mg/ml (active) G418. Individual colonies were screened for HNK-1 expression or glycosaminoglycan production, and positive ones were isolated with glass cloning rings and expanded.
Immunofluorescence Staining of Cells by Anti-HNK-1 Antibody-Cells were transfected with pcDNA3-GlcUAT-I and/or pcDNA3-HNK-1 3OST (where 3OST is 3-O-sulfotransferase; M. Fukuda). Two days later, the monolayers were washed twice with cold PBS and fixed at 4°C for 15 min with 4% (v/v) paraformaldehyde. Cells were washed twice with PBS, blocked at room temperature for 10 min with 2% (w/v) bovine serum albumin in PBS, and incubated at room temperature for 30 -40 min with mouse monoclonal anti-HNK-1 antibody (Becton Dickinson Advanced Cellular Biology) diluted 1:100 in buffer. Primary an-tibody was removed, and the cells were incubated at room temperature for 20 -30 min with fluorescein isothiocyanate-conjugated goat antimouse IgM antibody (Sigma) diluted 1:100 with buffer. Cells were washed again with PBS, mounted with antifade reagent (Molecular Probes, Inc.), and examined by fluorescence microscopy using a Zeiss epifluorescence microscope equipped with a fluorescein isothiocyanate filter.
Western Analysis of Transfected Cells Expressing HNK-1 Carbohydrate Epitopes-Cells were transiently transfected with pcDNA3-HNK-1 3OST. Two days later, they were scraped from the dishes, and 100 g of cell protein was separated by electrophoresis on a 10% SDS-polyacrylamide gel. After transfer onto nitrocellulose membranes (Bio-Rad), the blot was blocked at 4°C overnight with 4% (w/v) skim milk in PBS and then incubated at room temperature for 4 h with anti-HNK-1 antibody diluted 1:500 and for 1 h with peroxidase-conjugated goat anti-mouse IgM antibody (Sigma) diluted 1:500. Bound antibody was visualized with diaminobenzidine reagent (Aldrich). Some samples were digested with peptide N-glycosidase F (22).

Synthesis of UDP-[ 3 H]GlcUA-UDP-[ 3 H
]GlcUA was synthesized as described previously with slight modification (23). All of the enzymes were purchased from Boehringer Mannheim. D-[1-3 H]Glucose (10 Ci/ mmol; NEN Life Science Products) was dried (3.5 mCi, ϳ180 nmol) and dissolved in a solution (500 l) containing 10 mM MgCl 2 , 1 mM ATP, 4 mM UTP, 2 mM NAD ϩ , 10 mM D-glucose 1,6 diphosphate, 2 units/ml hexokinase, 4 units/ml inorganic pyrophosphatase, 1 unit/ml UDPglucose pyrophosphorylase, 0.3 units/ml UDP-glucose dehydrogenase, and 50 mM Tris-HCl, pH 8.35. After overnight incubation at 25°C, the samples were boiled for 5 min and centrifuged to remove precipitated protein. The products in the supernatant were separated for 10 h by descending paper chromatography using ethanol and 1 M sodium acetate in water (7:3, v/v) as the solvent. Material that comigrated with the authentic UDP-GlcUA was eluted from the paper with 10% (v/v) ethanol/water, lyophilized, and dissolved in 100 l of buffer containing 10 units of calf intestinal phosphatase (Life Technologies, Inc.). After 45 min at 37°C, the sample was heated to 75°C for 10 min, diluted with 1 ml of 0.1 N NH 4 HCO 3 , and loaded onto a 0.5-ml column of AG 1-X2 acetate resin (Bio-Rad). The column was washed with 3 ml of water and 5 ml of 0.25 N NH 4 HCO 3 , and UDP-GlcUA was eluted with 3 ml of 1 M NH 4 HCO 3 . Enough Dowex 50-X8 (H ϩ form, Bio-Rad) was added to neutralize the sample. After centrifugation, the supernatant was dried by lyophilization, dissolved in deionized water, and dried again. The final product was dissolved in 70% ethanol/water and stored at Ϫ20°C for future use.
In Vitro Assay of GlcUAT-I, GlcUAT-P, and HNK- and Gal␤1,3Gal␤-O-benzyl were chemically synthesized from individual monosaccharide units using standard orthogonal blocking chemistry, coupling, and deblocking strategies. The details of their synthesis, purification, and chemical characterization will be published elsewhere. 4 All compounds were Ͼ98% pure by 1 H NMR, 13 C NMR, and thin-layer chromatography. Gal␤1,3GalNAc␣-O-benzyl, monosaccharide glycosides, phosphatidylinositol, phosphatidylserine, asialofetuin (type II), and asialomucin were purchased from Sigma. N-Acetylheparosan oligosaccharides were prepared previously from Escherichia coli K5 capsular polysaccharide (24). Neolactotetraosylceramide was kindly provided by F. Jungalwala (Harvard Medical School), and neolactohexaose was provided by M. Fukuda. Unless otherwise indicated, the standard reaction (25 l) contained 10 l of IgG bead slurry (50%) containing immobilized enzyme, 0.1 Ci of UDP-[ 3 H]GlcUA, 100 M UDP-GlcUA, 0.03-10 mM acceptor, 100 mM HEPES, pH 6.5, 10 mM MnCl 2 , and 2.5 mM ATP. For substrates containing Gal␤1,3Gal␤-O-NM, only 2 l of the beads was used, and the reaction time was reduced to 1 h in order to work in the linear range (GlcUAT-I only). Assays for GlcUAT-P also contained only 2 l of enzyme immobilized on beads. In repetitions of the experiments when different batches of the enzyme were used, an aliquot was first analyzed by SDS-PAGE and silver staining to obtain comparable amounts of enzyme. After incubation at 37°C, the reaction products were diluted with 1 ml of 0.5 M NaCl and applied to a Sep-Pak C 18 cartridge (100 mg; Waters). After washing the cartridge with 2 ml of water, the product was eluted with 50% methanol, dried, and counted by liquid scintillation. Glycoprotein acceptors were precipitated with 10% (w/v) trichloroacetic acid, and the pellets were washed twice with 5% trichloroacetic acid. Transferase activity using N-acetylheparosan acceptors was assayed using anion-exchange chromatography as described previously (24).
␤-Glucuronidase Digestion-Radioactive product was dried and dissolved in 60 l of solution containing 1.33 g/l sodium taurodeoxycholate, 0.1 M sodium acetate, pH 5.0, and 2.5 units/l ␤-glucuronidase from limpets or Helix pomatia (Sigma), and the mixture was incubated overnight at 37°C. The sample was diluted with 1 ml of 0.5 M NaCl and separated on a Sep-Pak C 18 cartridge as described above. The flowthrough fraction, wash, and eluant were collected, and radioactivity was measured by liquid scintillation spectrometry.
Purification of Glycosaminoglycan Chains-Cells were labeled for 24 h with 10 Ci/ml H 2 35 SO 4 (1600 Ci/mmol; NEN Life Science Products) in sulfate-free medium. 35 S-Labeled glycosaminoglycan chains were isolated by anion-exchange chromatography as described previously (26) and analyzed by anion-exchange HPLC using a 7.5 mm (inner diameter) ϫ 7.5-cm column of DEAE-3SW (TosoHaas, Montgomeryville, PA). The column was equilibrated in 10 mM KH 2 PO 4 buffer, pH 6.0, containing 0.2% (w/v) Zwittergent 3-12 and 0.2 M NaCl. The glycosaminoglycans were eluted with a linear gradient of NaCl (0.2-1 M) in the same buffer using a flow rate of 1 ml/min and by increasing the NaCl concentration by 10 mM/min. The effluent from the column was monitored for radioactivity with an in-line radioactivity detector (Radiomatic Flo One/beta, Packard Instrument Co.) with sampling rates every 6 s. The data were averaged over 1-min intervals.

Isolation of a cDNA Clone for a Glucuronosyltransferase from
Chinese Hamster Ovary Cells-During a series of experiments to identify genes involved in glycosaminoglycan synthesis (see "Experimental Procedures"), we found a cDNA clone that showed high homology to the recently cloned rat glycoprotein glucuronosyltransferase involved in the assembly of HNK-1 epitopes (GlcUAT-P) (10). This clone, originally designated GlcUAT-X, contained sequence homologous to the C-terminal ectodomain of GlcUAT-P, but lacked an ATG start codon. Using the insert as a probe, we screened a commercial CHO-K1 cDNA library and isolated six different clones. Two of the longest cDNAs were potentially full-length based on the presence of putative start and stop codons. DNA sequencing revealed a single open reading frame with a potential Kozak consensus sequence for ribosome recognition just upstream from the ATG start codon and a polyadenylation signal located close to a poly(A) run (Fig. 1). The open reading frame encoded a 335amino acid protein. Kyte-Doolittle hydropathy analysis (38) indicated one potential transmembrane domain consisting of 18 hydrophobic amino acid residues located 7 amino acids from the initiating Met residue. A relatively proline-rich segment followed the hydrophobic section, and one potential N-glycosylation site in the putative ectodomain was present at Asn-299 (NCT, marked by an asterisk). These characteristics are common to type II transmembrane proteins and to many known Golgi glycosyltransferases (27). Overall, the sequence exhibited 65% identity to rat GlcUAT-P, and therefore, we tentatively characterized the cDNA as a homologue of this enzyme.
GlcUAT-X was expressed in various adult mouse tissues as measured by Northern blot analysis using the full-length cDNA as a probe. As shown in Fig. 2, a single transcript of ϳ1.8 kilobase pairs was obvious in adult liver, brain, and heart; was moderately expressed in lung, skeletal muscle, kidney, and testis; and was barely present in spleen. This distribution differs significantly from the expression of GlcUAT-P, which has two transcripts, one at 4.0 kilobase pairs and a minor one at 9.0 kilobase pairs, expressed strongly in brain (10). Furthermore, most non-neural tissues do not normally express HNK-1 epitopes, suggesting that GlcUAT-X might not participate in HNK-1 formation. As these studies were under way, Kitagawa et al. (18) cloned a GlcUA-transferase cDNA from a human placenta cDNA library using PCR and degenerate primers based on the sequence of GlcUAT-P. Comparing its sequence (GenBank TM /EBI accession number AB009598) with that of GlcUAT-X showed that they were 95% identical at the amino acid level (Fig. 3), suggesting that they most likely represent the same enzyme from different species. Based on the ability of the transferase to attach GlcUA to Gal␤1,3Gal␤1,4Xyl␤-O-Ser and its lack of activity with glycoprotein substrates, Kitagawa et al. concluded that the enzyme was involved in forming the linkage tetrasaccharide present in glycosaminoglycans, such as heparan sulfate and chondroitin sulfate. This enzyme is known as GlcUA-transferase I (GlcUAT-I) to distinguish it from other GlcUA-transferases involved in polymerization of glycosaminoglycans (13). The high homology of GlcUAT-X (which is designated GlcUAT-I below) and GlcUAT-P raised the question of whether these transferases might participate in both HNK-1 synthesis as well as glycosaminoglycan formation.

GlcUAT-I Can Generate HNK-1 Determinants in COS-7 and Lec2
Cells-To test if GlcUAT-I can produce HNK-1 epitopes, we cotransfected the cDNA encoding CHO GlcUAT-I and the human HNK-1 3-O-sulfotransferase (25) into COS-7 and Lec2 cells. Lec2 cells lack the Golgi CMP-sialic acid transporter, and therefore, the cells accumulate oligosaccharides that terminate with Gal residues that can serve as acceptors for GlcUA addition (10,28). Like wild-type CHO cells, Lec2 also does not express HNK-1 3-O-sulfotransferase. Therefore, control Lec2 cells and those transfected with only GlcUAT-I or HNK-1 3-Osulfotransferase did not stain with anti-HNK-1 antibody (Fig.  4, A, C, and D, respectively). However, when Lec2 cells were cotransfected with both GlcUAT-I and HNK-1 3-O-sulfotransferase, the cells stained with anti-HNK-1 antibody, suggesting that GlcUAT-I induces HNK-1 expression on cell-surface glycoconjugates (Fig. 4B). GlcUAT-I was also introduced into COS-7 cells, which have endogenous HNK-1 3-O-sulfotransferase (25). Therefore, transfection with GlcUAT-I with or without sulfotransferase resulted in HNK-1 expression (Fig. 4,  F and H). In this regard, GlcUAT-I behaved exactly the same as GlcUAT-P (25), but the staining of cells transfected with Gl-cUAT-I was somewhat weaker (data not shown). Nevertheless, these findings showed that GlcUAT-I can produce HNK-1 epitopes in vivo in different cell lines.
Conceivably, GlcUAT-I and GlcUAT-P may form HNK-1 determinants on different glycoproteins or glycolipids. To test this idea, we analyzed HNK-1-reactive glycoconjugates in Lec2 cell lines stably expressing GlcUAT-I or GlcUAT-P and transiently transfected with HNK-1 3-O-sulfotransferase. Analysis of the glycolipid fraction by thin-layer chromatography did not reveal any reactive lipids in either cell type (data not shown). In contrast, HNK-1-reactive proteins were present in both cell lines based on Western blotting of SDS-polyacrylamide gels of  total cell extracts (Fig. 5). Several prominent bands ranging from 65 to 100 kDa were present in cells stably transfected with GlcUAT-P, whereas only a minor band at ϳ110 kDa was seen in cells containing GlcUAT-I. The difference in reactivity was not due to variation in the level of expression of the enzymes since in vitro assays indicated that they did not vary significantly (7.2 Ϯ 0.2 pmol/min/mg of cell protein for GlcUAT-P, 7.9 Ϯ 2.2 pmol/min/mg for GlcUAT-I, and 0.42 Ϯ 0.01 pmol/min/mg for 3-O-sulfotransferase activity). All of the bands were peptide N-glycosidase F-sensitive, indicating that the epitope was assembled on N-linked oligosaccharides found on glycoproteins. Further characterization of the reactive bands has not yet been done.
Comparison of Substrate Specificity of GlcUAT-I and GlcUAT-P-To further analyze the substrate specificity of these GlcUA-transferases, the ectodomains of GlcUAT-I and GlcUAT-P were fused to the IgG-binding domain of protein A, and the chimeras were expressed in COS-7 cells. The secreted enzymes were absorbed to IgG-agarose, washed, and assayed with various acceptors and UDP-[ 3 H]GlcUA. As shown in Table  I, both glucuronosyltransferases can act on a variety of synthetic substrates containing terminal ␤-linked galactose residues. In general, disaccharides were better substrates than monosaccharides, but the individual enzymes showed strong differences in substrate utilization. As expected, GlcUAT-P could transfer GlcUA to asialofetuin and asialomucin containing terminal Gal residues, whereas GlcUAT-I had no detectable activity toward asialoglycoproteins and negligible activity with asialoglycosphingolipid substrates with or without added phospholipids (8). GlcUAT-I had the highest activity with Gal␤1, 3Gal␤-O-R (where R ϭ naphthalenemethanol or benzyl alco- hol), which resembles the linkage region of glycosaminoglycans. In contrast, GlcUAT-P was highly reactive with all disaccharides, including Gal␤1,4GlcNAc␤-O-NM, the acceptor sequence in glycoproteins, as well as those related to the linkage region. In addition, GlcUAT-P had high activity with naphthol ␤-galactosides, but reacted poorly with substrates contain-ing ␣-linked galactose residue.
Kinetic analysis of GlcUAT-I activity showed that the recombinant enzyme formed products in proportion to time for up to 5 h with all of the tested substrates (Fig. 6A). The apparent K m values for Gal␤1,4GlcNAc␤-O-NM, Gal␤1,3GlcNAc␤-O-NM, and Gal␤1,3GalNAc␤-O-NM were 1.8, 2.9. and 3.2 mM, respectively, whereas the K m for Gal␤1,3Gal␤-O-NM was significantly lower (0.67 mM) (Fig. 6B). Furthermore, the apparent V max was much greater for Gal␤1,3Gal␤-O-NM (Ͼ10-fold) than for the other substrates, suggesting that GlcUAT-I is much less reactive with substrates related to glycoproteins and is more reactive with those involved in glycosaminoglycan assembly. In contrast, GlcUAT-P was much more promiscuous, reacting with many substrates.

Restoration of Glycosaminoglycan Synthesis by Both Glucuronosyltransferases in a Glycosaminoglycan-deficient Mutant-
The high reactivity of GlcUAT-I with Gal␤1,3Gal-containing disaccharides supported the idea that this enzyme is involved glycosaminoglycan biosynthesis (18). To test this hypothesis directly, we transfected a pair of glycosaminoglycan-deficient mutants of CHO cells defective in GlcUAT-I. 2 Introduction of the cDNA for GlcUAT-I resulted in restoration of glycosaminoglycan biosynthesis as measured by the incorporation of 35 SO 4 (Table II). Analysis of the radioactive material by anionexchange HPLC confirmed that it consisted of a mixture of heparan sulfate and chondroitin sulfate chains (Fig. 7). Interestingly, transfection by GlcUAT-P also corrected the deficiency in the mutant (Table II). The relative level of the transfected GlcUA-transferases varied from 0.3 to 70 times of the endogenous value for GlcUAT-I in the wild-type, but glycosami-

Enzyme activity
GlcUAT-I GlcUAT-P pmol/min/ml medium) noglycan synthesis was restored under all conditions. These results suggested that both enzymes could facilitate formation of glycosaminoglycan chains. DISCUSSION In this report, we have described the isolation and characterization of a cDNA clone encoding a novel hamster glucuronosyltransferase (GlcUAT-I). Our initial characterization of the enzyme suggested that it might be a homologue of GlcUAT-P based on the high degree of homology (65% identity) (Fig. 3) and its ability to produce HNK-1-reactive material in the presence of HNK-1 3-O-sulfotransferase (Fig. 4). However, Northern blot analysis showed that the enzyme had a markedly different tissue distribution from HNK-1, with expression in virtually all tissues tested (Fig. 2). In contrast, HNK-1 (and GlcUAT-P) is restricted to brain and neurons (10). Furthermore, much less material contained HNK-1 determinants in cells transfected with hamster GlcUAT-I compared with GlcUAT-P (Fig. 5). These findings suggested that hamster GlcUAT-I might be involved in the formation of another type of GlcUA-containing glycoconjugate. To test this hypothesis, we examined a number of synthetic and modified natural substrates as acceptors using recombinant forms of the enzymes. Hamster GlcUAT-I transferred GlcUA efficiently to compounds containing Gal␤1,3Gal, which resembles the linkage tetrasaccharide found in heparan sulfate and chondroitin sulfate ( Table  I). As these studies were under way, Kitagawa et al. (18) reported a cDNA encoding an enzyme thought to be GlcUAT-I, which is involved in glycosaminoglycan biosynthesis. Hamster GlcUAT-I shows 95% identity to human GlcUAT-I (Fig. 3), suggesting that it most likely represents the same gene, but from a different species. Correction of a mutant defective in GlcUAT-I by transfection with hamster GlcUAT-I supported this idea ( Fig. 7 and Table II). Interestingly, GlcUAT-P also corrected the mutant, suggesting that it might work equally well in forming HNK-1 determinants as well as the linkage region of glycosaminoglycans.
The substrate specificity of GlcUAT-I has been debated ever since the enzyme activity was first described in cartilage and brain extracts (11,12). In these early studies, relatively crude enzyme preparations were found to catalyze the transfer of GlcUA not only to glycosaminoglycan linkage region fragments (Gal␤1,3Gal), but also to related compounds terminating in galactose, such as lactose (Gal␤1,4Glc) and Gal␤1,4GlcNAc, the precursor of HNK-1. HNK-1 had not yet been described when these studies were done (1), and therefore, it was not appreciated that the apparent reactivity with Gal␤1,4GlcNAc was most likely due to another GlcUA-transferase in the crude tissue extracts (GlcUAT-P). Several years ago, Curenton et al. (17) studied whether the glucuronosyltransferases for making HNK-1 epitopes and the glycosaminoglycan linkage region were the same enzyme using embryonic chick brain as an enzyme source. They found that the activity related to glycosaminoglycan assembly was firmly membrane-associated, whereas the activity related to HNK-1 formation was readily solubilized, suggesting that they were separate entities. Furthermore, no activity toward Gal␤1,4GlcNAc acceptors was detected in embryonic chick cartilage extract, which is a rich source of GlcUAT-I, but not GlcUAT-P. Based on these results, they concluded that two different enzymes catalyze the formation of linkage region fragments and HNK-1 determinants. More recent molecular cloning experiments support the idea that multiple enzymes exist, but the data presented here using recombinant forms of the enzymes suggest that they may be more promiscuous with respect to substrate utilization than previously appreciated.
Given the apparent overlap in behavior of the enzymes, especially GlcUAT-P, what can we say about their relative contribution to glycosaminoglycan and HNK-1 synthesis? Under normal conditions, GlcUAT-I probably does not give rise to HNK-1 determinants since the enzyme reacts relatively poorly with precursor glycoproteins, glycolipids, and synthetic disaccharides related to HNK-1 (Table I). Furthermore, transfection of wild-type CHO or COS-7 cells with only HNK-1 3-O-sulfotransferase does not result in expression of glycoproteins reactive with HNK-1 antibodies, yet these cells express endogenous GlcUAT-I activity. The inability of GlcUAT-I to form HNK-1 under these conditions might be due to differences in subcellu-

TABLE II
GlcUAT-I and GlcUAT-P restore glycosaminoglycan synthesis in pgsG mutants Two glycosaminoglycan-deficient cell lines and wild-type CHO cells were stably transfected with GlcUAT-I or GlcUAT-P. Some cells were labeled with 35 SO 4 , and after 2 days, the 35 S-labeled glycosaminoglycans were isolated from the cells and the medium and quantitated relative to cell protein. GlcUA-transferase activities were assayed using Gal␤1,3Gal␤-O-NM as substrate (see "Experimental Procedures").

Cell line Glycosaminoglycans
GlcUAT-transferase activity Control ϩGlcUAT-I ϩGlcUAT-P Control ϩGlcUAT-I ϩGlcUAT-P lar location of the enzyme and macromolecular precursors of HNK-1. However, Sugumaran et al. (29,30) have suggested that GlcUAT-I may be located in medial-and trans-Golgi fractions based on sucrose density gradient fractionation of chondrocytes. HNK-1 precursors are also likely to arise in these compartments of the Golgi since the terminal Gal residue on glycoprotein substrates is attached by ␤1,4-galactosyltransferase (lactose synthase), which has been immunocytochemically located in the medial-and trans-aspects of the Golgi (31). Thus, it is more likely that the poor reactivity of GlcUAT-I with precursors of HNK-1 explains why the endogenous enzyme does not give rise to HNK-1 determinants under normal conditions.
How do we explain the expression of HNK-1 epitopes after introduction of GlcUAT-I into cells (Figs. 4 and 5)? Expression of HNK-1 determinants in CHO and COS-7 cells under these conditions may merely reflect the high level of enzyme activity achieved by plasmid amplification and strong expression from the cytomegalovirus promoter, which overpowers the weak activity of the enzyme with glycoprotein substrates. Analysis of the enzyme activity in extracts prepared from stable transfectants indicates that the enzyme is enhanced, but not in all cases (Table II). Curiously, SDS-PAGE analysis of modified proteins indicates that those bearing HNK-1 determinants differ in cells transfected with GlcUAT-I and GlcUAT-P. Perhaps this reflects differences in protein substrate recognition or subcellular localization of the transfected enzyme and substrates. Identification of the reactive glycoproteins and more precise localization studies of the transferases may shed light on this issue.
The fact that GlcUAT-P can bypass a mutation in glycosaminoglycan biosynthesis (Table II) presumably reflects the promiscuous behavior of this enzyme with respect to acceptor substrates (Table I). These findings suggest that GlcUAT-P may actually catalyze glycosaminoglycan assembly in tissues expressing the enzyme, most notably brain (10). Perhaps, Gl-cUAT-P has two roles: one to generate HNK-1 determinants on glycoproteins and another to act as a failsafe system to ensure completion of glycosaminoglycan chains on proteoglycans that have somehow escaped the action of GlcUAT-I.
The similarity of GlcUAT-I and GlcUAT-P suggests an evolutionary relationship between the genes. However, it is difficult to determine which enzyme came first since glycosaminoglycans arose at about the same time as the nervous system in metazoan evolution (cf. Hydra). Furthermore, it is difficult to know if the more promiscuous enzyme (GlcUAT-P) arose from the more specific one (GlcUAT-I) or vice versa. Other members of the GlcUA-transferase family need to be analyzed to complete the comparison. Other family members include the transferases involved in elongation of heparan sulfate and chondroitin sulfate (15,30); hyaluronan synthases (32); GlcUAT-L, the enzyme involved in forming HNK-1 determinants on glycolipids (7,8); and possibly other enzymes inferred from GlcUA-containing products isolated from cells (33,34). Overall, GlcUAT-P and GlcUAT-I do not show any homology to the hyaluronan synthases and the putative heparan sulfate copolymerase recently reported by Lind et al. (35).
The difference in substrate selectivity of GlcUAT-I and GlcUAT-P presumably reflects variation in the acceptor sub-strate-binding sites of the proteins. Unfortunately, Ͼ40% of the residues differ between the two enzymes, making it difficult to pinpoint specific residues that impart selectivity by merely comparing sequences. However, it should be possible to study the enzyme structure by swapping contiguous blocks of residues. This approach helped define sites in lysosomal enzymes that are recognized by the phosphotransferase that adds Glc-NAc-P to terminal mannose residues only on lysosomal enzymes (36,37). A similar strategy applied to GlcUAT-I and GlcUAT-P might provide insight into active-site residues as well as features of the proteins that confer substrate specificity.