Molecular Cloning and Expression of a Novel Human β-Gal-3-O-sulfotransferase That Acts Preferentially onN-Acetyllactosamine in N- andO-Glycans*

A novel cDNA-encoding galactose 3-O-sulfotransferase was cloned by screening the expressed sequence tag data base using the previously cloned cDNA encoding a galactosyl ceramide 3-O-sulfotransferase, which we term Gal3ST-1. The newly isolated cDNA encodes a novel 3-O-sulfotransferase, termed Gal3ST-3, that acts exclusively on N-acetyllactosamine present inN-glycans and core2-branched O-glycans. These conclusions were confirmed by analyzing CD43 chimeric proteins in Chinese hamster ovary cells expressing core2 β1,6-N-acetylglucosaminyltransferase. The acceptor specificity of Gal3ST-3 contrasts with that of the recently cloned galactose 3-O-sulfotransferase (Honke, K., Tsuda, M., Koyota, S., Wada, Y., Iida-Tanaka, N., Ishizuka, I., Nakayama, J., and Taniguchi, N. (2001) J. Biol. Chem. 276, 267–274), which we term Gal3ST-2 in the present study because the latter enzyme can also act on core1 O-glycan and type 1 oligosaccharides, Galβ1→3GlcNAc. Moreover, Gal3ST-3 but not Gal3ST-2 can act on Galβ1→4(sulfo→6)GlcNAc, indicating that disulfated sulfo→3Galβ1→4(sulfo→6) GlcNAc→R may be formed by Gal3ST-3 in combination with GlcNAc 6-O-sulfotransferase. Although both Gal3ST-2 and Gal3ST-3 do not act on galactosyl ceramide, Gal3ST-3 is only moderately more homologous to Gal3ST-2 (40.1%) than to Gal3ST-1 (38.0%) at the amino acid level. Northern blot analysis demonstrated that transcripts for Gal3ST-3 are predominantly expressed in the brain, kidney, and thyroid where the presence of 3′-sulfation ofN-acetyllactosamine has been reported. These results indicate that the newly cloned Gal3ST-3 plays a critical role in 3′-sulfation of N-acetyllactosamine in both O- and N-glycans.

type 1 or type 2, fucosylated oligosaccharide, functions as an E-selectin ligand (4). When oligosaccharides were released from ovarian cystadenoma glycoprotein and conjugated to lipids, sulfo33Gal␤133/4(Fuc␣134/3)GlcNAc␤133Gal was found to bind to Chinese hamster ovary (CHO) 1 cells expressing E-selectin (4). On the other hand, sulfo33Gal-␤134(Fuc␣133)GlcNAc acted as a P-selectin ligand when a synthetic oligosaccharide with this structure was transferred to cell surface glycoproteins through a fucose residue by ␣1,3fucosyltransferase III (5). These studies suggest that 3Ј-sulfo galactose on the cell surface plays a role in carbohydrate-protein interactions, including those involved with selectin.
Recently, a comparison of the amino acid sequences of cloned sulfotransferases demonstrated that there is a weak but discernible homologous sequence motif among Golgi-associated sulfotransferases (6 -9). In particular, the amino acid sequences that are responsible for binding 5Ј-phosphosulfate and 3Ј-phosphate groups of the donor substrate, 3Ј-phosphoadenosine 5Ј-phosphosulfate (PAPS), are well conserved and are often highly homologous to each other among those that share the same acceptor specificity (10 -15). Previously, galactosyl ceramide 3Ј-sulfotransferase, which forms a sulfatide, sulfo33Gal3ceramide, has been cloned based on the amino acid sequence of purified protein (16). Because this enzyme, which we now term Gal3ST-1, is thought not to add a sulfate to glycoproteins (17), galactose 3-O-sulfotransferase was molecularly cloned by searching for an enzyme homologous to Gal3ST-1. This reported enzyme, which we now term Gal3ST-2, has the unique property of adding a sulfate on both type 1 and type 2 oligosaccharides and core1 O-glycans, Gal␤133GalNAc␣13 R (18). On the other hand, the structures of O-linked oligosaccharides containing 3Ј-sulfo galactose reported to date show that 3Ј-sulfo galactose is present in N-acetyllactosamine in core2 O-glycans, sulfo33Gal␤134GlcNAc␤136(Gal␤133)GalNAc-␣13 R (19); core3 O-glycans, sulfo33Gal␤134GlcNAc␤13 3GalNAc (20); and core1 extended structures, sulfo3 3Gal␤134GlcNAc␤133Ga␤133GalNAc (21). Moreover, no 3Јsulfo galactose in core1 O-glycans such as sulfo33Gal␤13 3GalNAc␣13 R has been previously reported. Similarly, galactose 3-O-sulfotransferase in human respiratory mucosa was found to act exclusively on N-acetyllactosamine in core2branched O-glycans and not on core1 O-glycans (17). The presence of 3Ј-sulfo galactose in N-glycans has been extensively studied in human, bovine, and porcine thyroglobulins, and these * This work was supported by Grants R01 CA48737 and P01 CA71932 awarded by the NCI, National Institutes of Health. 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  studies showed that 3Ј-sulfo galactose is present in N-acetyllactosamine in complex N-glycans (22)(23)(24). Notably, galactose 3-Osulfotransferase present in the thyroid was found to act only on N-acetyllactosamine but not on type 1 oligosaccharides, which differs from the properties of Gal3ST-2 (25). Although no glycoprotein acceptor was tested for Gal3ST-2, these results suggested that there is another galactose 3-O-sulfotransferase (Gal3ST) yet to be identified.
In the present study, we first identified a novel cDNA by screening the EST data base for cDNAs related to human Gal3ST-1 (16). The expression of a full-length cDNA revealed that this cDNA encodes a novel galactose 3-O-sulfotransferase, termed Gal3ST-3, that adds a sulfate exclusively in the 3Јposition of galactose in N-acetyllactosamine in both N-and O-glycans but not on a type 1 Gal␤133GlcNAc or core1 Gal␤133GalNAc structure. When CD43 (leukosialin) was tested as an acceptor, Gal3ST-3 preferentially acted on Nacetyllactosamine present in core2-branched O-glycans, whereas Gal3ST-2 acted on both core1-and core2-branched O-glycans. Moreover, we show that this novel enzyme is expressed almost exclusively in the thyroid, kidney, and brain, in contrast to the previously reported ubiquitous expression of Gal3ST-2 (18).

Isolation of cDNAs Encoding Galactose 3-O-Sulfotransferase-Galac-
tosyl ceramide 3Ј-sulfotransferase (Gal3ST-1) shares homologous sequences to the binding sites for 3Ј-phosphate and 5Ј-phosphosulfate with other Golgi-associated sulfotransferases. The amino acid sequences of residues 76 -100 and 156 -179, which include the above motifs, were used as probes to search the dbEST data base using the TBLASTN program. Initially two ESTs, AI860920 and AA010976, were identified that exhibited a significant homology to the 3Ј-phosphate binding site and 5Ј-phosphosulfate binding site of Gal3ST-1. After sequence analysis of both cDNAs obtained from Incyte Genomics (St. Louis, MO), a BLAST search of the working draft of the human genome with both cDNAs revealed that the ESTs encode a continuous portion of chromosome 11 linked at a NotI site. The cDNAs from both ESTs were excised with EcoRI and NotI and then cloned into the EcoRI site of pcDNA3/Neo (Invitrogen). After confirmation of the correct orientation by sequencing, pcDNA3-Gal3ST-3 was established. The sequence of the Gal3ST-3 transcript was confirmed by sequencing the 5Ј-RACE and 3Ј-RACE products obtained using human brain Marathon Ready cDNA (CLONTECH) as a template. Human Gal3ST-1 cDNA was cloned by polymerase chain reaction using transcribed cDNAs from kidney (CLONTECH) as a template. The 5Ј-and 3Ј-primers for the PCR were 5Ј-CCACGCCTGGTGTCTGA-3Ј (corresponding to nucleotides Ϫ18 to Ϫ2, where nucleotides 1-3 encode the initiation methionine) and 5Ј-CCTTTACTTCTGAGGTCT-3Ј (corresponding to nucleotides 1751-1769) (16). PCR was carried out using Taq polymerase and the resultant PCR products were cloned into a pCR2.1-TOPO TA cloning vector (Invitrogen). Correctly amplified cDNAs were confirmed by sequencing, excised from the EcoRI sites of the pCR2.1-TOPO vector, and ligated into the EcoRI site of pcDNA3/Neo, resulting in pcDNA3-Gal3ST-1. Gal3ST-2 cDNA was cloned in a similar manner using primers 5Ј-CCAGAGGCCAAGATGATGTC-3Ј (corresponding to nucleotides Ϫ12 to 3) and 5Ј-CGGAGAGAGGAGCTGGTGT-3Ј (corresponding to nucleotides 1361-1380) (18) and reverse-transcribed cDNAs from human colon (CLONTECH) as a template.
Sulfotransferase activities of Gal3ST-3 were assayed as described previously with modification (16). Briefly, the reaction mixture (20 l) contained 50 mM MES (2-(N-morpholino)ethanesulfonic acid) buffer (pH 7.0), 10 mM MgCl 2 , 0.1% Triton X-100, 10 mM NaF, 2 mM ATP, 500 M acceptor, 0.08 nmol of [ 35 S]PAPS, and 10 l of enzyme solution. After incubation at 37°C for 2 h, the reaction was terminated by boiling for 2 min. The reaction products were then adjusted to 0.25 M ammonium formate, pH 4.0, and applied to C18 reverse phase columns (Alltech) as described previously (6). After washing the column with the same solution, the product was eluted with 30% acetonitrile. Radioactivity was measured by scintillation counting. Gal3ST-2 activity was assayed according to the previously described reaction mixture (16).
Northern Blot Analysis-Northern blots of multiple human tissues or human multiple tissue expression array (both from CLONTECH) were hybridized with cDNA fragments isolated from pcDNA3-Gal3ST-3 after 32 P labeling using a nick translation kit (Prime-It RmT, Stratagene). Transient Transfection and Metabolic Cell Labeling-cDNA encoding a soluble form of CD43 or NCAM was ligated to human IgG hinge plus constant region as described previously (36,37). These cDNAs were cloned into pcDM8 and pIG vector, respectively, resulting in pcDM8-CD43⅐IgG and pIG-NCAM⅐IgG. CHO cells and CHO mutants Lec1 (38), Lec2 (39), and Lec8 (40) cells were transiently transfected with pcDNA3.1-Gal3ST-1, pcDNA3.1-Gal3ST-2, or pcDNA3.1-Gal3ST-3 together with pcDM8-CD43⅐IgG or pIG-NCAM⅐IgG as described previously (11). In some of the experiments, pcDNAI-C2GnT-1 (41) was transfected into cells, together with the above cDNA vectors. 24 h after transfection, the medium was replaced with sulfate-free minimal essential medium (Life Technologies, Inc.) containing 10% dialyzed fetal bovine serum plus 0.1 mM non-essential amino acids solutions (Life Technologies, Inc.), supplemented with sodium [ 35 S]sulfate (100 Ci/ml, PerkinElmer Life Sciences). After an additional 48 h of culture, CD43⅐IgG or NCAM⅐IgG in the culture medium was purified by protein A-Sepharose as described previously (37).
Structural Analysis of Oligosaccharides Attached to CD43⅐IgG-To elucidate the structures of mucin-type O-glycans attached to CD43⅐IgG, CD43⅐IgG was isolated from CHO cells, as described above, after cells were metabolically labeled with [ 3 H]glucosamine (20 Ci/ml) or [ 3 H]galactose (20 Ci/ml), together with sodium [ 35 S]sulfate (100 Ci/ ml) in the sulfate-free medium, similarly described above.
Purified CD43⅐IgG was first digested with Pronase and subjected to Sephadex G-50 gel filtration. Glycopeptides that eluted close to the void volume contained mucin-type O-glycans and were subjected to alkaline borohydride treatment (42). The released O-glycans were recovered after Sephadex G-50 gel filtration.
Analysis of Sulfated Products-To determine the product sulfated by Gal3ST-3, Gal␤134GlcNAc␤13p-nitrophenol was incubated with 35 S-PAPS and a soluble form of Gal3ST-3 as prepared above, and the product was purified by Sep-Pak cartridge column chromatography. The purified product was partially hydrolyzed in 40 ml of HCl at 100°C for 2 h (15). The acid hydrolysate was then purified by Bio-Gel P-4 gel filtration and analyzed by HPLC using a Whatman Partisil SAX-10 column (4.6 ϫ 250 mm, Whatman, Clifton, NJ) equilibrated with 10 mM KH 2 PO 4 at room temperature. The column was eluted with an isocratic elution of 10 mM KH 2 PO 4 in the first 40 min, then linearly increased to 40 mM KH 2 PO 4 in the next 20 min. After elution for an additional 20 min at 40 mM KH 2 PO 4 , the column was then re-equilibrated at 10 mM KH 2 PO 4 . The flow rate was 0.5 ml/min, and each fraction contained 0.25 ml. The position of the standard sulfo36Gal (Sigma) was determined by boiling aliquots of the eluate in 4% orcinol (w/v) and 80% H 2 SO 4 (16.1 M) for 1 min and measuring the A 520 level. Sulfo36GlcNAc and sulfo33GlcNAc (both from Sigma) were detected at A 214 and used as internal standards. Sulfo36Gal and sulfo36GlcNAc eluted at the same position, whereas sulfo33Gal eluted slightly earlier than sulfo36Gal under these conditions.

RESULTS
Isolation of cDNA Encoding Novel Gal3ST-By searching the EST data base for a novel cDNA related to Gal3ST-1, two novel cDNAs were found to have homology to Gal3ST-1. The cDNAs (AA010976 and AI860920) thus represent the 5Ј-portion and 3Ј-portion of the novel cDNA separated by a NotI site. The full-length cDNA obtained by ligation of the two cDNAs encodes an open reading frame of 1293 base pairs, predicting a protein of 431 amino acid residues (molecular mass 48,878 Da), which we subsequently termed Gal3ST-3 (Fig. 1). 2 Although two additional methionine residues are present at residues 14 and 15, the nucleotide sequence surrounding the first methionine is more consistent with Kozak's consensus sequence for translation initiation (43), suggesting that the first methionine is likely the initiation methionine. The cDNA encoding Gal3ST-3 was cloned into pcDNA3/Neo, resulting in pcDNA3-Gal3ST-3. Although we searched the new human genome data base for the Gal3ST-3 sequence, the 3Ј-region, including the 3Ј-untranslated sequence has not been deposited yet. We thus still have limited knowledge on the genomic structure of Gal3ST-3 other than the fact that it is located on chromosome 11.
The comparison of the amino acid sequences of Gal3ST-3 with those of Gal3ST-1 and Gal3ST-2 reveals the following points (Fig. 2). The sequences corresponding to the binding sites for the 5Ј-phosphosulfate and 3Ј-phosphate groups are highly homologous among these enzymes. Moreover, these binding sites are much closer to the transmembrane/anchoring domains in the Gal3ST gene family than in all of the other sulfotransferases cloned to date; the transmembrane/anchoring domain and 5Ј-phosphosulfate binding site are separated by only 21-38 residues, whereas in the majority of Golgi-associated sulfotransferases, the distance is more than 50 residues. On the other hand, the size of the entire amino acid sequence of the Gal3ST gene family is relatively large among Golgi-associated sulfotransferases, indicating that the size of the polypeptide from the 3Ј-phosphate binding site to the COOH terminus is larger than that of the other Golgi-associated sulfotransferases cloned to date. Overall, the amino acid sequence of Gal3ST-3 is slightly more homologous to that of Gal3ST-2 (40.1%) than that of Gal3ST-1 (38.0%). None of the other amino acid sequences in the data base showed significant homology to these three sulfotransferases.
Acceptor Specificity of Gal3ST-3-To determine the acceptor specificity of Gal3ST-3, a soluble form of Gal3ST-3 was incubated with 35 S-PAPS and various acceptor oligosaccharides. For comparison, a soluble form of Gal3ST-2 was prepared in the same manner. Gal3ST-3 was assayed at pH 7.0, because the enzymatic activity was found to be optimal between pH 7.0 and 8.0 in MES buffer (data not shown). The enzyme activity was 1.7-fold higher in 10 mM Mg 2ϩ than Mn 2ϩ , and thus 10 mM MgCl 2 was added to the assay solution (data not shown).

Incorporation of [ 35 S]Sulfate to NCAM and CD43 (Leukosialin) Chimeric Proteins by Gal3ST-3-
The substrate specificity of Gal3ST-3 was examined using NCAM⅐IgG and CD-43⅐IgG chimeric proteins in the absence and presence of core2 ␤1,6-Nacetylglucosaminyltransferase (C2GnT-1). NCAM contains almost exclusively N-glycans, whereas CD43 contains one Nglycan and ϳ70 O-glycans (37,42). As the first of a series of experiments, CHO cells were transfected with vectors encoding NCAM⅐IgG or CD43⅐IgG chimeric protein with or without expression of full-length cDNA encoding Gal3ST-3. Because CHO cells lack C2GnT activity (41,42), core2 O-glycans were absent when C2GnT was not transfected. Fig. 4A illustrates that the incorporation of [ 35 S]sulfate into CD43 chimeric protein was significantly increased by the ex-pression of Gal3ST-3. The incorporated [ 35 S]sulfate was mostly removed by N-glycanase treatment. [ 35 S]sulfate incorporation into NCAM was moderately increased by Gal3ST-3, and the majority of the incorporated radioactivity was removed by Nglycanase treatment. These results indicate that [ 35 S]sulfate was incorporated by Gal3ST-3 mainly into N-glycans in the absence of C2GnT. In contrast, there was only a slight increase in [ 35 S]sulfate incorporation into NCAM by Gal3ST-2. Moreover, [ 35 S]sulfate incorporated by Gal3ST-2 into CD43 was less susceptible to N-glycanase treatment (Fig. 4A).
To further evaluate the acceptor specificity of Gal3ST-3, CD43⅐IgG, Gal3ST-3, and C2GnT-1 were expressed in CHO cells. As shown in Fig. 4B, [ 35 S]sulfate incorporated into CD43 in the presence of C2GnT-1 was only slightly removed by Nglycanase, whereas the radioactivity incorporated in the absence of C2GnT-1 was significantly removed by the same treatment. These results indicate that [ 35 S]sulfate incorporation shifted from N-glycan sulfation to O-glycan sulfation with core2 branch formation.
To further examine the dependence of C2GnT on Gal3ST-3 and Gal3ST-2, Lec2 cells, which are defective in Golgi sialylation, were used as recipient cells. In Lec2 cells, it was expected that [ 35 S]sulfate incorporation would be enhanced, because there is no competition between 3Ј-sulfation and sialylation. The results shown in Fig. 5A illustrate that [ 35 S]sulfate incorporation into CD43 by Gal3ST-3 was significantly increased in the presence of C2GnT-1 and that the majority of the incorporated radioactivity was resistant to N-glycanase treatment. On the other hand, [ 35 S]sulfate incorporation by Gal3ST-2 was significantly increased even in the absence of C2GnT-1 and the incorporated label was only slightly removed by N-glycanase treatment. These results indicate that Gal3ST-2 acts on CD43 O-glycans in the absence of core2-branched O-glycans, whereas Gal3ST-3 preferentially acts on core2-branched O-glycans. These results are consistent with the above results that Gal3ST-2 but not Gal3ST-3 can act on a GlcNAc␤136(Gal- ␤133)GalNAc structure (Fig. 3).
To further corroborate these findings, CD43 chimeric protein together with Gal3ST-3 were expressed in Lec1 and Lec8 cells, respectively. The results shown in Fig. 5B indicate that 35 S incorporation by Gal3ST-3 is entirely dependent on the presence of core2-branched O-glycans when the recipient cells lack complex N-glycans as do Lec1 cells. Furthermore, sulfation is entirely dependent on galactose residues because no incorporation by Gal3ST-3 was observed in Lec8 cells, which lack Golgi galactosylation (Fig. 5C).
Structural Analysis of O-Glycans Sulfated by Gal3ST-3-To determine the structure of sulfated O-glycans synthesized by Gal3ST-3, CD43 chimeric protein was produced in CHO cells that express Gal3ST-3 and C2GnT-1. These transfected cells were metabolically labeled with [ 3 H]glucosamine or [ 3 H]galactose and [ 35 S]sulfate, and the CD43 chimeric protein released into the medium was collected. Sephadex G-50 gel filtration of glycopeptides obtained by Pronase digestion of Gal3ST-3-labeled CD43 chimeric protein showed that the majority of glycopeptides eluted close to the void volume, but a minority of glycopeptides eluted at fractions 43-60, which contain N-glycans (Fig. 6A). In contrast, Gal3ST-2-labeled CD43 produced only glycopeptides that eluted at the void volume, which contain multiple O-glycans in a peptide (Fig. 6B). These glycopeptides containing multiple O-glycans in each peptide, which were derived from Gal3ST-3-labeled cells, were then subjected to alkaline borohydride treatment to release O-glycans and then subjected to the same Sephadex G-50 gel filtration (Fig.  6C). After QAE-Sephadex column chromatography, the majority of the isolated O-glycans eluted as those containing two anionic charges (Fig. 6D), which were converted to a monosulfated form after desialylation (Fig. 6E). Upon Bio-Gel P-4 gel filtration, this oligosaccharide eluted at the elution position for core2-branched glycans, Gal␤134GlcNAc␤136(Gal␤133)-GalNAcOH (Fig. 6F). Previous studies demonstrated that a monosulfated form of core2-branched O-glycans elutes at almost the same position as non-sulfated O-glycans (11). Almost identical results were obtained for O-glycans derived from Gal3ST-2-labeled CD43 in the presence of C2GnT-1 (data not shown).
After desulfation by solvolysis, the obtained O-glycans eluted at the same position where Gal␤134GlcNAc␤136(Gal␤133)-GalNAcOH eluted upon HPLC analysis (Fig. 7A). Sulfate was judged to be attached to galactose in an N-acetyllactosaminyl core2 branch, because the oligosaccharide was resistant to ␤-galactosidase treatment (data not shown). If the sulfate had been attached to a core1 structure such as Gal␤134GlcNAc-␤136(sulfo33Gal␤133)GalNAcOH, one galactose would have been released after ␤-galactosidase treatment. To confirm that a sulfate group was attached to the C-3 position of galactose, sulfated monosaccharides were released by acid hydrolysis and subjected to SAX-10 column chromatography. Monosulfated galactose eluted at the same position as sulfo33galactose re-leased from sulfo33Gal␤134GlcNAc␤13p-nitrophenol prepared by Gal3ST-2 (Fig. 7, B and C), demonstrating that Gal3ST-3 adds a sulfate to the C-3 position of galactose in N-acetyllactosamine. These results indicate that the majority of sulfated O-glycans conferred by Gal3ST-3 on CD43 is a sialylated and sulfated core2 O-glycan, sulfo33Gal␤134GlcNAc-␤136(NeuNAc␣233Gal␤133)GalNAc.
Expression of Gal3ST-3 Is Highly Restricted to Brain, Kid- ney, and Thyroid-Northern blot analysis showed a highly specific Gal3ST-3 transcript of ϳ2.4 kilobases expressed in fetal brain and kidney, and adult brain, kidney, and thyroid (Fig. 8). Multiple tissue expression array analysis showed almost identical results and further demonstrated that expression is highly pronounced in the putamen, caudate nucleus, and pituitary gland of the brain (Fig. 9). These results contrast with the reported expression pattern of Gal3ST-2 showing ubiquitous expression in almost all tissues tested (18). DISCUSSION The present study describes the isolation of a novel cDNA encoding galactose 3-O-sulfotransferase by searching the EST data base for cDNAs homologous to the human galactosyl ceramide 3-O-sulfotransferase, Gal3ST-1 (16). Gal3ST-1 adds a sulfate to a ␤-galactose residue linked to ceramide, whereas Gal3ST-3 adds a sulfate to a ␤-galactose linked to N-acetylglucosamine through a 1,4-linkage. Previously, Gal3ST-2 was also cloned based on its similarity to Gal3ST-1, but the acceptor specificities of Gal3ST-2 and Gal3ST-3 differ substantially. Although Gal3ST-3 acts exclusively on N-acetyllactosamine, Gal3ST-2 can act also on type 1 oligosaccharide, Gal␤13 3GlcNAc, and core1 oligosaccharide, Gal␤133(GlcNAc␤13 6)GalNAc. On the other hand, Gal3ST-3 can act on Gal␤13 4(sulfo36)GlcNAc␤13 R, but Gal3ST-2 cannot act on the same acceptor (Fig. 3). Interestingly, the best acceptor for Gal3ST-3 is poly-N-acetyllactosamine attached to a side chain derived from C-6 of ␣-mannose, whereas the best acceptor for Gal3ST-2 is N-acetyllactosamine attached to a side chain derived from C-2 of ␣-mannose (Fig. 3). Moreover, Gal3ST-3 preferentially acts on N-acetyllactosamine attached to an N-glycan mannose core, such as Gal␤134GlcNAc␤136Man␣136Man␤13octyl, than on N-acetyllactosamine itself, whereas Gal3ST-2 does not possess this preference (Fig. 3). These results indicate that Gal3ST-2 and Gal3ST-3 act differentially on various acceptor glycoproteins.
In the present study, the transcripts of Gal3ST-3 were found to be expressed selectively in the brain, kidney, and thyroid. Previously, it was reported that human thyroglobulin contains a sulfo33Gal␤134GlcNAc␤13 R structure in the majority of N-glycans (23). This 3Ј-O-sulfated N-acetyllactosamine side chain was shown to exist in both bi-antennary and highly branched tri-and tetra-antennary N-glycans. In porcine thyroglobulin, the majority of sulfo33Gal␤134GlcNAc␤13 R are present in side chains derived from the 6-position of ␣-mannose. Moreover, the same glycoprotein contains a Gal␤134-(sulfo36)GlcNAc side chain in a portion of the N-glycans (24). As shown previously, sulfo36GlcNAc3 R is first formed from a GlcNAc3 R structure and then converted to Gal␤134-(sulfo36)GlcNAc3 R (11,12). Gal␤134(sulfo36)GlcNAc3 R can then be converted by Gal3ST-3 to sulfo33Gal␤134-(sulfo36)GlcNAc3 R, considering that Gal3ST-3 can act on Gal␤134(sulfo36)GlcNAc3 R. In the structural studies described above, no multisulfated N-glycans were analyzed (24). It is possible that sulfo33Gal␤134(sulfo36)GlcNAc3 R may be found with further analysis of highly sulfated N-glycans in thyroglobulin.
It has been reported that a major glycoprotein in calf thyroid contains core2-branched O-glycans (44) and that core2 branches in the thyroid are likely synthesized by C2GnT-1 and C2GnT-2 (45,46). Core2-branched oligosaccharides from calf thyroid apparently lack 3Ј-sulfate galactose. In contrast to human thyroglobulin, the presence of 3Ј-sulfate galactose in calf thyroglobulin is minimal, presumably due to the competition with the strong ␣1,3-galactosyltransferase activity in calf thyroid (44). It has also been reported that galactose 3-O-sulfotransferase in the thyroid acts only on N-acetyllactosamine as does Gal3ST-3 (25). These results indicate that Gal3ST-3 is most likely responsible for the formation of sulfo3 3Gal␤134GlcNAc attached to both N-glycans and core2branched O-glycans synthesized in the thyroid.
Although it was reported that the transcripts of Gal3ST-2 are expressed in various tissues, the amount of transcripts was relatively low in brain and kidney when estimated by reverse transcription PCR (18). Because the transcripts of Gal3ST-3, on the other hand, are highly expressed in brain and kidney, it is likely that GalST-3 plays a major role in the brain and kidney in addition to the thyroid. Previously, the existence of two different galactosyl 3-O-sulfotransferases has been reported: one (A) acts on core1 O-glycans, whereas the other (B) acts  on N-acetyllactosamine (47,48). Although it is not straightforward to correlate these findings with the acceptor specificities of the cloned enzymes, it appears that A and B correspond to Gal3ST-2 and Gal3ST-3, respectively. On the other hand, the acceptor specificity of Gal3ST-3 appears to be identical to the enzyme described in human airways (17), whereas Gal3ST-2, which acts on type 1 oligosaccharides, is most likely responsible for the formation of sulfo33Gal␤133(Fuc␣134)GlcNAc␤13 R (4,49).
In this context, it is noteworthy that Gal3ST-3 has a relatively narrow acceptor specificity compared with Gal3ST-2. Due to this distinct acceptor specificity, it is likely that the expression of Gal3ST-3 results in the formation of sulfo33Gal␤134GlcNAc3 R and related structures in a welldefined set of carbohydrates attached to glycoproteins. It is thus expected that the cDNA encoding Gal3ST-3, cloned in the present study, will be a powerful tool to determine the structure/function of sulfo33Gal␤134GlcNAc3 R and related structures.