Control of O-Glycan Branch Formation

A novel human UDP-GlcNAc:Gal/GlcNAcβ1–3GalNAcα β1,6GlcNAc-transferase, designated C2/4GnT, was identified by BLAST analysis of expressed sequence tags. The sequence of C2/4GnT encoded a putative type II transmembrane protein with significant sequence similarity to human C2GnT and IGnT. Expression of the secreted form of C2/4GnT in insect cells showed that the gene product had UDP-N-acetyl-α-d-glucosamine:acceptor β1,6-N-acetylglucosaminyltransferase (β1,6GlcNAc-transferase) activity. Analysis of substrate specificity revealed that the enzyme catalyzed O-glycan branch formation of the core 2 and core 4 type. NMR analyses of the product formed with core 3-para-nitrophenyl confirmed the product core 4-para-nitrophenyl. The coding region of C2/4GnT was contained in a single exon and located to chromosome 15q21.3. Northern analysis revealed a restricted expression pattern of C2/4GnT mainly in colon, kidney, pancreas, and small intestine. No expression of C2/4GnT was detected in brain, heart, liver, ovary, placenta, spleen, thymus, and peripheral blood leukocytes. The expression of core 2O-glycans has been correlated with cell differentiation processes and cancer. The results confirm the predicted existence of a β1,6GlcNAc-transferase that functions in both core 2 and core 4O-glycan branch formation. The redundancy in β1,6GlcNAc-transferases capable of forming core 2O-glycans is important for understanding the mechanisms leading to specific changes in core 2 branching during cell development and malignant transformation.

Mucin-type O-glycosylation is initiated by a large family of UDP-GalNAc:polypeptide GalNAc-transferases that add Gal-NAc to selected Ser and Thr residues (1). Further assembly of O-glycan chains involves different biosynthetic pathways: (i) formation of simple mucin-type core 1 structures by UDP-Gal: GalNAc␣ ␤1,3Gal-transferase activity; (ii) conversion of core 1 to complex-type core 2 structures by UDP-GlcNAc:Gal␤1-3GalNAc␣ ␤1,6GlcNAc-transferase 1 activities; (iii) direct formation of complex mucin-type core 3 by UDP-GlcNAc:GalNAc␣ ␤1,3GlcNAc-transferase activities; and (iv) conversion of core 3 to core 4 by UDP-GlcNAc:GlcNAc␤1-3GalNAc␣ ␤1,6GlcNActransferase activity (for an overview see Ref. 2; see also Fig. 1). Elongation and termination of complex oligosaccharide structures involves a large number of glycosyltransferases. Synthesis of the different O-glycan structures is cell-and tissuespecific. Different core structures are produced upon differentiation and malignant transformation (3)(4)(5)(6)(7). For example, increased formation of GlcNAc␤1-6GalNAc branching in O-glycans has been demonstrated during T-cell activation, during the development of leukemia, and for immunodeficiencies like Wiskott-Aldrich syndrome and AIDS (3,8,9). Core 2 branching may play a role in tumor progression and metastasis (10). In contrast, many carcinomas show changes from complex O-glycans found in normal cell types to immaturely processed simple mucin-type O-glycans such as T (Thomsen-Friedenreich antigen; Gal␤1-3GalNAc␣1-R), Tn (GalNAc␣1-R), and sialosyl-Tn (NeuAc␣2-6GalNAc␣1-R) (see Fig. 1) (11). The molecular basis for this has been extensively studied in breast cancer, where it was shown that specific down-regulation of core 2 ␤6GlcNAc-transferase was responsible for the observed lack of complex type O-glycans on the mucin MUC1 (7). Interestingly, the metastatic potential of tumors has been correlated with increased expression of core 2 ␤6GlcNAc-transferase activity (6). The increase in core 2 ␤6GlcNAc-transferase activity was associated with increased levels of poly-N-acetyllactosamine chains carrying sialyl-Le x , which may contribute to tumor metastasis by altering selectin-mediated adhesion (5,12). The control of O-glycan core assembly is regulated by the expression of key enzyme activities outlined in Fig. 1; however, epigenetic factors including post-translational modification, topology, or competition for substrates may also play a role in this process (13).
The kinetic properties and substrate specificities of the enzymes responsible for O-glycan core formation have been extensively studied (14 -20), and one of these enzymes, a core 2 ␤6GlcNAc-transferase (C2GnT), 2 has been cloned and charac-terized to date (21). Human C2GnT was identified by a transfection cloning strategy, and it shows similarity to another ␤6GlcNAc-transferase (IGnT) that is responsible for poly-Nacetyllactosamine branching (22). Surprisingly, these two genes were not similar in sequence to GnTV, which is responsible for formation of the ␤6GlcNAc branch in tetraantennary N-linked glycans (23). Studies of the kinetic properties and acceptor substrate specificities of ␤6GlcNAc-transferases involved in O-glycosylation from different cell lines and organs have suggested that multiple ␤6GlcNAc-transferases exist and that one of these would function in a manner that is distinct from the cloned C2GnT enzyme, in being capable of forming core 2, core 4, and possibly the branched I antigen structure (17,18). The existence of a homologous ␤6GlcNAc-transferase family containing C2GnT and IGnT suggests that additional ␤6GlcNAc-transferase activities may be encoded by homologous genes (22). Recently, similar genes for large families of galactosyltransferases (24 -26) have been identified by using the expressed sequence tag (EST) information for similarity searches. This report describes the use of this strategy for identification of a novel member of the ␤6GlcNAc-transferase family that forms core 2 as well as core 4 structures.

EXPERIMENTAL PROCEDURES
Identification of C2/4GnT-The BLASTn and tBLASTn were used with the reported coding sequence of human C2GnT (GenBank TM accession number M97347) to search the dbEST data base at The National Center for Biotechnology Information as described previously (24). Human EST and genomic bovine sequences similar but not identical to known members of the ␤6GlcNAc-transferase family were identified.
Cloning and Sequencing of the Gene Encoding C2/4GnT-EST clone 178656 (5Ј EST GenBank TM accession number AA307800), derived from a putative homologue to C2GnT, was obtained from the American Type Culture Collection. Sequencing of this clone revealed a partial open reading frame with significant sequence similarity to C2GnT. The coding region of human C2GnT and a bovine homologue was previously found to be organized in one exon (27). 3 Because the 5Ј and 3Ј sequence available from the C2/4GnT EST was incomplete but likely to be located in a single exon, the missing 5Ј and 3Ј portions of the open reading frame were obtained by sequencing genomic P1 clones. P1 clones were obtained from a human foreskin genomic P1 library (DuPont Merck Pharmaceutical Co. Human Foreskin Fibroblast P1 Library) by screening with the primer pair TSHC27 (5Ј-GGAAGTTCATACAGTTCCCAC-3Ј) and TSHC28 (5Ј-CCTCCCATTCAACATCTTGAG-3Ј). Two genomic clones for C2/4GnT, DPMC-HFF#1-1026(E2) and DPMC-HFF#1-1091(F1), were obtained from Genome Systems Inc. DNA from P1 phage was prepared as recommended by Genome Systems Inc. The entire coding sequence of the C2/4GnT gene was represented in both clones and sequenced in full using automated sequencing (ABI377, Perkin-Elmer). Confirmatory sequencing was performed on a cDNA zen et al.   4 The sequence of the 5Ј end of C2/4GnT mRNA including the translational start site and 5Ј-UTR was obtained by 5Ј rapid amplification of cDNA ends (35 cycles at 94°C for 20 s, 52°C for 15 s, and 72°C for 2 min) using total cDNA from the human COLO 205 cancer cell line with the antisense primer TSHC 48 (5Ј-GTGG-GAACTGTATGAACTTCC-3Ј) (see Fig. 2).
Expression of C2/4GnT and C2GnT in Insect Cells-An expression construct designed to encode amino acid residues 31-438 of C2/4GnT was prepared by PCR using P1 DNA and the primer pair TSHC55 (5Ј-CGAGAATTCAGGTTGAAGTGTGACTC-3Ј) and TSHC45 (see Fig.  2). The PCR product was cloned into the EcoRI site of pAcGP67A (PharMingen), and the insert was fully sequenced. An expression construct encoding amino acid residues 40 -428 of human C2GnT was prepared by PCR on genomic DNA of a healthy male blood donor using the primer pair EBCT1 (5Ј-AGCGGATCCTTTGTAAGTGTCAGA-CACTTGGAG-3Ј) and EBCT2 (5Ј-AGCGGATCCAAAATTGCCCGTA-ATGGTCAGTG-3Ј). The PCR product was cloned into the BamHI site of pAcGP67B (PharMingen), and the sequence was found to be identical to the sequence originally reported by Bierhuizen and Fukuda (21). Plasmids pAcGP67-C2/4GnT-sol and pAcGP67-C2GnT-sol were co-transfected with Baculo-Gold TM DNA (PharMingen) as described previously (24). Recombinant Baculo-virus were obtained after two successive amplifications in Sf9 cells grown in serum-containing medium, and titers of virus were estimated by titration in 24-well plates with monitoring of enzyme activities. Controls included the pAcGP67-GalNAc-T3-sol (28). The kinetic properties were determined with partially purified enzymes expressed in High Five TM cells. Partial purification was performed by consecutive chromatography on Amberlite IRA-95, DEAE-Sephacryl, and CM-Sepharose essentially as described (29). Protein concentrations were determined using the Bio-Rad reagent with bovine serum albumin as standard protein.  Table I for structures). Semi-purified C2GnT and C2/4GnT were assayed in 50-l reaction mixtures containing 100 mM MES (pH 7), 5 mM (C2/4GnT) or 2 mM EDTA (C2GnT), 1 mM dithiothreitol (C2GnT), 90 M UDP-[ 14 C]-GlcNAc (3,050 cpm/nmol) (Amersham Pharmacia Biotech), and the indicated concentrations of acceptor substrates. Reaction products were quantified by chromatography on Dowex AG1-X8. Complete glycosylation of core3-pNph was performed in a reaction mixture consisting of 6.9 milliunits C2/4GnT (specific activity determined with core1-pNph), 2 mg core3-pNph, 100 mM MES (pH 7.0), 5 mM EDTA, 4.6 mol UDP-GlcNAc, and 100 milliunits alkaline phosphatase in a final volume of 200 l. The glycosylation of core3-pNph was monitored by thin layer chromatography and run for 8 h until completed. The reaction product was purified on an octadecylsilica cartridge (Bakerbond; J. T. Baker), deuterium exchanged by repeated lyophilization from D 2 O and then dissolved in 0.5 ml of D 2 O for NMR analysis. One-dimensional 1 H NMR, two-dimensional 1 H-1 H TOCSY (30,31) and ROESY (32, 33), 1 H-detected, 13 C-decoupled, phase sensitive, gradient (34) 13 C-1 H HSQC (35), and HMBC (36, 37) experiments were performed at 298 K on a Varian Unity Inova 600 MHz spectrometer using standard acquisition software available in the Varian VNMR software package. A 2-mg sample of core3-pNph was prepared in similar fashion and analyzed under identical conditions for comparison. Chemical shifts are referenced to internal acetone (2.225 and 29.92 ppm for 1 H and 13 C, respectively).

Enzymatic Assays and Product Characterization-Standard
Northern Analysis-Total RNA was isolated from human colon and pancreatic adenocarcinoma cell lines AsPC-1, BxPC-3, Capan-1, Capan-2, COLO 357, HT-29, and PANC-1 essentially as described (38). 25 g of total RNA was subjected to electrophoresis on a 1% denaturing agarose gel and transferred to nitrocellulose as described previously (38). The cDNA fragment of soluble C2/4GnT was used as a probe for hybridization. The probe was random primer-labeled using [␣ 32 P]dCTP and an oligonucleotide labeling kit (Amersham Pharmacia Biotech). The membrane was probed overnight at 42°C as described previously (28) and washed twice for 30 min each at 42°C with 2 ϫ SSC, 0.1% SDS and twice for 30 min each at 52°C with 0.1 ϫ SSC, 0.1% SDS. Human multiple tissue Northern blots, MTN I and MTN II (CLONTECH), were 4 The website at the Swiss Institute for Experimental Cancer Research is http://www.isrec.isb-sib.ch/software/TMPRED_form.html). probed as described above and washed twice for 10 min each at room temperature with 2 ϫ SSC, 0.1% SDS; twice for 10 min each at 55°C with 1 ϫ SSC, 0.1% SDS; and once for 10 min with 0.1 ϫ SSC, 0.1% SDS at 55°C.
Chromosomal Localization of C2/4GnT: in Situ Hybridization to Metaphase Chromosomes-Fluorescence in situ hybridization was performed on normal human lymphocyte metaphase chromosomes using procedures described previously (24). For evaluation of the chromosomal slides a Zeiss epifluorescence microscope equipped with appropriate filters for visualization of fluorescein isothiocyanate was used. Hybridization signals and 4,6-diamidino-2-phenylindole-counterstained chromosomes were transformed into pseudo-colored images using image analysis software. For precise localization and chromosome identification 4,6-diamidino-2-phenylindole-converted banding patterns were generated using the BDS-image TM software package (ONCOR).

RESULTS
Isolation and Characterization of Human C2/4GnT-Analysis of the GenBank TM and dbEST data bases suggested the existence of additional members of the ␤6GlcNAc-transferase family. EST clones encoding the putative enzymes were obtained, and the full coding sequence of a novel gene with significant sequence similarity to human C2GnT (21) and blood group IGnT (22) was obtained by 5Ј-rapid amplification of cDNA ends and analysis of P1 DNA. The sequence encoded a type II transmembrane protein of 438 amino acids with an N-terminal cytoplasmic domain of 9 residues, a transmembrane segment of 18 residues, a stem region, and catalytic domain of 411 residues with two potential N-linked glycosylation sites (Fig. 2). A Kyte and Doolittle (39) hydropathy plot of C2/4GnT showed similarity to C2GnT and indicated that the putative stem region was hydrophilic, similar to other Golgilocalized glycosyltransferases (data not shown). The cloned cDNA included a single initiation codon according to the Kozak rule (40). The 3Ј-UTR of 510 base pairs was confirmed in additional EST clones and contains a polyadenylation signal at base pair 1749 (ϩ435) (Fig. 2).
A multiple sequence alignment (ClustalW) of three human ␤6GlcNAc-transferases is shown in Fig. 3. C2/C4GnT shows a higher overall amino acid sequence identity to human C2GnT (52%) than to human IGnT (41%). Sequence similarities among the three human proteins are found predominantly in the putative catalytic domains, and no significant sequence similarities were detected in the N-terminal regions. Eight cysteine residues are conserved in all three ␤6GlcNAc-transferases; a ninth cysteine residue in the N-terminal region of C2GnT and C2/C4GnT was not found in IGnT (Fig. 3). N-linked glycosylation consensus sequence sites are not generally conserved in homologous as well as orthologous glycosyltransferases, although one site is conserved in the central region of four ␤3Galtransferases (25) and one site is conserved in the C-terminal region of five of six ␤4Gal-transferases (41). Interestingly, a single potential N-linked site located in the stem region of C2/4GnT, C2GnT and IGnT is conserved (Fig. 3). The two N-glycosylation sites found in C2GnT are utilized, and the N-Glycan in the conserved position was shown to be essential for function of the recombinant enzyme (42).
Genomic Organization and Chromosomal Localization-The coding region of the human C2/4GnT gene was found to be organized in a single exon, similar to the genomic structure of the human C2GnT gene (27). The human C2GnT gene was previously found to be localized in a gene cluster with IGnT localized on chromosome 9q21 (22). However, analysis of the Human Gene Map at GenBank TM (The National Center for Biotechnology Information) indicates that the IGnT gene is located on chromosome 6p24 between microsatellite markers D6S1674 and D6S470 (13-17 cM, SHGC-12039). Bierhuizen et al. (22) observed weak binding of a genomic IGnT clone to p23 of chromosome 6 as well as a major hybridization signal at 9q21. This may suggest the existence of an additional highly similar gene or pseudogene at band q21 of chromosome 9. The existence of two IGnT enzyme forms has recently been suggested by Leppä nen et al. (43). By fluorescence in situ hybridization of the genomic clone DPMC-HFF#1-1091(F1) to human metaphase chromosomes, the C2/4GnT gene was found to reside on chromosome 15q21.3 (Fig. 4).
Characterization of Core 4 Product by 1 H and 13 C NMR Spectroscopy-The product derived from reaction of the puta-   tive Core 4 ␤6GlcNAc-transferase with ␤-D-GlcNAc-(1-3)-␣-D-GalNAc-1-para-nitrophenyl was characterized by NMR spectroscopy to confirm that the proper linkage was formed between the donor sugar and the acceptor substrate. Comparison of a one-dimensional 1 H NMR spectrum of the product (Fig. 5) with that of the substrate (data not shown) clearly showed an additional H-1 resonance (4.454 ppm) from a sugar residue linked in the ␤-configuration ( 3 J 1,2 ϭ 7-9 Hz). Because we were unable to find NMR data for the para-nitrophenyl glycosides of either the core 3 substrate or the expected core 4 product in the literature or in glycoconjugate NMR data bases and because the substantial anisotropic effects of the paranitrophenyl group obviate direct comparison of chemical shift data with those of the benzyl-glycosides (44), a de novo sequence analysis of the product was undertaken by consecutive application of two-dimensional 1 (44)); then the linkages were unambiguously established by observation of interglycosidic H1ϪC1ϪO1ϪCx and C1ϪO1ϪCxϪHx correlations in the HMBC spectrum. As shown in Fig. 6, the newly formed GlcNAc␤1 3 6GalNAc␣ linkage in the product is clearly demonstrated by strong cross-peaks correlating the  ␤-GlcNAc H-1 at 4.454 ppm with ␣-GalNAc C6 and the corresponding ␤-GlcNAc C-1 at 100.93 ppm with both ␣-GalNAc H-6 resonances. Consistent with this, rotating frame Overhauser enhancements were observed between ␤-GlcNAc H-1 and ␣-GalNAc H-6 in a ROESY spectrum (data not shown).
Expression Pattern of C2/4GnT-ESTs from C2/4GnT were derived from colonic and pancreatic cancer tissues as well as germ cell tumors. Northern analysis with mRNA from 16 healthy human adult organs showed expression of C2/4GnT in organs of the gastrointestinal tract with high transcription levels observed in colon and kidney and lower levels in small intestine and pancreas (Fig. 7A). To investigate changes in expression of C2/4GnT in cancer cells derived from tissues normally expressing C2/4GnT, mRNA levels in a panel of human adenocarcinoma cell lines were determined. Analyses of C2/4GnT transcription levels revealed differential expression in pancreatic cell lines; Capan-1 and AsPC-1 expressed the transcript, whereas PANC-1, Capan-2, BxPC-3, and COL0357 did not (Fig. 7B). The colonic cell line HT-29 expressed transcripts of C2/4GnT. The size of the predominant transcript was approximately 2.4 kilobases, which correlates to the transcript size of 2.1 kilobases of the smallest of three transcripts of human C2GnT (21). Additionally, transcripts of approximately 3.4 and 6 kilobases were obtained in mRNA from healthy colonic mucosa (Fig. 7A). The two additional transcripts may resemble the 3.3-and 5.4-kilobase transcripts of C2GnT, which have not yet been characterized. Multiple transcripts of C2GnT have been suggested to be caused by differential usage of polyadenylation signals, which affects the length of the 3Ј-UTR (21). DISCUSSION Additional members of a human ␤6GlcNAc-transferase gene family were previously predicted from analysis of enzyme activities as well as by low stringency Southern blot analysis (17,22,46). An EST cloning strategy produced a novel ␤6GlcNActransferase, designated C2/4GnT. The C2/4GnT enzyme is most similar in sequence to C2GnT, and it has the same simple genomic organization of the coding region as C2GnT. The recombinant forms of these enzymes have partly overlapping functions. The IGnT is apparently a more distant member of the family, and it has a different function (22). Three regions of extensive homology have been identified in C2GnT and IGnT: region A in the N-terminal and regions B and C in the Cterminal portion of the catalytic domain. C2/4GnT shows a particularly high degree of homology to C2GnT in region A, supporting the hypothesis that these domains could be directly involved in acceptor binding (22).
Previous analysis of enzyme activities in cell extracts suggested that there exists an enzyme capable of synthesizing core 4 as well as core 2. Brockhausen et al. (17) found that core 2 and core 4 ␤6GlcNAc-transferase activities of colon had similar properties and functioned independently of bivalent cations. Competition experiments with the two substrates yielded no additive effect; it therefore was suggested that in colon one enzyme was responsible for core 2 and core 4 biosynthesis. Ropp et al. (47) found that a bovine tracheal ␤6GlcNAc-transferase purified to apparent homogeneity with a molecular weight of 69,000 (specific activity, 70 units/mg with Gal␤1-3GalNAc␣-benzyl) utilized both core 1 and 3 acceptors. In addition, activity was detected with GlcNAc␤1-3Gal␤1-R for I antigen biosynthesis. This ␤6GlcNAc-transferase activity has been designated the M-form for the mucin-secreting tissue form of C2GnT in contrast to the recently cloned leukocyte form or L-form, which is restricted to the formation of core 2 by accepting Gal␤1-3GalNAc␣-R substrates exclusively (4,20,21). The C2/4GnT enzyme reported here shows broader acceptor substrate specificity in accepting core 1 as well as core 3 but does not resemble the mucin-secreting tissue form because it lacks IGnT activity. C2/4GnT showed approximately 3-4-fold better activity with core 1-para-nitrophenyl than core 3-paranitrophenyl, which is similar to the differences observed for both the bovine and porcine tracheal enzymes (47,48).
Species-specific differences in core 2 ␤6GlcNAc-transferase activities may exist. Sekine et al. (49) reported that a purified mouse kidney ␤6GlcNAc-transferase activity utilized the Gal-Gb 4 glycosphingolipid as well as core 1 substrates. Subsequently, evidence was presented that the mouse enzyme was encoded by the orthologous gene of human C2GnT (50). Interestingly, mouse kidney cells express a 5Ј-UTR spliced version of C2GnT that is distinct from other organs. The novel 5Ј-UTR does not affect the coding region (50). Whereas the normal C2GnT version was ubiquitously expressed at very low levels, a kidney form was highly expressed in Balb/c mice. DBA/2 mice, which lack the branched Gal-Gb 4 , showed no expression in kidney. To the best of our knowledge, studies have not determined whether human C2GnT or IGnT can use Gal-Gb 4 as substrate. This structure was not available for testing in the present study.
The existence of ␤6GlcNAc-transferases with unique functions in either core 2 or blood group I synthesis was predicted from analysis of activities in different organs. The cloning of C2GnT and IGnT confirmed this prediction (21,22). The C2GnT gene is widely expressed, although the levels may be quite low. 3 In contrast, C2/4GnT showed a restricted expression pattern with high expression in colon (Fig. 7A). The mucintype core 4 structure is less commonly found in mucins than core 1 and 2 (51-53) and has been found predominantly in gastric, respiratory, and colonic mucin preparations (54,55). However, in some studies colonic mucins have been found to carry exclusively core 3-based structures (56,57). C2GnT and C2/4GnT may provide redundancy for core 2 synthesis in some cells and tissues. This may be relevant to studies that demonstrated altered expression of core 2 in carcinomas (4, 6, 7). Interestingly, the major core 2 ␤6GlcNAc-transferase activity in normal colonic mucosa may be C2/4GnT, because competition experiments for core 1 and core 3 acceptors revealed no additive effect (17). This is further supported by recent reports showing that the mucin-type core 2 ␤6GlcNAc-transferase activity, which is expressed in healthy colonic mucosa, is replaced by the leukocyte form in colon cancer tissues (4). Further studies of the expression of the two enzymes are required for understanding their individual roles in cellular regulation of core 2 biosynthesis under normal physiological conditions and during progression to cancer (58).
The existence of several additional genes in the ␤6GlcNActransferase gene family may be predicted. Leppä nen et al. (43) provided evidence for an additional I ␤6GlcNAc-transferase activity and found that this activity differs from the cloned IGnT by acting on the penultimate galactose in poly-Nacetyllactosamine chains and not on galactose residues located centrally in an oligosaccharide chain. Hybrid globo-lactoseries glycosphingolipids may have ␤6GlcNAc branching either to the ␣Gal found in Ref. 59 or ␤GalNAc of galactosyl-Gb 4 or Gb 4 (49).
Originally the IGnT and C2GnT genes were mapped by isotopic in situ hybridization to a gene cluster at 9q21. Genomic DNA clones of C2GnT and IGnT showed a major hybridization signal at 9q21. Additionally weak binding of the IGnT clone to 6p23 was observed (22). C2/4GnT was shown in the present study by in situ hybridization to be located at 15q21.3. The localization of the IGnT gene may require reevaluation, because linkage analysis of 3Ј ESTs (SHGC-12039) for this gene indicates that it is located at 6p24 (60). Most glycosyltransferase gene families are not arranged in gene clusters (24 -26, 61, 62). It is possible that only the ␣2and ␣3/4-fucosyltransferase families have members clustered at one locus (63). Interestingly, the three members of the ␣3/4-fucosyltransferase family co-localized to chromosome 19 are highly similar in sequence. These appear to have been duplicated very recently, because only one gene for this cluster was found in the cow (64). Two more divergent members of this family are localized differently.
In summary the present data confirm the existence of a ␤6GlcNAc-transferase capable of forming both core 2 and core 4 and establish that core 2 O-glycan branching is controlled by multiple enzymes. This may have important implications for interpreting the molecular events underlying characteristic changes in O-glycan branching during cell development and malignant transformation. The in vitro biosynthesis of O-glycopeptide structures is presently hampered by lack of availability of the key enzymes adding either galactose or N-acetylglucosamine in a ␤1-3 linkage to GalNAc␣1-O-Ser/Thr to form core 1 and core 3, because these are required for the enzymes responsible for the build-up of complex type structures (Fig. 1). Most other enzymes required for elongation of branched Oglycans are available, and the core 2/4 enzyme described herein now makes the synthesis of core 4-based structures possible.