Differential Roles of TwoN-Acetylgalactosaminyltransferases, CSGalNAcT-1, and a Novel Enzyme, CSGalNAcT-2

By a tblastn search with β1,4-galactosyltransferases as query sequences, we found an expressed sequence tag that showed similarity in β1,4-glycosyltransferase motifs. The full-length complementary DNA was obtained by a method of 5′-rapid amplification of complementary DNA ends. The predicted open reading frame encodes a typical type II membrane protein comprising 543 amino acids, the sequence of which was highly homologous to chondroitin sulfate N-acetylgalactosaminyltransferase (CSGalNAcT-1), and we designated this novel enzyme CSGalNAcT-2. CSGalNAcT-2 showed much strongerN-acetylgalactosaminyltransferase activity toward glucuronic acid of chondroitin poly- and oligosaccharides, and chondroitin sulfate poly- and oligosaccharides with a β1–4 linkage,i.e. elongation activity for chondroitin and chondroitin sulfate, but showed much weaker activity toward a tetrasaccharide of the glycosaminoglycan linkage structure (GlcA-Gal-Gal-Xyl-O-methoxyphenyl), i.e.initiation activity, than CSGalNAcT-1. Transfection of theCSGalNAcT-1 gene into Chinese hamster ovary cells yielded a change of glycosaminoglycan composition, i.e. the replacement of heparan sulfate on a syndecan-4/fibroblast growth factor-1 chimera protein by chondroitin sulfate, however, transfection of the CSGalNAcT-2 gene did not. The above results indicated that CSGalNAcT-1 is involved in the initiation of chondroitin sulfate synthesis, whereas CSGalNAcT-2 participates mainly in the elongation, not initiation. Quantitative real-time PCR analysis revealed that CSGalNAcT-2 transcripts were highly expressed in the small intestine, leukocytes, and spleen, however, both CSGalNAcTs were ubiquitously expressed in various tissues.

By a tblastn search with ␤1,4-galactosyltransferases as query sequences, we found an expressed sequence tag that showed similarity in ␤1,4-glycosyltransferase motifs. The full-length complementary DNA was obtained by a method of 5-rapid amplification of complementary DNA ends. The predicted open reading frame encodes a typical type II membrane protein comprising 543 amino acids, the sequence of which was highly homologous to chondroitin sulfate N-acetylgalactosaminyltransferase (CSGalNAcT-1), and we designated this novel enzyme CSGalNAcT-2. CSGalNAcT-2 showed much stronger Nacetylgalactosaminyltransferase activity toward glucuronic acid of chondroitin poly-and oligosaccharides, and chondroitin sulfate poly-and oligosaccharides with a ␤1-4 linkage, i.e. elongation activity for chondroitin and chondroitin sulfate, but showed much weaker activity toward a tetrasaccharide of the glycosaminoglycan linkage structure (GlcA-Gal-Gal-Xyl-O-methoxyphenyl), i.e. initiation activity, than CSGalNAcT-1. Transfection of the CSGalNAcT-1 gene into Chinese hamster ovary cells yielded a change of glycosaminoglycan composition, i.e. the replacement of heparan sulfate on a syndecan-4/fibroblast growth factor-1 chimera protein by chondroitin sulfate, however, transfection of the CSGal-NAcT-2 gene did not. The above results indicated that CSGalNAcT-1 is involved in the initiation of chondroitin sulfate synthesis, whereas CSGalNAcT-2 participates mainly in the elongation, not initiation. Quantitative real-time PCR analysis revealed that CSGalNAcT-2 tran-scripts were highly expressed in the small intestine, leukocytes, and spleen, however, both CSGalNAcTs were ubiquitously expressed in various tissues.
Proteoglycans (PGs), 1 molecules consisting of a core protein and at least one glycosaminoglycan (GAG) chain, exist as one of the major components of extracellular matrix and on the cell surface. A variety of proteoglycan functions are exerted depending on the GAG chains. These chains are usually highly sulfated, and can be classified into several groups including chondroitin sulfate (CS)/dermatan sulfate, heparan sulfate (HS)/heparin, and keratan sulfate based on the GAG composition. HS has been shown to be involved in signal transduction and development together with certain growth factors, cytokines, and extracellular matrices. In Drosophila melanogaster, HS deficiencies caused by mutations in the genes encoding enzymes involved in the synthesis of HS result in abnormal developmental phenotypes (1). HS has been demonstrated to bind to a variety of cell growth factors such as the fibroblast growth factor (FGF) family molecules and to modulate their activities in various ways (2). Hepatocyte growth factor (HGF) and some interleukins (interleukins 3 and 7) bind to HS for efficient signal transduction (3,4). Recent studies have demonstrated that CS also plays various important roles in cell adhesion, migration, and recognition, especially of neuronal cells (5)(6)(7). The sulfation profiles of CS vary with aging in the cartilage (8,9). Disulfated disaccharide units, CS-D * This work was performed as part of the R&D Project of Industrial Science and Technology Frontier Program (R&D for Establishment and Utilization of a Technical Infrastructure for Japanese Industry) supported by the New Energy and Industrial Technology Development Organization (NEDO). 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) AB079252.
¶ ¶ To whom correspondence should be addressed: Glycogene Function Team, Research Center for Glycoscience (RCG), National Institute of Advanced Industrial Science and Technology (AIST), Central-2 C-2, (GlcA(2S)␤1-3GalNAc(6S)) and CS-E (GlcA␤1-3GalNAc-(4S,6S)), promote the outgrowth of neurites in rat brain (10). These reports suggest that HS and CS have different functions. Thus, it is of interest to clarify the mechanism for the biosynthesis of HS and CS, the molecules responsible for these diverse biological phenotypes.
The initial stage in the biosynthesis of both CS and HS involves a linkage tetrasaccharide structure (GlcA␤1-4Gal␤1-3Gal␤1-4Xyl␤1-), which binds covalently to serine residues of core proteins. To initiate the synthesis of HS, a GlcNAc residue is transferred to GlcA of the linkage tetrasaccharide with an ␣1-4 linkage. On the other hand, GalNAc is transferred to the same acceptor with a ␤1-4 linkage for the initiation of CS synthesis. Thus, because the acceptor substrate is identical, it is possible that the initiation enzyme for HS or CS synthesis compete with the acceptor substrate, i.e. the linkage tetrasaccharide bound to core proteins, in the cells. If this is the case, the initiation enzymes may play a key role in determining the species of the GAG chain, HS or CS, on the core proteins. After the initiation reaction, the addition of disaccharide units of GlcNAc␣1-4GlcA␤1-4 are repeated for elongation of the HS chain, whereas GalNAc␤1-4GlcA␤1-3 units are repeatedly added for polymerization of the CS chain.
For the synthesis of HS, five glycosyltransferase genes, EXT1, EXT2, EXTL1, EXTL2, and EXTL3, were cloned and their products characterized. They retain conserved motifs of short amino acid stretches in their COOH terminus and belong to one family. The EXT1 and EXT2 genes, which have been identified as tumor suppressor genes and were implicated in hereditary multiple exostoses, were found to encode HS polymerases having the activity of both ␤1,4-glucuronyltransferase (␤4GlcAT) and ␣1,4-N-acetylglucosaminyltransferase (␣4Glc-NAcT) (11). Both enzymes are responsible for the elongation of HS chains. Three enzymes, EXTL1-3, exhibit only ␣4GlcNAcT activity, but their substrate specificities were different. EXTL1 showed ␣4GlcNAcT activity toward GlcA in elongation. EXTL2 showed activity in initiation. EXTL3 showed activity in both initiation and elongation for the synthesis of HS (12,13).
Very recently, three glycosyltransferases, chondroitin synthase (CSS) (14), CS glucuronyltransferase (CSGlcAT) (15) and CSGalNAcT (16,17), which are involved in the synthesis of CS, have been reported. CSS was found to be a CS polymerase having the activity of both ␤1,3-glucuronyltransferase (␤3Glc-AT) and ␤4GalNAcT. CSGlcAT or CSGalNAcT exhibits the activity of only one glycosyltransferase, ␤3GlcAT and ␤4GalN-AcT, respectively. Furthermore, CSGalNAcT was found to exhibit apparent ␤4GalNAcT activity toward the linkage tetrasaccharide for the initiation of CS synthesis. Given the many enzymes involved in the synthesis of HS, it is easy to speculate that multiple glycosyltransferases would also participate in the synthesis of CS.
In this study, we report the cloning and characterization of a novel N-acetylgalactosaminyltransferase, CSGalNAcT-2, that is the fourth member participating in the synthesis of CS. In considering the substrate specificities of these two CSGal-NAcTs in vitro, we suggest their differential roles in vivo.

EXPERIMENTAL PROCEDURES
Isolation of Human CSGalNAcT-2 cDNA-We performed a tblastn search of the GenBank TM data base using ␤1,4-glycosyltransferase motifs as queries and identified an expressed sequence tag with Gen-Bank TM accession number NM_018590, which contained a partial open reading frame (ORF), but showed high homology to the carboxyl-terminal region of CSGalNAcT-1. An additional search of the Human Genome Project data base revealed that the genome sequence with Gen-Bank TM accession number NT_008776 was identical to the expressed sequence tag. To obtain the complete ORF, the 5Ј-rapid amplification of complementary DNA (cDNA) ends method was employed using a Mar-athon Ready TM cDNA Amplification Kit (Clontech, Palo Alto, CA). Two reverse primers were designed for the 1st PCR: GP245, 5Ј-GTCAG-GAAATCTGAACGATGCTGA-3Ј, and for the nested PCR, GP244, 5Ј-GCAGCTGTTAAGGAATTCGGCTGA-3Ј. The sequence of the DNA fragment obtained by the 5Ј-rapid amplification of complementary DNA method was determined using a DYEnamic ET Terminator Cycle Sequencing Kit (Amersham plc, Amersham Place, UK). Finally, a cDNA sequence encoding the ORF was obtained by PCR using the Marathon Ready TM cDNA of human bone marrow tissue (Clontech) as a template.
Construction and Purification of CSGalNAcT Proteins Fused with FLAG Peptide-The putative catalytic domain of CSGalNAcT-2 (amino acids 37 to 542) was expressed as a secreted protein fused with a FLAG peptide in insect cells according to the instruction manual of GATE-WAY TM Cloning Technology (Invitrogen, Groningen, Netherlands). An ϳ1.6-kb DNA fragment was amplified by PCR using the Marathon Ready TM cDNA derived from human bone marrow (Clontech) as a template, and two primers, 5Ј-GGGGACAAGTTTGTACAAAAAAGCA-GGCTTCAAAGGTGACGAGGAGCAGCTGGCAC-3Ј and 5Ј-GGGGAC-CACTTTGTACAAGAAAGCTGGGTCTCATGTTTTTTTGCTACTTGT-CTTCTGT-3Ј. The amplified fragment was inserted into the vector pDONR TM 201 (Invitrogen), then transferred into the expression vector pFBIF to construct pFBIF-CSGalNAcT-2 as described previously (18). A CSGalNAcT-1 expression vector, pFBIF-CSGalNAcT-1, was also constructed as reported elsewhere (17). The catalytic domains of CSGal-NAcT-1 and CSGalNAcT-2 were expressed in Sf21 insect cells. A 50-ml volume of culture medium was mixed and incubated with anti-FLAG M1 antibody resin (SIGMA). The resin was washed twice with 50 mM Tris-buffered saline (50 mM Tris-HCl, pH 7.4, and 150 mM NaCl) containing 1 mM CaCl 2 and suspended in 100 l of each assay buffer described below.
Construction of ␤4GalT-7, ␤3GalT-6, and GlcAT-I Fused with FLAG Peptides-The putative catalytic domain of ␤4GalT-7, ␤3GalT-6, or GlcAT-I was expressed as secreted protein fused with FLAG peptide in insect cells or COS-1 cells as described in detail previously (17). Each enzyme was purified as described (17) and suspended in 100 l of the glycosylation buffer described below.
For the GalT assay, 14 mM Hepes buffer (pH 7.4) containing 250 M UDP-Gal, 12.5 mM MnCl 2 , and 500 M of each acceptor substrate was used. For the GalNAcT assay, 50 mM MES buffer (pH 6.5) containing 0.1% Triton X-100, 1 mM UDP-GalNAc, 10 mM MnCl 2 , and 500 M of each acceptor substrate was used. A 10-l volume of enzyme solution for 20 l of each reaction mixture was added and incubated at 37°C for 2 h for the GalT and 16 h for the GalNAcT assay. After incubation, the mixture was filtrated with an Ultrafree-MC column (Millipore, Bedford, MA) and a 10-l aliquot was subjected to reversed-phase high performance liquid chromatography on an ODS-80Ts QA column (4.6 ϫ 250 mm; TOSOH, Tokyo, Japan). 0.1% Trifluoroacetic acid/H 2 O was used as a running solution and the products were eluted with a 0 -15 (for GlcA-␤-pNP) or 7-10% (for linkage tetrasaccharide-para-methoxypheny) acetonitrile gradient at a flow rate of 1.0 ml/min at 50°C. For glycosylated peptides, H 2 O containing 0.1% trifluoroacetic acid and 21% acetonitrile was utilized as the running solution. An ultraviolet spectrophotometer (absorbance at 210 nm), SPD-10A VP (Shimadzu, Kyoto, Japan) was used for detection of the peaks. In the analysis of glycosylated peptide, labeling was done with Cy5 (Amersham Biosciences) and fluorescence was detected with a fluorescence detector, RF-10A XL (Shimadzu). For the analysis of elongation activity, the CSGalNAcTs reaction mixture containing 100 g of chondroitin or CS and 40,000 -55,500 dpm of UDP-[ 14 C]GalNAc was used. After a 1-h incubation at 37°C, the reaction mixture was filtrated and fractionated with a G2500PW column (TOSOH, Tokyo, Japan) or Superdex 30 pg column (Amersham Biosciences). The radioactivity of each fraction was monitored by liquid scintillation spectrophotometry.
Assay of CSGalNAcT Activity with a Tetrasaccharide-bikunin Peptide-A Xyl-bikunin peptide (VLPQEEEGS(-Xyl)GGGQLVT) was pur-chased from the Peptide Institute Inc. (Osaka, Japan). The Cy5 (Amersham Biosciences)-labeled Xyl-peptide was incubated with 5 l of three glycosyltransferases, ␤4GalT-7, ␤3GalT-6, and GlcAT-I, and 1 mM donor substrates, UDP-Gal and UDP-GlcA, at 37°°C for 16 h as described in detail previously (17). A 50 mM MES buffer (pH 6.5) containing 0.1% Triton X-100, 1 mM UDP-Gal, 1 mM UDP-GlcA, 10 mM MnCl 2 , and 500 M Xyl-bikunin peptide was used for the reaction. The glycosyltransferases for the synthesis of the tetrasaccharide-bikunin peptide were inactivated by heating at 95°C for 5 min. Then, the reaction mixture was filtrated with an Ultrafree-MC column (Millipore), and a 10-l aliquot was incubated with 1 mM donor substrate, UDP-GalNAc, and each CSGalNAcT at 37°C for 8 h for the assay of initiation activity of each CSGalNAcT. The reaction products of CSGalNAcTs were filtrated with an Ultrafree-MC column and a 10-l aliquot was subjected to reversed-phase high performance liquid chromatography on an ODS-80Ts QA column as described above.
Quantitative Analysis of CSGalNAcT-1 and CSGalNAcT-2 Transcripts in Human Tissues by Real-time PCR-For quantification of the two CSGalNAcT transcripts, we employed the real-time PCR method, as described in detail previously (18,22,23). Marathon Ready TM cDNAs derived from various human tissues and cells were purchased from Clontech. Standard curves for the endogenous control, glyceraldehyde-3-phosphate dehydrogenase (GAPDH) cDNAs, were generated by serial dilution of a pCR2.1 (Invitrogen) DNA containing the GAPDH cDNA. The primer sets and the probes for CSGalNAcT-1 and CSGalNAcT-2 were as follows. The forward primer for CSGalNAcT-1 was 5Ј-GACT-TCATCAATATAGGTGGGTTTGAT-3Ј, the reverse primer, 5Ј-GTCCG-TACCACTATGAGGTTGCT-3Ј, and the probe, 5Ј-ACCTTTATCG-CAAGTATCT-3Ј with a minor groove binder (24). The forward primer for CSGalNAc-T2 was 5Ј-CTGACCATTGGTGGATTTGACAT-3Ј, the reverse primer, 5Ј-AACCGGAGTCCGAATCACAA-3Ј, and the probe, 5Ј-CATCTTTATCGAAAATACTTACATGG-3Ј with a minor groove binder. PCR products were continuously measured with an ABI PRISM 7700 Sequence Detection System (Applied Biosystems, Foster City, CA). The relative amount of each CSGalNAcT transcript was normalized to the amount of GAPDH transcript in the same cDNA.

Determination of Nucleotide and Amino Acid Sequence of
CSGalNAcT-2-We determined a novel full-length cDNA se-quence by the 5Ј-rapid amplification of cDNA ends method and registered it in the GenBank TM data base with accession number AB079252. The nucleotide sequence and the putative amino acid sequence are shown in Fig. 1. This cDNA sequence consisted of a 254-bp 5Ј-untranslated region, 1629-bp coding region, and 1791-bp 3Ј-untranslated region that contained a poly-A tail (Fig. 1). Although the original expressed sequence tag was obtained in a GenBank TM data base search with ␤1,4glycosyltransferase motifs as query sequences, the full-length ORF sequence was identified as highly homologous to CSGal-NAcT as previously reported by us (17). We designated this novel cDNA as CSGalNAcT-2, and renamed the previous CS-GalNAcT, CSGalNAcT-1. An alignment of CSGalNAcT-1 and CSGalNAcT-2 is shown in Fig. 2. A hydropathy profile of the putative amino acid sequence based on Kyte and Doolittle hydrophobicity plots indicates that the ORF of CSGalNAcT-2 encodes a typical type II membrane protein, which is consistent with the topology of other glycosyltransferases, with a cytoplasmic tail of 14 amino acids, a transmembrane domain of 20 amino acids, and a large catalytic portion of 508 amino acids. CSGalNAcT-2 contains a DXD motif, which is conserved in many glycosyltransferases and functions as a key sequence for divalent cation binding, and a GWGGED motif, which is highly conserved among some of the ␤4GalT family.
A genome sequence identical to that of the CSGalNAcT-2 cDNA was found in a genome clone (GenBank TM AC011890), which is localized on chromosome 10. The genomic structure of the CSGalNAcT-2 gene was determined to be composed of at least 7 exons by comparison between the cDNA and genome sequences (Fig. 1). The exon/intron junctions of CSGalNAcT-2 were identical to those of CSGalNAcT-1 (data not shown).
Determination of CSGalNAcT-2 Activity in Elongation and Initiation of Chondroitin Poly-and Oligosaccharides-The truncated soluble forms of CSGalNAcT-1 and CSGalNAcT-2 were expressed in insect cells and employed for all experiments. On Western blotting using anti-FLAG antibody, each enzyme was detected as a single band corresponding to the predicted size (Fig. 3B). An additional band appeared by Coomassie staining at ϳ50 kDa shared by all lanes (Fig. 3A), however, the protein recovered from mock transfectants showed no activity for any donors and acceptor substrates (data not shown). At first, we screened the transfer activity of CS-GalNAcT-2 using nine donor substrates and multiple monosaccharide-pNP acceptors. In screening of donor and acceptor substrates, CSGalNAcT-2 transferred a GalNAc residue to GlcA-pNP, however, no activity was observed with other combinations of donor and acceptor substrates (data not shown). It was identified in a previous report that CSGalN-AcT-1 is ␤4GalNAcT, which transfers GalNAc to GlcA for initiation and elongation in the synthesis of CS (17). So far, the linkage of GalNAc with GlcA has been identified only in CS. We considered that CSGalNAcT-2 was also a GalNAcT involved in the synthesis of CS, and performed further analysis using CS-related substrates as acceptors.
Two kinds of GalNAc-GlcA linkages are known in CS, one in its polymer structure (3GalNAc␤1-4GlcA␤1-) n and the other between the polymer CS and the linkage tetrasaccharide (GlcA␤1-3Gal␤1-3Gal␤1-4Xyl). At first, chondroitin was utilized as an acceptor substrate to examine the elongation activity of CSGalNAcT-2. As shown in Fig. 3C, CSGalNAcT-2 apparently transferred GalNAc to chondroitin to produce an additional peak (indicated by an arrow in Fig. 3C) as a reaction product. This peak was isolated and identified to have GalNAc on its nonreducing terminus with chondroitinase ACII treatment (data not shown), the method used having been described in a previous study (17). Second, a linkage tetrasaccharide (GlcA␤1-3Gal␤1-3Gal␤1-4Xyl␤-O-methoxyphenyl) synthesized chemically was utilized as an acceptor substrate to identify the initiation activity of CSGalNAcT-2. As shown in Fig.  3D, the peak P appeared at a 30.3-min retention time in addition to the acceptor substrate peak (peak S) at 31.1 min. Peaks S and P were isolated by reversed-phase chromatography and their molecular weights were determined by MALDI-TOF MS. Peak S gave a molecular mass of 779.14 m/z, the same as that of the linkage tetrasaccharide with Na ϩ (Fig. 3E). Peak P gave two peaks of 982.28 and 998.25 m/z as shown in Fig. 3F. The molecular mass of 982.28 or 998.25 m/z was the same molecular weight as a GalNAc-added linkage tetrasaccharide-O-methoxy-phenyl with Na ϩ or with K ϩ , respectively. Moreover, the peak P was identified to have GalNAc at its nonreducing terminus with chondroitinase ACII treatment (data not shown). These results suggested that CSGalNAcT-2 has two GalNAcT activities, i.e. elongation of chondroitin and initiation of the CS synthesis by transferring GalNAc to the linkage tetrasaccharide.
Comparison of Acceptor Substrate Specificity between CSGal-NAcT-1 and CSGalNAcT-2-In view of these results, CSGalN-AcT-2 was suggested to be GalNAcT involved in the synthesis of CS as well as CSGalNAcT-1. To distinguish the functions of the two CSGalNAcTs, we compared GalNAcT activity toward GlcA found at the nonreducing terminus of various kinds of chondroitin-related acceptor substrates. The amount of enzyme was estimated by Western blotting, and approximately equal amounts of enzyme were used for the GalNAcT reaction. The results are summarized in Table I. Regarding the elongation activity, CSGalNAcT-2 utilized chondroitin polysaccharide as an acceptor more than any other substrate examined, and showed higher levels of enzymatic activity than CSGalNAcT-1 toward all substrates except for the linkage tetrasaccharide (Table I). Furthermore, CSGalNAcT-2 showed remarkably strong activity, compared with CSGalNAcT-1, toward sulfated substrates such as CS poly-and oligosaccharides (Table I). CS-A and CS-B, both of which are sulfated at position C-4 of GalNAc, were better substrates for CSGalNAcT-2 than CS-C, which is sulfated at position C-6 of GalNAc and CS-D, which is sulfated at C-6 of GalNAc and C-2 of GlcA (Fig. 4). CSGal-NAcT-2 exhibited much stronger activity toward the longer oligosaccharides, prepared from chondroitin and CS, than toward the shorter ones. These results strongly indicated that CSGalNAcT-2 is much more active in the elongation of chondroitin and CS than is CSGalNAcT-1.
Regarding the initiation activity, CSGalNAcT-1 preferred the linkage tetrasaccharide as substrate and showed a much higher level of activity for the linkage tetrasaccharide than did CSGalNAcT-2. This indicated that CSGalNAcT-1 is the enzyme mainly responsible for the initiation of chondroitin and CS synthesis, not for elongation. To examine whether two GalNAcTs show a synergistic effect for initiation and elongation activities, two enzymes were mixed in the enzyme reaction toward the linkage tetrasaccharide or chondroitin. We observed additional effects of GalNAcT activity, but not synergistic effect, to any acceptor substrates (data not shown).
Comparison of Initiation Activity with Linkage Tetrasaccharide Peptide between Two CSGalNAcTs-We performed an in vitro enzymatic synthesis of a chondroitin pentasaccharidebikunin peptide (VLPQEEEGS*GGGQLVT; the peptide sequence encodes the NH 2 -terminal sequence of bikunin and the serine residue with an asterisk is attached with a saccharide chain) for determination of the initiation activity by CSGalN-AcTs. Peak S in Fig. 5A, which was identified to possess the linkage tetrasaccharide (GlcA-Gal-Gal-Xyl-bikunin peptide) in a previous study, was used as the acceptor substrate. To obtain the relative activity of each enzyme, equal amounts of enzyme were employed, and the incubation was terminated before a large amount of acceptor substrate remained. As seen in Fig. 5, B and C, the reaction product (Peak P) appeared at 30.1 min with CSGalNAcT-1 or -2, respectively. Peak P was identified as GalNAc-GlcA-Gal-Gal-Xyl-bikunin peptide in a previous study (17). The peak area of the CSGalNAcT-1 product (Fig. 5C) was 3-fold larger than that of the CSGalNAcT-2 product (Fig. 5B). The value of this relative activity was consistent with that toward the methoxyphenyl-linkage tetrasaccharide in Table I, i.e. the activity of CSGalNAcT-1 was 2.3-fold that of CSGalN-AcT-2. These results indicated that the presence of bikunin peptide does not influence the relative activity of the two CS-GalNAcTs in vitro.
Change of GAG Composition of PG-FGF-1 by CSGalNAcT-1 or CSGalNAcT-2 Transfection-Previously, the gene construct encoding a chimera protein (PG-FGF-1) consisting of syndecan-4 and FGF-1 was transfected into CHO-K1 cells to express the chimera protein (21). Syndecan-4 is a well analyzed proteoglycan possessing chains of HS and/or CS. The GAG composi-

FIG. 3. Determination of CS elongation and initiation activities of CSGalNAcT-2.
CSGalNAcT-1 and -2 were purified with anti-FLAG M1 antibody resin and were applied to SDS-PAGE. The detection of proteins was performed with Coomassie staining (A) and Western blotting using anti-FLAG antibody (B). C, chondroitin was used as an acceptor substrate for CSGalNAcT-2. The purified reaction mixtures were applied to a G2500PW column with 0.2 M NaCl as a running buffer at a flow rate of 0.6 ml/min. The radioactivity of each fraction (340 l each) was quantified by liquid scintillation spectrophotometry. The peak of the reaction product is indicated with an arrow. D, linkage tetrasaccharide-O-methoxyphenyl was used as an acceptor, and the product was analyzed by reversed-phase chromatography. Each peak of substrate (peak S) and reaction product (peak P) was isolated and subjected to MALDI-TOF-MS analysis (E and F).
tion of PG-FGF-1 secreted from the stable transfectant cells with PG-FGF-1 had been analyzed. It was found that GAG on PG-FGF-1 consisted of large populations of HS and small populations of CS. The initiation of the synthesis of HS or CS is determined by ␣4GlcNAcT or ␤4GalNAcT, respectively. If the two enzymes compete with the common linkage tetrasaccharide (GlcA-Gal-Gal-Xyl-Ser) as an acceptor in the cells, they would be key enzymes differentially regulating the number of HS or CS chains. To test this hypothesis, the CSGalNAcT-1 or CSGalNAcT-2 genes were introduced into CHO-K1 cells that were producing PG-FGF-1. The transcripts for CSGalNAcT-1 and CSGalNAcT-2 in the stable transfectant cells were quantified by real-time PCR. Their expression levels were almost 20 times higher than the physiological level in human placenta, in which the transcripts for the genes involved in the synthesis of CS are expressed at the highest level among many tissues (data not shown).
PG-FGF-1 secreted into the medium conditioned by the transfectants was detected by immunoblotting with anti-FGF-1 monoclonal antibody (mAb1) (20). PG-FGF-1 from CHO-K1 mock transfectants showed a major band with a large molecular mass of more than 200 kDa (Fig. 6, lane 1). The composition of GAGs on PG-FGF-1 was analyzed by glycosidase digestion. Digestion with chondroitinase ABC (CSase) yielded a protein band at 32 kDa (lane 2), which corresponds to the molecular mass of PG-FGF-1 with linkage saccharides, although most proteins remained at a large molecular mass. The ratio of CS to CS plus HS on PG-FGF-1 was obtained by measurement of the intensity of the 32-kDa band, for example, the ratio of CS/CS ϩ HS on PG-FGF-1 produced by the mock transfectants was 9.5%, which was calculated as the intensity of the 32-kDa band in lane 2 divided by that in lane 4. This indicated that PG-FGF-1 containing only CS without HS occupied only 9.5% of total PG-FGF-1, and HS occupied most of the PG-FGF-1 produced by the mock-transfected CHO-K1 cells. Digestion of PG-FGF-1 (lane 3) with a mixture of heparitinase (HSase) and heparanase (HPase) yielded a strong 32-kDa band, however, faint smear bands, which probably carry CS, were detected at a molecular mass of approximately 40 -70 kDa. CSase treatment of PG-FGF-1 recovered from the CSGal-NAcT-2 transfectants did not change significantly the intensity of the 32-kDa band, compared with the control (9.5 in lane 2 versus 4.3% in lane 6). In contrast, CSase treatment of PG-FGF-1 recovered from the CSGalNAcT-1 transfectants appar-

FIG. 5. Comparison of initiation activity between two CSGal-NAcTs.
A, the GlcA-Gal-Gal-Xyl-bikunin peptide (peak S) was synthesized enzymatically with three glycosyltransferases, ␤4GalT-7, ␤3GalT-6, and GlcAT-I, and detected by a reversed-phase chromatography. The product (peak S) was used as an acceptor substrate for CSGalNAcT-2 (B) or CSGalNAcT-1 (C). The products (peak P) were identified as described previously (17). ently increased the intensity of the 32-kDa band (9.5 in lane 2 versus 39.1% in lane 10). The above results indicated that CSGalNAcT-1 effectively initiates the synthesis of CS and increases the number of CS chains by transferring GalNAc to the linkage tetrasaccharide on syndecan-4, but CSGalNAcT-2 did not exert initiation activity in the cells.
Tissue Distribution of CSGalNAcT-2 Transcripts-We determined the tissue distribution and expression levels of CSGal-NAcT-2 transcripts, in comparison with CSGalNAcT-1 transcripts, by the real-time PCR method. The expression levels of both genes in various human tissues were shown as the rela-tive amount versus GAPDH transcripts (Fig. 7). Both genes were expressed ubiquitously in all tissues examined. However, the expression profiles differed from each other, i.e. CSGalN-AcT-2 was highly expressed in the small intestine, leukocyte, and spleen, whereas CSGalNAcT-1 was highly expressed in the placenta and thyroid gland. DISCUSSION The investigation of CS has progressed to the relationship between its structural properties including saccharide composition and sulfation, and its biological function. However, the mechanisms of its biosynthesis, such as the number of enzymes participating and their roles, still remained to be elucidated. To date, three glycosyltransferases, CSS (14), CSGlcAT (15), and CSGalNAcT-1 (16,17), which participate in the biosynthesis of CS, have been reported. In this study, a novel enzyme CSGal-NAcT-2 was found to be a fourth member involved in the synthesis, and identified to be the second GalNAcT transferring a GalNAc to GlcA.
Regarding the initiation activity, CSGalNAcT-1 exhibited much stronger GalNAcT activity toward the linkage tetrasaccharide than CSGalNAcT-2. In a previous study, we reported that CSGalNAcT-1 has effective initiation activity, i.e. it transferred GalNAc to the linkage tetrasaccharide conjugated with a bikunin peptide (GlcA-Gal-Gal-Xyl-bikunin peptide) in vitro (17). The initiation activity of CSGalNAcT-1 relative to CSGal-NAcT-2 was 2.3-and 3.0-fold higher toward methoxyphenyllinkage tetrasaccharide and tetrasaccharide-bikunin peptide, respectively. The presence of a peptide backbone did not affect the relative initiation activity. The in vivo assay system using the syndecan-4/FGF-1 chimera protein (PG-FGF-1) demonstrated that CSGalNAcT-1 exhibits initiation activity for syndecan-4 by competing with the synthesis of HS (Fig. 6). Syndecan-4 is a complex-type PG that has four putative GAG attachment sites, and both CS and HS can bind to these sites (36). In our construction of PG-FGF-1, a part of syndecan-4 containing three GAG attachment sites was ligated to the NH 2 terminus of FGF-1 (21). The intensity of the 32-kDa band after CSase or HSase/HPase treatment of PG-FGF-1 reflects the CS and HS composition of PG-FGF-1 as demonstrated in our previous study (21). In this assay system, we identified an increase in CS on PG-FGF-1 only in CSGalNAcT-1-transfected CHO-K1 cells, not in CSGalNAcT-2 transfectants. The increase of CS is probably directed by the initiation activity of CSGalNAcT-1. This in vivo observation is consistent with previous results in vitro of the effective initiation activity of CSGalNAcT-1 as we reported (17). Taken together with the previous findings in vitro, the in vivo results of the present study strongly support that CSGalNAcT-1 is the enzyme most responsible for the initiation of CS synthesis in the cells.
On the other hand, CSGalNAcT-2 transfected into CHO-K1 cells showed no increase in CS, although it had some initiation activity in vitro toward the linkage tetrasaccharide. This contradiction between the in vitro and in vivo activities of CSGal-NAcT-2 suggests that the initiation of CS synthesis might be controlled by some unknown mechanisms. One possibility is the effect of sulfation on the linkage tetrasaccharide. In vertebrates, CS is specifically sulfated at position C-6 of the inner Gal (Gal␤1-4) and at C-4 and C-6 of the outer Gal (Gal␤1-3) in the linkage tetrasaccharide (37,38). These sulfations might influence the CSGalNAcTs in terms of recognition. Another possibility is the effect of the peptide sequence of the core protein on the enzyme recognition. A currently proposed GAG attachment motif is the Ser-Gly-Ser-Gly sequence and surrounding acidic amino acids (39,40). CSGalNAcT-2 might recognize CS binding peptide sequences of core proteins other than syndecan-4.
The tissue distribution of transcripts for three glycosyltransferase genes, i.e. CSS, CSGlcAT, and CSGalNAcT-1, have been reported to be ubiquitously expressed in all tissues examined (14 -17). CSGalNAcT-2 also showed a ubiquitous expression like the others. However, characteristic of that CSGalNAcT-2 was a high level of expression in the small intestine, leukocytes, and spleen, whereas the others are not so highly expressed in these tissues. Hiraoka et al. (31) reported that the transcripts of C4ST-1 and C4ST-2, which produce CS-A, are also highly expressed in the small intestine, spleen, and leukocytes. In contrast, C6ST-2, which produces CS-C, is expressed significantly in the spleen, but not so in small intestine (34). The preference of CSGalNAcT-2 for CS-A and CS-B to CS-C and CS-D, as demonstrated in Table I, led us to speculate of a cooperative CS synthesis by CSGalNAcT-2 and C4STs in the cells. The similar profile of the tissue distribution of CSGalN-AcT-2 to that of C4STs further suggested this cooperative synthesis. In the case of HS, it has been proposed that EXTL2 and EXTL3 are involved in the initiation of the synthesis (12,13), however, the exact role of each enzyme in vivo remains to be elucidated.
Finally, the catalytic activities of each enzyme involved in the synthesis of HS and CS are schematically summarized in Fig. 8. Four glycosyltransferases responsible for the synthesis of the linkage tetrasaccharide have been cloned and analyzed (41)(42)(43)(44)(45)(46). Using these enzymes, it is now possible to synthesize enzymatically the linkage tetrasaccharide bound to the peptide (17). The initiation of the CS chain is also feasible using CS-GalNAcT-1 (17). Three enzymes, i.e. CSS, CSGlcAT, and CS-GalNAcT-2, have been identified as involved in the polymerization of the CS chain. In addition to the four enzymes, there still may be unknown enzymes, which remain to be cloned and analyzed, for the synthesis of CS.