Altered Golgi Localization of Core 2 β-1,6-N-Acetylglucosaminyltransferase Leads to Decreased Synthesis of Branched O-Glycans*

Mucin type O-glycans with core 2 branches are distinct from nonbranched O-glycans, and the amount of core 2 branched O-glycans changes dramatically during T cell differentiation. This oligosaccharide is synthesized only when core 2 β-1,6-N-acetylglucosaminyltransferase (C2GnT) is present, and the expression of this glycosyltransferase is highly regulated. To understand how O-glycan synthesis is regulated by the orderly appearance of glycosyltransferases that form core 2 branched O-glycans, the subcellular localization of C2GnT was determined by using antibodies generated that are specific to C2GnT. The studies using confocal light microscopy demonstrated that C2GnT was localized mainly in cis tomedial-cisternae of the Golgi. We then converted C2GnT to atrans-Golgi enzyme by replacing its Golgi retention signal with that of α-2,6-sialyltransferase, which resides intrans-Golgi. Chinese hamster ovary cells expressing wild type C2GnT and the chimeric C2GnT were then subjected to oligosaccharide analysis. The results obtained clearly indicate that the conversion of C2GnT into a trans-Golgi enzyme resulted in a substantial decrease of core 2 branched oligosaccharides. These results, taken together, strongly suggest that the predominance of core 2 branched oligosaccharides in those cells expressing C2GnT is due to the fact that C2GnT is located earlier in the Golgi than α-2,3-sialyltransferase that competes with C2GnT for the common substrate. Furthermore, alteration of Golgi localization renders the chimeric C2GnT much less efficient in synthesizing core 2 branched oligosaccharides, indicating the critical role of orderly subcellular localization of glycosyltransferases.

The conversion of O-glycan biosynthesis is due to the turning on or off of core 2 ␤-1,6-N-acetylglucosaminyltransferase (C2GnT). 1 It has been demonstrated that activated T lymphocytes express a substantial amount of C2GnT activity, while resting T lymphocytes express negligible C2GnT activity (8). By in situ hybridization of the transcript, it has been shown that immature cortical thymocytes express a substantial amount of C2GnT mRNA, while it was not detected in medullary thymocytes (9). The conversion of O-glycan structures during thymocyte development may be critical for the apoptotic process in thymus, since such a process is modulated by the presence of O-glycans on thymocytes (10).
Expression of the branched hexasaccharide in peripheral blood T lymphocytes has been also observed in patients with immunodeficient syndromes such as Wiskott-Aldrich syndrome (11,12). It has also been shown that AIDS patients express substantially increased amounts of the hexasaccharide or its monosialylated forms (13,14). AIDS patients produce antibodies against leukosialin expressing those oligosaccharides, and such antibodies are implicated in causing T lymphocyte depletion, which may be a cause of pathological conditions in these diseases (15). These combined results indicate that it is critical to understand how core 2 branchings are synthesized.
The biosynthesis of oligosaccharides is also controlled by specific localization of glycosyltransferases that add a specific monosaccharide in each reaction (16). If a glycosyltransferase is misplaced, sequential reactions would not take place, since a given glycosyltransferase adds a monosaccharide to a particular acceptor that was formed by another glycosyltransferase that resides in an earlier compartment(s). Although subcellular localization of glycosyltransferases that form N-glycans is relatively well studied (17), very little is known about subcellular distribution of glycosyltransferases that form O-glycans (see Ref. 18).
In the present study, we have first determined that C2GnT is localized in cis to medial-Golgi using antibodies specific for C2GnT. We then converted C2GnT into a trans-Golgi enzyme by replacing its domain responsible for Golgi retention with that of Gal␤134GlcNAc ␣-2,6-sialyltransferase, ST6Gal I (19). Such altered localization of C2GnT was found to result in altered synthesis of oligosaccharides, demonstrating the importance of the orderly presence of glycosyltransferases.

EXPERIMENTAL PROCEDURES
Construction of pGEX-C2GnT-To prepare antibodies specific to C2GnT, a cDNA encoding the catalytic domain of C2GnT was amplified by PCR using C2GnT cDNA (20) as a template and fused with GST protein. The 5Ј-primer for PCR is 5Ј-aaacgtggatcc CATCATCATCAT-CATCAT ccc ggg TCTTCTTTCATC, (BamHI site and 6-His linker are singly and doubly underlined, respectively, while the italic type corresponds to residues 101-104 of C2GnT). The 3Ј-primer is 5Ј-aaaacggaat-tccccgggTCAGTGTTTTAATGT-3Ј (the last 15 nucleotides correspond to residue 425 to the stop codon). PCR was carried out as described (21), and the amplified DNA was digested with BamHI and EcoRI and cloned into the same sites of pGEX-KG expression vector (Pharmacia). The resultant cDNA encodes a fusion protein composed of GST and a thrombin cleavage site, six histidines, and the catalytic domain (residues 101-428) of C2GnT. Escherichia coli HB101 was transformed with this plasmid vector, and a GST fusion protein was produced after isopropyl-1-thio-␤-D-galactopyranoside induction.
Purification of C2GnT Protein-HB101 cells were recovered by centrifugation and frozen at Ϫ80°C. After thawing on ice, the pellet was digested with 5 mg/ml lysozyme in 25 mM Tris-HCl, pH 8.0, containing 10 mM EDTA and 1% Triton X-100 (buffer A). After the addition of DNase I (Amersham Corp.), the sample was then sonicated and centrifuged. The resulting pellet was resuspended in 25 mM Tris-HCl, pH 8.0, containing 10 mM EDTA and 1.5% N-lauroylsarcosine (Sigma) (buffer B). The suspended residue was then centrifuged, and the sarcosyl extract was obtained as described (22). Glutathione-Sepharose beads were equilibrated with buffer B and added to the sarcosyl extract. The suspension was mixed gently at 4°C for 90 min using a rotary mixer and then briefly centrifuged to recover the beads. After washing the beads with buffer C (50 mM Tris-HCl, pH 8.0, containing 150 mM NaCl and 2.5 mM CaCl 2 ), the beads were suspended in 2 ml of buffer C containing 20 units of thrombin and mixed overnight at room temperature.
Thrombin-released material was recovered in the supernatant after centrifugation of the above mixture. The proteins that remained on beads were then released by SDS-polyacrylamide gel electrophoresis sample buffer, which contained no reducing reagent, and the C2GnT protein fragment was recovered in this extraction and separated from other contaminating proteins by SDS-polyacrylamide gel electrophoresis. The purified protein sample, extracted from polyacrylamide gels by electroelution, was immunized in rabbits. The antiserum was applied to a protein A-Sepharose column, bound antibodies were eluted with glycine-HCl, pH 2.5, and the eluent was immediately neutralized by the addition of 1.0 M Tris-HCl buffer, pH 8.0. The partially purified antibodies were further applied to a column of Sepharose 4B conjugated to E. coli proteins, and the unbound fraction was used as a purified antibody sample.
Construction of Vectors Expressing a Chimeric Protein Consisting of C2GnT and ST6Gal I-A cDNA encoding ST6Gal I was cloned by PCR using a human HL-60 cDNA library (20) as the template. The 5Ј-primer for this PCR corresponds to nucleotides Ϫ15 to 15 with respect to the translation initiation site (23) plus SmaI site. The 3Ј-primer is 5Ј-aaacccggctcgagTGCTTAGCAGTGAATGGTCC-3Ј. The XhoI site is underlined, while the italic type corresponds to nucleotides 1201-1221 (nucleotides 1216 -1218 encode the stop codon). The PCR product was digested with SmaI and XhoI and cloned into the same sites in the pMSG vector (Pharmacia). A cDNA encoding the cytoplasmic, transmembrane, and stem regions of ST6Gal I was amplified by PCR using the above plasmid vector as a template. The 5Ј-primer, DS23, corresponds to nucleotides Ϫ9 to 11 in relation to the translation initiation site of ST6Gal I, with the BamHI site at the 5Ј-end. The 3Ј-primer sequence was 5Ј-ATCACTACTAGGGTCCTGGGTGCTGCTT-3Ј. The first 12 nucleotides of this primer (shown by italics) correspond in antisense to residues 53-56 of C2GnT, and the last 16 nucleotides correspond in antisense to nucleotides 195-210 of ST6Gal I (nucleotides 196 -210 encode codons 66 -70). This PCR product encodes the first 70 amino acid residues of ST6Gal I plus 4 amino acids in the stem region of C2GnT.
A cDNA encoding the catalytic domain of C2GnT was amplified by PCR. The 5Ј-primer sequence was 5Ј-AGCACCCAGGACCCTAGTAGT-GATATTAATTG-3Ј. In this sequence, the first 12 nucleotides encode residues 67-70 of ST6Gal I, and the following 20 nucleotides encode residues 53-58 plus the portion of residue 59 of C2GnT. The 3Ј-primer, DS26, encodes the stop codon plus the following fifteen 3Ј-untranslated nucleotides of C2GnT sequence with the addition of the XhoI site.
The PCR products of the C2GnT catalytic domain and ST6Gal I sequence overlap at sequences corresponding to Ser-Thr-Gln-Asp-Pro-Ser-Ser-Asp, in which Ser-Thr-Gln-Asp comes from ST6Gal I and Pro-Ser-Ser-Asp comes from C2GnT. To make a chimera of the NH 2 -terminal region of ST6Gal I and the catalytic domain of C2GnT, PCR was carried out using DS23 and DS26 (shown above) as primers and a mixture of the above two PCR products as templates (21). After amplification under the same conditions as described, the PCR product was digested with BamHI and XhoI and then ligated into the same sites of pcDNAI, yielding pcDNAI-ST6Gal I/C2GnT.

Establishment of CHO Cells Stably Expressing C2GnT and C2GnT
Chimeric Protein-CHO DG44 cells were transfected with pZIPNEOleu alone, with pZIPNEO-leu and pcDNAI-C2GnT, or with pZIPNEOleu and pcDNAI-ST6Gal I/C2GnT using LipofectAMINE and were subsequently selected for G418 resistance. Clonal cell lines expressing a substantial amount of either leukosialin (CHO-leu) or both leukosialin and core 2 branched oligosaccharides (CHO-leu⅐C2GnT, CHO-leu⅐ST6Gal I/C2GnT) were selected as described (24).
Double Immunofluorescent Staining and Confocal Microscopy-CHO-leu⅐C2GnT and CHO-leu⅐ST6Gal I/C2GnT cells were grown on coverslips and fixed in 4% paraformaldehyde in PBS and immersed in 0.05% saponin, 0.1% bovine serum albumin solution in PBS for 10 min at room temperature. They were then incubated with rabbit anti-C2GnT antibodies followed by rhodamine-conjugated goat anti-rabbit IgG as described previously (25). After washing with PBS containing 0.1% bovine serum albumin, they were sequentially washed with PBS containing 1% normal goat serum, 10 g/ml unconjugated secondary antibody (goat anti-rabbit IgG), and then 100 g of unconjugated protein A/ml of PBS for 10 min each. The cells were then incubated with rabbit anti-␣-mannosidase II antibodies (26) followed by fluorescein isothiocyanate-conjugated goat F(abЈ) 2 fragment of IgG that is specific to the Fc portion of rabbit IgG (Axell). After washing with PBS containing 0.1% bovine serum albumin followed by PBS, the samples were visualized with a Zeiss Axioplan microscope (25) or Zeiss CSM410 confocal laser scanning microscope (27) as described. To detect C2GnT and ␤-1,4-galactosyltransferase in the same sample, pcDNAI-GalT (28) was transiently transfected in the above CHO cells. Simple double immunofluorescent staining was then carried out as described (25), since a mouse monoclonal antibody specific to human ␤-1,4-galactosyltransferase (29) was available. Controls were performed by omitting the primary antibodies.
Indirect Immunoperoxidase Staining for Electron Microscopy-Staining of specimens was performed as described previously (30,31). Briefly, frozen sections of human kidney specimens, CHO-leu⅐C2GnT and CHO-leu⅐ST6GalI/C2GnT cells, were fixed with paraformaldehydelysine-periodate, and incubation with primary antibody (rabbit antihuman C2GnT IgG, absorbed against E. coli proteins) was performed overnight at 4°C, followed by three washes in PBS containing 1% egg albumin and 0.075% saponin and incubation with secondary antibodies (goat horseradish peroxidase-conjugated anti-rabbit IgG, Amersham) for 1 h at room temperature. After three washes, bound antibodies were visualized with 0.05% diaminobenzidine (Sigma) and 3% H 2 O 2 in 20 mM Tris-HCl, pH 7.4. After fixation in 2.5% glutaraldehyde in 20 mM phosphate buffer, pH 7.4, and embedding in epon, ultra thin sections were cut and examined in a Jeol 1200 microscope. Controls were performed by either omitting the primary antibody or by replacing it with rabbit preimmune serum.
The CHO cells (ϳ1 ϫ 10 7 cells) were metabolically labeled with [ 3 H]glucosamine (10 Ci/ml), and the cell residues were subjected to Pronase digestion as described (24). The glycopeptides obtained were applied to a Sephadex G-50 (superfine) column (1.0 ϫ 110 cm) equilibrated with 0.1 M NH 4 HCO 3 . Higher molecular weight glycopeptides were subjected to ␤-elimination as described (24), and released Oglycans were separated from remaining glycopeptides using the same Sephadex G-50 gel filtration. The obtained O-glycans were analyzed by Bio-Gel P-4 gel filtration using the same conditions as described (24). Oligosaccharide peaks were desialyzed by clostridial neuraminidase, and the digest was again analyzed by Bio-Gel P-4 gel filtration. The standard oligosaccharides used were obtained as described previously (24). To obtain the ratio of the oligosaccharides synthesized, the amount of the radioactivity was determined after converting all of these oligosaccharides to Gal␤133GalNAcOH by various exoglycosidase treatments as described (24). After correcting the yield during chromatography, the relative amount of each oligosaccharide could be calculated (32). The ratio of the specific radioactivity of GlcNAc and GalNAcOH was found to be 1:0.7, as seen before (32).
Western Blot Analysis of C2GnT in Human Tissues-Membrane proteins were extracted in 200 mM Na 2 CO 3 , pH 10.5, and Triton X-114 phase partition as described previously (31). Twenty g of the membrane proteins were separated on SDS-polyacrylamide (10%) gel electrophoresis, transferred onto nitrocellulose, and probed with anti-C2GnT antibodies. Alkaline phosphatase-conjugated anti-rabbit IgG (Promega) was used as a secondary antibody and detected by nitro blue tetrazolium-5-bromo-4-chloro-3-indolyl phosphate chromogenic sub-strate (Kierkergard and Perry, Gaithersberg, MD). As a control, the duplicate blot was probed with IgG purified from rabbit preimmune sera.
Metabolic Labeling and Immunoprecipitation-Cells were metabolically labeled with Tran 35 S-label and immunoprecipitated as described (21).

Preparation of Antibodies Specific to
C2GnT-To determine the subcellular distribution of C2GnT, it was essential to produce antibodies specific to C2GnT. First, the catalytic domain of C2GnT was fused with GST protein and expressed in E. coli. The produced protein was then immunized in rabbits. After two additional boost immunizations, the titer of the antibodies was increased enough to detect C2GnT in CHO cells expressing C2GnT (Fig. 1A). The same antibodies also reacted with COS-1 cells, which transfected with C2GnT cDNA (Fig. 2B) but not with untransfected COS-1 cells (data not shown). Moreover, the antibodies did not react with CHO cells expressing I-branching ␤-1,6-N-acetylglucosaminyltransferase, which shares homology with C2GnT (33) (Fig. 1C). To confirm that the antibodies reacted with C2GnT, Western blot analysis was performed on the protein products used for immunization. Fig.  2A shows that the antibodies reacted with a fusion protein of ϳ68 kDa before thrombin digestion (lanes 1 and 2) and reacted with ϳ36-kDa protein after the digestion (lanes 3 and 4). The results are consistent with the calculated molecular mass for the GST-C2GnT fusion protein (ϳ68 kDa) and C2GnT catalytic domain (36 kDa).
Our preliminary studies on rat tissues showed that kidney had the highest activity of C2GnT. Western blot analysis of human kidney membrane proteins demonstrated that a ϳ57-kDa protein strongly reacted with the antibodies, while the control experiment gave negative results (Fig. 2B, lanes 5 and  6). Finally, immunoprecipitation of [ 35 S]methionine-labeled CHO cells stably expressing C2GnT produced a specific band at ϳ60 kDa, which was absent in wild type CHO cells (Fig. 2C,  lanes 7 and 8). These results combined clearly indicate that the antibodies generated are specific to C2GnT.
Localization of C2GnT in Golgi Complex-To determine the subcellular distribution of C2GnT, CHO cells were transfected with pcDNAI-C2GnT and pZipNeo-leu, and those cells stably expressing C2GnT and leukosialin (CHO-leu⅐C2GnT) were established (24). As shown previously, C2GnT is absent in CHO cells (24,34); thus, only introduced C2GnT can be detected in CHO cells. When C2GnT is localized differently by the replacement of the Golgi retention signal, such change should be clearly detected. Leukosialin was co-transfected, since those cells expressing core 2 oligosaccharides on leukosialin can be detected by T305 antibody (24).
CHO-leu⅐C2GnT cells (clone 1) were stained by rabbit antibodies specific to C2GnT followed by rhodamine-conjugated goat anti-rabbit IgG. After chasing the remaining antibodies by protein A, as detailed under "Experimental Procedures," the same specimens were incubated with rabbit antibodies specific to ␣-ManII, a glycosidase normally found in cis to medial-Golgi (26), followed by fluorescein isothiocyanate-conjugated goat anti-rabbit IgG. Preliminary experiments showed that the anti-bodies raised against mouse ␣-ManII cross-reacted with CHO ␣-ManII. As shown in Fig. 3 (top left), the majority of C2GnT and ␣-ManII are overlapping in their distributions, showing strong yellow staining.
In the second set of experiments, pcDNAI-human GalT cDNA (28) was transiently introduced into CHO-leu⅐C2GnT cells, and the expressed GalT, trans-Golgi enzyme, was similarly visualized by immunofluorescent staining. Since a mouse monoclonal antibody specific to human ␤-galactosyltransferase was available (29), the transfected cells were stained with rabbit anti-C2GnT antibodies and rhodamine-conjugated goat anti-rabbit IgG, followed by mouse monoclonal anti-GalT antibodies and goat FITC-conjugated anti-mouse IgG.
The results, as shown in Fig. 3 (bottom left) demonstrated that there was almost no overlap in the distribution of C2GnT and GalT. These results, combined, established that C2GnT is present in the cis to medial-Golgi. The results also demonstrated that two-step immunostaining was specific, since staining for only C2GnT or ␣-ManII can be seen in certain cells (Fig.  3). These results were confirmed by immunoelectron microscopy using specimens of human kidney and stably transfected CHO cells (data not shown).
Replacement of the Transmembrane Domain of C2GnT with That of ST6Gal I-The next question we asked was whether or not we could shift the Golgi localization of C2GnT by replacing the Golgi retention signal in C2GnT with that of a glycosyltransferase present in the trans-Golgi. For this, we utilized the sequence of ST6Gal I, which was shown to be in the trans-Golgi (35). We thus replaced the transmembrane portion and its flanking sequence of C2GnT with the corresponding sequence of ST6Gal I, resulting in ST6Gal I/C2GnT, as schematically shown in Fig. 4.
After transfecting CHO cells with ST6Gal I/C2GnT cDNA, different clones expressing various activities of C2GnT were chosen. We reasoned that overexpression of a chimeric protein may obscure the change in the subcellular localization brought about by replacing the Golgi retention signal. Fig. 5A shows that CHO cells containing the chimeric protein expressed amounts of C2GnT comparable with those expressed by CHO-leu⅐C2GnT cells as established previously (see clone M.B., es- It was possible, however that CHO cells expressing the chimeric protein expressed a substantially higher amount of C2GnT that had a lower specific activity. To exclude that possibility, the cell lysates derived from CHO-leu⅐C2GnT and CHO-leu⅐ST6Gal I/C2GnT were subjected to immunoprecipitation, and the residual C2GnT activity that remained in the supernatant was measured. As shown in Fig. 5B, the enzymatic activity in CHO cells expressing the wild type or chimeric C2GnT was precipitated in an almost identical manner, indicating that no detectable difference in the specific activity was present in these different CHO cells. The expression of the chimeric protein was then examined in CHO-leu⅐ST6Gal I/C2GnT by immunofluorescent staining. As seen in Fig. 3 (top right), the chimeric protein did not overlap with ␣-Man II. In contrast, the distribution of the chimeric protein and ␤-galactosyltransferase overlapped appreciably, as shown in Fig. 3 (bottom right). These results indicate that the replacement of the domain responsible for Golgi retention allowed the shift in Golgi localization of C2GnT to the trans-side.
Although we attempted to localize C2GnT and the chimeric protein in CHO cells by immunoelectron microscopy, only C2GnT could be strongly detected in single Golgi cisternae, whereas for the chimeric protein a diffused, weak signal was obtained (data not shown). Similarly, ST6Gal I introduced into CHO cells was not detected by immunoelectron microscopy in the previous studies, although its product was detected by a lectin, Sambucus nigra agglutinin (36). These results were obtained most likely due to the insufficient amounts of the expressed glycosyltransferases.
Effect of Altered Golgi Localization on Oligosaccharide Synthesis-We then tested if the altered localization of C2GnT affected the biosynthesis of O-glycans, resulting in altered O-glycan products. Mucin-type O-glycans produced from wild-type CHO cells were eluted in two peaks (a and b) after Bio-Gel P-4 gel filtration (Fig. 6A). After desialylation, peak a produced almost exclusively Gal␤133GalNAcOH (peak 3), and Gal␤133(Gal␤134GlcNAc␤136)GalNAcOH (peak 2) was barely produced (Fig. 6B). Moreover, peak b produced only Gal␤133GalNAcOH (Fig. 6C). The structures of these oligosaccharides were confirmed by exoglycosidase digestion followed by gel filtration as done previously (24). The analysis of peak a, which was derived from CHO-C2GnT cells (clone 8 in Fig. 5), exclusively produced Gal␤133(Gal␤134GlcNAc␤136)GalNAcOH, while peak b produced Gal␤133GalNAcOH as well (Fig. 6, E and F) ( Table I).
These results are consistent with the previous findings that wild-type CHO cells almost exclusively synthesize disialo or monosialo derivatives of Gal␤133GalNAc (Fig. 7, left). A large proportion of O-glycans were shifted to those containing core 2 branchings in CHO-C2GnT cells (Fig. 7, right). In contrast, CHO cells expressing the chimeric C2GnT expressed only a small amount of core 2 branched oligosaccharides (see peak 2 in Fig. 6H, which was derived from peak a, Fig. 6G). The majority of the oligosaccharides was either disialo (peak a in Fig. 6G) or monosialo (peak b in Fig. 6G) derivative of Gal␤133GalNAcOH (peak 3 in Fig. 6, H and I) (see Table I).
These results clearly indicate that the shifting of C2GnT from the cis-Golgi to the trans-Golgi converted the oligosaccharide biosynthesis as if the oligosaccharides were synthesized in the cells expressing a minimum amount of C2GnT. DISCUSSION In the present study, we have prepared polyclonal antibodies specific to C2GnT and studied the subcellular localization of C2GnT and its altered form by using the obtained antibodies. The antibodies reacted with GST-C2GnT fusion protein produced in E. coli and reacted with a 57-kDa protein in kidney cells and CHO-C2GnT cells (Fig. 2). This molecular mass is consistent with the calculated molecular mass (49,790 kDa), assuming that three N-glycan sites are utilized. The antibodies also stained the Golgi complex in CHO cells where C2GnT was expressed, but not with those expressing IGnT, demonstrating that the prepared antibodies are specific to C2GnT (Fig. 1).
By using the specific antibodies prepared, the subcellular localization of C2GnT was examined by light microscopy and immunoelectron microscopy. Both methods revealed that C2GnT was present in a relatively confined area, which appears to be cis to medial-cisternae of the Golgi complex (Fig. 3). Among glycosyltransferases involved in O-glycan synthesis, this is only the second enzyme of which subcellular localization was determined. Previously, ␣-N-acetylgalactosaminyltransferase, which adds N-acetylgalactosamine to a polypeptide precursor, was shown to be localized in cis-Golgi (18).
The localization of C2GnT in the cis to medial-Golgi is consistent with the previous findings on the biosynthesis of core 2 branched oligosaccharides. It has been shown that T cell activation is associated with the shifting of O-glycans from sialylated Gal␤133GalNAc to sialylated Gal␤133(Gal␤13 4GlcNAc␤136)GalNAc. In this particular system, activated T cells contain an increased amount of C2GnT activity and synthesize almost exclusively the hexasaccharides (see Fig. 7 also). The same cells synthesize only small amounts of the tetrasaccharide, which is the dominant O-glycan in resting T lymphocytes (8). This shift from the tetrasaccharide to the hexasaccharide in activated T cells can be well explained if C2GnT is present in earlier compartments of the Golgi complex than ␣-2,3-sialyltransferase that competes with C2GnT for the same acceptor (Fig. 7). If both enzymes are present in the same compartment, simple competition of the same acceptor site would provide a mixture of the tetrasaccharide and hexasac- In the present study, we then tested the possibility of altering the localization in the Golgi complex by altering the amino acid sequence responsible for the Golgi retention. It has been demonstrated in several laboratories that the transmembrane and its flanking sequences are critical as Golgi retention signals (28,(37)(38)(39)(40)(41). We thus replaced those sequences in C2GnT with ST6Gal I, since ST6Gal I was previously shown to be localized in the trans-Golgi (35). Examination of the chimeric protein, ST6Gal I/C2GnT, by confocal microscopy demon-strated that the chimeric protein resided in a relatively broad range of the Golgi complex, and its highest concentration could be observed in trans-Golgi (Fig. 3).
These results indicate that the transmembrane domain and its flanking sequence actually dictate the localization within the Golgi complex. Slightly broader distribution of the chimeric protein indicates, however, that other parts of the molecule such as a catalytic domain might also contribute to a proper localization within the Golgi complex. This may be related to the recent report that the kin recognition between medial-Golgi enzymes is dependent also on the catalytic domains (42). Our Resting T lymphocytes synthesize the tetrasaccharide (bottom left), while activated T lymphocytes synthesize the hexasaccharide due to the presence of core 2 ␤-1,6-N-acetylglucosaminyltransferase (boxed) (8). ␣-2,6-Sialyltransferase is also boxed. The subcellular distribution of ␣-N-acetylgalactosaminyltransferase (18) and core 2 ␤-1,6-N-acetylglucosaminyltransferases (the present study) has been established. The intra-Golgi compartmentation of the other glycosyltransferases needs to be established. It appears that chimeric ST6Gal I/C2GnT protein resides mostly in the trans-Golgi, similar to ST6Gal I. a These oligosaccharides were derived from peak a or b in Fig. 6, A, D, and G, and their sources are indicated in parentheses. b The ratio of the oligosaccharides were determined after exoglycosidase treatment to convert all of these oligosaccharides to Gal␤133GalNAcOH. chimeric protein consists of two segments derived from two glycosyltransferases that reside in either the cis-or the trans-Golgi. In the chimeric protein molecules, the combined signal derived from two enzymes is probably less optimal than that derived from two enzymes residing in the same location within the Golgi complex.
The structural analysis of O-glycans synthesized in CHO cells indicates that O-glycans containing ␣-2,6-linked sialic acid also contain ␣-2,3-sialic acid, and no structures such as NeuNAc␣236GalNAc and NeuNAc␣236(Gal␤133)GalNAc have been reported (24,34). These results strongly suggest that ␣-2,6-sialyltransferase in CHO cells requires NeuNAc␣23 3Gal␤133GalNAc as an acceptor (43,44), suggesting that ␣-2,6-sialyltransferase probably resides in the trans-cisternae of the Golgi complex. Consistent with the altered localization of the chimeric protein in the Golgi, CHO cells expressing the chimeric protein produced much less core 2 branched oligosaccharide than CHO cells expressing wild-type C2GnT (Table I). Before alteration, C2GnT was present in the earlier Golgi compartments than ␣-2,3-sialyltransferase, which forms the acceptor for ␣-2,6-sialyltransferase. This ␣-2,6-sialyltransferase competes for the same acceptor, C-6 of GalNAc (Fig. 7). After the alteration of the Golgi retention signal of C2GnT, the chimeric protein needs to directly compete with ␣-2,6-sialyltransferase, since C2GnT and ␣-2,3-sialyltransferase are present in overlapping regions of the Golgi (Fig. 7). Because of this situation, the amount of core 2 branched oligosaccharides was substantially reduced in CHO cells expressing the chimeric protein (Table I).
It has been shown repeatedly that tumor cells express aberrant O-glycans that are not synthesized under normal conditions (45,46). These aberrant expressions are often associated with the increase of a particular glycosyltransferase that is absent or present in low quantities under normal conditions. It is possible that overexpression of a particular glycosyltransferase under such pathological conditions may result in the mislocalization or wider distribution of the enzyme than under normal conditions. Further studies will be of significance to determine if some of the aberrant glycosylation under pathological conditions are due to aberrant localization of such a glycosyltransferase.