Ubiquitous 9-O-acetylation of sialoglycoproteins restricted to the Golgi complex.

9-O-Acetylation of sialic acid is known as a cell type-specific modification of secretory and plasma membrane glycoconjugates of higher vertebrates with important functions in modulating cell-cell recognition. Using a recombinant probe derived from influenza C virus hemagglutinin, we discovered 9-O-acetylated protein in the Golgi complex of various cell lines, most of which did not display 9-O-acetylated sialic acid on the cell surface. All cell lines expressed a sulfated glycoprotein of 50 kDa (sgp50) carrying 9-O-acetylated sialic acids, which was used as a model substrate. Like gp40, the major receptor for influenza C virus of Madin-Darby canine kidney I cells, sgp50 is 9-O-acetylated on O-linked glycans. However, gp40 was not 9-O-acetylated when expressed in Madin-Darby canine kidney II or COS-7 cells. The results demonstrate the existence of two 9-O-acetylation machineries for O-glycosylated proteins with distinct substrate specificities. The widespread occurrence of 9-O-acetylated protein in the Golgi furthermore suggests an additional intracellular role for this modification.

Sialic acids constitute a family of acidic monosaccharides typically found at terminal positions of glycoconjugates. Their diversity is mainly the result of enzymatic acetylation of hydroxyl groups at carbons 4, 7, 8, and 9 (1). 9-O-Acetylation is the most prevalent modification of sialic acids of glycoproteins and gangliosides in higher vertebrates. It is a tissue-specific and tightly regulated modification involved in various important processes such as cell-cell interaction and development (2)(3)(4). For example, binding of the B-cell adhesion molecule Siglec-2 (CD22␤) to sialylated ligands on the surface of thymocytes is inhibited by 9-O-acetylation (5). Furthermore, 9-Oacetylation was found to be crucial at the two-cell stage of murine development and later in the organization of the retina (6). 9-O-Acetyl sialic acid also serves as the receptor for influenza C virus (7), a group of corona viruses (8), and hemagglutinating encephalomyelitis virus (9) and thus appears to be a major determinant of their cell tropism (10).
O-Acetyl transferases for sialic acids have resisted purification or cloning thus far. However, their activity was localized to the trans side of the Golgi apparatus (11). By "freeze-frame" analysis with isolated membranes of rat hepatocytes, 9-Oacetylation was shown to occur in a compartment of the trans-Golgi network separate from that of sialylation and galactosy-lation (11). In this system, 9-O-acetylation was enriched on membrane-bound but not soluble cargo proteins leaving the trans-Golgi network.
9-O-Acetylation is specific for only a subset of sialoglycoproteins expressed in a given cell type, indicating that substrate recognition by 9-O-acetyl transferase(s) extends beyond the sialic acid itself. The 9-O-acetylation patterns of different cell types are thus likely to be the result of differential expression of substrates and/or transferase(s). Very few protein substrates for 9-O-acetylation of sialic acids have been identified at the molecular level, and all of them are sialomucins: bovine submaxillary mucin, CD43 and CD45RB of T lymphocytes (12), and gp40, the major receptor of influenza C virus in Madin-Darby canine kidney (MDCK) 1 type I cells (13,14).
Influenza C virus was found to be extremely useful in detecting and characterizing 9-O-acetylated sialoglycoproteins and gangliosides. The viral 9-O-acetyl sialic acid binding activity (the hemagglutinin, H) is part of a multifunctional spike protein, HEF, which also contains a receptor destroying acetylesterase activity (E) and a fusion function (F) (15). The entire virus has been used to detect 9-O-acetyl sialic acid in tissues and cell lines and on Western blots (13, 14, 16 -19). In addition, a chimeric protein, CHE-Fc, consisting of the influenza C virus hemagglutinin-esterase domains fused to the Fc region of human IgG1, was constructed as a specific tool for the detection of 9-O-acetylated sialoglycoconjugates in histochemistry, Western blots, thin-layer chromatography overlays (20), enzymelinked immunosorbent assay, flow cytometry (12), and even immunogold electron microscopy (11).
Here, we have used CHE-Fc as a probe to analyze the subcellular distribution of 9-O-acetylated glycoproteins in a variety of cell lines by fluorescence microscopy. We discovered that all cell lines tested, including some that were previously shown to be resistant to infection by influenza C virus and considered to lack 9-O-acetyl sialic acid, stained positive intracellularly in the Golgi apparatus. A sulfated glycoprotein of 50 kDa could be precipitated by CHE-Fc in all cases and was useful as a model substrate of this ubiquitous, intracellular 9-O-acetylation. Our results indicate that differential 9-O-acetylation patterns in various cell types depend on differences in both the substrates and the acetylation machineries.
CHO-Lec1 cells, were grown in Eagle's minimal essential medium or Dulbecco's modified minimal essential medium with 10% fetal calf serum at 37°C with 7.5% CO 2 . The media were supplemented with 2 mM L-glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin.
CHE-Fc Production-CHE-Fc, the chimera of influenza C virus hemagglutinin-esterase and the Fc region of human IgG, was produced by stably transfected HEK-293 cells provided by A. Varki (University of California, San Diego, CA). The cells were adapted to the protein-free medium CHO-S-SFM II (Invitrogen) and grown in a CELLine CL 350 device (Integra Bioscience). Conditioned medium was collected twice a week and kept at 4°C until purification. CHE-Fc was isolated using protein A-Sepharose as described previously (20). The probe was concentrated by centrifugation in a Centricon-50 filter device (Amicon). To inactivate the esterase activity, CHE-Fc was incubated on ice for 30 min with 1 mM diisopropyl fluorophosphate, producing CHE-FcD. To improve the binding efficiency of the probe, inactivation was repeated, followed by dialysis against PBS. CHE-FcD was stable at 4°C for several months.
Immunofluorescence-For indirect immunofluorescence staining, cells were grown on 14-mm glass coverslips, fixed with PBS containing 3% paraformaldehyde for 20 min at room temperature, washed in PBS, and quenched with 50 mM NH 4 Cl in PBS. To permeabilize the cells, the cells were then incubated with 0.1% Triton X-100 for 10 min at room temperature. Alternatively, cells were fixed with methanol and acetone according to Schwarz and Futerman (21), i.e. using conditions shown to extract glycolipids. Nonspecific antibody binding was blocked with PBS containing 1% bovine serum albumin and 10% goat serum (PBSB). The fixed cells were incubated with 10 g/ml CHE-FcD in PBSB for 1 h, washed with PBSB, and incubated with fluorescein isothiocyanatelabeled anti-human immunoglobulin secondary antibody diluted 1:400 in PBSB for 30 min. After additional washes with PBSB, PBS, and water, the coverslips were mounted in Mowiol 4 -88 (Hoechst) containing 2.5% 1,4-diazobicyclo-(2,2,2)-octane and analyzed using a Zeiss Axiophot microscope. For double immunofluorescence, the fixed cells were also incubated with mouse monoclonal antibodies against giantin (a gift from H.-P. Hauri, Biozentrum) followed by incubation with Cy3-labeled anti-mouse IgG secondary antibodies.
Western Blot Analysis-Cell lysates were split in two halves, one of which was incubated with 0.1 M NaOH for 15 min at 37°C and neutralized with HCl. Both samples were then separated by SDS-PAGE and electroblotted onto polyvinylidene difluoride membrane (Millipore). The membrane was blocked with 5% bovine serum albumin in PBS and incubated at 4°C with 3 g/ml CHE-FcD in PBS/1% bovine serum albumin. After several washes in PBS/0.05% Tween 20, horseradish peroxidase-conjugated anti-human IgG was added in PBS/1% bovine serum albumin. The blot was developed using the enhanced chemiluminescence kit from Amersham Biosciences.
Construction and Expression of gp40 HMY -To aid characterization, gp40 (provided by G. Herrler) was fused to the tag sequence HMY, which is composed of a His 6 sequence for Ni 2ϩ -NTA-agarose purification, a c-myc epitope, and a nonapeptide consensus sequence for tyrosine sulfation (22). The cDNA sequence encoding the signal peptide of gp40 was exchanged by the sequence encoding the signal and the HMY tag from MPR46 HMY (23) (provided by C. Itin, Stanford University). For this purpose, an in-frame KpnI site was introduced at the 3Ј end of the triple tag sequence and at the 5Ј end of the sequence encoding the mature sequence of gp40 by PCR. The two segments were ligated at the KpnI site and introduced into the expression plasmid pCB6 and the retroviral shuttle vector pMX-IRES (24) (from J. Bogan and H. F. Lodish, Whitehead Institute, Cambridge, MA). MDCK I cells were transfected with LipofectAMINE (Invitrogen). Clonal cell lines resistant to 0.8 mg/ml G418-sulfate (Geneticin; Invitrogen) were isolated and screened for expression of gp40 HMY by metabolic labeling with [ 35 S]methionine followed by Ni 2ϩ -NTA-agarose purification according to the manufacturer's protocol (Qiagen), SDS-gel electrophoresis, and fluorography or phosphorimaging. For expression in MDCK II cells, we took advantage of a MDCK II cell line (MDCK IIeco) stably expressing the murine ecotropic receptor for Moloney murine leukemia virus particles, which was prepared by transfecting MDCK II cells with pIRESeco-neo (from J. Bogan) and selecting for resistance to G418-sulfate (Invitrogen). Virus particles were produced by transfection of ⌽NX-Eco cells, virus-free helper cells obtained form J. Nolan (Stanford University), with pMX-gp40 HMY .
Metabolic Labeling and Analysis of 9-O-Acetylated Proteins-Metabolic labeling with [ 35 S]methionine or [ 35 S]sulfate was performed as described previously (22). Cells were labeled for 1 h at 37°C, lysed, and subjected to precipitation with Ni 2ϩ -NTA-agarose, lectin-agarose, or CHE-FcD on protein A-Sepharose. Five g of CHE-FcD prebound to 20 l of packed protein A-Sepharose beads were added per sample. The suspensions were incubated overnight at 4°C with end-over-end rotation. The beads were washed twice in lysis buffer and once in PBS and finally boiled in gel loading buffer containing reducing agent. The eluate was separated by SDS-gel electrophoresis and analyzed by fluorography. The biotinylated lectins Maackia amurensis hemagglutinin (MAH, sold as MAL-II by Vector Laboratories; 5 g/sample) and Sambucus nigra agglutinin (SNA; Vector Laboratories; 10 g/sample) were coupled to avidin-agarose (Pierce; 10 l packed beads/sample) for lectin precipitation.
As a control, the 9-O-acetyl modification was hydrolyzed by mild base treatment by the addition of 0.1 M NaOH to the samples, incubation for 15 min at 37°C, and neutralization with HCl. To test for membrane integration of sgp50, [ 35 S]sulfate-labeled cells were subjected to saponin or alkaline extraction (25,26) before precipitation with CHE-FcD. To block N-glycosylation, cells were preincubated for 4 h and [ 35 S]sulfatelabeled in the presence of 5 g/ml tunicamycin (Sigma). Complete inhibition of N-glycosylation was confirmed in a parallel analysis of MDCK II cells expressing the asialoglycoprotein receptor H1 (data not shown).
To remove N-glycans, CHE-FcD precipitates were suspended in 100 l of 0.1 M sodium phosphate, pH 6.8, containing 50 mM EDTA, 1% ␤-mercaptoethanol, and 0.1% SDS and boiled for 5 min. After the addition of Nonidet P-40 to a final concentration of 1%, samples were incubated with 0.25 unit of endoglycosidase F/N-glycosidase F (Roche Biochemicals) for 3 h at 37°C. Samples were then analyzed directly or after a second precipitation with CHE-FcD. To test for O-glycosylation of sgp50, CHE-FcD precipitates were incubated with 5 milliunits of neuraminidase from Arthrobacter ureafasciens for 4 h at 37°C in 50 mM sodium acetate, pH 5.5, followed by incubation with 1.25 milliunits of O-glycosidase (both from Roche Biochemicals) after the addition of 75 mM potassium phosphate, pH 7.4, overnight at 37°C. To test for the presence of chondroitin sulfate chains, [ 35 S]sulfate-labeled gp40 HMY isolated from a 35-mm well of gp HMY expressing MDCK I cells was purified on Ni 2ϩ -NTA-agarose and incubated on the beads with or without 0.2 unit of chondroitinase ABC (Fluka) in 100 l of 33 mM Tris acetate, 66 mM potassium acetate, 10 mM magnesium acetate, and 0.5 mM dithiothreitol, pH 7.9, for 3 h at 37°C. The samples were then boiled with SDS-gel sample buffer and analyzed.

Intracellular 9-O-Acetylated Sialic Acid Detected with CHE-FcD Fluorescence
Microscopy-CHE-Fc, the fusion protein of influenza C virus hemagglutinin-esterase with the constant region of human IgG, efficiently binds to 9-O-acetylated sialic acid if the esterase is inactivated by diisopropyl fluorophosphate treatment (20). We have used the resulting CHE-FcD as a probe for fluorescence microscopy in combination with a fluorescein isothiocyanate-labeled anti-human IgG antibody on MDCK I and II cells (Fig. 1). Previous studies using whole influenza C virions have shown that MDCK I cells contain abundant 9-O-acetylated sialoglycoconjugates on their surface, mediating efficient infection, whereas the closely related MDCK II cells do not (13). As shown in Fig. 1A, unpermeabilized MDCK I cells were strongly labeled by CHE-FcD on the cell surface. Upon permeabilization, additional perinuclear staining became apparent (Fig. 1B), most likely corresponding to the trans-Golgi, where 9-O-acetylation takes place. The specificity of staining for 9-O-acetyl sialic acids was demonstrated using CHE-Fc with the active esterase ( Fig. 1C): as the esterase hydrolyzed the 9-O-acetyl esters, staining was eliminated. Consistent with the experiments using viral infection, MDCK II cells were completely negative for CHE-FcD staining at the plasma membrane (Fig. 1E). In permeabilized cells, however, a strong perinuclear signal was discovered (Fig. 1F), which was specific for 9-O-acetylation because it was eliminated when 9-O-Acetylation of Sialoglycoproteins untreated CHE-Fc was used (Fig. 1G). Even when the cells were fixed with methanol and acetone under conditions that have been shown to extract glycolipids (21), the labeling pattern was retained, indicating that the molecules detected by CHE-FcD are primarily glycoproteins.
A similar staining pattern was observed in a variety of different cell lines derived from different tissues and species (Table I and Fig. 2). None of these cell lines showed any detectable CHE-FcD labeling at the cell surface, yet all of them showed perinuclear staining. Intracellular 9-O-acetylation of sialoglycoconjugates is thus a very common, if not ubiquitous, modification.
The intracellular staining pattern is reminiscent of the Golgi apparatus. To test this directly, calf primary aortic endothelial cells were colabeled with CHE-FcD and an antibody against the Golgi marker giantin (29). Both probes stained the same structures (Fig. 3, A and B). Even upon incubation with brefeldin A, which causes a dramatic redistribution of Golgi elements (30), the labeling pattern of both probes was affected in an identical manner (Fig. 3, C and D).
Identification of a Sulfated 9-O-Acetylated Glycoprotein of 50 kDa-To visualize 9-O-acetylated glycoproteins, proteins in cell lysates were subjected to SDS-gel electrophoresis, transferred to polyvinylidene difluoride membranes, and probed with CHE-FcD (Fig. 4). As a control for specificity, half the samples were treated with mild base (0.1 M NaOH for 15 min at 37°C) before analysis, which is sufficient to hydrolyze the O-acetyl groups on sialic acid. Accordingly, bovine submaxillary mucin, an established 9-O-acetylated sialoglycoprotein, was recognized by CHE-FcD before but not after mild base treatment (Fig. 4, lanes 1 and 2). In MDCK I cells (Fig. 4, lanes  7 and 8), the major 9-O-acetylated protein has a molecular mass of ϳ40 kDa, corresponding to gp40, the major influenza C virus receptor at the cell surface (13,14). In addition, a second specific band with an apparent molecular mass of ϳ50 kDa could be detected. Proteins with the same electrophoretic mobility were also observed in MDCK II and HEK-293 cells (  Ϫ ϩ ϩ CHO-Lec1 (ovary, hamster) Ϫ ϩ ϩ COS-7 (kidney, monkey) Ϫ ϩ ϩ CPAE (endothelia, bovine) Ϫ ϩ ϩ HEK-293 (kidney, human) Ϫ ϩ ϩ HeLa (cervix, human) Ϫ ϩ ϩ HepG2 (liver, human) Ϫ ϩ ϩ NIH-3T3 (fibroblast, mouse) Ϫ ϩ ϩ a 9-O-Acetylated sialic acid (9-O-AcSia) detected at the cell surface or in the Golgi by fluorescence microscopy of nonpermeabilized or permeabilized cells, respectively, using CHE-FcD as a probe visualized by fluorescein isothiocyanate-conjugated protein A as in Fig. 1. As an alternative detection method, CHE-FcD coupled to protein A-Sepharose was used to directly precipitate 9-O-acetylated proteins from metabolically labeled cells. Using [ 35 S]methionine for labeling led to considerable nonspecific background (data not shown). However, when cells were labeled with [ 35 S]sulfate, the pattern of radioactive proteins specifically precipitated with CHE-FcD was strikingly simple (Fig. 5). In all cell lines tested, a single major sulfated protein of ϳ50 kDa was detected (Fig. 5, lanes 1-12; Table I), which may correspond to the major 50-kDa protein detected by CHE-FcD on Western blots (Fig. 4). This protein was called sgp50 (for sulfated glycoprotein of 50 kDa).
To test whether sgp50 is a membrane protein, [ 35 S]sulfatelabeled HEK-293 cells were extracted with 0.1% saponin (Fig.  5, lanes 15 and 16), which allows soluble polypeptides to be released into the medium, whereas the membranes remain sufficiently intact to retain integrated proteins (26). Alternatively, the labeled cells were subjected to alkaline extraction, which solublilizes even peripheral proteins, whereas integral membrane proteins remain pelletable with the membranes (Fig. 5, lanes 13 and 14). With both procedures, sgp50 remained with the membranes and was not released into the supernatant, indicating that it is an integral membrane protein.
gp40 and sgp50 Are 9-O-Acetylated by Different Machineries-MDCK II cells and other cell lines may lack 9-O-acetyl sialic acids in the plasma membrane because they do not express plasma membrane substrates for 9-O-acetyl transferase. To test this possibility, gp40, the major 9-O-acetylated glycoprotein of MDCK I cells, was expressed in some of these cell lines. To allow simultaneous analysis of gp40 and endogenous sgp50 by 35 S sulfation, we modified the amino terminus of gp40 on the DNA level by adding a triple tag sequence consisting of a His 6 sequence for Ni 2ϩ -NTA-agarose purification, a c-myc epitope, and a nonapeptide consensus sequence for tyrosine sulfation (22,23). The tagged protein, gp40 HMY , was expressed in MDCK I and MDCK II cells. The cells were labeled with [ 35 S]sulfate and subjected to different isolation procedures followed by gel electrophoresis and fluorography (Fig. 6A). gp40 HMY was expressed in both cell lines to a similar extent as shown by Ni 2ϩ -NTA-agarose precipitation (Fig. 6A, lanes 1 and  2). As expected, 35 S-sulfated gp40 HMY was 9-O-acetylated in MDCK I cells and thus isolated with CHE-FcD like sgp50 (Fig.   6A, lane 3). In MDCK II cells, however, only sgp50, and not gp40 HMY , was recognized by CHE-FcD (Fig. 6A, lane 4). The machinery for 9-O-acetylation of sgp50 therefore does not recognize gp40. The same negative result was also obtained in transfected COS-7 cells (Fig 6A, lanes 5 and 6), where gp40 HMY  lanes 1-12). 9-O-Acetylated sialoglycoproteins were then isolated using CHE-FcD and protein A-Sepharose and analyzed by SDS-gel electrophoresis and fluorography. To test for membrane integration, labeled HEK-293 cells were subjected to alkaline extraction (Alk.) and separation into a membrane pellet (P) and a supernatant (S) (lanes 13 and 14). Alternatively, the labeled cells were extracted with 0.1% saponine (Sap.) and separated into a soluble saponin extract (S) and the rest of the cells (C). The fractions were analyzed as described above. The arrowhead indicates the 50 kDa position.

FIG. 6. gp40 HMY is 9-O-acetylated in MDCK I and CHO cells, but not in MDCK II or COS-7 cells.
A, MDCK I and II, CHO-K1, and COS-7 cells expressing gp40 HMY were labeled with [ 35 S]sulfate, lysed, and incubated with Ni 2ϩ -NTA-agarose (Ni) or CHE-FcD/protein A-Sepharose (CHE) to isolate gp40 HMY or glycoproteins with 9-O-acetyl sialic acids, respectively (lanes 1-6). In the case of COS-7 cells, Ni 2ϩ -NTA-agarose and CHE-FcD purification were performed successively. The positions of sgp50 and gp40 are indicated. The arrow points at the heterogeneous high molecular weight forms of gp40 HMY . B, to analyze the nature of the high molecular weight forms of gp40 HMY , [ 35 S]sulfatelabeled gp40 HMY from gp40 HMY -expressing MDCK II cells was isolated with Ni 2ϩ -NTA-agarose (Ni) and then incubated with (Ch) or without (-) chondroitinase before SDS-gel electrophoresis and fluorography. The ϳ90-kDa band (asterisk) most likely represents covalent dimers of gp40 HMY generated by oxidation during chondroitinase incubation. The positions of sgp50 and gp40 are indicated.

9-O-Acetylation of Sialoglycoproteins
runs with slightly lower mobility, probably due to different glycosylation.
Does differential 9-O-acetylation of gp40 and sgp50 correlate with a major difference in glycosylation of the two substrates? To answer this question, we analyzed the two model substrates with respect to the presence of N-or O-glycosylation and the linkages of their sialic acids. gp40 is a mucin-like protein with extensive O-linked glycosylation, but without N-linked glycans (13,14). On our gels, we always observed a diffuse signal around 70 -120 kDa (arrow in Fig. 6, A and B) in addition to the compact band of 40 kDa when gp40 or gp40 HMY was isolated. When gp40 HMY expressed in MDCK I cells was labeled with [ 35 S]sulfate, purified using Ni 2ϩ -NTA-agarose, and incubated with chondroitinase, the diffuse band collapsed into the 40-kDa band (Fig. 6B, lane 2), indicating that a fraction of the proteins carried chondroitin sulfate chains. However, 9-O-acetylation of gp40 HMY was obviously independent of glycosaminoglycan addition (Fig. 6A, lane 3).
To test sgp50 for O-linked glycans, sgp50 was first precipitated with CHE-FcD from [ 35 S]sulfate-labeled MDCK I cells (Fig. 7A, lane 1), desialylated (lane 2), and then incubated with O-glycosidase (lane 3). A shift in mobility upon SDS-gel electrophoresis indicated the presence of O-glycans. Similarly, labeled sgp50 precipitated with CHE-FcD was incubated with or without endoglycosidase F and analyzed either directly (Fig.  7A, lanes 4 and 5) or after a second CHE-FcD precipitation (lanes 6 and 7). Endoglycosidase F digestion produced two forms with increased electrophoretic mobility, indicating the presence of at least two N-linked glycans in sgp50. Even the lowest of the two forms was still recognized by CHE-FcD and thus carried 9-O-acetyl sialic acid (Fig. 7A, lane 7). As an alternative approach, MDCK II cells were labeled with [ 35 S]sulfate in the presence of tunicamycin under conditions that completely inhibit N-glycosylation. Using CHE-FcD, a 9-O-acetylated form of sgp50 was precipitated with a mobility identical to that of the lowest endoglycosidase F digestion product (Fig. 7A, lane 9). These results show that sgp50 carries both N-and O-linked glycans and suggest that the O-linked glycans contain 9-O-acetylated sialic acid.
To analyze gp40 and sgp50 with respect to the most common linkages of sialic acids, MDCK I and MDCK II cells expressing gp40 HMY were labeled with [ 35 S]sulfate, lysed, and subjected to lectin precipitation. We used either MAH, which specifically recognizes sialic acids linked ␣2-3 to Gal in the O-linked sequence Sia␣2-3Gal␤1-3(ϮSia␣2-6)GalNAc (31,32), or SNA, which recognizes Sia␣2-6Gal of N-glycans or O-linked Sia␣2-6GalNAc (33). gp40 HMY and sgp50 were both precipitated with MAH, either directly from cell lysates (Fig. 7B, lanes 1 and 2) or after initial purification with CHE-FcD (lane 6). In contrast, a labeled protein of 50 kDa, but none of 40 kDa, was precipitated with SNA along with several proteins of other sizes from both MDCK I and MDCK II cells (Fig. 7B, lanes 3 and 4). gp40 in MDCK I cells was previously shown not to be recognized by SNA (13). The identity of the 50-kDa protein with the 9-Oacetylated sgp50 was confirmed by successive precipitation with CHE-FcD and SNA (Fig. 7B, lane 7). sgp50 synthesized in MDCK II cells treated with tunicamycin and thus lacking N-glycans was still recognized by MAH (Fig. 7A, lane 11) but was not precipitated with SNA (lane 13). Likewise, all other SNA-positive sulfate-labeled glycoproteins were not precipitated any more after tunicamycin treatment, demonstrating the presence of ␣2-6-linked sialic acid in N-linked glycans.
These findings could be confirmed using the CHO cell lines CHO-K1 and CHO-Lec1. CHO cells are known to lack ␣2-6linked sialic acid in N-glycans (34,35). In addition, CHO-Lec1 mutant cells are unable to synthesize hybrid-type and complex N-glycans due to a defect in N-acetylglucosaminyltransferase I (36,37) and stop at Man 9 GlcNAc 2 structures. Upon [ 35 S]sulfate labeling and precipitation using CHE-FcD, sgp50 of its normal size and an additional sulfated protein of slightly less than 40 kDa were detected in CHO-K1 cells (Fig. 8, lane 1). In CHO-Lec1 cells, sgp50 was still recognized but shifted to a position of reduced molecular mass (lane 2).
To show more directly that 9-O-acetylation occurred on Oglycans in both gp40 and sgp50, cells were treated with either benzyl-GalNAc to inhibit sialylation on O-glycans (27) or ␤-Dxyloside to block glycosaminoglycan synthesis on proteins (28). Treatment of MDCK I cells expressing gp40 HMY with benzyl-GalNAc did not affect synthesis of gp40 HMY or its mobility upon gel electrophoresis, as judged by precipitation with Ni 2ϩ -NTAagarose (Fig. 9, lanes 1 and 2). Removal of sialic acid has previously been observed not to alter its electrophoretic mobil- ity (13). However, both the 40-kDa band and the form carrying chondroitin sulfate chains were not recognized by CHE-FcD any more (Fig. 9, lane 4 versus lane 3), indicating that normally 9-O-acetylation takes place on sialic acids of O-glycans. ␤-D-Xyloside, in contrast, strongly reduced the chondroitin sulfate chains of gp40 HMY , but not the ability of the 40-kDa protein to be recognized by CHE-FcD (Fig. 9, lanes 5-8). The same result was obtained for sgp50 in MDCK I cells that were treated in addition with tunicamycin before and during [ 35 S]sulfate labeling to analyze sgp50 in the absence of N-glycans and shift the newly synthesized proteins away from preexisting ones that might be further sulfated during the labeling period (Fig. 9,  lanes 9 -12). Recognition by CHE-FcD was strongly inhibited by benzyl-GalNAc (Fig. 9, lane 10) but was not affected by ␤-D-xyloside (lane 12).
Together, the results demonstrate the presence of ␣2-3linked sialic acid on O-linked glycans of both gp40 and sgp50, whereas only sgp50 contains Sia␣2-6Gal. However, both gp40 and sgp50 are 9-O-acetylated on sialic acids of O-linked gly-cans. The difference between gp40 and sgp50 that is responsible for their differential 9-O-acetylation is therefore more subtle than the presence or absence of N-linked glycosylation. DISCUSSION CHE-FcD has proven to be a valuable tool to detect 9-Oacetylated glycoconjugates in a variety of methods. In the present study, we have used it for the first time in fluorescence microscopy and discovered significant intracellular 9-O-acetylation of sialoglycoproteins in cell lines previously considered to lack this modification. MDCK II and BHK-21 cells were shown to be devoid of 9-O-acetylated cell surface proteins by influenza C virus binding or infection assays (10,13), whereas CHO and BHK-21 cells scored negative in flow cytometry experiments using CHE-FcD (35). Our results are consistent with these studies because in all four cell lines, CHE-FcD staining is restricted to a perinuclear structure corresponding to the Golgi, most likely the trans-Golgi network, and is absent from the plasma membrane. This intracellular 9-O-acetylation is remarkably ubiquitous because it was detected in all 12 cell lines we tested, which were derived from different species and tissues (Table I).
It is not surprising per se that 9-O-acetylated proteins can be detected in the Golgi because this modification has previously been shown to occur in Golgi membranes of rat liver (38,39), more precisely in a compartment of the trans-Golgi network separate from the compartment of sialylation (11). Secretory and plasma membrane proteins thus pass through this compartment and may acquire 9-O-acetylation as one of the last modifications before arrival at the cell surface. The known functions of 9-O-acetylation in cell-cell and host-pathogen interactions as well as the relative abundance of 9-O-acetyl sialic acids in secretory proteins (20) pointed to a predominantly extracellular role of this modification. The striking presence of 9-O-acetylated proteins in the Golgi and their absence from the cell surface also suggest an intracellular function.
Western analysis revealed a major 9-O-acetylated glycoprotein of ϳ50 kDa in HEK-293 and MDCK II cells, which was also prominently present in MDCK I cells besides gp40. Using sulfate labeling, a sulfated membrane protein of the same size, sgp50, could be precipitated with CHE-FcD in all cell lines analyzed. Because sulfation yielded good sensitivity and low background, sgp50 and gp40 tagged with a sulfation site (gp40 HMY ) were useful tools to analyze the glycosylation characteristics of these two 9-O-acetylation substrates.
The simplest explanation for the absence of 9-O-acetylated proteins on the surface of a cell line is that the cells do not express substrates that are transported to the plasma membrane. However, gp40 HMY , which is modified in MDCK I cells, was not acetylated when expressed in MDCK II and COS-7 cells. The 9-O-acetyl transferase responsible for acetylation of the Golgi-resident substrate(s) apparently does not recognize gp40 as a substrate. There are thus two distinct 9-O-acetylation machineries: a ubiquitous one with mainly intracellular substrates including sgp50, and a second one present in MDCK I cells that recognizes gp40. In MDCK II and COS-7 cells, this latter machinery and its substrate gp40 are missing.
Previously, it has been shown that 9-O-acetyl transferases can have clear linkage specificity. CHO-K1 cells, which lack ␣2-6-linked sialic acids on N-linked glycans (34) and do not bind CHE-FcD at the plasma membrane, presented 9-O-acetyl sialic acid on the cell surface upon expression of the ␣2-6 sialyltransferase for N-glycans (ST6Gal I), but not when the competing ␣2-3 sialyltransferase (ST3Gal III) was transfected (35). The parental CHO cells thus have a 9-O-acetyl transferase for N-linked Sia␣2-6Gal but essentially no corresponding substrates. This transferase is therefore not a candidate en-  1-8) or normal MDCK I cells (lanes 9 -12) were incubated for 18 h with or without benzyl-GalNAc (G) or 4-methylumbelliferyl xyloside (X) to inhibit sialylation of O-glycans or synthesis of glycosaminoglycans, respectively. The cells were then labeled with [ 35 S]sulfate, lysed, and subjected to precipitation with CHE-FcD/protein A-Sepharose or Ni 2ϩ -NTA-agarose. To test the effect on sgp50 in the absence of N-glycans (sgp50-N), the cells in lanes 9 -12 were treated with tunicamycin (Tu) before and during labeling. zyme for the 9-O-acetylation of sgp50 or gp40. There are thus at least three distinct 9-O-acetylation machineries for glycoproteins, modifying gp40, sgp50, or N-linked Sia␣2-6Gal, respectively.
gp40 and sgp50 are 9-O-acetylated on O-linked glycans. Because both proteins are recognized by MAH, they positively contain sialylated core 1 structures, i.e. Sia␣2-3Gal␤1-3GalNAc and/or Sia␣2-3Gal␤1-3(Sia␣2-6)GalNAc, most probably in addition to other structures for which there is no direct evidence at present. Considering the high specificity of 9-Oacetylation as observed for example in MDCK I and II cells, substrate recognition of 9-O-acetyl transferases must reach beyond the sialic acid and its linkage and must include more extended, possibly rare glycan structures and/or contributions from the protein core. Our findings indicate that 9-O-acetylation is more widespread than previously known and may also have intracellular functions.