Specificities of three distinct human chondroitin/dermatan N-acetylgalactosamine 4-O-sulfotransferases demonstrated using partially desulfated dermatan sulfate as an acceptor: implication of differential roles in dermatan sulfate biosynthesis.

4-O-Sulfation of GalNAc is a high frequency modification of chondroitin sulfate and dermatan sulfate (DS), and three major GalNAc 4-O-sulfotransferases including dermatan 4-O-sulfotransferase-1 (D4ST-1) and chondroitin 4-O-sulfotransferases-1 and -2 (C4ST-1 and -2) have been identified. 4-O-Sulfation of GalNAc during DS biosynthesis had long been postulated to be a prerequisite for iduronic acid (IdoUA) formation by C5-epimerization of GlcUA. This hypothesis has recently been argued based on enzymological studies using microsomes that C5-epimerization precedes 4-O-sulfation, which was further supported by the specificity of the cloned D4ST-1 with predominant preference for IdoUA-GalNAc flanked by GlcUA-GalNAc over IdoUA-GalNAc flanked by IdoUA-GalNAc in exhaustively desulfated dermatan. Whereas the counterproposal explains the initial reactions, apparently it cannot rationalize the synthetic mechanism of IdoUA-GalNAc(4-O-sulfate)-rich clusters typical of mature DS chains. In this study, we examined detailed specificities of the three recombinant human 4-O-sulfotransferases using partially desulfated DS as an acceptor. Enzymatic analysis of the transferase reaction products showed that D4ST-1 far more efficiently transferred sulfate to GalNAc residues in -IdoUA-Gal-NAc-IdoUA-than in -GlcUA-GalNAc-GlcUA-sequences. In contrast, C4ST-1 showed the opposite preference, and C4ST-2 used GalNAc residues in both sequences to comparable degrees, being consistent with its phylogenetic relations to D4ST-1 and C4ST-1. Structural analysis of the oligosaccharides, which were isolated after chondroitinase AC-I digestion of the 35S-labeled transferase reaction products, revealed for the first time that D4ST-1, as compared with C4ST-1 and C4ST-2, most efficiently utilized GalNAc residues located not only in the sequence -IdoUA-GalNAc-IdoUA- but also in -GlcUA-Gal-NAc-IdoUA- and -IdoUA-GalNAc-GlcUA-. The isolated oligosaccharide structures also suggest that 4-O-sulfation promotes subsequent 4-O-sulfation of GalNAc in the neighboring disaccharide unit.

Dermatan sulfate (DS) 1 glycosaminoglycan chains are covalently bound to a variety of core proteins, forming various proteoglycans with widespread distribution in mammalian tissues (see Ref. 1 and references therein). DS proteoglycans are detected at cell surfaces in addition to extracellular matrices and intracellular storage granules of mast cells and macrophages (2), where they participate in various biological events including extracellular matrix assembly (3,4), cell adhesion (5), migration (6 -8), proliferation (9 -11), neurite outgrowth (12,13), wound repair (14,15), and anticoagulant processes (16 -18). Although biological activities of DS chains involved in these events are expressed through the interactions between DS proteoglycans and particular target molecules such as growth factors and heparin cofactor II, accumulating evidence suggests that some of the functions depend on DS glycosaminoglycan moieties rather than core proteins (4, 5, 11-13, 16, 17, 19 -22).
DS glycosaminoglycan corresponds to a stereoisomeric form of chondroitin sulfate (CS) composed of the repeating disaccharide units containing GlcUA and GalNAc residue, with varying proportions of IdoUA in place of GlcUA. The general copolymeric hybrid structure tends to exhibit a periodic and wavelike distribution of GlcUA-containing disaccharide repeats and IdoUA-containing disaccharide repeats in a tissue-specific manner (23). The former region can be sulfated mainly at the C-4 or C-6 position of GalNAc residues. In addition, the C-2 or C-3 position of GlcUA residues may also be sulfated (24). In contrast, the latter region is largely sulfated at the C-4 position of GalNAc and contains occasional 6-O-sulfated GalNAc and 2-O-sulfated IdoUA residues (23).
In biosynthesis, the structural variability of CS/DS is generated under the control of multiple sulfotransferases and glucuronyl C5-epimerase converting GlcUA to IdoUA (for reviews, * This work was supported in part by the Scientific Research Promotion Fund of the Japan Private School Promotion Foundation, by Grants-in-aid for Encouragement of Young Scientists 14771299 (to T. M.) and Scientific Research (B) 13470493 (to K. S.), and by the Sasagawa Scientific Research Grant from the Japan Science Society (to T. M.). 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  see Refs. [25][26][27]. 4-O-Sulfation of GalNAc residues is a typical modification found in CS/DS at higher frequency and was once postulated to be a prerequisite for the adjacent IdoUA formation by glucuronyl C5-epimerase, and GalNAc 4-O-sulfotransferase(s) had been suggested to be a critical and rate-limiting factor during DS biosynthesis (28). However, it was recently demonstrated that human skin fibroblast microsomes exhibited GalNAc 4-O-sulfotransferase activity with a marked substrate preference to dermatan, chemically desulfated DS, rather than chondroitin (29), suggesting that C5-epimerization of GlcUA residues can precede 4-O-sulfation of GalNAc residues. Indeed, recent molecular cloning studies revealed that two chondroitin 4-O-sulfotransferases termed C4ST-1 and C4ST-2 also catalyzed the 4-O-sulfation of GalNAc in dermatan as well as in chondroitin (30 -32). However, the detailed properties of C4STs and the precise spatial arrangement of IdoUA residues and 4-O-sulfation sites have not been investigated. In addition to C4ST-1 and C4ST-2, several other sulfotransferases that belong to the same HNK-1 sulfotransferase (HNK-1ST) family have also been cloned, although they catalyze different regioselective sulfations of other types of glycan chains: HNK-1ST, which facilitates 3-O-sulfation of a GlcUA residue in HNK-1 antigen precursor oligosaccharide chains of glycoproteins and/or glycolipids (33,34), and N-acetylgalactosamine 4-O-sulfotransferases-1 and -2 (GalNAc4ST-1 and GalNAc4ST-2), which catalyze 4-O-sulfation at the nonreducing terminal GalNAc residues in the GalNAc␤1-4GlcNAc␤1-containing sequences in both N-and O-glycans (35)(36)(37)(38).
In an attempt to identify additional sulfotransferase(s) by public data base searches, we found and identified another human GalNAc 4-O-sulfotransferase, which acted primarily on DS. During characterization of this enzyme, the cDNA and the catalytic activity of the identical gene product were reported, and the enzyme was designated as D4ST-1 (39). The acceptor specificity was also characterized to some extent using dermatan, nearly exhaustively desulfated DS, as an acceptor. In this study, we found that partially desulfated DS also serves as an excellent acceptor, which allowed us to investigate more detailed acceptor specificity toward the recognition sequences of D4ST-1 compared with those of C4ST-1 and -2. 35 S-Labeled 3Ј-phosphoadenosine 5Ј-phosphosulfate (PAPS) (1.69 mCi/mmol) was purchased from PerkinElmer Life Sciences. Unlabeled PAPS and GalNAc 4-O-sulfate were obtained from Sigma. The following sugars and enzymes were purchased from Seikagaku Corp. (Tokyo, Japan): chondroitin (a chemically desulfated derivative of whale cartilage CS-A); various CS preparations including CS-A, CS-B (porcine skin DS), CS-C, CS-D, and CS-E; six unsaturated standard disaccharides derived from CS (i.e. ⌬Di-0S, ⌬Di-4S, ⌬Di-6S, ⌬Di-diS D, ⌬Di-diS E , and ⌬Di-triS); conventional and highly purified (protease-free) chondroitinase ABC ( (40) for varying time periods using porcine skin DS, which contains ϳ90% and less than 1% 4-O-sulfated and nonsulfated disaccharide units, respectively. The disaccharide composition analysis by chondroitinase ABC digestion of the preparations, which were obtained after 1, 2.5, 5, 7, or 12 h of solvolysis, showed that they contained 4-O-sulfated disaccharide units at 80, 50, 31, 21, and 0.6%, respectively. For most assays, the preparation that contained 4-O-sulfated disaccharide units at 31% was used. The K m values of D4ST-1 for this partially desulfated DS and the exhaustively desulfated dermatan were 11 and 13 g/ml, respectively.

Materials-
Cloning of D4ST-1 cDNA-A TBLASTN search of the human EST data base at GenBank™ using the query sequence of human C4ST-1 (31, 32) retrieved a human EST (GenBank™ accession number R50932) corresponding to the IMAGE Consortium cDNA clone ID number 37420, which was obtained from Genome System, Inc. (St. Louis, MO).
The insert cDNA was sequenced and consisted of a partial open reading frame with significant sequence similarity to human C4ST-1, followed by a 3Ј-untranslated region containing a potential polyadenylation signal. Further BLASTN searches against a human genome data base using the obtained EST sequence as a query identified a human genomic clone (GenBank™ accession number AC013356) including one contig with 100% identity to the EST sequence. The genomic sequence was analyzed using a gene prediction program, GENSCAN (41), and the predicted complete coding sequence was obtained. The putative complete open reading frame encoding a putative sulfotransferase was amplified from human placenta cDNA library (gt10) (Clontech, Palo Alto, CA, USA) by two-round PCR using specific primer sets corresponding to the 5Ј-and 3Ј-noncoding regions, based on the human genomic clone sequence. The first PCR was performed with a forward primer, 5Ј-CAA CTA CCC CCG GTC CCA GA-3Ј, and a reverse primer, 5Ј-CAT CCA GGA TCC TGA GAA GC-3Ј, followed by nested PCR with nested primers: a forward primer, 5Ј-ACC CCT TGA GCA CCA TGT TCC-3Ј, and a reverse primer, 5Ј-AGC TGG CGT TGA AAC CAG TC-3Ј. Each PCR was carried out with KOD-Plus (TOYOBO, Tokyo) in the presence of 5% (v/v) dimethyl sulfoxide by 30 cycles at 94°C for 30 s, 55°C for 42 s, and 68°C for 2.5 min. The amplified cDNA fragment of an expected size (ϳ1.1 kbp) was subcloned into a pGEM®-T Easy vector (Promega, Tokyo) and sequenced in a 377 DNA sequencer (PerkinElmer Life Sciences), which identified the clone as the recently reported D4ST-1 (39). The additional 5Ј-noncoding sequence was also amplified using a human placenta cDNA library as a template by PCR with a gene-specific reverse primer and a gt10 insert screen amplimer (Clontech).

Construction of Expression Vectors Encoding Soluble Forms of Sulfotransferases-
The cDNA encoding a truncated form of D4ST-1 lacking the first NH 2 -terminal 62 amino acid residues was amplified by PCR with the pGEM®-T Easy vector containing the full coding sequence of the protein using a 5Ј primer containing an in-frame BamHI site (5Ј-GCG GAT CCG GCA TCC TGG CCG AGA TG-3Ј) and a 3Ј primer containing a BamHI site located 27 bp downstream from the stop codon (5Ј-GCG GAT CCG GCG TTG AAA CCA GTC CC-3Ј). PCR was carried out with Pfu polymerase (Promega) by 30 cycles at 95°C for 42 s, 65°C for 42 s, and 72°C for 3.5 min. The PCR products of the expected size were digested with BamHI, cloned into the BamHI site of an expression vector, pEF-BOS/IP (42), and sequenced. The resultant vector contained the cDNA encoding a fusion protein that had an NH 2 -terminal cleavable insulin leader peptide and a protein A IgG-binding domain followed by a truncated form of D4ST-1.
Likewise, other expression constructs of the following sulfotransferases were designed to exclude the membrane-spanning segment and to include the stem region of the longest possible size. In the case of C4ST-1 (31,32), the cDNA fragment encoding a truncated form of the protein, lacking the NH 2 -terminal first 57 amino acid residues, which contained a putative proteolytic cleavage site in mouse C4ST-1 (30), was amplified by PCR using a human placenta cDNA library (Clontech) as the template with a 5Ј primer containing an in-frame BamHI site (5Ј-CGG GAT CCC TGC AGG AAC TCT ACA AC-3Ј) and a 3Ј primer corresponding to 16 bp downstream from the stop codon (5Ј-CGG GAT CCT AAA AAG CAT GAT TCT CTC-3Ј).
In the case of human C4ST-2 (31), the cDNA encoding a truncated form of the protein, lacking the NH 2 -terminal first 76 amino acid residues, was amplified by PCR using the EST clone (GenBank™ accession number AA182540, IMAGE Consortium cDNA clone, ID number 613430) obtained from Genome System, Inc. as the template with a 5Ј primer containing an in-frame BamHI site (5Ј-CGG GAT CCG GCG TGA AGC AGA GCG AC-3Ј) and a 3Ј primer corresponding to 21 bp downstream from the stop codon (5Ј-CGG GAT CCC GTC AGG TTC CAG GCA CG-3Ј).
Expression of the Soluble Forms of the Recombinant Sulfotransferases-Each expression plasmid (6.7 g) was transfected into COS-1 cells in 100-mm plates using FuGENE™6 (Roche Applied Science) according to the manufacturer's instructions. Two days after transfection, a 1-ml aliquot of the culture medium was incubated with 10 l of IgG-Sepharose beads (Amersham Biosciences) for 1 h at 4°C. The enzyme-bound beads were washed with and resuspended in each assay buffer described below and used as enzyme sources for sulfotransferase assays.
Sulfotransferase activities toward CS/DS variants were assayed by the method described previously (43) with slight modifications. Briefly, the standard reaction mixture (60 l) contained 10 l of the resuspended beads, 50 mM imidazole-HCl, pH 6.8, 2 mM dithiothreitol, 10 M [ 35 S]PAPS (ϳ1 or 3 ϫ 10 5 dpm), and an acceptor polysaccharide preparation (10 nmol as GlcUA). The reaction mixtures were incubated at 37°C for 1 h and subjected to gel filtration using a syringe column packed with Sephadex G-25 (superfine) (44). [ 35 S]Sulfate incorporation into polysaccharides was quantified by determination of the radioactivity in the flow-through fractions by liquid scintillation counting.
Identification of the Transferase Reaction Products-The 35 S-labeled chondroitin and partially desulfated DS were isolated by gel filtration as described above, dried, and subjected to exhaustive digestion with chondroitinase ABC, AC-I, or B (45). The digest was analyzed by anion exchange HPLC on an amine-bound silica PA03 column (46) or by gel filtration on a Superdex™ Peptide column (Amersham Biosciences) equilibrated with 0.2 M NH 4 HCO 3 (47) as described previously. Chondro-4-O-sulfatase digestion was conducted as previously reported (48).
A large scale D4ST-1 reaction product, required for a detailed analysis of the sulfation sites, was prepared using partially desulfated DS (8 mol as GlcUA) as an acceptor. The reaction sample was exhaustively digested with chondroitinase AC-I, and the digest was fractionated by gel filtration on a Superdex™ Peptide column as described above.
Radioactive fractions corresponding to tetra-and hexasaccharides were pooled and further fractionated by anion exchange HPLC on an aminebound silica column as described above. Eluates were monitored by absorption at 232 nm. Each radioactive tetra-or hexasaccharide fraction was pooled and desalted through a column (0.8 ϫ 57 cm) of Sephadex G-25 (fine) (Amersham Biosciences). The disaccharide composition of each isolated fraction was determined by HPLC analysis of the chondroitinase ABC digest. Digestion of the isolated hexasaccharide fractions with a highly purified chondroitinase ABC preparation was carried out as described previously (49).

Delayed Extraction Matrix-assisted Laser Desorption Ionization Time-of-flight (DE MALDI-TOF) Mass Spectrometry-DE MALDI-TOF
mass spectra in the positive ion mode of the isolated tetra-and hexasaccharide fractions were recorded in the linear mode in a Voyager DE-RP/Pro (PerSeptive Biosystems, Framingham, MA) (50). Each oligosaccharide fraction (30 pmol) was mixed with a small volume of an aqueous solution (10 mg/ml) of a matrix, 2,5-dihydroxybenzoic acid. An aliquot (1 l) of this sample-matrix mixture was placed on the sample plate well, dried under an air stream, and analyzed. 500-MHz 1 H NMR-The isolated oligosaccharide fractions for NMR analysis were repeatedly exchanged in 2 H 2 O with intermittent lyophilization. The 500-MHz 1 H NMR spectra of the oligosaccharide fractions 4-III and 6-II were recorded in a Varian VXR-500 spectrometer at a probe temperature of 26°C as reported previously (45,51,52). Chemical shifts are given relative to sodium 4,4-dimethyl-4-silapentane-1sulfonate but were actually measured indirectly relative to acetone (␦ 2.225) in 2 H 2 O (53).

Evolutionary Relations of Chondroitin/Dermatan 4-O-Sulfo-
transferases-A TBLASTN search, using the catalytic domain of a previously cloned human C4ST-1 (31,32), showed two human cDNAs encoding putative sulfotransferases with PAPS binding motifs (54) and a type II transmembrane topology. The deduced amino acid sequence of one of the two cDNAs (Gen-Bank™ accession number AB066595) was identical to D4ST-1, which catalyzes preferentially 4-O-sulfation of GalNAc residues in dermatan rather than chondroitin (39). It showed a significant homology to the HNK-1ST family members of human origin (C4ST-1 (31, 32), C4ST-2 (31), GalNAc4ST-1 (35,36,38), GalNAc4ST-2 (37,38), and HNK-1ST (34)) and is the sixth family member. The other cDNA also shared significant sequence identities and structural similarities with the same family members and was identified as chondroitin 4-O-sulfotransferase-3 (C4ST-3) recently reported as the seventh member (55). As shown in Fig. 1, the evolutionary relations of these gene products showed that D4ST-1 was the closest to C4ST-2, and C4ST-3 was most similar to C4ST-1, suggesting that the sulfotransferases responsible for 4-O-sulfation of CS/DS should be categorized as a "CS/DS 4-O-sulfotransferase subfamily." Differential Preferences of D4ST-1, C4ST-1, and C4ST-2 toward Chondroitin and Partially Desulfated DS-To facilitate the functional analysis of D4ST-1, a soluble form of the protein was generated by replacing the first 62 amino acids with a cleavable insulin signal sequence and a protein A IgG-binding domain as described under "Experimental Procedures," and then the soluble chimeric sulfotransferase was transiently expressed in COS-1 cells. The recombinant enzyme secreted into the culture medium was adsorbed to IgG-Sepharose beads for elimination of endogenous sulfotransferases, and then the enzyme-bound beads were used for enzyme assays.
In view of its evolutionary relation to the CS/DS 4-O-sulfotransferase subfamily members ( Fig. 1), sulfotransferase activity of the purified fusion protein was examined using various CS/DS acceptor substrates. As shown in Table I, the enzyme preferentially transferred [ 35 S]sulfate from [ 35 S]PAPS to partially desulfated DS, chondroitin, and DS. The sulfate incorporation into partially desulfated DS was ϳ10-fold greater than that into chondroitin, with negligible activity toward several CS variants. The activities toward partially desulfated DS preparations, which contained 4-O-sulfated disaccharide units at 21, 31, or 50%, were comparable (Ϯ15%) with that toward exhaustively desulfated DS, when measured at an acceptor concentration of 10 g/ml, whereas the activities toward less desulfated DS containing 4-O-sulfated disaccharide units at 80% were significantly (30 -50%) lower than that toward exhaustively desulfated DS (data not shown). Therefore, subsequent assays were carried out using the partially desulfated DS The recombinant D4ST-1, C4ST-1, C4ST-2, and C4ST-3 were assayed using each of various GAG polysaccharides as the acceptor (10 nmol as GlcUA) under the reaction conditions described under "Experimental Procedures." The reaction products were separated from [ 35 S]PAPS by gel filtration on a syringe column packed with Sephadex G-25 (superfine). The radioactivity was measured by liquid scintillation counting. Acceptor a The values represent the averages of two independent experiments. b No detectable sulfotransferase activities were detected towards CS/DS variants tested when the vector-control was used as an enzyme source.
c When sulfotransferase activity of C4ST-3 was assessed at 28°C as reported previously (55) for 12 h, the appreciable incorporation of sulfate into chondroitin (4.7 pmol) and partially desulfated DS (2.9 pmol), respectively, was detected. Under the same conditions, negligible activities were detected toward natural CS variants A-E. containing 4-O-sulfated units at 31%. No detectable sulfotransferase activity was observed when a control sample prepared from the pEF-BOS/IP expression vector-transfected cells was used as an enzyme source.
We also examined sulfotransferase activities of C4ST-1 and C4ST-2 toward a series of CS/DS substrates using truncated forms of recombinant C4ST-1 and C4ST-2 under the same assay conditions as those for D4ST-1 for comparison. As shown in Table I, a soluble form of the recombinant C4ST-1 efficiently transferred sulfate to both chondroitin and partially desulfated DS with ϳ2-fold greater incorporation into the former than into the latter with far less activity toward other CS variants. Although a soluble form of C4ST-2 exhibited lower activity compared with D4ST-1 and C4ST-1, it also exhibited appreciable activity toward CS variants in addition to partially desulfated DS as in the case of C4ST-1. A soluble form of C4ST-3 showed negligible activity toward all the tested acceptors when compared with the other enzymes in our assay system (Table I). However, when assayed at 28°C as reported previously (55), appreciable activities could be detected with chondroitin and partially desulfated DS but not with other CS variants (see Footnote c for Table I). Due to the extremely low activity near the detection limit, further characterization of the substrate specificity of C4ST-3 was not carried out.

4-O-Sulfation
Sites within DS-To further distinguish the acceptor specificity of D4ST-1 from those of C4ST-1 and C4ST-2, an aliquot of 35 S-labeled reaction products, which were obtained by incubation of a truncated form of D4ST-1, C4ST-1, or C4ST-2 with partially desulfated DS, was structurally characterized. The 35 S-labeled products, obtained from each transferase reaction, were digested exhaustively with either chondroitinase AC-I or chondroitinase B, which cleaves GalNAc␤1-4GlcUA linkages or GalNAc␤1-4IdoUA linkages, respectively. It should be noted that the sulfation of GalNAc residues flanking a target IdoUA residue is essential for the action of chondroitinase B (45,56), although chondroitinase AC-I reactions are not much influenced by the sulfation pattern of neighboring sugar residues (58). One-half of each digest was analyzed by gel filtration (Fig. 2) or anion exchange HPLC (Fig. 3).
In the analysis of radiolabeled products from the D4ST-1 reaction, only a small proportion (8%) of the total radioactivity was detected at the position of ⌬Di-4S for the chondroitinase AC-I digest, whereas as much as 32% of the total radioactivity was identified as ⌬Di-4S for the chondroitinase B digest (Fig. 2,  A and B), suggesting the strong preference of D4ST-1 for the sequence -IdoUA-GalNAc(4S)-IdoUA-over -GlcUA-GalNAc(4S)-GlcUA-. In contrast, chondroitinase AC-I or B digestion of the 35 S-labeled products from the C4ST-1 reaction gave rise to the radiolabeled ⌬Di-4S, which accounted for 50 or 2% of the total radioactivity, respectively (Fig. 2, C and D), suggesting the marked preference of C4ST-1 for the sequence -GlcUA-GalNAc(4S)-GlcUA-over -IdoUA-GalNAc(4S)-IdoUA-. Interestingly, in the case of the C4ST-2 reaction products, comparable proportions of radioactivity (nearly 20 -25% of the total) were recovered as ⌬Di-4S after digestion with chondroitinase AC-I or B (Fig. 2, E and F). These results are summarized in Fig. 4, which revealed marked differences among the three sulfotransferases in terms of the preference for the isomeric structures of the uronic acids flanking the target GalNAc residue.
C4ST-1 showed a marked preference for GalNAc residues in the GlcUA-rich regions typical of CS, which are also dispersed in DS chains, being consistent with previous findings (30 -32). In contrast, D4ST-1 was found to utilize mainly GalNAc residues in the IdoUA-rich repeating region. C4ST-2, whose substrate preference was not fully characterized previously (31), catalyzes the sulfation of GalNAc residues in the IdoUA-rich region in addition to those in the GlcUA-rich regions.
To locate the sulfation sites within the oligosaccharides, the chondroitinase AC-I digest from large scale D4ST-1 reaction products was fractionated by gel filtration. The tetra-and hexasaccharide fractions were further subfractionated by anion exchange HPLC. Five (4-I to 4-V) and four (6-I to 6-IV) major UV-absorbing peaks were isolated from the tetra-and hexasaccharide fraction, respectively (Fig. 5). Among the isolated fractions, major radioactive fractions 4-III, 4-V, 6-I, 6-II, 6-III, and 6-IV were subjected to structural analysis as described below. Fractions 4-I and 4-II were not radiolabeled and therefore were not analyzed.
The disaccharide composition of each isolated tetra-and hexasaccharide fraction was determined by digestion with chondroitinase ABC in conjunction with HPLC (46), and the results are summarized in Table III. The digestion of each oligosaccharide fraction gave only two kinds of unsaturated disaccharide units, ⌬Di-0S and/or ⌬Di-4S, suggesting that the sulfation positions were restricted to C-4 positions of GalNAc residues. Since the oligosaccharide fractions were isolated from the radiolabeled partially desulfated DS after exhaustive digestion with chondroitinase AC-I, the internal uronic acid residues in the major component(s) in each fraction were IdoUA residues, as expected.  The digestion of fractions 4-V and 6-IV with chondroitinase ABC gave a single peak at the elution position of ⌬Di-4S (Table III). Based on these and the results from the DE MALDI-TOF mass spectrometry analysis (Table II), the following structures are proposed as the major compounds in fractions 4-V and 6-IV, respectively: fraction 4-V, ⌬HexUA␣1-3GalNAc(4S)␤1-4IdoUA␣1-3GalNAc(4S); fraction 6-IV, ⌬HexUA␣1-3GalNAc(4S)␤1-4IdoUA␣1-3GalNAc-(4S)␤1-4IdoUA␣1-3GalNAc(4S).

MALDI-TOF/MS analysis of the tetra-and hexasaccharide fractions isolated from D4ST-1 reaction products
The digestion of fraction 4-III or 6-II with chondroitinase ABC gave two major products, ⌬Di-0S and ⌬Di-4S, in a molar ratio of 1.0:1.0 or 1.0:2.1, respectively, as determined by HPLC (Table III). Hence, fractions 4-III and 6-II contained monosulfated tetrasaccharide and disulfated hexasaccharide, respectively, as major components, being consistent with the results from the DE MALDI-TOF analysis. The oligosaccharide fractions 4-III and 6-II were then analyzed using 500-MHz 1 H NMR spectroscopy. The one-dimensional and two-dimensional correlation spectroscopy (COSY) spectra of these fractions (4-III and 6-II) are depicted in Figs. 7 and 8, respectively. The proton chemical shifts were assigned by the analysis of the two-dimensional COSY spectra. These NMR findings are summarized in Table IV.
The chemical shifts of most of the protons of the tetrasaccharide in fraction 4-III and the hexasaccharide in fraction 6-II were identical with those reported for the corresponding protons of analogous tetra-and hexasaccharides (52). All of the internal uronic acid residue(s) of the oligosaccharides in fractions 4-III and 6-II were confirmed as IdoUA residues based upon the chemical shifts (␦ 4.926 and 4.881 for fraction 4-III; ␦ 4.911, 4.867, and 4.881 for fraction 6-II) and the coupling constants J 1,2 (2.5 and 2.0 Hz for fraction 4-I; 3.5, 3.5, and 3.5 Hz for fraction 6-II) of the anomeric proton signals. The coupling constants J 1,2 of ␣IdoUA and ␤GlcUA of CS/DS have been reported to be 3.0 and 8.0 Hz, respectively (59,60). Compared with the proton chemical shifts reported for the reference compound, ⌬HexUA␣1-3GalNAc(4S)␤1-4IdoUA␣1-3GalNAc(4S) (52), a large upfield shift of H-4 of GalNAc-3 (⌬0.534 ppm) was observed on NMR of fraction 4-I, suggesting that the GalNAc-1 was 4-O-sulfated, whereas the C-4 position of the GalNAc-3 residue was not. Likewise, compared with the proton chemical shifts reported for the reference compound, ⌬HexUA␣1-
Fractions 4-V and 6-III contained disulfated structures with 4-O-sulfated GalNAc residues on the reducing and nonreducing side of IdoUA residue(s). Interestingly, the specific radioactivity of fraction 4-V was ϳ6.8-fold higher than that of fraction 4-III (Table V), suggesting that the GalNAc residues on the nonreducing side of IdoUA residues (i.e. in -GlcUA␤1-3Gal-NAc␤1-4IdoUA-sequences) may be sulfated by D4ST-1, although the opposite possibility also exists. To clarify this point, the trisulfated hexasaccharide (6-IV) was digested with highly purified chondroitinase ABC, taking advantage of its exolytic action on a hexasaccharide to produce a disaccharide and a tetrasaccharide from the nonreducing and reducing sides of the parent hexasaccharide, respectively. The enzyme yielded two radioactive peaks corresponding to the elution positions of ⌬Di-4S and disulfated tetrasaccharide [⌬HexUA␣1-3GalNAc-(4S)␤1-4IdoUA␣1-3GalNAc(4S)] with ϳ33 and 67% of the total radioactivity of the parent oligosaccharide, respectively (Fig. 10), suggesting that D4ST-1 at least catalyzed to a considerable degree the 4-O-sulfation of the GalNAc residue located at the nonreducing side of the hexasaccharide sequence GlcUA␣1-3GalNAc(4S)␤1-4IdoUA␣1-3GalNAc(4S)␤1-4IdoUA␣1-3GalNAc(4S) that had been embedded in the parent DS polysaccharide. Namely, D4ST-1 was involved in 4-Osulfation of a GalNAc residue in the sequence -GlcUA␤1-3GalNAc␤1-4IdoUA-. In the case of disulfated oligosaccharide fractions (6-II and 6-III), the positions of radioactive sulfate group(s) in their sequences could not be determined in this study due to the low incorporated radioactivity. Although fraction 4-V had a high specific radioactivity, the location of the radiolabeled sulfate was not analyzed due to the difficulty in discriminating the two sulfate groups.
In strong contrast to the radioactive peaks corresponding to the sulfated tetra-and hexasaccharides with retention times between 24 and 55 min, which were generated by chondroitinase AC-I (Fig. 3A) or B (Fig. 3B) digestion of the D4ST-1 transferase reaction products, the C4ST-1 reaction products, prepared with partially desulfated DS as an acceptor, yielded radiolabeled ⌬Di-4S as a predominant peak along with much smaller proportions of radioactive tetra-and hexasaccharide peaks upon digestion with chondroitinase AC-I (Fig. 3C), and the chondroitinase B digest gave no significant radioactive peaks (Fig. 3D). These results suggested that C4ST-1 utilized predominantly the sequences containing consecutive GlcUA residues but not those adjacent to IdoUA. Chromatographic patterns of the chondroitinase AC-I and B digests of the Chemical shifts are given in ppm downfield from internal sodium 4,4-dimethyl-4-silapentane-1-sulfonate, but actually measured indirectly in reference to acetone (␦ 2.225 ppm) in 2 H 2 O at 26°C. The estimated error for the values to two decimal places was only Ϯ0.01 ppm because of partial overlap of signals. That for the values to three decimal places was Ϯ0.002 ppm. Coupling constants J 1,2 (in Hz) of GalNAc, IdoUA, and ⌬HexUA residues are given in parentheses. ND, not determined; -, not occurring. Sugar  C4ST-2 reaction products were very similar to those obtained from the D4ST-1 reaction products, except for the disaccharide peak (Figs. 2, A, B, E, F, and 3, A, B, E, and F). These results suggest that C4ST-2 may possess both properties of D4ST-1 and C4ST-1. Hence, C4ST-2 can be considered as C4ST-2/D4ST-2.

DISCUSSION
The evolutionary relations shown in Fig. 1 suggest that D4ST-1 and C4ST-1, -2, and -3 constitute a CS/DS 4-O-sulfotransferase subfamily in the HNK-1ST family as proposed in this study. The close evolutionary relation between D4ST-1 and C4ST-2/D4ST-2 is consistent with their substrate specificities as discussed below. D4ST-1, C4ST-1, C4ST-2, and C4ST-3, cloned from mouse and human (30 -32, 39, 55), exhibit GalNAc 4-O-sulfotransferase activity toward dermatan and chondroitin. However, their specificities, especially toward DS lineage, have not been fully clarified. In this study, we investigated the acceptor specificities of these enzymes except for C4ST-3, which was phylogenetically closest to C4ST-1 but showed negligible catalytic activity toward the above mentioned substrates. Although the low activity was reportedly due to its thermal instability (55), specific substrates other than chondroitin may exist for C4ST-3 in view of its restricted expression profile in tissues such as liver and kidney (55).
The acceptor specificity of D4ST-1 toward CS/DS variants was clearly distinct from those of C4ST-1 and -2 (Table I). Among natural CS/DS variants, only CS-B (DS) was utilized by D4ST-1, whereas C4ST-1 and -2 showed broader specificity, exhibiting significant activities toward various natural CS/DS variants. With chondroitin and partially desulfated DS as an acceptor, significantly higher activities were detected for all three enzymes compared with those obtained with natural CS/DS as acceptors. D4ST-1 exhibited a stronger preference to partially desulfated DS. In contrast, C4ST-1 showed a greater preference for chondroitin, and C4ST-2 exhibited comparable preferences for both acceptors, although C4ST-2 activity especially toward partially desulfated DS was significantly lower than those of D4ST-1 and C4ST-1. Whereas these results are basically consistent with those reported recently by Evers et al. (39), the novel acceptor specificity of D4ST-1 was also revealed in this study, as described below, when the reaction products were analyzed in detail.
Analysis of the transferase reaction products clearly showed that the observed differential specificities of the three enzymes represent differences in their preferences for the GalNAc residues flanked by either IdoUA or GlcUA residue(s) in the polymers, as evidenced by the relative abundance of 35 S-labeled disaccharide (⌬Di-4S) in the chondroitinase AC-I or B digest of the 35 S-labeled DS product (Fig. 4). D4ST-1 predominantly utilized GalNAc residues in the -IdoUA␣1-3GalNAc␤1-4Ido-UA-sequence, whereas C4ST-1 efficiently worked on GalNAc residues in -GlcUA␤1-3GalNAc␤1-4GlcUA-. C4ST-2 used Gal-NAc residues in both sequences to comparable extents. Chondroitinase AC-I digestion of the D4ST-1 reaction products, obtained with partially desulfated DS as acceptor, yielded multiple radioactive tetra-and larger oligosaccharides containing internal IdoUA residue(s) in addition to disaccharides. The structural determination of the isolated tetra-and hexasaccharides provided detailed information about the 4-O-sulfation sites within the sugar chain. It was clearly demonstrated for the first time that D4ST-1 could catalyze 4-O-sulfation of Gal-NAc residues not only in the sequence -IdoUA␣1-3GalNAc␤1-4IdoUA-but also in -IdoUA␣1-3GalNAc␤1-4GlcUA-and -Glc-UA␤1-3GalNAc␤1-4IdoUA-. The demonstrated sequence preferences of D4ST-1, C4ST-1, and C4ST-2 are illustrated in Fig. 11.
During the course of this study, Evers et al. (39) reported the acceptor specificity of the same enzyme, which acted preferentially on nearly exhaustively desulfated dermatan prepared from porcine intestine. After digestion of the transferase reaction products with chondroitinases AC-I and B, a tetrasaccharide ⌬HexUA-GalNAc-IdoUA-GalNAc(4S) and a hexasaccharide ⌬HexUA-GalNAc(4S)-GlcUA-GalNAc-GlcUA-GalNAc were isolated, respectively, as major fragments, suggesting that they were derived from a parent sequence, -GlcUA-GalNAc-IdoUA-GalNAc(4S)-GlcUA-GalNAc-GlcUA-GalNAc-IdoUA-GalNAc-. Thus, they concluded that D4ST-1 shows a strong preference for IdoUA-GalNAc flanked by GlcUA-GalNAc but not for IdoUA-GalNAc flanked by IdoUA-GalNAc, IdoUA-rich repeating regions containing -IdoUA␣1-3Gal-NAc␤1-4IdoUA-sequences that were scarcely utilized as sulfation sites. The contradictory findings between their and our studies are probably due to the structural differences of the acceptors. In contrast to the nonsulfated dermatan used in their study, our partially desulfated DS containing 4-Osulfated GalNAc at 31% also served as an excellent acceptor for D4ST-1, and the GalNAc residues located on both reducing and nonreducing sides of internal IdoUA residues were efficiently sulfated by the enzyme, despite its lower frequency of free GalNAc residues in the polymer compared with the exhaustively desulfated form. Although the reactions observed with nonsulfated dermatan probably represent initial sulfation reactions, they probably do not display an overall picture of the DS synthetic processes involving D4ST-1 to form a mature highly sulfated chain. In contrast, the reactions observed with partially desulfated DS appear to include multiple distinct sulfation reactions of D4ST-1 in terms of the sulfate acceptor sequence, which take place at various stages of the biosynthetic process required for the maturation of DS chains as discussed below. FIG. 9. HPLC analysis of the enzyme digests of the fractions 6-I (A) and 6-III (B) prepared using a highly purified preparation of chondroitinase ABC. The enzyme digests of fractions 6-I (A) and 6-III (B) (1 nmol each) prepared using a highly purified preparation of chondroitinase ABC were analyzed by HPLC on an amine-bound silica column using a liner gradient of NaH 2 PO 4 as described under "Experimental Procedures." Fractions were monitored by absorbance at 232 nm. Arrow 1, ⌬Di-0S; arrow 2, ⌬HexUA␣1-3GalNAc␤1-4IdoUA␣1-3GalNAc(4S); arrow 3, ⌬HexUA␣1-3GalNAc(4S)␤1-4IdoUA␣1-3Gal-NAc; arrow 4, ⌬Di-4S. The peaks marked by asterisks are derived from the reaction buffer and represent unidentified impurities.
Comparison of the specific radioactivity of the isolated tetraand hexasaccharide fractions revealed the effects of preexisting sulfate groups on the subsequent sulfation reactions. The specific radioactivity of fraction 4-V was 6.8-fold higher than that of fraction 4-III, and that of 6-IV was 5.2-fold higher than that of fraction 6-III (Table V). These results suggest that preexisting sulfate group(s) promoted the following sulfation of the GalNAc residue immediately adjacent to the sulfated disaccharide unit, although the order of incorporation of the sulfate groups to different GalNAc residues, for example, in fractions 4-V and 6-IV was not determined. Thus, initial 4-O-sulfation of GalNAc residues either at random or at stringently defined positions promotes further 4-O-sulfation of the GalNAc residues in the -IdoUA-GalNAc-IdoUA-GalNAc(4S)-and/or -Gal-NAc(4S)-IdoUA-GalNAc-IdoUA-sequences by D4ST-1, consequently forming IdoUA-GalNAc(4S)-rich regions.
Another role of sulfation in DS biosynthesis appears to inhibit possible back-epimerization of IdoUA to GlcUA. Although the epimerization reaction in the nonsulfated chondroitin and dermatan is reversible, the epimerase hardly works on CS or DS (61). In the absence of PAPS, the epimerization reaction in the nonsulfated chondroitin is reversible and reaches a particular equilibrium state, where the D-gluco configuration dominates over the L-ido configuration (61). The degree of interconversion to IdoUA is much lower than actual IdoUA contents observed in native DS (62). A marked 4-O-sulfation modification of dermatan has been observed using human skin fibroblast microsomes as an enzyme source (29). Since the microsome preparation utilizes dermatan more efficiently than chondroitin, it has been proposed that IdoUA residues may enhance the sulfation of their neighboring GalNAc residues (29). Furthermore, based on differential enzyme activities of C5-epimerase reflecting different IdoUA/GlcUA ratios in contrast to high 4-O-sulfotransferase activities in the used tissues, it was recently postulated that C5-epimerase is the rate-limiting factor for DS synthesis (63). D4ST-1, which exhibits a striking preference for DS lineage, probably contributes substantially to IdoUA formation in vivo, effectively fixing the L-ido configuration and consequently protecting it from the back epimerization as recently discussed for the reactions of the recombinant D4ST-1 (39).
Although the D4ST-1 message is ubiquitously expressed in adult tissues (39), the IdoUA-containing disaccharide cluster shows a tissue-specific nature in abundance and distribution in the polysaccharide backbone (23, 63) (see Ref. 64 and references therein). Native DS is composed of a periodic and wavelike arrangement of IdoUA-GalNAc(4S)-containing disaccharide repeats and GlcUA-GalNAc(4S)-containing disaccharide repeats in a tissue-specific manner (23). Although the specificity of D4ST-1 reported by Evers et al. (39) suitably explains the  10. HPLC analysis of an enzyme digest of fraction 6-IV using a highly purified preparation of chondroitinase ABC. Fraction 6-IV (3 nmol) was digested with a highly purified preparation of chondroitinase ABC and analyzed by HPLC on an amine-bound silica column as described under "Experimental Procedures." The eluates were monitored by absorption at 232 nm (A) and by 35  The arrows indicate the potential sulfation sites in the schematic repeating disaccharide sequence of a CS/DS hybrid structure catalyzed by D4ST-1, C4ST-1, and C4ST-2/ D4ST-2. The thickness of each arrow represents the relative preference of each enzyme for -IdoUA-GalNAc-IdoUA-, -IdoUA-GalNAc-GlcUA-, and -GlcUA-GalNAc-IdoUA-. Possible effects of sulfate groups on the depicted sulfation reactions are not considered here. The activity of C4ST-1 toward the -GlcUA-GalNAc-IdoUA-sequence has been reported in Ref. 57. initial reactions for the formation of isolated IdoUA-GalNAc(4S) units, it appears that the synthetic mechanism of the above mentioned consecutive IdoUA-GalNAc(4S)-rich clusters typical of mature DS chains cannot be explained only by the proposal of successive actions of C5-epimerase and D4ST-1 (39,63), provided that D4ST-1 barely works on IdoUA-GalNAc flanked by IdoUA-GalNAc as reported previously (39). In contrast, the strong preference of D4ST-1 for IdoUA-GalNAc-Id-oUA revealed in the present study is ideal for rationalizing the formation of IdoUA-GalNAc(4S)-rich clusters. The application of PAPS to the epimerase assay in a cell-free system increases the IdoUA content (28), and DS formation is strongly enhanced by sulfate concentrations in a fibroblast culture (64). These findings suggest tight coupling of sulfation and epimerization. A possibility remains that 4-O-sulfation stimulates the epimerization of GlcUA to IdoUA in addition to 4-O-sulfation, which would facilitate the formation of highly sulfated IdoUA-rich clusters, as previously discussed by Hiraoka et al. (31) for possible roles of C4ST-1 and C4ST-2 in DS synthesis.
The microsome fractions prepared from three different tissues, which contain distinct proportions of IdoUA and GlcUA in the DS chains of decorin and biglycan proteoglycans, exhibited adequate GalNAc 4-O-sulfotransferase activities toward both dermatan and chondroitin (63). Thus, multiple CS and DS sulfotransferases characterized in this study are probably expressed in these tissues, and their concerted actions with C5epimerase are most likely involved in the formation of the highly sulfated IdoUA-GalNAc(4S) cluster in DS. In addition to D4ST-1, C4ST-1 may be essential for the sulfation modification of GlcUA-rich regions (30 -32) but also in IdoUA-containing regions in DS. It may trigger modification reactions if 4-Osulfation stimulates C5-epimerase as discussed above. In fact, Habuchi et al. (57) reported that C4ST-1 also transferred sulfate to C-4 positions located on the nonreducing sides of IdoUA residues in partially desulfated DS (Fig. 11), supporting the involvement of C4ST-1 in the synthesis of the IdoUA-containing region of DS. Furthermore, the present study demonstrated that C4ST-2/D4ST-2 might also be required for the sulfation of both IdoUA-and GlcUA-rich regions in DS (Fig. 11). Taken together, the collaborative or competitive contributions of these three sulfotransferases in different ratios in distinct tissues probably takes part in tissue-specific structural expression of DS, although the expression level of C5-epimerase is probably the rate-limiting factor for the DS formation. Molecular cloning of C5-epimerase(s) will make it possible to investigate possible concerted actions between the C5-epimerization and 4-O-sulfation reactions, which will allow us to depict an overall picture of the modification reactions required for the synthesis of mature DS chains.