Molecular cloning and characterization of a novel human galactose 3-O-sulfotransferase that transfers sulfate to gal beta 1-->3galNAc residue in O-glycans.

We have identified a novel galactose 3-O-sulfotransferase, termed Gal3ST-4, by analysis of an expression sequence tag using the amino acid sequence of human cerebroside 3'-sulfotransferase (Gal3ST-1). The isolated cDNA contains a single open reading frame coding for a protein of 486 amino acids with a type II transmembrane topology. The amino acid sequence of Gal3ST-4 revealed 33%, 39%, and 30% identity to human Gal3ST-1, Gal beta 1-->3/4GlcNAc:-->3'-sulfotransferase (Gal3ST-2) and Gal beta 1-->4GlcNAc:-->3'-sulfotransferase (Gal3ST-3), respectively. The Gal3ST-4 gene comprised at least four exons and was located on human chromosome 7q22. Expression of Gal3ST-4 in COS-7 cells produced a sulfotransferase activity that catalyzes the transfer of [(35)S]sulfate to the C-3' position of Gal beta 1-->3GalNAc alpha 1-O-Bn. Gal3ST-4 recognizes Gal beta 1-->3GalNAc and Gal beta 1-->3(GlcNAc beta 1-->6)GalNAc as good substrates, but not Gal beta 1-->3GalNAc(OH) or Gal beta 1-->3/4GlcNAc. Asialofetuin is also a good substrate, and the sulfation was found exclusively in O-linked glycans that consist of the Gal beta 1-->3GalNAc moiety, suggesting that the enzyme is specific for O-linked glycans. Northern blot analysis revealed that 2.5-kilobase mRNA for the enzyme is expressed extensively in various tissues. These results suggest that Gal3ST-4 is the fourth member of a Gal:-->3-sulfotransferase family and that the four members, Gal3ST-1, Gal3ST-2, Gal3ST-3, and Gal3ST-4, are responsible for sulfation of different acceptor substrates.

Sulfation is one of the most extensive modifications for glycan chains in various glycoconjugates. Sulfated glycans are associated with the physiological functions of glycoproteins in the mucosal barrier system, regulation of their half-life, and cell-to-cell interaction. The content of sulfated glycans in mucus glycoproteins is modified in cystic fibrosis and colon cancer (for review, see Refs. [1][2][3]. However, the precise molecular mechanisms in these phenomena and other significances of sulfation in various glycoproteins remain unclear. The SO 3 Ϫ 33Gal structure in O-linked and N-linked glycans has been found in various glycoproteins including thyroglobulin (4 -6), meconium glycoproteins (7), respiratory mucous glycoproteins from patients with cystic fibrosis (8 -10) and chronic bronchitis (11), an ovarian cystadenoma glycoprotein (12), LS174T-HM7 colon carcinoma mucin (13), Tamm-Horsfall glycoprotein (14), sulfomucins (15), and oviducal mucins (16). Sulfated residues in these glycoproteins are attached to C-3Ј of Gal␤133/4GlcNAc, Gal␤133GalNAc, or Gal␤133Gal structure. The occurrence of the SO 3 Ϫ 33Gal structure has not been fully elucidated to date because of the difficulty of structural studies for sulfated glycans. However, the sulfated structure seems to be distributed extensively because the glycoproteinspecific ␤-Gal-3Ј-sulfotransferase (Gal3ST-2) gene is expressed ubiquitously in various human tissues (17).
cDNA Cloning of Gal3ST-4 -Based on the amino acid sequence of human Gal3ST-1 (19), we found one sequence (GenBank AW961058) with high similarity in the EST data bases at the National Center for Biotechnology Information (National Institutes of Health, Bethesda, MD). We used GeneTrapper ® cDNA Positive Selection System (Life Technologies, Inc.) to obtain the cDNA clone according to the manufacturer's instructions. Briefly, an oligonucleotide, 5Ј-GCCCTAGCGAAA-CATTGTCTGGTA-3Ј (the nucleotide sequence corresponding to 1440 -1463; see Fig. 1A) was biotinylated and then used as a probe. SuperScript TM human testis cDNA library (Life Technologies, Inc.) in which testis cDNA was cloned into the eukaryotic expression vector pCMV-SPORT, was degraded to single-stranded cDNA by Gene II and exonuclease III digestion. The single-stranded cDNA was hybridized with the biotinylated probe, and the target cDNA was captured to streptavidin-conjugated paramagnetic beads. The target cDNA was released, re-double-stranded, and then transformed into DH5␣ cells.
Expression of Gal3ST-4 in COS-7 Cells-The plasmid (1 g) was transfected into COS-7 cells on 35-mm dishes using Lipofectin Reagent (Life Technologies, Inc.) according to the manufacturer's instructions. After 48 h, the cells were washed once with phosphate-buffered saline, scraped off from the dishes in 10 mM HEPES-NaOH (pH 7.2), 0.25 M sucrose, and homogenized. The homogenate was ultracentrifuged at 100,000 ϫ g for 1 h. The precipitated crude membranes were suspended in 20 mM HEPES-NaOH (pH 7.2) and kept at Ϫ80°C until use.
Assay of Sulfotransferase Activity-Twenty l of reaction mixture consisting of 0.1 M sodium cacodylate (pH 6.3), 10 mM MnCl 2 , 0.1% (v/v) Triton X-100, 0.1 M NaF, 2 mM ATP-Na 2 , 6.5 M [ 35 S]PAPS (2.8 ϫ 10 5 dpm), 1 mM Gal␤133GalNAc␣1-O-pNP, and the crude membrane fraction appropriately diluted, was incubated at 37°C for 1 h. The 35 Slabeled products were purified by paper electrophoresis (pyridine/acetic acid/water, 3:1:387, pH 5.4). The R F values of 35 S-sulfated Gal␤133GalNAc␣1-O-pNP and PAPS are 0.69 and 1.89, respectively, when the R F value of bromphenol blue is taken as 1.0. After extraction with water, the radioactivities were counted. In the case of glycolipids used as acceptor substrates, the detection of the 35 S-labeled products was performed according to the methods reported by Kawano et al. (27).
Characterization of the 35 S-Labeled Product-The 35 S-labeled product was subjected to periodate oxidation (28). The labeled oligosaccharides were dissolved in 20 l of 75 mM sodium metaperiodate, 75 mM sodium acetate (pH 5.3) and incubated at 4°C for 24 h in the dark. Excess periodate was destroyed by adding 2 l of 20% ethylene glycol. After 1 h at room temperature, 300 l of 0.1 M sodium borate (pH 9.0) containing 0.1 M sodium borohydride was added, and the solutions stood for 1 h at room temperature. The solutions were acidified by adding acetic acid and passed through a column (0.5 ϫ 3 cm) of Bio-Rad AG-50W-X8 (H ϩ form). The eluates were evaporated, and residual boric acid was removed by repeated evaporation with methanol. The residues were hydrolyzed in 100 l of 0.05 N H 2 SO 4 at 80°C for 1 h. After being neutralized with NaOH, the mixtures underwent paper electrophoresis. The 35 S-labeled compounds were extracted with water, applied on a thin layer plate (Kieselgel 60F 254 , Merck, Darmstadt, Germany), and developed with solvents, pyridine/ethyl acetate/acetic acid/water (5:5: 1:3) or 1-butanol/ethanol/water (4:1:1). The radioactivities were monitored by a radiochromatogram scanner.
[ 3 H]Gal␤133GalNAc␣1-O-Bn as a positive control for periodate oxidation was prepared as follows. Twenty l of solution containing 50 mM HEPES-NaOH (pH 7.2), 10 mM MnCl 2 , 0.5% (v/v) Triton X-100, 5 mM GalNAc␣1-O-Bn, 2.5 M UDP-[ 3 H]Gal (3.3 ϫ 10 6 dpm), 250 M UDP-Gal, and crude membrane fractions from porcine colonic mucosa was incubated at 37°C for 1 h. The reaction mixture underwent paper electrophoresis, then the neutral fraction further underwent paper chromatography, which was developed with the solvent pyridine/ethyl acetate/acetic acid/water (5:5:1:3). A radioactive fraction ([ 3 H]Gal␤133GalNAc␣1-O-Bn) with an R F value of 0.72 was extracted with water. Linkage position of [ 3 H]Gal residue was confirmed by its binding to a peanut agglutinin-agarose column because peanut agglutinin specifically recognizes the Gal␤133GalNAc structure (29). Synthesis of 6-[ 35 S]sulfo-GlcNAc␤1-O-Bn was performed using human Gl-cNAc:36-sulfotransferase as described previously (30). 35 S Sulfation of Asialofetuin by Gal3ST-4 -Forty g of asialofetuin (Sigma) was dissolved in the enzyme reaction solution described above without Gal␤133GalNAc␣1 -O-pNP and incubated at 37°C for 16 h. Half of the reactant was subjected to mild alkaline treatment in 50 l of 1 M NaBH 4 , 0.05 N NaOH at 37°C for 24 h. After acidification by acetic acid, the solution was applied on an AG-50W-X8 column, and the eluate was evaporated as described above. The other half of the reactant was subjected to N-glycanase digestion by the Glycopeptidase F De-Nglycosylation set (Takara Shuzo Co., Kyoto, Japan). Liberated 35

Molecular Cloning of a cDNA Homologous to Human
Gal3ST-1-We found small sequences (GenBank AW961058) similar to the sequence of human Gal3ST-1 (19) in the EST data bases. We prepared a sense oligonucleotide, 24 nucleotides in length, the sequence of which is present in the EST, and used it as a probe for GeneTrapper ® cDNA positive selection system to screen a human testis cDNA library. One clone was obtained, and the nucleotide sequence was determined (Fig.  1A). The 2,460-bp cDNA had a 5Ј-untranslated region of 236 bp, a single open reading frame of 1,458 bp, and a 3Ј-untranslated region of 766 bp including a poly(A) ϩ tail. The translation initiation site conformed to Kozak's rule (31), and the upstream region contained an in-frame stop codon. The open reading frame predicts a protein of 486 amino acid residues with a molecular mass of 54,173 Da with one potential N-linked glycosylation site. Hydropathy plot analysis of the deduced amino acid sequence revealed one prominent hydrophobic segment, 22 amino acid residues in length in the N-terminal region, predicting that the protein has a type II transmembrane topology (Fig. 1B). The cDNA sequence was compared with the Human  Genome Project Data Base, and the genomic organization and the chromosomal localization were revealed (Fig. 1C). The gene comprises at least four exons and spans about 10 kilobases in human chromosome 7q22. The intron/exon junctions followed the GT/AG rule (33). The coding region is located in three exons (exons 2, 3, and 4), and two introns were inserted between nucleotides 361 and 362 (at Arg-42) and nucleotide 665 and 666 (between Glu-143 and Val-144) of the cDNA (Fig. 1A).
Characterization of the Putative Sulfotransferase as Gal␤133GalNAc:33Ј-Sulfotransferase-The putative sulfotransferase was expressed in COS-7 cells, and the crude membrane fraction was prepared as an enzyme source. The membrane fraction from the cells transfected with pCMV-SPORT-Gal3ST-4, pCMV-SPORT vector containing the cDNA for Gal3ST-4, had a sulfotransferase activity (5.6 pmol/ min/mg of protein) using Gal␤133GalNAc␣1-O-Bn as acceptor. The membrane fractions derived from the transfectant with pCMV-SPORT and wild type had no sulfotransferase activity. The 35

S-sulfated product, [ 35 S]SO 3
Ϫ 3 (Gal␤133GalNAc␣1-O-Bn), purified by paper electrophoresis, was resistant to 6646K ␤-galactosidase digestion and passed through a R. communis agglutinin I-agarose column (data not shown). Because authentic Gal␤133GalNAc␣1-O-Bn was retarded on the R. communis agglutinin I-agarose column, these results suggested that the [ 35 S]sulfate was transferred to the galactose residue. The migrating position of the 35 S-sulfated product in paper electrophoresis was close to that of authentic [ 35 S]SO 3 Ϫ 36GlcNAc␤132Man, suggesting that the product is monosulfated disaccharide. To determine the linkage position of [ 35 S]sulfate residue, the 35 Ssulfated product was subjected to periodate oxidation. Ϫ 3 (Gal␤13 3Gal-NAc␣1-O-Bn) underwent paper electrophoresis (Fig. 3A). It migrated to the same position as the untreated 35 S-labeled product. If [ 35 S]sulfate is transferred to the C-2, C-4, or C-6 position of galactose residue, or N-acetylgalactosamine residue, CH 2 4. Effect of pH on Gal3ST-4 activity. Buffers used were 0.1 M sodium acetate (q), sodium cacodylate (E), and HEPES-NaOH (OE). Ion strength was adjusted to 0.1 with NaCl.

TABLE I Effect of various compounds on Gal3ST-4 activity
a Relative ratios are taken with the value of Gal␤1 3 3GalNAc␣1-O-pNP as 100.

S]SO 3
Ϫ -OCH 2 -CHOH-CH 2 OH, migrated faster in paper electrophoresis than (Fig. 3A). The result suggests that [ 35 S]sulfate does not bind to GalNAc residue or the C-2, 4, or 6 position of Gal, but the C-3 position of Gal. Moreover, the reaction product was subjected to TLC using different solvent systems (Fig. 3, B and C). The reaction product was developed to the same position as the untreated product in both solvent systems. The structure of the reaction product was further confirmed by TLC with authentic SO 3 Ϫ 33Gal␤133GalNAc␣1-O-Bn (Fig. 3D). The reaction product was developed to the same position as the authentic oligosaccharide and the oligosaccharide treated with periodate oxidation. These results suggest that [ 35 S]sulfate binds to the C-3 position of Gal residue and that the enzyme cloned here is a Gal:33-sulfotransferase.
The optimal pH of the enzyme was 6 -7, using Gal␤133GalNAc␣1-O-pNP as acceptor substrate (Fig. 4). In the presence of MnCl 2 , the activity increased about 2.2-fold (Table I). EDTA had no effect on the activity, suggesting that Gal3ST-4 does not essentially require divalent cations. N-ethylmaleimide and dithiothreitol had a weak inhibitory effect on the activity when these assays were performed in the presence of MnCl 2 .
Gal3ST-4 activity was inhibited by high concentrations of acceptor substrates (Fig. 5). The activities for core 2-O-pNP and Gal␤133GalNAc␣1-O-pNP were maximum at 1.5 and 1 mM and decreased significantly above 3 and 1.5 mM, respectively. The K m values for Gal␤133GalNAc␣1-O-pNP and core 2-O-pNP, which were estimated using only the data obtained with lower concentrations of the two substrates, were 0.24 and 0.23 mM, respectively (Table III).

Incorporation of [ 35 S]Sulfate into Asialofetuin by
Gal3ST-4 -We investigated whether Gal3ST-4 can specifically add sulfate to Gal␤133GalNAc in asialofetuin, which contains both bi-and tri-antennary N-glycans and O-glycans consisting of Gal␤133GalNAc (35,36). After incubating asialofetuin with Gal3ST-4, half of the 35 S-sulfated products were subjected to mild alkaline treatment, which specifically releases O-glycans, and the other half of the products were digested with N-glycanase. As shown in Table IV, 35 S-sulfated oligosaccharides were released by alkaline treatment, but not by N-glycanase digestion. These results also support that Gal3ST-4 specifically acts on Gal␤133GalNAc residue in O-glycans.
The Gal3ST-4 gene comprises at least four exons (Fig. 1C). The coding region is inserted with two introns at Arg-42 and between Glu-143 and Val-144. The putative transmembrane domain and two putative PAPS binding domains (5Ј-PSB and 3Ј-PB) (34) are localized in exons 2, 3, and 4, respectively. We investigated the intron/exon alignment of Gal3ST-2 gene using the Human Genome Project Data Base and found that the coding region for Gal3ST-2 is also inserted with two introns, at Pro-40 and between Gln-125 and Val-126. The insertion positions of the Gal3ST-2 gene are very close to those of Gal3ST-4 (Fig. 1C). In contrast, the coding region for Gal3ST-1 is inserted with one intron at Thr-44 (37); the position is close to Arg-42 of Gal3ST-4 and Pro-40 of Gal3ST-2. As for the 5Ј-untranslated region, it has been shown that there exist at least seven exons for the 5Ј-untranslated region of the human Gal3ST-1 gene and that these exons are alternatively utilized in a cancer-associated manner (37). Whether or not there exist alternative forms of mRNA for Gal3ST-4 remains unclear, but this is an important issue for elucidating transcriptional regulation of the Gal3ST-4 gene.
Gal3ST-4 is highly specific for the Gal␤133GalNAc␣13 structure. Gal␤133(GlcNAc␤136)GalNAc␣1-O-pNP is also a good substrate, and the K m and V max values for the core 2 oligosaccharide are similar to those for Gal␤133GalNAc␣1-O-pNP (Table III), indicating that the substitution of ␤-GlcNAc at the C-6 of GalNAc does not affect the substrate recognition of Gal3ST-4. Kuhns et al. (38) showed that Gal␤133GalNAc: ␤136 N-acetylglucosaminyltransferases in acute myeloid leukemia cells and rat colon can act on Gal␤133GalNAc␣1-O-Bn, but not on SO 3 Ϫ 33Gal␤133GalNAc␣1-O-Bn and suggested that the substitution of ␤-GlcNAc at the C-6 of Gal␤133GalNAc should precede sulfation at the C-3Ј. Our result that Gal3ST-4 can utilize core 2 oligosaccharide as a good substrate is consistent with their results with regard to biosynthesis of sulfated core 2 glycans.
The enzyme activity of Gal3ST-4 is inhibited by higher concentrations of acceptor substrates (Fig. 5). Similar inhibitory effects have been reported for ␤134-galactosyltransferase I, II, and III (39), ␤134-galactosyltransferase from human colonic mucosa (40), and GlcNAc:36-sulfotransferase (30). These transferase activities are inhibited at concentrations in excess of 2-5 mM, except for ␤134-galactosyltransferase II, the activity of which is inhibited even at ϳ0.6 mM GlcNAc␤1-O-Bn (39). The inhibitory effects of the two substrates for Gal3ST-4 ( Fig.  5) appear at similar concentrations as those for the transferases described above, although the molecular mechanism and biological significance remain unclear.
The result in Fig. 6 showed a relatively extensive expression of the mRNA for Gal3ST-4. It is important whether or not the SO 3 Ϫ 33Gal␤133GalNAc␣13 structure is present in the tissues examined in Fig. 6. Although Chance and Mawhinney (10) showed the occurrence of SO 3 Ϫ 33Gal␤133(R3 GlcNAc␤13 6)GalNAc␣13 structure in tracheobronchial mucous glycoproteins from a patient with cystic fibrosis, information about the existence of the sulfated glycan has so far been scarce. On the other hand, Gal:33-sulfotransferase activities for the Gal␤133GalNAc␣13 structure have been reported in rat colonic mucosa (38), human breast, colon, and several tumor tissues (41). To assess whether or not the SO 3 Ϫ 3 3Gal␤133GalNAc␣13 structure is widely distributed, a lectin or antibody that specifically recognizes the sulfated glycan needs to be explored.
Chandrasekaran et al. (41) reported the occurrence of two distinct Gal:3-O-sulfotransferases (groups A and B) in human various tumors and normal tissues with a tissue-dependent distribution. Group A sulfotransferase recognizes Gal␤13 3GalNAc␣-O-allyl and 3-O-MeGal␤134GlcNAc␤13 6(Gal␤13 3)GalNAc␣-O-Bn as good acceptors, but not Gal␤134GlcNAc␤-O-allyl and Gal␤133GlcNAc␤-O-allyl, whereas group B sulfotransferase has rather broad substrate specificity (41). The substrate specificity of group A sulfotransferase is similar to that of Gal3ST-4. They also showed that group A Gal:33sulfotransferase is dominant in breast tumor, some ovarian tumor, and some metastatic ovary, and that the specific activities in breast tumor are higher than those in breast normal tissues (41). In contrast, Brockhausen et al. (42) showed that a sulfotransferase activity for Gal␤133GalNAc␣1-O-Bn in a human mammary epithelial cell line, MTSV1-7, is higher than those in three human breast cancer cell lines. We are in the process of investigating changes in expression level of mRNA for Gal3ST-4 in these cancerous tissues.