Location and Mechanism of α2,6-Sialyltransferase Dimer Formation

A significant proportion of the α2,6-sialyltransferase of protein Asn-linked glycosylation (ST6Gal I) forms disulfide-bonded dimers that exhibit decreased activity, but retain the ability to bind asialoglycoprotein substrates. Here, we have investigated the subcellular location and mechanism of ST6Gal I dimer formation, as well as the role of Cys residues in the enzyme's trafficking, localization, and catalytic activity. Pulse-chase analysis demonstrated that the ST6Gal I disulfide-bonded dimer forms in the endoplasmic reticulum. Mutagenesis experiments showed that Cys-24 in the transmembrane region is required for dimerization, while catalytic domain Cys residues are required for trafficking and catalytic activity. Replacement of Cys-181 and Cys-332 generated proteins that are largely retained in the endoplasmic reticulum and minimally active or inactive, respectively. Replacement of Cys-350 or Cys-361 inactivated the enzyme without compromising its localization or processing, suggesting that these amino acids are part of the enzyme's active site. Replacement of Cys-139 or Cys-403 generated proteins that are catalytically active and appear to be more stably localized in the Golgi, since they exhibited decreased cleavage and secretion. The Cys-139 mutant also exhibited increased dimer formation suggesting that ST6Gal I dimers may be critical in the oligomerization process involved in stable ST6Gal I Golgi localization.

The sialyltransferases are a large family of glycosyltransferases that function in the Golgi apparatus to transfer NeuAc from the sugar nucleotide donor, CMP-NeuAc, to terminal positions of N-linked and O-linked oligosaccharides of glycoproteins and the oligosaccharides of glycolipids. The sialylated oligosaccharides that are products of these enzymes' action have a variety of important roles in the normal cell and during development and disease (reviewed in Ref. 1). For example, sialylated oligosaccharides function as selectin ligands during inflammation and the homing of lymphocytes; act as receptors for viruses, toxins, and parasites; function in B cell maturation and activation; maintain glycoproteins in the circulation; and play roles in negatively modulating cell adhesion during development and oncogenesis (1).
Using ␣2,6-sialyltransferase (ST6Gal I) 1 as a model sialyltransferase, we have been investigating post-translational mechanisms that control sialyltransferase activity and function. For the ST6Gal I, these include the expression of two enzyme isoforms that differ in their localization and processing by cells (2) and the formation of a disulfide-bonded dimer (3). This disulfide-bonded dimer comprises ϳ20 -30% of the ST6Gal I found in liver Golgi. It exhibits reduced catalytic activity because it has little affinity for its sugar nucleotide donor, CMP-NeuAc (3). It does, however, retain the ability to bind galactose and asialoglycoprotein substrates and may be acting as a galactose-specific lectin in the Golgi.
The ST6Gal I isoforms appear to be generated by an RNA editing event in which an A is changed to a G, resulting in the replacement of a Tyr residue by a Cys residue at position 123 in the catalytic domain (2). The STtyr isoform is transiently localized in the Golgi and moves to a post-Golgi compartment, where it is cleaved and secreted into the extracellular space. The STcys isoform is stably localized in the Golgi and is not cleaved and secreted in most cell types. When highly expressed, the STtyr isoform is observed at low levels on the cell surface, whereas the STcys isoform is found in ER as well as the Golgi. Both isoforms are catalytically active when transfected into Chinese hamster ovary (CHO) cells that lack endogenous ST6Gal I enzyme. Both are also able to form disulfide-bonded dimers, suggesting that the "extra" Cys in the lumenal domain of the STcys isoform is not responsible for dimerization observed in the liver and in tissue culture (4). Interestingly, the difference in isoform Golgi retention reflects a difference in the ability of these two proteins to form insoluble oligomers (4). The stably localized STcys is quantitatively recovered as insoluble oligomers when Golgi membranes are solubilized at pH 6.3, the pH of the late Golgi, whereas only 13% of the transiently localized STtyr is found as insoluble oligomers under these conditions. A similar analysis of rat liver Golgi membranes, which contain both ST6Gal I isoforms, reveals that the disulfide-bonded dimer represents a significant proportion of the enzyme in the insoluble oligomer pellet when these membranes are solubilized at pH 5.8 -6.3 (4). This suggests that the disulfide-bonded dimer of either isoform may be a critical component in oligomer formation and Golgi localization. Three regions of homology or "sialyl motifs" have been identified in the sialyltransferase family (5)(6)(7)(8). Using a mutagenesis approach, Datta and Paulson (9) identified amino acids in the sialyl motif L (amino acids 178 -225 in rat ST6Gal I), including Cys-181, which are required for CMP-NeuAc donor binding. In addition, a second Cys residue, Cys-332, is found in the sialyl motif S (amino acids 318 -340 in rat ST6Gal I), a region shown by Datta et al. (10) to be involved in both asialoglycoprotein substrate and donor binding. More recently, these researchers demonstrated that these two Cys residues form a necessary intramolecular disulfide bond in the ST6Gal I monomer (11). This suggests that these Cys residues would not be involved in an intermolecular disulfide bond that would lead to dimer formation. However, the decreased ability of the ST6Gal I dimer to bind CMP-NeuAc suggests that disulfide bond formation between two monomers may occur between other Cys residues in the catalytic domain.
In this study, we investigated the location and mechanism of ST6Gal I dimer formation. Analysis of the time course of STtyr dimerization demonstrated that the dimer forms in the ER. To determine which Cys residues in the STtyr play roles in disulfide-bonded dimer formation, we have created a series of Cys mutants in this protein and have evaluated their cellular localization, in vivo activity, cleavage, and secretion, as well as their dimerization. We have found that mutagenesis of the single Cys residue in the transmembrane domain of the ST6Gal I (Cys-24) abolishes dimerization of the enzyme. Additional experiments rule out the possibility that secondary disulfide bonds form between Cys residues in the catalytic domains of two monomers. These results suggest that the close packing of two catalytic domains, rather than a direct obstruction of sialyl motifs L and S by the formation of an intermolecular disulfide bond, leads to the reduced ability of the ST6Gal I dimer to bind CMP-NeuAc.

Materials
Tissue culture media and reagents, including minimal essential medium (␣-MEM), Dulbecco's modified Eagle's medium (DMEM), Lipofectin, and LipofectAMINE were purchased from Life Technologies, Inc. Fetal bovine serum was obtained from Atlanta Biologicals (Norcross, GA). Sequenase enzyme was obtained from U. S. Biochemical Corp. Vent DNA polymerase was purchased from New England Biolabs (Beverly, MA). Fluorescein isothiocyanate (FITC)-conjugated goat anti-rabbit IgG was purchased from EY Laboratories (San Mateo, CA). FITCconjugated Sambucas nigra agglutinin (SNA) lectin was purchased from Vector Laboratories (Burlingame, CA). The QuickChange sitedirected mutagenesis kit was purchased from Stratagene (La Jolla, CA). Horseradish peroxidase-conjugated goat anti-rabbit or anti-mouse IgGs were purchased from Jackson Immunoresearch Laboratories, Inc. (West Grove, PA
Protein Expression in Tissue Culture Cells-COS-1 and CHO cells maintained in either DMEM, 10% fetal bovine serum (COS-1) or in ␣-MEM, 10% fetal bovine serum (CHO) were plated on 100-mm tissue culture dishes and grown in a 37°C, 5% CO 2 incubator until 50 -70% confluent. Lipofectin or LipofectAMINE transfections were performed according to protocols provided by Life Technologies, Inc. Briefly, 30 l of Lipofectin or LipofectAMINE was incubated with 1.5 ml of Opti-MEM medium for 40 min at room temperature in a polystyrene tube. Twenty micrograms of ST6Gal I DNA was mixed with 1.5 ml of Opti-MEM medium, added to the Lipofectin or LipofectAMINE solutions, and incubated at room temperature for 15 min. Cells to be transfected were washed with Opti-MEM medium and the transfection mixture added to the tissue culture dishes. Cells were incubated with the DNA-Lipofectin/LipofectAMINE transfection solution (3 ml) for 6 h at 37°C in a 5% CO 2 incubator. Ten milliliters of DMEM or ␣-MEM containing 10% fetal bovine serum was directly added to the cells following the 6-h incubation and cells were incubated for another 16 -48 h. Transfection of cells plated on coverslips was essentially the same, except 250 -500 ng of DNA, 2 l of Lipofectin, or LipofectAMINE in 250 l of Opti-MEM were used per coverslip.
Immunofluorescence Localization and SNA Lectin Staining-COS-1 or CHO cells were plated onto glass coverslips and transfected with cDNAs encoding wild type or mutant forms of the STtyr isoform of ST6Gal I subcloned into the pSVL (Amersham Pharmacia Biotech) expression vector and processed for immunofluorescence microscopy, as described previously (2,12). For immunolocalization of ST6Gal I proteins in COS-1 cells or lectin/immunolocalization double staining in CHO cells, transfected cells were fixed and permeabilized in Ϫ20°C methanol for 10 min. For cell surface SNA lectin staining of CHO cells, transfected cells were fixed using 3% paraformaldehyde in PBS for 8 min. For immunolocalization, cells were washed with PBS, blocked with 5% goat serum in PBS for 1 h at room temperature, and incubated with a 1:100 dilution of an affinity-purified anti-ST6Gal I antibody in 5% goat serum in PBS for 1 h at room temperature. Following four PBS washes, cells were incubated with a 1:100 dilution of FITC-conjugated goat anti-rabbit IgG antibody for 45 min at room temperature. Cells were again washed four times with PBS before mounting on glass slides using ϳ20 l of mounting media (15% (w/v) Vinol 205 polyvinyl alcohol, 33% (v/v) glycerol, 0.1% sodium azide, pH 8.5, in PBS). Cells were visualized and photographed using a Nikon Axiophot microscope equipped with epifluorescence illumination and a 60ϫ oil immersion Plan Apochomat objective. For SNA staining, cells were not subjected to a blocking step but directly incubated with a 1:200 dilution of FITCconjugated SNA in PBS for 1 h at room temperature. Following four PBS washes, coverslips were mounted, visualized, and photographed as described above. For double lectin staining/immunolocalization in CHO cells, following fixation and permeabilization, cells were washed with PBS, incubated with a 1:200 dilution of FITC-SNA for 1 h, washed four times in PBS, and then processed for immunolocalization as described above beginning with the 5% goat serum in PBS blocking step.
Metabolic Labeling of Cells and Immunoprecipitation of ST Proteins-Pulse-chase analysis and immunoprecipitation were performed as described previously (2,12). Transfected COS-1 cells were incubated with methionine-and cysteine-free DMEM for 1 h. Medium was removed and replaced with 3 ml of fresh methionine-free DMEM containing 100 Ci/ml EXPRE 35 S 35 S protein labeling mix, and cells labeled in a 37°C, 5% CO 2 incubator for 30 min to 1 h. Radioactive medium was then removed, cells were extensively washed, 4 ml of DMEM plus 10% fetal bovine serum was added, and cells were incubated in the CO 2 incubator for 6 h. Cell medium was collected, and cells were extensively washed with PBS and lysed in immunoprecipitation buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM EDTA, 0.5% Nonidet P-40, 0.1% SDS, 100 mM iodoacetamide) containing protease inhibitors. ST6Gal I proteins were immunoprecipitated from both cell lysates and media using the anti-ST6Gal I antibody and protein A-Sepharose as described previously (2,3). Immunoprecipitated proteins were analyzed by SDSpolyacrylamide gel electrophoresis and fluorography (2,3). Bio-Rad prestained broad range gel standards were used to estimate molecular mass: myosin, 203 kDa; ␤-galactosidase, 116 kDa; bovine serum albumin, 83 kDa; ovalbumin, 48.7 kDa; carbonic anhydrase, 33.4 kDa; soybean trypsin inhibitor, 28.2 kDa; lysozyme, 20.7 kDa; aprotinin, 7.6 kDa.
Dimerization Assay-COS-1 cells were transfected with wild type STtyr-pSVL or Cys mutants as described above. The cells were then trypsinized, washed once with PBS, and suspended in 0.25 M sucrose, 10 mM Tris-HCl, pH 7.4, 1 mM magnesium acetate, and 100 mM iodoacetamide (one volume of cell pellet per five volumes of homogenizing medium). Equilibrium sucrose density gradients were used to isolate a Golgi-enriched fraction according to the method described by Xu and Shields (13). All the sucrose gradient solutions contain 100 mM iodoacetamide to avoid aberrant disulfide bond formation during the membrane preparation procedures (see Ref. 3). The Golgi membrane-enriched fraction was collected, washed once with 100 mM iodoacetamide and pelleted by centrifugation at 39,000 rpm for 1 h. The pellets were then resuspended in phosphate buffer, pH 7.0, 10% glycerol and stored at Ϫ80°C. To analyze dimerization of the STtyr and its Cys mutants, the Golgi enriched membrane fractions from cells expressing these proteins were treated with Laemmli sample buffer with or without 10% ␤-mercaptoethanol at 100°C for 10 min. Samples were electrophoresed on a 10% SDS-polyacrylamide gel and then transferred to a nitrocellulose membrane for immunoblotting. Immunoblot analysis was performed using the rabbit anti-rat ST6Gal I antibody and a goat anti-rabbit horseradish peroxidase-conjugated secondary antibody. Visualization of bands was achieved using the SuperSignal West Pico chemiluminescence reagent (Pierce).
Affinity Chromatography-Asialofetuin-Sepharose 4 FF was prepared by cross-linking asialofetuin (Sigma) with CNBr-activated Sepharose 4 FF (Amersham Pharmacia Biotech) according to the protocol described by the manufacturer. About 20 mg of asialofetuin was cross-linked to 1 ml of CNBr-activated Sepharose 4 FF. CDP-hexanolamine-agarose was a gift Dr. Gerald Hart (The Johns Hopkins University, Baltimore, MD). Myc-tagged versions of STtyr, C350S, and C361S mutants in the pcDNA 3.1 vector were expressed in COS-1 cells and tested for their affinity for donor and substrate using the affinity chromatography methods described previously (3) and described below.
CDP-hexanolamine-agarose Affinity Chromatography-Each plate of transfected COS cells was lysed with 1 ml of 0.5% Triton X-100 and homogenized by passing through a 25-gauge needle four times. Cell lysates were diluted in 9 ml of Buffer E (10 mM sodium cacodylate, pH 6.5, 0.1% Triton CF-54, 0.25 M NaCl) and applied to a 1-ml CDPhexanolamine-agarose affinity column. The column was washed with 2 ml of Buffer E and 4 ml of Buffer H (20 mM sodium cacodylate, pH 5.3, 0.1% Triton CF-54, 0.15 M NaCl), and eluted with Buffer H containing 5 mM CDP. Fractions of 800 l were collected.
Asialofetuin-Sepharose FF Affinity Chromatography-Medium of transfected CHO cells was collected for the asialofetuin binding assay. The conditioned medium was incubated with asialofetuin-Sepharose FF beads for 1 h at 4°C. The beads were washed extensively with Buffer E and Buffer H, and ST6Gal I proteins were eluted with 0.5 M galactose in Buffer H. Proteins were recovered by immunoprecipitation using the rabbit anti-rat ST6Gal I and then detected by immunoblotting with a mouse anti-Myc antibody, a goat anti-mouse horseradish peroxidaseconjugated antibody, and the SuperSignal West Pico chemiluminescence reagent, as described above.

ST6Gal I Disulfide-bonded Dimer Formation Occurs in the
ER-Under steady state conditions, we detect the ST6Gal I disulfide-bonded dimer in liver Golgi membranes, but not in significant quantities in liver ER membranes (3). Since disulfide bond formation usually occurs in the ER, this could reflect the formation of the disulfide-bonded dimer in the ER and its rapid transport to the Golgi. Interestingly, Bosshart and Berger (14) showed that they were first able to detect a higher molecular weight form (ϳ97 kDa) of the endogenous ST6Gal I in H-4-II-E rat hepatoma cells after 2 h of chase, suggesting that a dimer of the enzyme is formed in the Golgi. However, these researchers commented that their recovery of this form of the enzyme was at times erratic. Their inability to consistently recover the ST6Gal I dimer could have compromised their ability to detect this form of the protein at earlier time points. To evaluate where the ST6Gal I disulfide-bonded dimer is formed in the COS-1 cell expression system, we performed a pulse-chase analysis in cells transiently expressing the STtyr isoform. Cells were labeled with EXPRE 35 S 35 S protein labeling mix for 30 min and chased for 0 -180 min. STtyr proteins were immunoprecipitated from cell lysates at each time point and analyzed under non-reducing and reducing conditions by SDSpolyacrylamide gel electrophoresis. In this experiment and the others described in this report, we included 100 mM iodoacetamide in all buffers to prevent aberrant disulfide bond formation (see Ref. 3). Under nonreducing conditions, in addition to the expected 47-50-kDa STtyr band, a band of ϳ100 kDa was observed after only 30 min of labeling and reflected 15% of the total STtyr protein (Fig. 1). This 100-kDa band disappeared when samples were treated with reducing agents, arguing that it represented a disulfide-bonded dimer of the STtyr. In addition, this dimer persisted throughout the chase period and only decreased in intensity at the longest chase time point (180 min), coincident with a decrease in the monomer, and likely reflecting the enzyme's cleavage and secretion. Increases in the molecular mass of the monomer band were observed between 60 and 180 min of chase and undoubtedly reflected the terminal glycosylation of the enzyme's N-linked oligosaccharides in the Golgi (Fig. 1). This was verified by a parallel experiment in which samples were digested with endoglycosidase H, an enzyme that specifically cleaves high mannose, but not complex N-linked oligosaccharides and can be used to monitor transport into the Golgi where N-linked oligosaccharides are modified to complex forms (15, 16) (data not shown). In a subsequent identical experiment, we were able to successfully repeat these results, and we also demonstrated that an ER retained/retrieved form of the STtyr, Iip33-STtyr (17), was capable of forming dimer, again suggesting that the STtyr dimer is forming within the ER. Taken together, these results suggest that the disulfide-bonded dimer of the STtyr forms in the ER of COS-1 cells and is rapidly transported to the Golgi along with the monomeric enzyme, where both forms of the enzyme are cleaved within the stem region and secreted.
Replacement of Cys-24, Cys-181, and Cys-332 Eliminates the Dimerization of the ST6Gal I Protein-To determine which Cys residues are involved in the dimerization of the ST6Gal I STtyr isoform, we individually replaced one Cys residue in the transmembrane domain (Cys-24) and six Cys residues in the catalytic domain (Cys-139, -181, -332, -350, -361, and -403) with either Ala or Ser residues (Fig. 2). Wild type and mutant proteins were transiently expressed in COS-1 cells and Golgienriched membranes were isolated in the presence of 100 mM iodoacetamide according to the method of Xu and Shields (13). Aliquots of membrane preparations were electrophoresed under reducing and non-reducing conditions on SDS-polyacrylamide gels and STtyr proteins detected by immunoblotting with the antibody specific for the ST6Gal I (3). In four separate experiments, we found that replacement of Cys-24 (C24A), Cys-181 (C181A (shown) or C181S (not shown)), and Cys-332 (C332A) eliminated detectable disulfide-bonded dimer formation (Fig. 3). Replacement of Cys-350 (C350S) and Cys-403 (C403S) did not significantly influence dimer formation, whereas replacement of Cys-361 (C361S) slightly decreased the amount of dimer. Notably, replacement of Cys-139 (C139S) appeared to significantly enhance dimer formation, and a higher molecular mass band was also observed in addition to the dimer when Cys-403 was replaced (C403S).
Our inability to detect dimers of the C24A, C181A/S, and C332A mutants could reflect the absence of Cys residues that are involved in dimer formation or an inability of the proteins to form dimers because they are generally misfolded. Strikingly, both the Cys-181 and Cys-332 mutants were found in the Golgi-enriched membranes in lower amounts relative to the other mutant STtyr proteins, suggesting that these proteins may be misfolded and retained to some extent in the ER (Fig.  3). In addition, the lower amounts of these proteins in the Golgi-enriched membranes could have also compromised our ability to detect dimers if they were formed. This latter possibility was ruled out by additional experiments in which no dimers of the C181 and C332 mutants were detected even when greater amounts of these two mutant proteins were analyzed (data not shown). However, to more definitively identify the Cys residue or residues involved in dimer formation, we needed to evaluate the possibility that some or all of the STtyr Cys mutants were misfolded and/or mislocalized. To do this, we evaluated the localization of these proteins, as well as their catalytic activity and protein processing as measures of proper folding.
Localization of ST6Gal I Cys Mutants: Replacement of Cys-181 and Cys-332 Leads to ER Retention-To determine whether C24A, C181S/A, and/or C332A mutants, as well as the other STtyr Cys mutants, were properly transported to the Golgi, we transiently expressed each mutant in COS-1 cells and performed indirect immunofluorescence microscopy using the anti-ST6Gal I antibody, as described previously (2). Immunoelectron microscopy with ST6Gal I-specific antibodies has demonstrated that the endogenous ST6Gal I is localized in the trans-cisternae of the Golgi and the trans-Golgi network of rat liver hepatocytes (18). In addition, we have previously demonstrated that the both isoforms of the ST6Gal I enzyme are localized to the Golgi when expressed in COS-1 cells (2,19). In these studies we found that, like the wild type STtyr, Ser, or Ala mutants in Cys-24, -139, -350, -361, and -403 were localized in the Golgi (Fig. 4). In contrast, and as suggested by the results in Fig. 3, mutants in Cys-181 and Cys-332 were predominately found in reticular structures reminiscent of ER. These results suggested that the C181A and C332A mutants may have folded into forms that are partially or completely transport incompetent. This is consistent with the results of Datta et al. (11), who showed that these mutant proteins are predominantly localized in the ER and inactive in an identical in vivo activity assays. Based on additional experiments by these investigators, it is likely that the behavior of the C181A and C332A mutants reflects the absence of a critical Cys-181-Cys-332 intramolecular disulfide bond. In contrast, the C24A mutant was transported efficiently to the Golgi and yet still was not capable of forming the disulfide-bonded ST6Gal I dimer, suggesting that it plays a direct role in dimer formation.
Activity of ST6Gal I Cys Mutants: Replacement of Cys-24 or Cys-181 Does Not Eliminate ST6Gal I Activity, whereas Replacement of Cys-332, Cys-350, or Cys-361 Inactivates the Enzyme-The results described above suggest that the disulfide bond that is responsible for the dimerization of the ST6Gal I protein forms between the Cys-24 residues in the transmembrane regions of two ST6Gal I monomers. Interestingly, dimer formation seems to be influenced by alterations of Cys residues in the catalytic domain. For example, replacement of Cys-181 or Cys-332 eliminates detectable disulfide bond formation, whereas replacement of Cys-139 enhances disulfide bond formation. This implies that catalytic domain conformation is critical for disulfide bond formation between Cys residues in two ST6Gal I transmembrane regions. Another possibility, although unlikely, is that a replacement of Cys-24 in the transmembrane domain causes a misfolding of the catalytic domain of the protein that does not prevent its transport between the ER and Golgi, but does compromise dimer formation. As another measure of the correct folding of the C24A mutant and other Cys mutants, we analyzed their activity in vivo. Catalytic activity was assessed by transiently transfecting CHO cells, which lack endogenous ST6Gal I enzyme, with cDNAs encoding FIG. 2. ST6Gal I structure and Cys mutants. The ST6Gal I is a type II membrane protein with a short 9-amino acid cytoplasmic tail, a 17-amino acid transmembrane region (TMD), an ϳ70-amino acid stem region, and a large lumenal carboxyl-terminal catalytic domain (12,31). Within the catalytic domain are three region of sialyltransferase homology termed sialylmotifs (L-SM, S-SM, and VS-SM) (5)(6)(7)(8). The Cys residues found in the rat ST6Gal I STtyr sequence are indicated. the wild type STtyr or the STtyr Cys mutants, and staining for the presence of cell surface ␣2,6-sialylated glycoconjugates using FITC-conjugated SNA lectin (2,12). This assay not only requires that the enzyme is catalytically active, but also that it is targeted to the correct cisternae of the Golgi in order to interact with the CMP-NeuAc donor and the appropriate asialoglycoprotein substrates. We found that mutants with replacements of Cys-24, -139, -181, and -403 were active when transfected into CHO cells (Fig. 5). The demonstrated in vivo activity of the C24A, C139S, and C403S mutants confirmed their correct folding as well as their Golgi localization. The inactivity of the C332A mutant in this in vivo assay was not surprising since this mutant protein appears to be retained in the ER. In addition, in vitro assays verified the inactivity of this mutant (data not shown). In contrast, the activity of C181A mutant was surprising, based on its predominant localization in the ER (see Figs. 4 and 5). Double staining for both SNA reactivity and protein localization showed that only a very few cells expressing the C181A mutant protein (ϳ1%) actually stained with SNA-FITC. This demonstrated that, although the majority of the C181A mutant was localized in the ER, a small proportion of this protein appeared to be transported into the Golgi where it was able to sialylate proteins (data not shown).
More striking was the lack of activity of the C350S and C361S mutants since these proteins appeared to be properly localized in the Golgi. One possibility is that these Cys residues are critical for enzyme activity since they are localized adjacent to the VS sialyl motif. Another possibility was that these mu-tant proteins were transported only as far as the early Golgi and never entered the later Golgi compartments, where they could encounter substrates and CMP-NeuAc donor.
Comparison of the Cleavage and Secretion of ST6Gal I Cys Mutants: Cys-139 and Cys-403 Are Required for Efficient Cleavage and Secretion-To evaluate the possibility that the C350S and C361S mutants and any of the other ST6Gal I Cys mutants were not trafficking like the wild type STtyr protein, we analyzed their patterns of cleavage and secretion. The ST6Gal I STtyr isoform is transiently retained in the Golgi and is ultimately cleaved and secreted from COS-1 cells with a half-time of 3-6 h (2). This cleavage event occurs after the enzyme moves through the Golgi into a post-Golgi compartment (2). If the Golgi-localized C350S and C361S mutants were not cleaved and secreted, this would suggest that they may be localized too early in the Golgi to be functional. To determine whether these and the other STtyr Cys mutants were transported to and through the Golgi like the wild type STtyr protein, we compared their rates of cleavage and secretion. STtyr, STcys, and STtyr Cys mutants were transiently expressed in COS-1 cells, labeled for 1 h with EXPRE 35 S 35 S protein labeling mix, and chased for 6 h with medium containing unlabeled amino acids. The sialyltransferase proteins were immunoprecipitated from both cell lysates and medium fractions and immunoprecipitates analyzed by SDS-polyacrylamide gel electrophoresis and fluorography. We found that C181A and C332A mutants were not cleaved and secreted as expected, based on their predominant localization in the ER (Fig. 6). The C24A, C350S, and

FIG. 4. Localization of ST6Gal I Cys mutants in COS-1 cells.
STtyr and the Cys mutants were transiently expressed in COS-1 cells. Cells were fixed and permeabilized with Ϫ20°C methanol and STtyr proteins detected using rabbit anti-rat ST6Gal I antibodies and FITCconjugated goat anti-rabbit IgG secondary antibodies, as described under "Experimental Procedures." Cells were visualized using a Nikon Axiophot microscope equipped with epifluorescence illumination. Shown are representative micrographs from four independent experiments. Original magnification, ϫ750.

FIG. 5. Comparison of the catalytic activity of the STtyr and its Cys mutants following expression in CHO cells. The STtyr and the
Cys mutants were expressed in CHO cells that lack endogenous ST6Gal I protein. Cells were fixed with 3% paraformaldehyde and incubated with a 1:200 dilution of FITC-conjugated SNA lectin in PBS to detect the presence of glycoconjugates terminating with ␣2,6-sialic acid on the cell surface. Cells were visualized using a Nikon Axiophot microscope equipped with epifluorescence illumination. These in vivo activity assessments were repeated a total of four times by two different investigators (C. Chen and R. Qian). Original magnification, ϫ750.
C361S proteins were cleaved and secreted to the same extent as the wild type STtyr (Fig. 6). Consequently, aberrant transport through the Golgi could not explain the lack of in vivo activity of the C350S and C361S mutant proteins. However, we did observe a 53% reduction in the cleavage and secretion of the C403S mutant (24% found in the medium versus 51% for the wild type STtyr), and a 64% reduction in the cleavage and secretion of the C139S mutant (18.5% found in the medium versus 51% for the wild type STtyr) (Fig. 6). The C403S and C139S mutant proteins were localized in the Golgi and functional in vivo, but they were not cleaved and secreted at the same rate as the wild type STtyr. These results suggest that these mutant proteins may be more stably localized in the Golgi compared with the wild type STtyr and consequently behave more like the STcys isoform. Notably, there is an additional high molecular weight band formed by the C403S mutant protein, and an increase in the amount of dimer formed by the C139S mutant protein (see Fig. 3), suggesting that additional disulfide-based interactions or conformational changes in the catalytic domain impact the Golgi localization of these proteins. Interestingly, the position of the Tyr/Cys change is found at residue 123 not far from Cys-139, and we have identified other mutations in this area that completely or partially block the cleavage and secretion of the STtyr protein (4). It is tempting to speculate that the formation of more stable oligomers by C139S and C403S mutant proteins may enhance their localization in the Golgi and prevent them from efficiently moving to a post-Golgi compartment for cleavage (see Ref. 4).
No Disulfide-bonded Forms of Cleaved and Secreted STtyr or Its Cys Mutants Are Observed in COS-1 Cell Medium-It is possible that the increase we observe in C139S dimer formation does not simply reflect a change in catalytic domain conformation, but the formation of additional intermolecular disulfide bonds between Cys residues in the catalytic domains of two C139S monomers. To evaluate this possibility and the possibility of secondary disulfide bonds occurring between Cys residues in the catalytic domains of any of the Cys mutants or the wild type STtyr, we investigated whether the cleaved and secreted forms of these proteins existed as dimers in the cell medium. If the only disulfide bond involved in dimer formation occurs between Cys-24 residues in the transmembrane regions of two ST6Gal I monomers, then one would expect that cleavage within the stem region of these molecules would lead to monomeric soluble forms. Conversely, if additional intermolecular disulfide bonding occurs between catalytic domain Cys residues, then higher molecular weight "dimerized" forms would be secreted. COS-1 cells expressing the STtyr and its Cys mutants were metabolically labeled for 1 h and chased with medium containing unlabeled amino acids for 6 h. Secreted proteins were immunoprecipitated from cell media and electrophoresed under non-reducing conditions on SDS-polyacrylamide gels. In two separate experiments, we found that only monomeric species were observed in the medium of cells expressing the wild type STtyr and all the Cys mutants (Fig. 7). Interestingly, C139S and C403S proteins, which demonstrate decreased rates of cleavage and secretion relative to the other proteins, were clearly migrating with a slightly higher molecular weight than the other secreted proteins. This suggested that the likely alterations in the conformation of these proteins' catalytic domains are not only impacting their localization in the Golgi, but also the site of cleavage in the stem region.
Taken together, the data described above demonstrate that the replacement of Cys-24 in the ST6Gal I transmembrane domain eliminates disulfide-bonded dimer formation without altering protein localization, activity, or processing (please see Table I for a summary of the results). The secondary participation of other Cys in the catalytic domain in the dimerization process is unlikely since no disulfide-bonded, cleaved, and secreted forms of the STtyr or any of the Cys mutants are observed in the cell culture medium. In addition, the data also suggest that the replacement of Cys residues in the catalytic domain of the STtyr does in some cases alter the efficiency of dimer formation, presumably by altering the conformation of the catalytic domain and the association of enzyme monomers.
The gested that the only disulfide bond formed in the ST6Gal I was found between Cys-181 and Cys-332. However, we found that replacing either Cys-350 or Cys-361 with Ser led to an inactive enzyme. This suggested that, if these residues did not form a critical intramolecular disulfide bond, then they may be present in the active site of the enzyme and critical for CMP-NeuAc donor or asialoglycoprotein substrate binding. To investigate this possibility, we used CDP-hexanolamine-agarose affinity chromatography and asialofetuin-Sepharose affinity chromatography to evaluate the relative affinity of wild type STtyr and the C350S and C361S mutants for sugar nucleotide donor and asialoglycoprotein substrate, respectively. STtyr, C350S, and C361S coding sequences were inserted into the pcDNA 3.1 vector that fuses a Myc epitope tag at the carboxyl terminus of each protein and these proteins expressed in either COS-1 or CHO cells. Proteins found in both cell lysates and medium fractions were applied to each affinity column and eluted with similar profiles. Shown in Fig. 8 (A and B) are a CDP-hexanolamine-agarose column profile for sialyltransferase proteins from COS-1 cell lysates and a asialofetuin-Sepharose column profile for sialyltransferase proteins from CHO cell medium (these are representative of two independent experiments for each protein from both cell lysates and medium of expressing cells). Following standard wash and elution conditions (described under "Experimental Procedures"), sialyltransferase proteins were immunoprecipitated using anti-ST6Gal I antibodies, separated by SDS-polyacrylamide gel electrophoresis, and detected by immunoblotting with the anti-Myc antibodies. It is clear that, although wild type STtyr bound strongly to CDP-hexanolamine-agarose and was specifically eluted with 5 mM CDP, the C350S and C361S mutants showed no affinity or weak affinity for this matrix (Fig. 8A). The C350S mutant protein was found predominantly in the flow-through fraction, demonstrating its lack of binding to the CDP-hexanolamineagarose. The C361S protein was progressively removed from the column by the two wash buffers, and any residual protein was ultimately eluted by the 5 mM CDP elution, demonstrating its very weak affinity for CDP-hexanolamine-agarose. Likewise, the C350S and C361S mutants demonstrated a weaker binding to asialofetuin-Sepharose relative to the wild type STtyr protein (Fig. 8B). These results suggested that the lack of catalytic activity of the C350S and C361S mutants was due to a combined lack of affinity for both donor and substrate and imply that these amino acids are positioned in or near the active site of the enzyme. DISCUSSION The experiments described in this report show that the disulfide-bonded dimer of the ST6Gal I is formed in the ER and that it forms between Cys-24 residues in the transmembrane regions of two adjacent ST6Gal I proteins. Interestingly, evaluation of the dimerization of other Cys mutants showed that replacement of selected Cys in the enzyme's catalytic domain leads to conformational changes that decrease/eliminate dimer formation (Cys-181 and Cys-332) or enhance dimer formation (C139S). We have also found that several of the ST6Gal I Cys residues are critical for its activity. First, the replacement of either Cys-181 or Cys-332 cause the enzyme to fold into trans-port incompetent forms that are either minimally active or inactive, respectively. Second, the replacement of either Cys-350 or Cys-361 with Ser appears to inactive the enzyme by decreasing the affinity for CMP-NeuAc donor and asialoglycoprotein substrates.
Datta and colleagues (9,10) have shown that the sialyl motif L, including Cys-181, is critical for CMP-NeuAc binding, whereas sialyl motif S, including Cys-332, is critical for CMP-NeuAc and asialoglycoprotein binding. Their recent work suggested that Cys-181 and Cys-332 form an essential disulfide  bond (11). They found no evidence for other disulfide bonds in the catalytic domain, implying that Cys-139, -350, -361, and -403 exist as free sulfhydryls. In contrast, Angata et al. (20) demonstrated that the PST polysialyltransferase (ST8Sia IV) has two critical intramolecular disulfide bonds. One is between Cys-142 and Cys-292 linking the L and S sialyl motifs and a second between Cys-156 and Cys-356, linking the L sialyl motif to the very C terminus of the enzyme. As we have confirmed in our own work, Datta et al. (11) found that replacement of Cys-181 or Cys-332 in the membrane-associated enzyme leads to misfolded proteins that were largely retained in the ER. Although Datta et al. (11) showed that C181A and C332A proteins were inactive in in vitro assays (isolated soluble mutant forms) and in the in vivo CHO/SNA staining assay (membrane-associated mutant forms), we detected some minimal activity in the latter assay for our C181S (data not shown) and C181A mutants. It is difficult to explain the differences in our results. We have shown previously, using glycosylation mutants of the ST6Gal I, that forms of the enzyme that exhibit no activity in vitro can be active in vivo (12). We attributed this to the aggregation of specific mutant forms in the in vitro assay. It is also possible that differences in the results of our in vivo assays and those of Datta et al. (11) are simply differences in level of detection of SNA positive material. In any case, the ST6Gal I Cys-181-Cys-332 disulfide bond is certainly required for efficient transport from ER to Golgi and maximum catalytic activity of the enzyme. Replacement of Cys-350 or Cys-361 with Ser inactivated the enzyme in the in vivo assay (Fig. 5). Additional analyses (Fig.  8) suggested that these mutants were compromised in binding to both CMP-NeuAc donor and asialoglycoprotein substrates. The analysis of Datta et al. (11) implies that these two relatively close Cys residues do not form a disulfide bond, as we had initially suspected from our results. Alternatively, they may be part of the active site of the enzyme. Recent preliminary results demonstrate that mutants in which these two Cys are individually replaced with Ala, a double C350S/C361S mutant, or a double C350A/C361A mutant, are all catalytically active in the in vivo assay. 2 This suggests that, if Cys-350 and Cys-361 are in a region that forms part of the active site, this region is exquisitely sensitive to the nature of the side chains at these two positions.
A number of other Golgi glycosyltransferases have been shown to form disulfide-bonded dimers. These include the GM 2 synthase (21,22), ␣1,2-mannosidase (23), ␣1,2-fucosyltransferase (H enzyme) (24), ␣1,3-fucosyltransferase VI (FucT VI) (25), and the Gal␤1,3-glucuronosyltransferase (26). Unlike the dimers formed by the ST6Gal I, the dimers of most of these enzymes are catalytically active and are generated by disulfide bonds between Cys residues in their stem regions and/or catalytic domains. For example, Young and colleagues (21,22) showed that disulfide-bonded dimers of the GM 2 synthase form in the ER, that the disulfide bonds are formed between the catalytic domains of two monomers, and that the active form of the enzyme is a disulfide-bonded dimer. More recent work from these researchers showed that these intercatalytic domain disulfide bonds are formed in an anti-parallel manner between Cys-80 and Cys-82 with Cys-412 and Cys-529 (27).
Like the ST6Gal I, FucT VI is found as a combination of disulfide-bonded dimers and monomers in cells (25). ST6Gal I dimers constitute from ϳ15-30% of the total enzyme in tissue culture cells and rat liver Golgi (see Ref. 3 and Fig. 1). Although the FucT VI dimers were not quantitated by Borsig et al. (25), they appear to constitute less than 50% of the total enzyme expressed in CHO cells. As reported here, the ST6Gal I STtyr isoform requires Cys-24 in the transmembrane domain for dimerization. Other experiments show that the dimerization of the STcys isoform of this enzyme also requires Cys-24. 3 The precise Cys residues involved in FucT VI dimerization have not been identified to our knowledge. However, like the STtyr, the absence of dimers of the soluble secreted form of FucT VI suggests that dimer formation involves disulfide bonds forming between Cys residues in its transmembrane and/or cytoplasmic tail regions (25).
What controls the partial dimerization of the ST6Gal I and the FucT VI enzymes? Previously, we suggested that the ST6Gal I dimer may form in the Golgi and that dimerization may be controlled by the levels of donor and substrate in this compartment (3). Our finding that the disulfide-bonded dimer forms in the ER eliminates that possibility. The most obvious alternative possibility is that the level of dimer is related to the level of enzyme expression and the "packing" of the enzyme in the ER. Preliminary experiments to address this possibility demonstrated that increasing expression levels of the STtyr isoform in an inducible expression system did increase the amount of ST6Gal I dimer formed. 4 Interestingly, even at very low levels of expression, ST6Gal I dimer could still be detected when sufficient quantities of Golgi membrane preparations were loaded on the gels.
What is the function of the ST6Gal I dimer? Previously, we demonstrated that the ST6Gal I dimer exhibited a reduced affinity for CMP-NeuAc relative to that of the monomer. The dimer did, however, retain its ability to bind to galactose and asialoglycoprotein substrates, suggesting that it may act as a galactose specific lectin in the secretory pathway (3). We were surprised to find that Cys-181 and Cys-332 in the L and S sialyl motifs were not involved in our dimerization. If they had been, it would have provided a straightforward explanation for the decreased affinity of the ST6Gal I dimer for CMP-NeuAc. At this point, we can only suppose that the dimer's CMP-NeuAc binding site is partially blocked by close packing of catalytic domains in the dimer. Until we are able to generate a form of the enzyme that is expressed exclusively as a disulfide-bonded dimer, the hypothesis that the ST6Gal I dimer acts as a lectin in the Golgi will be difficult to test.
Another possible function for the ST6Gal I dimer is suggested by this work and previous work from our laboratory concerning the mechanism of stable ST6Gal I Golgi localization. Chen et al. (4) showed that the stable localization of the STcys isoform in the Golgi and its subsequent lack of cleavage and secretion correlated with this isoform's ability to quantitatively form insoluble oligomers when membranes were solubilized at pH 6.3-the pH of the late Golgi. In contrast, the STtyr that is only transiently localized in the Golgi and moves to a post-Golgi compartment, where it is cleaved and secreted, does not quantitatively form these oligomers (only 13% was found in the insoluble oligomer pellet at pH 6.3). When this assay was performed using rat liver Golgi membranes, the insoluble oligomers isolated when membranes were solubilized at pH 5.8 -6.3 consisted of predominantly disulfide-bonded dimer (4). For technical reasons, we are unable to quantitate the amounts of STtyr and STcys isoforms in these membranes. However, since the STtyr is the form encoded by genomic DNA, and is likely to predominate in most tissues (2), we suspect that the dimer form of the STtyr is the form that is found in the oligomer pellet. In Figs. 3  the wild type STtyr, and this correlates with a decrease in the cleavage and secretion of this protein. Like the STcys, this mutant form of the STtyr may assemble into more stable oligomers, which would lead to an increased retention time in the Golgi and a decreased rate of cleavage and secretion.
The correlation between the increased dimerization of the STtyr C139S mutant and its more stable Golgi localization, coupled with the predominance of the dimer in the oligomer pellet of liver Golgi membranes, makes it tempting to speculate that dimerization may play a role in the oligomerization process. For both ST6Gal I isoforms, the dimer form of the enzyme may "seed" oligomerization, inducing monomers to also form oligomers. In the case of the STcys, we would predict that the conformation of the STcys catalytic domain would confer stability to these oligomers leading to more stable localization in the ER-Golgi system. In the case of the STtyr, the conformation of the STtyr catalytic domain would lead to the formation of unstable oligomers that ultimately dissociate, allowing the STtyr to move beyond the Golgi to a post-Golgi compartment where it is cleaved and secreted.
Work by Yamaguchi and Fukuda (28) suggested that dimers of the ␤1,4-galactosyltransferase (GalT I) are involved in oligomer formation and Golgi localization. These researchers, and others (24,29), showed that the GalT I also forms dimers (as well as higher molecular weight oligomers). These dimers are dependent upon Cys and His residues in the enzyme's transmembrane domain (28). However, these GalT I dimers, which appear to constitute less than one half of the total enzyme, are quite resistant to standard reducing conditions, suggesting that dimerization may not necessarily involve a disulfide bond (28). Intriguingly, previous work by Aoki et al. (30) showed that the replacement of this Cys and His residue in the GalT I transmembrane domain increased cell surface expression of the GalT I. Analysis of the C24A mutant that eliminates the dimerization of the STtyr isoform suggests that dimerization of this form of the enzyme does not significantly alter its exit from the Golgi, as measured by its cleavage and secretion (Fig. 6) or cell surface expression (data not shown). Since this form of the ST6Gal I is only transiently localized in the Golgi, we may not be able to detect subtle differences in its trafficking. However, in the future, the C24A mutant form of the more stably localized STcys isoform should provide us with more definitive information concerning the role of dimers in oligomer formation and Golgi localization.