Expression of Human Glycophorin A in Wild Type and Glycosylation-deficient Chinese Hamster Ovary Cells ROLE OF N- AND 0-LINKED GLYCOSYLATION IN CELL SURFACE EXPRESSION*

, Glycophorin A, the most abundant sialoglycoprotein on human red blood cells, carries several medically important blood group antigens, To study the role of glycosylation in surface expression and antigenicity of this highly glycosylated protein (1 N-linked and 15 0-linked oligosaccharides), glycophorin A cDNA (M-al-lele) was expressed in Chinese hamster ovary (CHO) cells. Both wild type CHO cells and mutant CHO cells with well defined glycosylation defects were used. Glycophorin A was well expressed on the surface of transfected wild type CHO cells. On immunoblots, the CHO cells expressed monomer (-38 kDa) and dimer forms of glycophorin A which co-migrated with human red blood cell glycophorin A. The transfected cells speci-ficially expressed the M blood group antigen when tested with mouse monoclonal antibodies. Tunicamycin treatment of these CHO cells did not block surface expression of glycophorin A, indicating that, in the presence of normal 0-linked glycosylation, the N-linked oligosaccharide is not required for surface expression. To

Glycophorin A, the most abundant sialoglycoprotein on human red blood cells, carries several medically important blood group antigens, To study the role of glycosylation in surface expression and antigenicity of this highly glycosylated protein (1 N-linked and 15 0linked oligosaccharides), glycophorin A cDNA (M-allele) was expressed in Chinese hamster ovary (CHO) cells. Both wild type CHO cells and mutant CHO cells with well defined glycosylation defects were used. Glycophorin A was well expressed on the surface of transfected wild type CHO cells. On immunoblots, the CHO cells expressed monomer (-38 kDa) and dimer forms of glycophorin A which co-migrated with human red blood cell glycophorin A. The transfected cells specificially expressed the M blood group antigen when tested with mouse monoclonal antibodies. Tunicamycin treatment of these CHO cells did not block surface expression of glycophorin A, indicating that, in the presence of normal 0-linked glycosylation, the Nlinked oligosaccharide is not required for surface expression. To study 0-linked glycosylation, glycophorin A cDNA was transfected into the Lec 2, Lec 8, and ldlD glycosylation-deficient CHO cell lines. Glycophorin A with truncated 0-linked oligosaccharides was well expressed on the surface of ldlD cells (cultured in the presence of N-acetylgalactosamine alone), Lec 2 cells, and Lec 8 cells with monomers of -25 kDa, -33 kDa, and -25 kDa, respectively. In contrast, non-0-glycosylated glycophorin A (-19-kDa monomers) was poorly expressed on the surface of ldlD cells cultured in the absence of both galactose and N-acetylgalactosamine. Thus, under these conditions, in the absence of 0-linked glycosylation, the N-linked oligosac-* This work was supported in part by National Institutes of Health Grants CA 45690 and CA 33000, American Cancer Society Grant JFRA-193, and the Kosciuszko Foundation. This work was presented in part at the 74th Annual Meeting of the Federation of American Societies for Experimental Biology ((1990) FASEB J. 4, A1929). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. charide itself is not able to support appropriate surface expression of glycophorin A in transfected CHO cells. Glycophorin A, an important constituent of the human red blood cell membrane (for reviews, see Refs. 1 and 2), is the most abundant sialoglycoprotein on the surface of these cells, numbering approximately 500,000 copies per cell. It is 131 amino acids in length, contains one transmembrane domain, is oriented with an extracytoplasmic amino terminus (3), and is highly homologous to another abundant glycoprotein on the red cell surface, glycophorin B (1,2). Glycophorin A is highly glycosylated, containing 1 N-linked and 15 0-linked oligosaccharides (1)(2)(3). Most of the 0-linked oligosaccharides have the classical structure, NeuAca2-3Galpl-3[NeuAca2-61GalNAc-R, found on many proteins (4); the N-linked oligosaccharide is predominantly a disialylated bisected, biantennary complex type structure (5). Due to this high degree of glycosylation, glycophorin A carries approximately 70% of the red blood cell sialic acid, although it comprises just 2-4% of the total membrane protein (1, 2). Glycophorin A encodes several blood group antigens important in transfusion medicine, including peptide antigens such as Mi.I(6,7), carbohydrate antigens such as Pr2 (8), T (9, lo), and T n ( l l ) , and complex glycopeptide antigens such as M and N (12)(13)(14). Glycophorin A is also involved in the pathogenesis of malaria since its presence on the red blood cell surface is required for invasion by Plasmodium falciparum merozoites (15)(16)(17)(18)  The effect of tunicamycin on glycophorin A expression was examined using Clone 26.1. In order to evaluate the effect of tunicamycin on newly synthesized glycophorin A, Clone 26.1 cells were first treated with neuraminidase followed by incubation in the presence of tunicamycin. The appearance of newly synthesized glycophorin A on the cell surface was monitored by immunofluorescence using the sialic acid-dependent monoclonal antibody 6A7. Antibody 6A7 bound to live, intact Clone 26.1 cells (Fig. 5A), but not if the live cells were first treated with neuraminidase (Fig. 523). If neuramin-   idase-treated cells were cultured for 24 h following enzyme treatment, 6A7 again bound to the cells (Fig. 5C), most likely to newly synthesized, sialylated glycophorin A. When live cells were treated with neuraminidase and recultured for 24 h in complete medium containing 5 pg/ml tunicamycin (371, intact, sialylated glycophorin A was again detected at the cell surface by antibody 6A7. This suggests that newly synthesized glycophorin A lacking the N-linked oligosaccharide is expressed on the cell surface. This experimental approach did not disrupt the integrity of the plasma membrane since monoclonal antibody Pep80 (recognizing an epitope on the cytoplasmic tail of glycophorin A) did not bind to live, neuraminidase, and tunicamycin-treated cells (Fig. 5E). However, if these cells were deliberately permeabilized before testing with Pep80, bright fluorescence was seen (Fig. 5F). Thus, the results in Fig. 5 0 were due to antibody 6A7 binding to cell surface antigen and not to binding to non-N-glycosylated glycophorin A aggregated inside disrupted cells.
T o exclude the possibility that the results in Fig. 5 were due to recycling of desialylated cell surface glycophorin A by intracellular vesicles through the Golgi and back to the plasma membrane, allowing resialylation (44), the experiment was repeated by culturing the cells with and without tunicamycin (5 pg/ml) in the presence of cycloheximide (10 pg/ml). Since cycloheximide blocks protein synthesis, this approach allows antibody 6A7 to distinguish between recycled and newly synthesized glycophorin A. The results (see Table I) demonstrate that there is no detectable recycling of desialylated cell surface glycophorin A with subsequent resialylation in this system.

The Role of 0-linked Glycosylation in Surface Expression of
Glycophorin A-To address the role of 0-linked oligosaccharides in glycophorin A transport and expression, three glycosylation-deficient CHO cell lines were used ldlD, Lec 2, and Lec 8. The 0-linked oligosaccharides present on human red blood cell glycophorin A have the disialylated tetrasaccharide structure illustrated in Fig. 6 (1, 2, 4); the biosynthetic pathway leading to the formation of this structure is also illustrated (45).
The ldlD cells have a defect in the 4-epimerase enzyme (26, 27), blocking conversion of either glucose into galactose or Nacetylglucosamine into N-acetylgalactosamine. Thus, when these cells are cultured in medium lacking both galactose and N-acetylgalactosamine, no 0-linked oligosaccharides are added to nascent glycoproteins. Similarly, when galactose alone is added to the medium, no 0-linked oligosaccharides are added to nascent glycoproteins. When cultured in the presence of N-acetylgalactosamine alone, only this monosaccharide (or, possibly, the NeuAca2-6GalNAc disaccharide) is found in an 0-glycosidic linkage on glycoproteins. When both galactose and N-acetylgalactosamine are added to the medium, the mature tetrasaccharide can be synthesized. Since galactose and N-acetylgalactosamine are also constituents of glycosphingolipids and of N-linked oligosaccharides on glycoproteins, these are also modified in ldlD cells grown under restrictive conditions. Following transfection of ldlD cells with pSG5gpa, several clonal cell lines were obtained which highly expressed glycophorin A when the cells were cultured in ldlD CM containing both galactose and N-acetylgalactosamine. One clone, Clone ldlD-5, was analyzed in detail by flow cytometry using glycophorin A-specific rabbit polyclonal antibody (Fig. 7). This antibody recognizes all forms of glycophorin A synthesized by ldlD cells under both permissive and restrictive conditions (see Fig. 8). If Clone ldlD-5 was cultured in ldlD CM containing galactose and N-acetylgalactosamine, high surface expression was seen (mean channel fluorescence 996), similar to Cell surface glycophorin A is not detectably recycled and resialylated in transfected CHO cells Clone 26.1 cells were treated with neuraminidase, cultured in the presence or absence of tunicamycin and cycloheximide, and analyzed by indirect immunofluorescence, as described under "Experimental Procedures." There was no effect of cycloheximide on cell viability when assayed by trypan blue exclusion. Monoclonal antibody 6A7 recognizes a sialic acid-dependent epitope on glycophorin. "Positive" results indicate bright fluorescence identical with that seen in Fig. 5, panel A. "Negative" results indicate background fluorescence identical with that seen in Fig. 5,

--ldlD (no sugars or Gal alone)
GalNAcal-Ser/Thr The cells were incubated with rabbit polyclonal anti-glycophorin A antibody followed by fluorescein-conjugated goat anti-rabbit Ig. Nonspecific background fluorescence was determined using normal rabbit serum followed by fluorescein-conjugated goat anti-rabbit Ig.
that found with Clone 26.1 cells probed with rabbit polyclonal antibody (mean channel fluorescence 896; data not shown). Somewhat diminished fluorescence intensity was observed when Clone ldlD-5 cells were cultured with N-acetylgalactosamine alone (mean channel fluorescence 434). This suggests that truncated 0-linked oligosaccharides are sufficient for ensuring adequate surface expression. In contrast, when cul- tured in ldlD CM lacking galactose and N-acetylgalactosamine, these cells exhibited dim fluorescence (mean channel fluorescence of ll), suggesting that the presence of 0-linked oligosaccharides is necessary for optimal surface expression of glycophorin A. Diminished, but not absent fluorescence was seen when cells were cultured with N-acetylgalactosamine alone (mean channel fluorescence 247), although glycophorin A synthesized by these cells should lack 0-linked oligosaccharides. Untransfected wild type CHO cells (Pro-5) exhibited dim fluorescence by this method (mean channel fluorescence 37; data not shown). When permeabilized Clone ldlD-5 cells were probed with monoclonal antibody Pep80, equivalent binding was seen by classical indirect immunofluorescence regardless of whether the cells were cultured in the presence or absence of galactose and N-acetylgalactosamine (data not shown).

-
To biochemically examine the glycophorin A variants synthesized by transfected ldlD cells under various culture conditions, detergent lysates of Clone ldlD-5 cells were separated by SDS-PAGE and analyzed by Western blotting (Fig. 8).
Using glycophorin A-specific rabbit polyclonal antibody, human red blood cells (Fig. 8, lane 7), transfected wild type CHO cells (Fig. 8, lane 6), and Clone ldlD-5 cells cultured in the presence of both galactose and N-acetylgalactosamine (Fig. 8, lane 5) all express glycophorin A monomers and dimers with virtually identical electrophoretic mobility. In contrast, Clone ldlD-5 cells grown under restrictive conditions express glycophorin A monomers and dimers of progressively faster electrophoretic mobility. This most likely results from the presence of truncated 0-linked oligosaccharides (Fig. 8,  lane 4 ) and the absence of 0-linked oligosaccharides (Fig. 8,  lanes 2 and 3 ) on glycophorin A. Interestingly, glycophorin A in lane 3 migrates slower than that in lane 2. There is also increased heterogeneity in the banding pattern. It is possible that in the absence of galactose and N-acetylgalactosamine, the N-linked oligosaccharide on glycophorin A is restricted to be only a high mannose or hybrid form (46). In contrast, galactose allows further processing to occur leading to larger and more heterogeneous complex type N-linked oligosaccharides (46). Alternatively, small amounts of residual N-acetylgalactosamine in the culture medium or in the cells may allow sufficient 0-linked glycosylation to permit surface expression.
To further confirm the results obtained with ldlD cells, glycophorin A cDNA was transfected into Lec 2 and Lec 8 cells. Lec 2 and Lec 8 cells have nonreversible defects in glycosylation owing to their inability to transport CMP-NeuAc and GDP-Gal, respectively (47, 48). Therefore, the sequences of 0-linked oligosaccharides on glycoproteins synthesized in Lec 2 and Lec 8 cells should be restricted to be Gal@l-3GalNAcal-Ser/Thr and GalNAcal-Ser/Thr (or, possibly, NeuAca2-6GalNAcal-Ser/Thr), respectively (see Fig.  6). Live, transfected Lec 2 and Lec 8 cells both expressed glycophorin A at the cell surface, by indirect immunofluorescence using rabbit polyclonal anti-glycophorin A antibody, in amounts comparable to transfected wild type CHO cells (data not shown). As expected, the sialic acid-dependent monoclonal antibody, 6A7, did not bind to either transfected Lec 2 or Lec 8 cells (data not shown), confirming a predicted biochemical alteration in glycophorin A expressed by these cells. Thus, similar to the results obtained with ldlD cells, smaller, nonsialylated 0-linked oligosaccharides allow the expression of glycophorin A at the cell surface.
The glycophorin A molecules synthesized by Lec 2 and Lec 8 cells were examined by Western blotting using rabbit polyclonal anti-glycophorin A antibody (Fig. 9). The monomer and dimer forms of glycophorin A produced by a clone of transfected Lec 2 cells (Fig. 9, lane 1 ) migrate faster than those synthesized by either transfected wild type CHO cells (Clone 26.1; Fig. 9, lane 2 ) or human red blood cells (Fig. 9,  lune 3 ) . In addition, if Clone 26.1 proteins were first treated with neuraminidase and then separated by SDS-PAGE and analyzed by Western blotting, then the neuraminidase-treated monomer and dimer forms of glycophorin A (Fig. 9, lane 9) co-migrated with those synthesized by Lec 2 cells (Fig. 9, lane  8). This result suggests that desialylated glycophorin A from wild type cells and untreated glycophorin A from Lec 2 cells should both have similar 0-linked oligosaccharides. Glycophorin A synthesized by transfected Lec 8 cells migrated even faster by SDS-PAGE (Fig. 9, lane 6) and co-migrated with that produced by ldlD cells cultured in the presence of Nacetylgalactosamine alone (Fig. 9, lane 5) suggesting that To examine the role in cell surface expression of glycophorin A of the N-linked oligosaccharide at AsnZ6, transfected wild type CHO cells were cultured with tunicamycin. Surface expression of the glycoprotein was not affected (Fig. 5). Analogous results were obtained by inactivating the N-linked glycosylation acceptor sequence by site-directed mutagenesis (Thr" + Met)? These results agree well with the finding that individuals carrying the mutant allele Mi.1 (ThrZ8 + Met) express the mutant protein on the surface of their red blood cells (7). Similarly, tunicamycin does not affect glycophorin A surface expression in K562 cells (21, 23). In addition, the homologous glycoprotein, glycophorin B, lacks an N-linked oligosaccharide but is expressed on human red blood cells (1,2).
The role of N-linked glycosylation in glycoprotein surface expression has been previously described (37,(52)(53)(54)(55)(56)(57). The oligosaccharide(s) may encourage proper folding of the nascent glycoprotein preventing aggregation or proteolysis in the endoplasmic reticulum (53, 58). However, N-linked glycosylation is not required for surface expression or secretion of all proteins (ie. Ref. 59). In the current case, the 0-linked oligosaccharides on glycophorin A may be sufficient to ensure its translocation in the absence of N-linked glycosylation.
To study the role of the 0-linked oligosaccharides in glycophorin A surface expression, the cDNA was transfected into the Lec 2, Lec 8, and ldlD glycosylation-deficient CHO cell lines. Surface expression was evaluated by indirect immunofluorescence using rabbit polyclonal antibody. Transfected glycophorin A was highly expressed on the plasma membrane of Lec 2 and Lec 8 cells (data not shown) and also on the surface of ldlD cells cultured in the presence of N-acetylgalactosamine alone (Fig. 7). This suggests that even truncated, nonsialylated 0-linked oligosaccharides promote high cell surface expression of glycophorin A. The results with Lec 2 and ldlD cells are analogous to those found with another heavily 0-glycosylated glycoprotein, leukosialin, synthesized by the human T cell Jurkat where the major 0-linked oligosaccharide consisted of N-acetylgalactosamine alone (60). In contrast, when transfected ldlD cells were cultured in the absence of both galactose and N-acetylgalactosamine, conditions which prevent 0-linked glycosylation, glycophorin A was poorly expressed at the cell surface (Fig. 7). Interestingly, intermediate levels of glycophorin A surface expression were seen when ldlD cells were cultured with galactose alone, although under these conditions 0-linked glycosylation should also be prevented (Fig. 7). In addition, the protein synthesized by ldlD cells cultured with galactose alone showed greater heterogeneity on Western blots (Fig. 8). These latter two findings together suggest that the presence of galactose alone may allow extensive processing of the N-linked oligosaccharide permitting some surface expression of non-0-glycosylated glycophorin A. Alternatively, due to variations in the experimental conditions, the presence of small amounts of residual N-acetylgalactosamine in the culture medium or in the cells may allow sufficient 0-linked glycosylation to permit surface expression. To distinguish between these two possibilities, it will be interesting to examine surface expression of the Mi.1 mutant (described above) in transfected ldlD cells. When cultured with galactose alone or in the absence of both galactose and N-acetylgalactosamine, the protein should not be glycosylated at all. Biochemical analysis of the 0-linked oligosaccharides on purified glycophorin A obtained from these cells should also be revealing. In a similar study, marked reduction in cell surface expression of the attachment glycoprotein of human respiratory syncytial virus has been found in cells in which both Nand 0-linked glycosylation is blocked (61).
Several mechanisms may lead to poor surface expression or secretion of abnormally non-0-glycosylated glycoproteins. The protein may be retained or aggregated in the endoplasmic reticulum (62). Alternatively, as with the IL-2 receptor, it may be missorted to the wrong cellular compartment and not appear on the plasma membrane (63). In contrast, it may be expressed on the surface but then undergo rapid proteolysis, as with low density lipoprotein receptor (26, 64), decay accelerating factor (65), and the Epstein-Barr virus major envelope antigen (63). Similarly, the abnormal protein may reach the surface but then be rapidly internalized and degraded. Interestingly, secretion of the 0-glycosylated secretory glycoproteins human chorionic gonadotropin (66) and apolipoprotein E (67) is not affected by blocking their acquisition of 0-linked oligosaccharides. The mechanism leading to diminished surface expression of non-0-glycosylated glycophorin A is under investigation.
By extending the approach outlined here, it will be possible to investigate the structural correlates of carbohydrate, glycopeptide, and peptide blood group antigens on glycophorin A. This is possible because transfected glycophorin A was highly expressed with or without N-linked oligosaccharides and with truncated, nonsialylated 0-linked oligosaccharides. For example, it is predicted that transfected Lec 8 cells will express the T antigen (9,10) and ldlD cells cultured with Nacetylgalactosamine alone will express the Tn antigen (11) on glycophorin A. In addition, since human anti" alloantibodies recognize complex glycopeptide epitopes (13,14,68), cell lines expressing variant oligosaccharides will permit a detailed analysis of this specificity. Finally, site-directed mutagenesis will enable study of the amino acid sequences important in glycopeptide and peptide epitopes such as M and Mi.1, respectively (6, 7, 12-14, 68). Similarly, a detailed analysis of the interaction between P. fakiparum merozoites and glycophorin A can be undertaken, particularly since the parasite recognizes signals encoded by both the amino acid sequence and oligosaccharides on glycophorin A (69-72). Thus, the availability of these cell lines will allow detailed investigations regarding the antigenicity, role in host-parasite interactions, and intracellular biosynthesis and trafficking of this interesting glycoprotein.  I~s u e culture chamber 518des (Nunc Inc.. Nape~IIe. IL). as described (37). However. lor qioplasmc In prcmary rabbsl or mouse a n l r w y diluted I" PES-BSA (phosphate buHered salrne pH 7.4. conlalnzng 1% BSA and 0 1% NaN,) at 1 x lb cellslml and lncubaled lor 1 h at 4%