A Unique Nonreducing Terminal Modification of Chondroitin Sulfate by N-Acetylgalactosamine 4-Sulfate 6-O-Sulfotransferase*

N-Acetylgalactosamine 4-sulfate 6-O-sulfotransferase (GalNAc4S-6ST) transfers sulfate from 3′-phosphoadenosine 5′-phosphosulfate (PAPS) to position 6 of N-acetylgalactosamine 4-sulfate (GalNAc(4SO4)). We previously identified human GalNAc4S-6ST cDNA and showed that the recombinant GalNAc4S-6ST could transfer sulfate efficiently to the nonreducing terminal GalNAc(4SO4) residues. We here present evidence that GalNAc4S-6ST should be involved in a unique nonreducing terminal modification of chondroitin sulfate A (CSA). From the nonreducing terminal of CS-A, a GlcA-containing oligosaccharide (Oligo I) that could serve as an acceptor for GalNAc4S-6ST was obtained after chondroitinase ACII digestion. Oligo I was found to be GalNAc(4SO4)-GlcA(2SO4)-GalNAc(6SO4) because GalNAc(4SO4) and ΔHexA(2SO4)-GalNAc(6SO4) were formed after chondroitinase ABC digestion. When Oligo I was used as the acceptor for GalNAc4S-6ST, sulfate was transferred to position 6 of GalNAc(4SO4) located at the nonreducing end of Oligo I. Oligo I was much better acceptor for GalNAc4S-6ST than GalNAc(4SO4)-GlcAGalNAc(6SO4). An oligosaccharide (Oligo II) whose structure is identical to that of the sulfated Oligo I was obtained from CS-A after chondroitinase ACII digestion, indicating that the terminal modification occurs under the physiological conditions. When CS-A was incubated with [35S]PAPS and GalNAc4S-6ST and the 35S-labeled product was digested with chondroitinase ACII, a 35S-labeled trisaccharide (Oligo III) containing [35S]GalNAc(4,6-SO4) residue at the nonreducing end was obtained. Oligo III behaved identically with the sulfated Oligos I and II. These results suggest that GalNAc4S-6ST may be involved in the terminal modification of CS-A, through which a highly sulfated nonreducing terminal sequence is generated.

of different positions of the sugar residues composing the repeating disaccharide units (1). The resulting sulfated chains show significant structural diversity depending on the type of tissues and cells or age of the animal from which CS was extracted. Among the sulfotransferases involved in the formation of the divergent structure, sulfotransferases belonging to the chondroitin 6-sulfotransferase family (2)(3)(4) and the chondroitin 4-sulfotransferase family (5-9) have been purified and cloned. Uronosyl 2-O-sulfotransferase was cloned as a sulfotransferase belonging to the heparan sulfate 2-sulfotransferase family (10). Chondroitin 6-sulfotransferase and chondroitin 4-sulfotransferase transfer sulfate to positions 6 and 4, respectively, of the GalNAc residue. On the other hand uronosyl 2-O-sulfotransferase transfers sulfate to position 2 of GlcA or IdoA residues (10). GalNAc4S-6ST transfers sulfate to position 6 of GalNAc(4SO 4 ) residues. We previously purified GalNAc4S-6ST from squid cartilage (11) and identified human GalNAc4S-6ST cDNA on the basis of amino acid sequences of the squid enzyme (12). Unlike squid GalNAc4S-6ST, human GalNAc4S-6ST exhibited high activity toward the nonreducing terminal GalNAc(4SO 4 ) residue of CS. Such specificity of human GalNAc4S-6ST suggests that human GalNAc4S-6ST may be involved in the modification of the nonreducing terminal of CS. The sulfotransferase activities capable of modifying the terminal end of CS-A have been found in quail oviduct (13) and human serum (14). It has been shown that, in CS of aggrecan obtained from various sources, GalNAc (4, residues are present at the nonreducing end much more abundantly than in the internal repeating units (15)(16)(17)(18). The proportion of the nonreducing terminal GalNAc(4,6-SO 4 ) contained in the aggrecan has been reported to decrease in osteoarthritis (19). CS of thrombomodulin, which is involved in the antithrombin-dependent anticoagulant activity, has been reported to bear Gal-NAc(di-SO 4 )-GlcA-GalNAc(di-SO 4 ) at the nonreducing end (20). These observations suggest that CS may participate in the various physiological processes through the nonreducing terminal structures containing GalNAc(4,6-SO 4 ) residues; however, the structure and biosynthesis of the nonreducing terminal regions of CS have not been fully understood. In this report, we investigated the structure of the nonreducing ends that * This work was supported by Grant-in-Aid for Scientific Research 5801 and Grant-in-Aid for Scientific Research on Priority Areas 10178102 from the Ministry of Education, Science, Sports and Culture of Japan and by a special research fund from Seikagaku Corporation. 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.
Preparation of the Recombinant Human GalNAc4S-6ST-Recombinant GalNAc4S-6ST was expressed as a fusion protein with FLAG peptide and was affinity-purified as described (12). The purified protein was visualized with Western blot as described below before or after N-glycosidase F digestion (Fig. 2). After N-glycosidase F digestion, a single protein band was detected at the migration position of 66 kDa that agreed well with the molecular weight, 66,160, calculated from the cDNA.
Western Blot Analysis-The affinity-purified GalNAc4S-6ST was precipitated with 10% trichloroacetic acid. The precipitate was washed with acetone and digested with recombinant N-glycosidase F (Roche Applied Science) by the methods recommended by the manufacturer. After digestion, the samples were separated by SDS-polyacrylamide gel electrophoresis as described by Laemmli (25). The separated proteins were electrophoretically transferred to a Hybond ECL membrane (Amersham Biosciences) and stained with anti-FLAG M2 monoclonal antibody (Sigma). The blot was developed with polyclonal anti-mouse IgG antibody coupled to horseradish peroxidase using an ECL detection kit and a Hyperfilm ECL (Amersham Biosciences).
Assay of Sulfotransferase Activity-GalNAc4S-6ST activity was assayed by the method described previously (12). The standard reaction mixture contained, in a final volume of 50 l, 2.5 mol of imidazole HCl, pH 6.8, 0.5 mol of CaCl 2 , 1 mol of reduced glutathione, 25 nmol (as galactosamine) of CS-A or trisaccharides, 50 pmol of [ 35 S]PAPS (about 5.0 ϫ 10 5 cpm), and enzyme. The reaction mixtures were incubated at 37°C for 20 min, and the reaction was stopped by immersing the FIG. 1. Chondroitinase ACII digestion of four trisaccharides prepared from chondroitin sulfate. Trisaccharides Tri-44 (B), Tri-46 (C), Tri-64 (D), and Tri-66 (E) were digested with chondroitinase AC II as described under "Experimental Procedures," and the degradation products were separated by SAX-HPLC as described under "Experimental Procedures." The column was developed with gradient B and monitored at 210 nm. The elution profile of the standard materials is shown in A. The number above each peak in A indicates the elution position of the standard material. Peak 1, ⌬Di-0S; peak 2, GalNAc(6SO 4 ); peak 3, GalNAc(4SO 4 ); peak 4, ⌬Di-6S; peak 5, ⌬Di-4S; peak 6, GalNAc(4,6-SO 4 ). The broad peaks observed around 20 min represent materials derived from the column because these peaks were detected in A. reaction tubes in a boiling water bath for 1 min. After the reaction was stopped, 35 S-labeled glycosaminoglycans were isolated by the precipitation with ethanol followed by gel chromatography with a fast desalting column as described previously (2), and the radioactivity was determined. When oligosaccharides were used as acceptors, the reaction mixtures were applied directly to the Superdex 30 column as described below, and the 35 S-labeled oligosaccharides were separated from 35

SO 4 and [ 35 S]PAPS.
Digestion with Hyaluronidase, ␤-Glucuronidase, Chondroitinase ACII, Chondroitinase ABC, and Chondro-6-sulfatase-Digestion with hyaluronidase was carried out for 24 h at 37°C in a reaction mixture containing, in a final volume of 0.7 ml, 35 mg of CS-C, 0.15 M NaCl in 0.1 M sodium acetate buffer, pH 5.0. Digestion with ␤-glucuronidase was carried out for 4 h at 37°C in a reaction mixture containing, in a final volume of 200 l, tetrasaccharide (ϳ1 mol as galactosamine), 10 mol of sodium acetate buffer, pH 4.5, 100 nmol of 2-acetamido-2deoxy-D-galactonic acid-1,4-lactone, 4 mol of sodium fluoride, and 18 units of ␤-glucuronidase. Under these conditions, removal of the nonreducing terminal GlcA was complete, and no release of inorganic sulfate was observed. Unless otherwise stated, digestion with chondroitinase ACII or chondroitinase ABC under the standard conditions was carried out for 4 h at 37°C in the reaction mixture containing, in a final volume of 25 l, 35 S-labeled glycosaminoglycans or 35 S-labeled trisaccharides, 1.25 mol of Tris acetate buffer, pH 7.5, 2.5 g of bovine serum albumin, and 30 milliunits of chondroitinase ACII or chondroitinase ABC. For degrading oligosaccharides derived from the nonreducing terminal of CS-A with chondroitinase ABC or chondroitinase ACII, a strong condition was used under which digestion with chondroitinase ABC or chondroitinase ACII was carried out in the reaction mixtures described above three times successively; first with 120 milliunits enzyme for 28 h, second with 100 milliunits enzyme for 18 h, and finally with 100 milliunits enzyme for 7 h. The new enzymes were added after heating the reaction mixtures at 100°C for 1 min. Digestion with chondro-6-sulfatase under the standard conditions was carried out for 5 h at 37°C in the reaction mixture containing, in a final volume of 25 l, trisaccharides or GalNAc(4,6-SO 4 ), 1.25 mol of Tris acetate buffer, pH 7.5, 2.5 g of bovine serum albumin, and 100 milliunits of chondro-6-sulfatase. After digestion of oligosaccharides with chondroitinase ABC or chondroitinase ACII under the strong conditions, digestion with chondro-6-sulfatase was carried out twice successively in the reaction mixtures described above; first with 100 milliunits enzyme for 17 h and second with 100 milliunits enzyme for 5 h.
Removal of Unsaturated Uronic Acid by Mercuric Acetate-Removal of unsaturated uronic acid was carried out as described (26). Oligosaccharides containing unsaturated uronic acid were dried and dissolved in 1 ml of 35 mM mercuric acetate in 25 mM Tris with 25 mM sodium acetate, pH 5.0. The reaction was carried out for 2 h at room temperature. After the reaction was over, the samples were applied to Dowex 50 (H ϩ ) column (bed volume, 1 ml). The column was washed with 3 ml of water. The flow-through fractions and the washings were combined and lyophilized. The lyophilized materials were further purified with Superdex 30 and SAX-HPLC.
Identification of Uronic Acid-Glycosaminoglycans or oligosaccharides (100 nmol as galactosamine) were hydrolyzed with 2 M trifluoroacetic acid at 100°C for 4 h. The hydrolysates were dried in a vacuum desiccator on P 2 O 5 and NaOH. The dried materials were dissolved in distilled water and applied to a column of Dowex 50 (H ϩ ) (bed volume, 0.5 ml). The flow-through fractions were concentrated to dryness and spotted on a cellulose thin layer plate (Funakoshi, Kyoto, Japan). The plate was developed with a solvent (pyridine:ethyl acetate:acetic acid: water, 5:5:1:3) for 4 h and stained with silver nitrate. To remove 2-O-sulfate, we treated Oligo I with 0.4 M NaOH before hydrolysis with trifluoroacetic acid (27). Oligo I was lyophilized in the presence of 0.4 M NaOH twice and then neutralized with acetic acid. The 2-O-desulfated products were isolated with Superdex 30 chromatography.
Superdex 30 Chromatography and HPLC-A Superdex 30 16/60 column was equilibrated with 0.2 M NH 4 HCO 3 and run at a flow rate of 2 ml/min. One-ml fractions were collected. Separation of the degradation products formed from 35 S-labeled glycosaminoglycans and 35 S-labeled oligosaccharides were carried out by HPLC using a Whatman Partisil-10 SAX column (4.6 mm ϫ 25 cm) equilibrated with 8 or 5 mM KH 2 PO 4 . The column was developed with gradient A (8 mM KH 2 PO 4 for 10 min followed by a linear gradient from 8 to 720 mM KH 2 PO 4 ) or gradient B (5 mM KH 2 PO 4 for 10 min followed by a linear gradient from 5 to 500 mM KH 2 PO 4 ) depending of the lot of the column; the gradient used is indicated in the legend for each figure. The fractions (0.5 ml) were collected at a flow rate of 1 ml/min and a column temperature of 40°C. The conditions used for HPLC using a YMC-Pack Polyamine II column (4.6 mm ϫ 25 cm) were the same as those for SAX-HPLC, except that the column was developed with 10 mM KH 2 PO 4 for 10 min followed by a linear gradient from 10 to 500 mM KH 2 PO 4 .

Isolation of an Oligosaccharide with the Acceptor Activity for GalNAc4S-6ST from the Nonreducing Terminal of Chondroitin
Sulfate-We previously showed that GalNAc4S-6ST efficiently transferred sulfate to the nonreducing terminal GalNAc(4SO 4 ) residue of CS-A to yield GalNAc(4,6-SO 4 ) (12). On the other hand, a highly sulfated trisaccharide bearing GalNAc(4,6-SO 4 ) at the nonreducing end was obtained from the nonreducing terminal of thrombomodulin CS by a partial digestion with chondroitinase ABC (20). GalNAc4S-6ST may possibly be involved in the synthesis of the highly sulfate trisaccharide structure from a corresponding precursor. To determine whether commercially available CS-A also contains a unique nonreducing terminal structure that could serve as the acceptor for GalNAc4S-6ST, we tried to isolate oligosaccharides from the nonreducing end of CS-A after chondroitinase ACII digestion. The strategy for detection of oligosaccharides released from the nonreducing end is based on the fact that oligosaccharides derived from the internal repeating units have unsaturated uronic acid and hence show the absorption at 232 nm, but those TABLE II Analysis of trisaccharides having sulfate groups at the position 6 or position 4 of GalNAc residues The purified trisaccharides (25 nmol as galactosamine) were digested with chondroitinase ACII. The digested materials were subjected to SAX-HPLC. Monosaccharides and unsaturated disaccharides were monitored by the absorbance at 210 nm (Fig. 1). From the elution profiles shown in Fig. 1, composition of monosaccharides and unsaturated disaccharides were calculated on the basis of our previous observation that the ratio of (molecular absorption of monosaccharides)/ (molecular absorption of unsaturated disaccharides) determined at 210 nm was 0.32 (11). The data represent molar ratios when the amount of monosaccharides were set at unity.  derived from the nonreducing end exhibit no absorption at this wavelength. Instead, oligosaccharides released from the nonreducing end of CS are able to be detected by the absorption at 210 nm.
CS-A (550 mol as galactosamine) was digested for 18.5 h at 37°C with chondroitinase ACII in the reaction mixture containing, in a final volume of 10 ml, 500 mol of Tris acetate buffer, pH 7.5, 1 mg of bovine serum albumin, and 2 units of chondroitinase ACII. The digest was subjected to gel filtration on Superdex 30 (Fig. 3A). Fractions containing trisaccharides to tetrasaccharides (indicated by a horizontal bar in Fig. 3A) were pooled and separated with SAX-HPLC (Fig. 3, B and C). When the elution profile from the SAX-HPLC detected by the absorbance at 210 nm (Fig. 3B) was compared with the elution profile detected at 232 nm (Fig. 3C), two peaks eluted around 29 and 33.5 min were found to have much higher absorption at 210 nm than at 232 nm. Of these two peaks, we analyzed the peak eluted at 33.5 min (indicated by arrowhead X in Fig. 3B), because the peak at 29 min was not obtained reproducibly. The materials eluted at 33.5 min were designated as Oligo I and were purified by the second SAX-HPLC and Superdex 30 chro-matography. About 0.57 mol (as galactosamine) of Oligo I was obtained from 550 mol of whale cartilage CS-A.
Oligo I showed absorption at 210 nm but not at 232 nm (Fig.  4, A and B). Oligo I was not degraded completely by chondroitinase ABC under the standard conditions but degraded completely under the strong conditions. When the completely degraded Oligo I was separated with SAX-HPLC, two peaks corresponding to GalNAc(4SO 4 ) and ⌬Di-diS D were detected (Fig. 4C). When Oligo I was digested with chondro-6-sulfatase after digestion with chondroitinase ABC, the second peak disappeared and shifted to the position of ⌬Di-2S (Fig. 4D). These observations clearly indicate that Oligo I is a trisaccharide with three sulfate groups, GalNAc(4SO 4 )-HexA(2SO 4 )-GalNAc (6SO 4 ). When Oligo I was digested with chondroitinase ACII under the standard conditions, no peaks corresponding to GalNAc(4SO 4 ) and ⌬Di-diS D were detected (data not shown). Even under the strong conditions, only 59% of Oligo I was degraded with chondroitinase ACII to yield GalNAc(4SO 4 ) and ⌬Di-diS D (Fig. 4E). To determine whether the observed resistance of Oligo I against chondroitinase ACII digestion might be due to the presence of IdoA in Oligo I, we analyzed the kind of HexA contained in Oligo I. Oligo I was hydrolyzed with 2 M trifluoroacetic acid at 100°C for 4 h, and the hydrolysates were separated with thin layer chromatography. Under the hydrolysis conditions, GlcA was released from CS-A and Tri-44 (Fig.  5, lanes 2 and 4), and IdoA was released from DS (Fig. 5, lane  3). No partial degradation products were detected when DS was hydrolyzed, indicating that the ␣-L-iduronosyl bonds were more labile than the ␤-D-glucuronosyl bonds as reported previously (28). From the hydrolysates of Oligo I, no IdoA was detected. Instead, GlcA was clearly detected after 2-O-desulfation by the treatment with sodium hydroxide (Fig. 5, lane 6). These observations indicate that the apparent resistance of Oligo I against chondroitinase ACII is not due to the presence of IdoA residue but due to the sulfation pattern of this oligosaccharide.
Sulfation of Oligo I with the Recombinant GalNAc4S-6ST-To determine whether Oligo I could serve as the acceptor for GalNAc4S-6ST and be converted to an oligosaccharide with GalNAc(4,6-SO 4 ) residues at the nonreducing end, we incubated Oligo I with the recombinant human GalNAc4S-6ST and [ 35 S]PAPS, and the resulting sulfated product was isolated by Superdex 30 chromatography (Fig. 6A) and the SAX-HPLC (Fig. 6B). When the sulfated Oligo I was separated by the SAX-HPLC after digestion with chondroitinase ABC under the strong conditions, the 35 S radioactivity was detected at the position of GalNAc(4,6-35 SO 4 ) (Fig. 6C), indicating that the structure of the sulfated Oligo I is GalNAc(4,6-SO 4 )-GlcA(2SO 4 )-GalNAc(6SO 4 ). The sulfated Oligo I was also degraded by chondroitinase ACII under the strong conditions to give rise to GalNAc(4,6-35 SO 4 ) (data not shown), although complete degradation of Oligo I could not be achieved under the same conditions as described above.
The relative rate of sulfation of Oligo I with GalNAc4S-6ST was compared with those of Tri-44 and Tri-46 (Fig. 7). Both Tri-44 and Tri-46 were sulfated at position 6 of the nonreducing terminal GalNAc(4SO 4 ) residues as observed previously (12). The sulfated Tri-44 and the sulfated Tri-46 were eluted 5.5 and 8.5 min earlier, respectively, than the sulfated Oligo I in the SAX-HPLC, supporting the idea that the sulfated Oligo I contains one more negative charge than the sulfated Tri-44 or the sulfated Tri-46. The K m for Oligo I, Tri-44, and Tri-46 determined by the double reciprocal plots were 13, 28, and 820 M, respectively, and the V max for Tri-44 and Tri-46 were 1.03 and 0.82, respectively, of the V max for Oligo I. These results indicate that the sulfation of position 6 of the GalNAc(4SO 4 ) residue at the nonreducing end by GalNAc4S-6ST was inhibited by the presence of the GalNAc(6SO 4 ) residue at the reducing side and stimulated by 2-O-sulfation of the penultimate GlcA residue.
Presence of the Nonreducing Terminal Structure Corresponding to the Sulfated Oligo I-Conversion of Oligo I to the sulfated product by GalNAc4S-6ST was observed in the in vitro reaction. To investigate whether such modification of the nonreducing terminal structure of CS occurs in the physiological conditions, we tried to isolate an oligosaccharide from CS-A, the structure of which is identical to that of the sulfated Oligo I. When Fig. 3 (B and C) was examined, a peak was observed at the position of the sulfated Oligo I (indicated by arrowhead Y in

FIG. 6. Isolation of the sulfated Oligo I and digestion with chondroitinase ABC. A, Oligo I was incubated with [ 35 S]PAPS and
GalNAc4S-6ST as described under "Experimental Procedures." The sulfated Oligo I was isolated by Superdex 30 chromatography. The peak fractions (indicated by a horizontal bar) were pooled, concentrated, and lyophilized. The standards were the same as those described in the legend to Fig. 3A. B and C, the isolated sulfated Oligo I was separated with SAX-HPLC before (B) or after digestion with chondroitinase ABC under the strong conditions. The column was developed with gradient A. The standards were the same as those described in the legend to  3B). This peak showed absorption at both 210 and 232 nm, suggesting that this peak contained both saturated and unsaturated oligosaccharides. To prepare the oligosaccharide eluted at the sulfated Oligo I, CS-A was digested with chondroitinase ACII as described above for the preparation of Oligo I except that the digestion was carried out for 4 h, because in the preliminary experiments, this oligosaccharide was found to be more sensitive to chondroitinase ACII than Oligo I. To remove contaminating unsaturated oligosaccharides, we treated the materials recovered in the peak Y with mercuric acetate as described under "Experimental Procedures," and the mercuric acetate-resistant component was further purified with Superdex 30 chromatography and SAX-HPLC. About 0.16 mol (as galactosamine) of the purified oligosaccharide (designated as Oligo II) was obtained from 1650 mol of whale cartilage CS-A. Oligo II was eluted at the same position as that of the sulfated Oligo I in the SAX-HPLC (Fig. 8A) and Superdex 30 chromatography (data not shown). The purified Oligo II showed no absorption at 232 nm (Fig. 8B). When Oligo II was digested with chondroitinase ABC under the strong conditions and subjected to SAX-HPLC, two peaks corresponding to GalNAc(4,6-SO 4 ) and ⌬Di-diS D were detected (Fig. 8C). When Oligo II was digested with chondro-6-sulfatase after digestion with chondroitinase ABC, the two peaks disappeared and shifted to the position of GalNAc(4SO 4 ) and ⌬Di-2S (Fig. 8D), respectively. When Oligo II was digested with chondroitinase ACII under the strong conditions, GalNAc(4,6-SO 4 ) and ⌬Di-diS D were detected (Fig. 8E). These observations clearly indicate that Oligo II is a trisaccharide with four sulfate groups, GalNAc(4,6-SO 4 )-GlcA(2SO 4 )-GalNAc(6SO 4 ). These observations strongly suggest that the terminal modification catalyzed by GalNAc4S-6ST should occur in physiological conditions.

Formation and Characterization of an Oligosaccharide from 35 S-Labeled Glycosaminoglycans Synthesized from CS-A after Incubation with [ 35 S]PAPS and Human GalNAc4S-6ST-As
shown above, the terminal modification of Oligo I occurred when the oligosaccharide was used as the acceptor for GalNAc4S-6ST. To demonstrate that such a terminal modification could occur in polysaccharide level, we analyzed the sulfated products formed from CS-A. When the 35 S-labeled glycosaminoglycans derived from CS-A after incubation with [ 35 S]PAPS and the recombinant human GalNAc4S-6ST were digested with chondroitinase ACII under the standard conditions, three radioactive peaks were obtained in SAX-HPLC (Fig. 9A). The peaks at 29 and 38.5 min corresponded to Gal-NAc(4,6-SO 4 ) and ⌬Di-diS E , respectively. The elution position of the third peak was exactly the same as that of the sulfated Oligo I. The third peak was not obtained when the 35 S-labeled glycosaminoglycan was digested with chondroitinase ABC (data not shown). The materials eluted at the position of the sulfated Oligo I (designated as Oligo III) were purified by Superdex 30 chromatography (Fig. 9B). To determined the structure of Oligo III, Oligo III was digested with chondroitinase ABC and applied to SAX-HPLC (Fig. 10A). The radioactivity appeared at the position of GalNAc(4,6-SO 4 ). To establish the position to which 35 SO 4 was transferred, we digested Oligo III with chondro-6-sulfatase after digestion with chondroitinase ABC and after being subjected to SAX-HPLC. The radioactivity of GalNAc(4,6-SO 4 ) disappeared and was shifted to the position of inorganic sulfate (Fig. 10B). These results indicate that Oligo III contained GalNAc(4,6-SO 4 ) residue at the nonreducing end. When Oligo III was digested with chondro-6sulfatase alone, the 35 S radioactivity was detected at the position of ⌬Di-diS E (Fig. 10C). However, this material was not ⌬Di-diS E but an oligosaccharide containing GalNAc(4,6-SO 4 ) at its nonreducing end, because GalNAc(4,6-SO 4 ) was formed after further digestion with chondroitinase ABC (Fig. 10D). Chondro-6-sulfatase was reported to remove sulfate groups attached to position 6 of GalNAc residue located at the reducing end of hexasaccharides containing unsaturated hexuronic acid at their nonreducing end (29). To determine whether chondro-6-sulfatase could act on oligosaccharides containing GalNAc(4SO 4 ) or GalNAc(6SO 4 ) at their nonreducing terminal in the same manner, we digested three trisaccharides, Tri-66, Tri-64, and Tri-46, with chondro-6-sulfatase and analyzed the reaction products by SAX-HPLC after chondroitinase ACII digestion (Fig. 11). After chondro-6-sulfatase digestion followed by chondroitinase ACII digestion, GalNAc(6SO 4 ) and ⌬Di-0S (Fig. 11B), GalNAc(6SO 4 ) and ⌬Di-4S (Fig. 11C), and GalNAc-(4SO 4 ) and ⌬Di-0S (Fig. 11D) were formed from Tri-66, Tri-64, and Tri-46, respectively. These results clearly indicate that chondro-6-sulfatase could remove sulfate from GalNAc(6SO 4 ) residue located exclusively at the reducing end of these trisaccharides. On the basis of the specificity of chondro-6-sulfatase indicated above, Oligo III should bear nonradioactive sulfate at position 6 of GalNAc residue located at the reducing end. When the isolated Oligo III was digested with chondroitinase ACII under the strong conditions, Oligo III was degraded to give rise to [ 35 S]GalNAc(4,6-SO 4 ) (data not shown), indicating that Oligo III contains GlcA.
Susceptibility of Oligosaccharides to Chondroitinase ACII-The susceptibility of Oligo I, the sulfated Oligo I, Oligo II, Oligo III, and Tri-46 to chondroitinase ACII under the standard conditions or strong conditions was summarized in Table III. Under the standard conditions, Oligo I was hardly degraded, but Tri-46 was completely degraded, indicating that the resistance of Oligo I against chondroitinase ACII digestion is attributable to the presence of 2-O-sulfate attached to the GlcA residue. Because Oligo II was more sensitive than Oligo I to chondroitinase ACII, the presence of nonreducing terminal GalNAc(4,6-SO 4 ) should promote the rate of reaction with chondroitinase ACII. These results indicate that the rate of degradation with chondroitinase ACII is markedly affected by the sulfation pattern of these oligosaccharides. The susceptibility of Oligo III was nearly the same as those of Oligo II and sulfated Oligo I. Because Oligo III was indistinguishable from the sulfated Oligo I in the chromatographic behaviors, the position to which sulfate was transferred, the existence of 6-sulfate on the reducing terminal GalNAc residue, and the susceptibility to chondroitinase ACII, the structure of Oligo III is most probably identical to that of the sulfated Oligo I. These results strongly suggest that the terminal modification could occur at the polysaccharide level. However, at present the possibility that Oligo III may contain GalNAc(4,6-SO 4 ) residues at the reducing end could not be excluded.

Formation of Oligo III from Various Chondroitin Sulfate Preparations
Derived from Different Sources-Oligo III was initially found in the chondroitinase ACII digests of the 35 Sglycosaminoglycan formed from whale cartilage CS-A after the reaction with GalNAc4S-6ST. To determine whether Oligo III could be formed from other CS preparations obtained from different sources, bovine cartilage CS, chick embryo cartilage CS, or sturgeon notochord CS were incubated with [ 35 S]PAPS and GalNAc4S-6ST. The 35 S-glycosaminoglycans formed from these CS preparations were digested with chondroitinase ACII and analyzed by SAX-HPLC (Fig. 12). Disaccharide compositions of these CS preparations are shown in Table I. The relative rates of incorporation of sulfate into CS from whale cartilage, bovine cartilage, chick embryo cartilage, and sturgeon notochord were 1.00, 1.81, 0.46, and 0.88, respectively. The proportion of the radioactivity recovered in the peak at the position of the sulfated Oligo I was highest when CS from chick embryo cartilage was used as the acceptor (Fig. 12C). In contrast, no peak was observed at the position of the sulfated Oligo I when sturgeon CS was used as the acceptor, although [ 35 S]GalNAc(4,6-SO 4 ) was formed (Fig. 12D). These observations indicate that the terminal structure from which Oligo III was produced is present at least in avian and mammalian CS. The ratio of Oligo III to the sum of Oligo III and GalNAc(4,6-SO 4 ) was found to be related to the contents of ⌬Di-6S in each CS (Table I); the higher the content of ⌬Di-6S was, the higher the ratio of Oligo III was. DISCUSSION In this report, we presented data that GalNAc4S-6ST could transfer sulfate to the unique nonreducing terminal sequence and catalyzed the formation of the highly sulfated structure. The highly sulfated nonreducing terminal structure produced by the reaction with GalNAc4S-6ST is present in native CS-A because Oligo II was obtained from CS-A. These observations suggest that the terminal modification catalyzed by GalNAc4S-6ST may occur in the physiological conditions. The 2-O-sulfation of the GlcA residue adjacent to GalNAc(6SO 4 ) may stimulate 6-sulfation of the nonreducing terminal GalNAc(4SO 4 ) residue, because Oligo I was much better acceptor for GalNAc4S-6ST than Tri-46.
At present the physiological role of the highly sulfated nonreducing terminal sequence is not known. Thrombomodulin with anticoagulant activity was reported to have CS as an essential functional domain (30). CS attached to thrombomodulin contained GalNAc(di-SO 4 ) at the nonreducing end (20). When thrombomodulin CS was partially digested with chondroitinase ABC, an oligosaccharide was obtained. After chondroitinase AC digestion, GalNAc(di-SO 4 ) and ⌬HexA-GalNAc(di-SO 4 ) were formed from the oligosaccharide. ⌬HexA-GalNAc(di-SO 4 ) migrated to the position of ⌬Di-diS E on paper electrophoresis at pH 1.7. However, under the conditions for the paper electro-   35 S-labeled glycosaminoglycans formed from various chondroitin sulfate preparations after the reaction with GalNAc4S-6ST. 35 S-Labeled glycosaminoglycans formed from whale cartilage CS-A (A), bovine nasal cartilage CS (B), 12-day-old chick embryo cartilage CS (C), and sturgeon notochord CS (D) after the reaction with GalNAc4S-6ST were digested with chondroitinase ACII, and the digests were separated with SAX-HPLC. The column was developed with gradient A. The standards were the same as those described in the legend to phoresis, ⌬Di-diS E could not be separated from ⌬Di-diS D ; therefore, it remains possible that the nonreducing terminal structure of thrombomodulin CS may be the same as that of Oligo II. Midkine, chemokines, and fibroblast growth factor family proteins have been reported to interact with squid cartilage CS-E (31-34). The highly sulfated nonreducing terminal sequence generated by the enzymatic reaction with GalNAc4S-6ST might interact with such molecules. Approximately 30% of the CS chains of proteochondroitin sulfate extracted from the cell matrix pool of the cultured chick embryo chondrocytes was found to have nonreducing terminal GalNAc(4,6-SO 4 ) residues, but none of the CS chains in the proteochondroitin sulfate recovered from the culture medium pools were terminated with these residues (16). These observations raise the possibility that the cell matrix proteoglycans might interact with some extracellular matrix or cell surface components through nonreducing terminal GalNAc(4,6-SO 4 ) residues of CS chains.
Under the standard conditions for chondroitinase ACII digestion, Tri-46 was degraded completely, but Oligo I was hardly affected. In contrast, Oligo II was more sensitive to chondroitinase ACII digestion than Oligo I. These observations suggest that chondroitinase ACII may recognize not only the kind of uronic acid but also sulfation pattern of the component sugar residues; the presence of 2-O-sulfate on GlcA residue may make the trisaccharide resistant to chondroitinase ACII, and the presence of nonreducing terminal GalNAc(4,6-SO 4 ) may relieve the inhibitory effect of the 2-O-sulfate. The observed resistance of Oligo I to chondroitinase ACII might be due to the presence of 3-O-sulfated GlcA (36); however, this possibility is not the case, because GalNAc(4SO 4 ) and ⌬Di-diS D were formed from Oligo I on chondroitinase ABC digestion, whereas the GlcA(3SO 4 ) residue has been reported to be degraded by chondroitinase ABC digestion (36). Because Oligo II was rather sensitive to chondroitinase ACII digestion, the amount of Oligo II obtained from CS-A after digestion with chondroitinase ACII may not necessarily reflect the amount of the nonreducing terminal sequence from which Oligo II was derived. Instead, the nonreducing terminal highly sulfated sequence may be present more abundantly than the yield of Oligo II. In this report, chondroitinase ACII digestion was carried out at pH 7.5. In contrast, the reaction with chondroitinase ACII was originally performed in acetate buffer, pH 6.0 (35). It may be possible that specificity of chondroitinase ACII might be altered by the pH of the reaction mixture, but this possibility is not the case because the same results were obtained when the reaction of chondroitinase ACII was carried out at pH 6.0 (data not shown). It might be possible that Oligos I and II were formed from the respective unsaturated tetrasaccharides by the reaction with a hypothetical unsaturated uronate-specific glycuronidase that might be included in chondroitinase ACII. However, this possibility is unlikely, because neither ⌬Di-4S (⌬HexA-GalNAc(4SO 4 )) nor ⌬Di-diS E (⌬HexA-GalNAc(4,6-SO 4 )) was degraded by chondroitinase ACII even under the strong conditions (data not shown).
The molecular weight of CS-A from whale cartilage is within 25,000 -50,000 according to the manufacturer's data. It is thus assumed that CS-A from whale cartilage contains 50 -100 repeating disaccharide units. The recovery of Oligo I was about 0.1% on the basis of the content of galactosamine. We found previously that the content of the nonreducing terminal GalNAc(4SO 4 ) of the same CS-A preparation was about 0.8% of the total repeating units (12). From these data, the contents of the nonreducing terminal structures from which Gal-NAc(4SO 4 ) and Oligo I were generated after chondroitinase ACII digestion could be roughly estimated to be 40 -80 and 2.5-5%, respectively, of the total nonreducing terminal. Be-cause the loss of Oligo I during the purification was not included in this consideration, the content of the nonreducing terminal structure from which Oligo I was generated may be higher than the calculated value. On the other hand, the content of the nonreducing terminal structure from which Oligo II was generated could not be determined from the recovery of Oligo II, because Oligo II was rather sensitive to chondroitinase ACII.
When trisaccharides were used as the acceptor for Gal-NAc4S-6ST, the sulfation pattern of the trisaccharides affected the rates of the sulfation of GalNAc(4SO 4 ) residues at the nonreducing terminal. The rate of sulfation of Tri-46 was much lower than that of Tri-44, indicating that the presence of the GalNAc(6SO 4 ) residue at the reducing side inhibits the sulfation of position 6 of the GalNAc(4SO 4 ) residue at the nonreducing end. On the other hand, the rate of sulfation of Oligo I was much higher than that of Tri-46. The K m for Tri-46 was 60-fold of the K m for Oligo I, indicating that 2-O-sulfate on the GlcA residue markedly augmented the affinity for the acceptor substrate. Thus, the 2-O-sulfation of the penultimate GlcA residue appears to forward the production of the highly sulfated terminal sequence.
When CS from various origins except for the sturgeon notochord were used as the acceptors for GalNAc4S-6ST, and the 35 S-labeled products formed were digested with chondroitinase ACII, the 35 S radioactivity was detected at the position of the sulfated Oligo I. These results suggest that the nonreducing terminal sequence from which Oligo I was derived is present in CS from avian and mammalian tissues. Among the CS used here, CS from chick embryo cartilage had the highest content of ⌬Di-6S, and CS from sturgeon notochord had the lowest one (Table I). On the other hand, the proportion of the radioactive peak detected at the position of the sulfated Oligo I was also highest when chick embryo CS was used as the acceptor. Taken together, the synthesis of the nonreducing terminal structure from which Oligo I was released by chondroitinase ACII digestion may depend on the synthesis of the GalNAc(6SO 4 ) residue adjacent to the reducing side of the penultimate GlcA. Uronosyl 2-O-sulfotransferase has been reported to transfer sulfate to position 2 of GlcA residue adjacent to GalNAc(6SO 4 ) residue (10). Such specificity of uronosyl 2-O-sulfotransferase seems to agree with the hypothetical requirement for the GalNAc(6SO 4 ) residue.