Biosynthesis of the Linkage Region of Glycosaminoglycans

A family of five β1,3-galactosyltransferases has been characterized that catalyze the formation of Galβ1,3GlcNAcβ and Galβ1,3GalNAcβ linkages present in glycoproteins and glycolipids (β3GalT1, -2, -3, -4, and -5). We now report a new member of the family (β3GalT6), involved in glycosaminoglycan biosynthesis. The human and mouse genes were located on chromosomes 1p36.3 and 4E2, respectively, and homologs are found inDrosophila melanogaster and Caenorhabditis elegans. Unlike other members of the family, β3GalT6 showed a broad mRNA expression pattern by Northern blot analysis. Although a high degree of homology across several subdomains exists among other members of the β3-galactosyltransferase family, recombinant enzyme did not utilize glucosamine- or galactosamine-containing acceptors. Instead, the enzyme transferred galactose from UDP-galactose to acceptors containing a terminal β-linked galactose residue. This product, Galβ1,3Galβ is found in the linkage region of heparan sulfate and chondroitin sulfate (GlcAβ1,3Galβ1,3Galβ1,4Xylβ-O-Ser), indicating that β3GalT6 is the so-called galactosyltransferase II involved in glycosaminoglycan biosynthesis. Its identity was confirmed in vivo by siRNA-mediated inhibition of glycosaminoglycan synthesis in HeLa S3 cells. Its localization in the medial Golgi indicates that this is the major site for assembly of the linkage region.

ducing end. Xylosyltransferase (2), galactosyltransferase I (GalTI) 1 (3,4), and glucuronosyltransferase (GlcATI) (5,6) have been cloned and partially characterized. In addition, mutant cell lines have been identified that contain defects in each of these enzymatic steps (7)(8)(9), and C. elegans mutants in GalTI (sqv3) and the GlcATI (sqv8) have been described, demonstrating the importance of these reactions and glycosaminoglycans in cellular processes and organismal development (10,11). The enzyme that transfers the second galactose unit, galactosyltransferase II (GalTII) has not yet been identified. Previous studies have described the appropriate enzyme activity in tissue extracts (12), but the enzyme has not yet been purified or cloned. Presumably, GalTII should be distinct from GalTI based on linkage (␤1,3 versus ␤1,4) and substrate differences (galactose versus xylose).
GalTII catalyzes the formation of Gal␤1,3Gal linkage; therefore, it belongs to a family of ␤1,3-galactosyltransferases. Five members of this family have already been described (13)(14)(15)(16)(17)(18)(19)(20), with a possible homolog in Drosophila (Braniac) (21). None of the ␤3GalT enzymes characterized to date have activity with galactose-terminated oligosaccharide acceptors, suggesting that another member of the family must exist in order to catalyze the formation of the linkage region of glycosaminoglycans.
In this report, we describe the identification of GalTII and show that it is the sixth member of confirmed ␤1,3-galactosyltransferases. This enzyme was actually identified in a previous publication but was designated a ␤1,3 N-acetylglucosaminyltransferase because of a clerical error (22). Here we demonstrate that the expressed protein catalyzes the formation of Gal␤1,3Gal linkages. The protein sequence is a close homolog of other ␤1,3-galactosyltransferase, and orthologs are present in Drosophila melanogaster and Caenorhabditis elegans genomes, consistent with the idea that this enzyme is broadly distributed as might be expected for an enzyme involved in glycosaminoglycan biosynthesis. tion (ATCC CCL61, CRL-1651, and CCL2.2). CHO cells were grown under an atmosphere of 5% CO 2 in air and 100% relative humidity in Ham's F-12 growth medium or F-12/Dulbecco's modified Eagle's medium (1:1, v/v) (Life Technologies, Inc.) supplemented with 10% (v/v) fetal bovine serum (HyClone Laboratories), 100 g/ml streptomycin sulfate, and 100 units/ml penicillin G. COS-7 and HeLa S3 cells were maintained in Dulbecco's modified Eagle's medium (Life Technologies) supplemented with 10% fetal bovine serum and antibiotics.
Cloning and Expression of the Human and Mouse GalTII cDNAs-The human and mouse GalTII cDNAs were isolated from -phage libraries of human fetal brain cDNA (CLONTECH) and mouse newborn brain cDNA (Stratagene) using as a probe a PCR fragment derived from the expressed sequence tag AA150140. The PCR fragment was amplified from human genomic DNA by primers 5Ј-GCGACTAC-TACCTGCCCTACG-3Ј and 5Ј-CTCCCTTCTCTGGCAAGCACT-3Ј. The probe was labeled with [␣-32 P]CTP (Hartmann Analytics, Braunschweig, Germany) by random priming (Amersham Pharmacia Biotech).
The full-length human and mouse cDNAs were expressed with the Bac-to-Bac (Life Technologies) baculovirus Sf9 insect cell system (15). The full-length cDNAs flanked by EcoRI adaptors derived from the -phage inserts were subcloned into the pFastbac1 vector (Life Technologies). Recombinant baculoviruses were generated by site-mediated transposition as recommended by the manufacturer. The full-length cDNA was also cloned in pcDNA3.1(ϩ) at the BamHI and XhoI sites and expressed as a soluble secreted form using the pFLAG-CMV-1 vector (Sigma). An EcoRI-XbaI fragment isolated from a partial human ␤GalTII cDNA truncated after the transmembrane domain was ligated into pFLAG-CMV-1 opened with the same restriction enzymes. The resulting plasmid was transiently transfected into COS-7 cells, and the secreted enzyme was purified by affinity chromatography using anti-FLAG M2 (Sigma) and protein A-agarose beads (Amersham Pharmacia Biotech). The conditioned media and beads were mixed end-over-end overnight at 4°C and centrifuged for 5 min, and the supernatant was aspirated. The beads were washed twice with 10 ml of 20% (v/v) glycerol in 50 mM Tris-HCl, pH 7.4, and resuspended in the same buffer containing protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 1 g/ml leupeptin, and 1 g/ml pepstatin A) to achieve a ϳ50% (v/v) slurry. The immobilized enzyme was stable at 4°C for at least 4 months.
Northern Blot Analysis-GalTII mRNAs were detected by Northern blot analysis using commercially available multiple tissues poly(A) ϩ RNA blots (CLONTECH). A probe, derived from the human GalTII cDNA between nucleotides 616 and 983, was labeled and hybridized to the poly(A) ϩ RNA blots. Blots were washed in 0.1ϫ SSC, 0.1% SDS at 55°C, dried, and exposed to film for 4 days at Ϫ70°C between intensifying screens. Glycosyltransferase Activity Assays-Glycosyltransferase activity was investigated with radioactive forms of UDP-galactose, UDP-Glc-NAc, and UDP-GalNAc and various acceptor saccharide substrates,   After incubation at 37°C for 1 h, the reaction products were diluted with 1 ml of 0.5 M NaCl and applied to a Sep-Pak C18 cartridge (100 mg; Waters). After washing the cartridge with 5 ml of water, the products were eluted with 50% methanol, dried, and counted by liquid scintillation.
Product Identification-Twenty small scale enzymatic assays were performed in parallel for 4 h using Gal␤1,4Xyl␤-O-Bn as substrate and nonradioactive UDP-galactose as donor. After the clean up on a Sep-Pak C18 cartridge, the product was separated from the large amounts of unreacted disaccharide acceptor by silica gel chromatography using dichloromethane/methanol (10:3, v/v). Samples were analyzed on silica gel 60 aluminum-backed high performance thin layer chromatography plates (Merck). A section of the plate was stained with ␣-naphthol reagent and heated at 110°C for 5 min to detect the product (26).
The disaccharide substrate and purified trisaccharide product were repeatedly exchanged in D 2 O (100%; Sigma) with intermediate lyophilization. The trisaccharide was dissolved in a final volume of 40 l of 100% D 2 O and transferred into a 40-l nanocell. The disaccharide standard was dissolved in a final volume of 0.6 -0.7 ml and transferred to a 5-mm tube (Wilmad 528-PP). 1 H NMR experiments were carried out on a Varian Unity Inova 500-MHz spectrometer at 25°C. A SUN Microsystems Ultra-10 computer running Varian's VNMR software (version 6.1B) controlled data acquisition. Chemical shifts are relative to 4,4-dimethyl-4-silapentane-1-sulfonate; they were typically measured relative to the residual acetate peak (␦ 1.908 ppm at 22-30°C and pD 6 -8) and compared with literature values (27)(28)(29)(30)(31).
Small Interfering RNA (siRNA) Inhibition of GalTII-Two small interfering RNAs (siRNA GalTII-A and -B) were designed to target GalTII based on the method of Elbashir and co-workers (32). siRNA GalTII-A targeted nucleotides 461-482 numbered from the start codon (AY050570, sense: 5Ј-GGCGGACGACGACUCCUUCTT-3Ј; antisense: 5Ј-GAAGGAGUCGUCGUCCGCCTT-3Ј). siRNA GalTII-B targeted nucleotides 65-87 (sense: 5Ј-GCACGCGACGCUGGCGCGCTT-3Ј; antisense: 5Ј-GCGCGCCAGCGUCGCGUGCTT-3Ј). As a control for nonspecific effects of siRNA, oligonucleotides were designed to target mouse lamin A (mLaminA) using nucleotides 260 -280 (NM019390; sense: 5Ј-GAAGCAGCUUCAGGAUGAGAU-3Ј; antisense: 5Ј-CUCAUCCU-GAAGCUGCUUCUU-3Ј). siRNAs were obtained from Genset Oligos (Paris), and the complementary oligonucleotides were annealed as described previously (32). For each 30-mm culture dish of cells, 12 l of the 20 M stock siRNA duplex was mixed with 200 l of Opti-MEM (Life Technologies, Inc.). This mixture was gently added to a solution containing 12 l of Oligofectamine (Life Technologies, Inc.) in 48 l of Opti-MEM. After 20 min at room temperature, 128 l of Opti-MEM was added. This solution was gently overlaid onto 10 -20% confluent HeLa S3 cells, which had been previously washed with Opti-MEM. After 5 h, 2 ml of Dulbecco's modified Eagle's medium with 30% fetal calf serum was added without removing the transfection media. Four days later, the cells were treated with trypsin and reseeded in 30-mm dishes. Three rounds of siRNA transfection were required to decrease glycosaminoglycan biosynthesis as determined by 35 SO 4

incorporation (33).
Fluorescence Microscopy-A green fluorescent protein (GFP)-fused form of GalTII was prepared by amplifying the full-length cDNA from the original pcDNA3.1(ϩ) clone. An EcoRI restriction site at the 5Ј-end (5Ј-GCAGGAATTCTATAGCCACATTCCC-3Ј) and BamHI restriction site at the 3Ј-end (5Ј-CTGGATCCTCCCTTCTCTGGCAGCAC-3Ј) were introduced. After digestion with the appropriate enzymes, the fragment was inserted in frame with GFP at the C terminus in pEGFP-N1. Cells grown on 24-well glass microscope slides were transfected with pEG-FPN1-GalT II using LipofectAMINE (Life Technologies) in accordance with the manufacturer's instructions. After 48 h, the cells were rinsed with phosphate-buffered saline (PBS) (34), fixed for 1 h with 2% paraformaldehyde in 75 mM phosphate buffer (pH 7.3), and permeabilized with 0.1% Triton-X in PBS containing 0.1% bovine serum albumin (BSA). Primary rabbit antibodies against ␣-mannosidase II and CAL-NUC (medial and cis Golgi markers, respectively) were a gift from M. G. Farquhar (University of California, San Diego). The antiserum was diluted 1:500 in PBS containing 1% BSA and incubated with fixed cells in a humid chamber for 1 h. After several washes, the cells were incubated with 4Ј,6Ј-diamidino-2-phenylindole to stain nuclei and a secondary antibody (Alexa Fluor 488 goat anti-rabbit IgG, Molecular Probes, Inc., Eugene, OR) diluted 1:200 in PBS containing 1% BSA. The coverslips were washed several times with PBS containing 0.1% BSA and mounted with Vectashield mounting medium (Vector Laboratories, CA). Images were captured with a Photometrics CCD mounted on a Nikon microscope adapted to a DeltaVision (Applied Precision, Inc.) deconvolution imaging system. The data sets were deconvolved and analyzed using SoftWorx software (Applied Precision, Inc.) on a Silicon Graphics Octane work station.

RESULTS AND DISCUSSION
A large family of ␤1,3-galactosyltransferases has been described based on expressed sequence tags exhibiting a high degree of homology and by use of PCR-based cloning strategies (13)(14)(15)(16)(17)(18)(19)(20). ␤3GalT1, -2, -3, and -5 utilize GlcNAc-terminated oligosaccharide as acceptor substrates, producing linkages characteristic of type 1 N-acetyllactosamine repeat units on N-and O-linked oligosaccharides and glycolipids, whereas ␤3GalT4 transfers galactose to the terminal GalNAc unit of the ganglioside series GM2, GD2, and GA2 acceptors (35). In addition to GlcNAc␤-based acceptors, ␤3GalT5 is also capable of transferring galactose to the terminal GalNAc unit of the globoside Gb4 (20). In a search for additional family members, we used the mouse ␤3GalT1, -2, and -3 (15) protein sequences to query the expressed sequence tag division of GenBank TM , and the fragment AA150140 was identified. This fragment was used as a probe to isolate the corresponding full-length cDNAs from human and mouse libraries, which corresponded to open reading frames encoding proteins of 329 and 325 amino acids, respectively. This cDNA was previously associated with a ␤1,3-Nacetylglucosaminyltransferase activity (22), but the actual sequence responsible for the glucosaminyltransferase activity reported previously is encoded by AF092050 and AF092051. These correct designations have been posted in the GenBank TM data base. As shown below the cDNA actually encodes a ␤1,3galactosyltransferase that participates in the formation of the linkage region of glycosaminoglycans. As the newest confirmed member of the ␤3GalT family, it is designated ␤3GalT6. However, in this paper it will be referred to as GAG GalTII after the nomenclature used in the proteoglycan field (1).
As shown in Fig. 1, GAG GalTII shows extensive homology to four other members of the human ␤1,3-galactosyltransferase family that have confirmed preferences for GlcNAc-terminated oligosaccharides. Like the other ␤1,3-galactosyltransferases, GAG GalTII has a typical type II transmembrane orientation. The tentative transmembrane domain consists of 19 amino acids starting at residue 12 from the N terminus. In the transmembrane domain, the GAG GalTII has a cysteine residue that is conserved in mouse, human, and C. elegans (Fig. 2). Interestingly, ␤3GalT1, ␤3GalT5, and the homologous Drosophila protein Brainiac also contain Cys residues in their putative transmembrane domains, but the function of these residues is unknown. One possibility is that they may affect the oligomerization of the protein as recently reported for ␣2,6-sialyltransferase I (36). The transmembrane domain is followed by a segment rich in Ser, Pro, Ala, and Gly, which has been described as the "SPLAG" domain in other transferases (37), and presumably represents a stem region that determines the distance between the catalytic domain and the membrane. All of the other motifs identified previously in the ␤3GalTs presum-ably define essential folds in the catalytic domain (13,15,22). These were also found in similar locations in GAG GalTII, but the conserved cysteine residues are either absent or located differently.
The chromosomal location of human and mouse GAG GalTII was determined using both the public genome data base and the partially annotated mouse genome. Human GalTII maps to chromosome 1p36.3 and occupies a similar position on the mouse chromosome at 4E2. Inspection of the sequence indicates that GalTII, like the other members of the ␤3GalTs, is encoded by a single exon in the human, but its organization in the mouse genome is unclear from the existing annotated data base. However, the gene appears to contain multiple exons in D. melanogaster and C. elegans. GalTII mRNA is expressed broadly across both human and mouse tissues (Fig. 3 and Ref. 22), revealing three different transcripts of ϳ1.6, 2.4, and 3.3 kilobases. This pattern differs from the expression profiles of the other ␤3GalTs, which tend to be restricted to specific tissues (13,15,16,18,19). The presence of multiple transcripts suggests that additional exons exist encoding 5Ј-and 3Ј-untranslated regions.
Differences in nucleotide and amino acid sequences were used to measure the evolutionary relationship of GalTII and the other members of the ␤3GalT family (Fig. 4). Both dendrograms indicated a "starburst" relationship among the human ␤3GalT genes; i.e. all members of the family are relatively equidistant from each other (the internodal distances have high bootstrap support). This observation suggests that the ␤3GalT family emerged early in evolution and has not undergone recent duplications and radiations. Human GalTII has undergone progressive variation in sequence, diverging as much from earlier phylogenetic forms as from other members in the family.
Enzymatic Activity-All of the previously described ␤3GalTs transfer galactose from UDP-galactose to glycans or glycoconjugates containing terminal ␤-linked GlcNAc or GalNAc residues. GAG GalTII, however, lacks this activity (Table I), and instead the recombinant enzyme prefers glycans with a terminal ␤-linked galactose residue or simple ␤-galactosides. In particular, GalTII reacted with Gal␤1,4Xyl␤-O-Bn more strongly than any of the other substrates. A plot of activity versus acceptor concentration yielded an apparent K m value of ϳ5.7 mM (Fig. 5). The enzyme also showed a strict requirement for UDP-galactose, failing to transfer sugar from UDP-GalNAc and UDP-GlcNAc.
GalTII Is Required for Glycosaminoglycan Biosynthesis-To determine whether GalTII was actually involved in glycosaminoglycan formation, siRNAs were generated to two different oligonucleotides segments of the human sequence. One round of transfection of HeLa cells only marginally inhibited glycosaminoglycan biosynthesis, as measured by 35   tion into the chains. However, a second round of transfection increased the level of inhibition, and by the third cycle a 6 -10fold reduction in synthesis was observed (207 cpm/g of protein versus 25 and 37 cpm/g of protein for GalTII-A and GalTII-B, respectively; Fig. 7). Digestion of the residual [ 35 S]glycosaminoglycans showed that siRNA to GalTII inhibited both heparan sulfate and chondroitin sulfate synthesis, consistent with the idea that siRNA had silenced a gene early in the pathway. GalTII Is Localized to the Medial Golgi-To determine the subcellular location of GAG GalTII, green fluorescent protein was fused to the C terminus, and the chimeric protein was expressed in Chinese hamster ovary cells (Fig. 8). By deconvolution microscopy, the tagged enzyme was detected in a punctate distribution in a perinuclear region overlapping to a greater extent with ␣-mannosidase II, a marker of the medial Golgi (36), than with CALNUC (nucleobindin), which has been localized to the cis-Golgi cisternae and cis-Golgi network (37). Prior studies indicate that GalTI (␤4GalT7) and glucuronosyltransferase I (the enzymes that generate the precursor for GalTII and that acts on its product, respectively) are localized similarly (40). Xylosyltransferases have recently been identified, but their localizations have not yet been established (2). Immunological evidence suggests that this reaction occurs in an earlier endoplasmic reticulum compartment (41), suggesting that GAG synthesis initiates in the endoplasmic reticulum and then linkage region assembly is completed in the medial Golgi.
Conclusions-In this report, we have identified the cDNA for GAG GalTII and report that it is the previously described ␤1,3-N-acetylglucosaminyltransferase (22). This enzyme is part of a gene family that encodes at least six ␤1,3-galactosyltransferases involved in glycoprotein, glycolipid, and proteoglycan processing. These enzymes in fact contain a motif in common with other ␤1,3-glycosyltransferases. One of the enzymes, ␤3GalT3, has been shown to also transfer GalNAc in ␤1,3linkage to globosides (42), and four structurally related ␤1,3-N-acetylglucosaminyltransferases have been recently cloned based on sequence homology to ␤3GalT1 (43)(44)(45). In contrast to these enzymes, GAG GalTII (␤3GalT6) preferentially acts on Gal␤1,4Xyl, which is found in the linkage region of glycosaminoglycans. Furthermore, silencing of the gene by siRNA blocked glycosaminoglycan assembly in vivo, verifying the identity of this cDNA as GAG GalTII. Its localization to the medial Golgi along with GalTI and GlcATI is consistent with the idea that linkage region formation takes place in this compartment.