Glycophorin A protects K562 cells from natural killer cell attack. Role of oligosaccharides.

Glycophorin A is a protein with an abundant glycosylation (60% carbohydrate by weight), and studies have suggested that resistance of target cells to natural killing may be correlated with the level of glycophorin A expression. To assess the role of glycophorin A and of its carbohydrates in sensitivity to lysis by natural killer (NK) cells, the glycoprotein was inserted into the membrane of K562 target cells using electropulsation. Peripheral blood lymphocytes were used as effector cells. When glycophorin A was inserted into the membrane, the level of resistance to NK cell attack increased with the number of glycophorin A molecules electroinserted. The resistance to lysis was not due to a defect in target cell-effector cell binding. Electroassociation of glycophorin A did not cause a decrease in the expression of either “positive signals” for NK cells (such as CD71, CD15, and CD32 antigens) or cellular adhesion molecules (CD18, CD29, CD54, and CD58). Furthermore, electroinsertion of glycophorin A did not trigger any “negative signals,” such as class I HLA antigen expression. Finally, it was shown that the sialic acid and O-linked oligosaccharides of glycophorin A did not play any role in its effect against NK cells. Conversely, the unique N-linked oligosaccharide was shown to be essential for resistance to occur.

Glycophorin A is a protein with an abundant glycosylation (60% carbohydrate by weight), and studies have suggested that resistance of target cells to natural killing may be correlated with the level of glycophorin A expression. To assess the role of glycophorin A and of its carbohydrates in sensitivity to lysis by natural killer (NK) cells, the glycoprotein was inserted into the membrane of K562 target cells using electropulsation. Peripheral blood lymphocytes were used as effector cells. When glycophorin A was inserted into the membrane, the level of resistance to NK cell attack increased with the number of glycophorin A molecules electroinserted. The resistance to lysis was not due to a defect in target cell-effector cell binding. Electroassociation of glycophorin A did not cause a decrease in the expression of either "positive signals" for NK cells (such as CD71, CD15, and CD32 antigens) or cellular adhesion molecules (CD18, CD29, CD54, and CD58). Furthermore, electroinsertion of glycophorin A did not trigger any "negative signals," such as class I HLA antigen expression. Finally, it was shown that the sialic acid and O-linked oligosaccharides of glycophorin A did not play any role in its effect against NK cells. Conversely, the unique N-linked oligosaccharide was shown to be essential for resistance to occur.
Natural killer (NK) 1 cells are generally considered part of the first host defenses against neoplastic and infectious diseases. NK cells are CD3 Ϫ large granular lymphocytes and can be functionally defined as cells that mediate non-histocompatibility-restricted killing of some target cells. These lymphocytes are spontaneously cytotoxic against some tumors and virally infected cells via nonspecific mechanisms. In addition, NK cells are able to kill certain normal cells in vitro (1).
Although there has been extensive characterization of many features of NK cells, the NK cell receptors and their ligands on the target cell surface have remained elusive for a long time. Thus, the sensitivity of target cells has been considered alternatively as being due to the expression of immature structures and viral antigens (2) or to an increased expression on the cell surface of normal glycoproteins such as transferrin receptor (CD71) and CD32 antigen (3)(4)(5), of glycolipids such as asialo-G M2 and G M2 (6,7), of oligosaccharides such as 3-fucosyl-Nacetyllactosamine (CD15 antigen) (8), and of peptides such as a 42-kDa polypeptide (9). These molecules could act as "positive signals" on NK cells. Conversely, it has been shown that a high level of cell membrane sialylation decreases the sensitivity of target cells (6,10). More recently, it has been demonstrated that the expression of MHC class I molecules protects tumor cells (predominantly those of lymphoid origin) against NK cell attack (11)(12)(13), acting as a "negative signal" for NK cell-mediated lysis. Moreover, inhibitory receptors for NK cell activation have been identified, and they bind MHC class I molecules (14,15). Finally, in the past year, a natural killer cell receptor proved to be specific for ubiquitous oligosaccharides and triggers the NK cell cytolytic mechanism (16). In conclusion, the target molecules involved in recognition of target cells by NK cells may contain peptidic or carbohydrate residues that can activate or inhibit the NK cell-mediated killing (17).
In addition to specific NK ligand(s), cell adhesion molecules (such as CD54, CD58, CD56, lymphocyte function-associated antigen 1, and CD2) are also involved in cellular cytotoxic mechanisms (18,19). However, these accessory molecules do not appear to represent NK ligands by themselves, but rather as strengthening target cell-effector cell conjugates after the initial cognate interaction between the target cell and the NK cell (19).
Recently, it has been established that resistance of K562 cells to NK cells can be correlated with an increase of glycophorin A on the cell surface (20,21). However, it has never been suggested that this glycoprotein was capable of protecting tumor cells against lysis by NK cells. The insertion of this molecule into the membrane of cells sensitive to NK cell attack gives a direct measure of its involvement in NK cell-mediated lysis. In addition, the abundant glycosylation of this molecule allows us to use an experimental approach to determine the role of the glycoprotein sugars in the mechanism of resistance to NK cell-mediated lysis.
In this report, glycophorin A was inserted into the K562 cell membrane. This was achieved by electroinsertion, which has recently been used to insert proteins with a membrane spanning sequence into mouse red blood cell membranes (22) and into nuclear cells (23,24). We demonstrate that such modified cells are resistant to NK cell-mediated lysis. The carbohydrate nature of the molecular entity that is involved in this modulation has also been determined.

MATERIALS AND METHODS
Cells-K562 cells were grown at 37°C in a humidified 5% CO 2 incubator. The culture medium was Eagle's minimum essential medium (MEM 0111, Eurobio, Paris) supplemented with 10% fetal calf serum (Boehringer, Mannheim, Germany), penicillin (100 IU/ml), streptomycin (100 mg/ml), and L-glutamine (0.58 mg/ml). The cell density was maintained between 2 ϫ 10 5 and 1.6 ϫ 10 6 cells/ml. Peripheral blood mononuclear cells from normal human volunteers were isolated on Ficoll-Hypaque (MSL, Eurobio). Peripheral blood lymphocytes (PBL) were obtained after peripheral blood mononuclear cell depletion of adherent cells by 1 h of incubation at 37°C in plastic Petri dishes. Chemicals-The pulsing buffer contained 125 mM sucrose, 69 mM KCl, 1 mM MgCl 2 , and 10 mM potassium phosphate buffer, pH 7.4. Its low ionic content reduced the joule heating associated with the electric field pulse. The washing phosphate-buffered saline (PBS) contained 42 mM K 2 HPO 4 , 8.3 mM NaH 2 PO 4 /H 2 O, and 125 mM NaCl, pH 7.4. All chemicals were from Sigma. Glycophorin A type MN (GPA) and asialoglycophorin were obtained from Sigma.
Electroinsertion-Electroinsertion was carried out as described previously for Chinese hamster ovary cells (24) using a CNRS electropulser (Jouan, St. Herblain, France) able to deliver square wave pulses, the parameters of which (voltage, pulse duration, and number and frequency of pulses) can all be adjusted separately. The pulses were monitored using a 15-MHz oscilloscope (Enertec, St. Etienne, France). Electrodes were parallel and flat with an anode-cathode distance of 1.5 mm. Since electroinsertion is a back-effect of electropermeabilization (24), various values of field intensity and pulse number and duration were applied to K562 cells to obtain both maximum viability and membrane permeabilization. Optimal conditions were one square wave pulse of 7-ms duration at 0.6 kV/cm. Consequently, electroinsertion was performed using K562 cells (10 7 ), which were incubated at 37°C for 15 min with glycophorin A or asialoglycophorin in a total volume of 18 l and then submitted to a permeabilization pulse. After pulse application, the cells were washed three times in PBS (400 ϫ g for 4 min at room temperature) and analyzed by immunofluorescence. A control sample was subjected to all these steps, except that no pulse was applied. Cell viability was checked by using the trypan blue exclusion test after pulsing (always Ͼ95%) and by observing cell growth and cell viability 24 h (always Ͼ98%) after pulsing.
Immunofluorescence Assay-GPA, asialoglycophorin, transferrin receptor (CD71), CD15 antigen, MHC class I antigen, and cellular adhesion molecules were detected by an indirect staining method using, respectively, an anti-GPA mouse monoclonal antibody (mAb) that also stains the asialoglycophorin and that binds to amino acids 27-39 of GPA (Immunotech, France), anti-CD71 mAb and anti-CD15 mAb (Becton Dickinson, Mountain View, CA), anti-MHC class I W6/32 (American Type Culture Collection, Rockville, MD), and anti-adhesion molecules, anti-CD54 mAb, anti-CD58 mAb, anti-CD29 mAb, anti-CD18 mAb, and anti-CD56 mAb (Immunotech, Marseille, France). Purified mouse IgG was used as a control reagent. Fluorescein isothiocyanate-labeled (F(abЈ) 2 ) goat antibodies against mouse immunoglobulins were used as a second-step reagent for indirect staining. The cells were then washed in PBS and analyzed by flow cytometry (Fac Scan, Becton Dickinson). Two parameters were used: the percentage of fluorescent cells after autofluorescence background subtraction and the mean of fluorescence intensity of positive cells expressed in arbitrary units.
After fluorescence intensity calibration of the cytofluorometer using quantitative fluorescent microbead standards, we used Simply Cellular TM microbeads (Becton Dickinson) to quantify the inserted glycophorin. These microbeads may be considered as "model lymphocytes" since they are approximately the size of lymphocytes. They bind mouse monoclonal antibodies (such as goat anti-mouse IgG antibodies) that are covalently bound to the microbead surface. The beads were calibrated in terms of the number of monoclonal mouse IgG molecules they bind, allowing determination of the effective fluorescence/protein ratio. We applied the same method of antibody labeling to these microbeads as the one used for cells.
Endoglycosidase Assays-Glycophorin A is composed of 60% carbohydrate by weight, with most of the sugars being 15 O-linked tetrasaccharides that are attached to serine or threonine and that have the structure shown in Structure 1 (25). A single complex N-linked oligosaccharide is also present. Its structure is as shown in Structure 2.
O-Glycanase (TEBU, Le Perray, France) catalyzes the hydrolysis of the Gal-GalNAc disaccharide core attached to serine or threonine residues of asialoglycoproteins. In this experiment, we used asialoglycophorin since sialic acid was observed to inhibit the enzyme activity (26). Cells (3 ϫ 10 6 ) were mixed with 25 milliunits/ml O-glycanase. The mixture was then incubated for 12 h at 37°C.
Endoglycosidase F (Sigma, St. Quentin, France) cleaves the link between the two N-acetylglucosamine residues linking the glycan moi-ety to the asparagine of the protein backbone (27). Cells (3 ϫ 10 6 ) were mixed with 60 milliunits/ml of enzyme. Incubation was then conducted at 37°C for 12 h.
These conditions were chosen to reach complete deglycosylation of susceptible asparagine-, serine-, or threonine-linked oligosaccharides. After enzyme treatment, the cells were washed, checked for staining with fluorescent lectins, and used in the cytotoxicity assay.
Checking Enzyme Treatments-From knowledge of GPA oligosaccharide structures, fluorescein isothiocyanate-conjugated lectins (Dolichos biflorus or Lens culinaris (Sigma)) were used to evaluate glycosylation patterns in control and endoglycosidase-treated cells. To evaluate the efficiency of endoglycosidase F treatment, K562 cells were incubated for 15 min with L. culinaris lectin (25 g/ml), which is known to react specifically with ␣-D-mannose. The cells were then washed in PBS and analyzed by flow cytometry. In the case of O-glycosylation, K562 cells were incubated for 15 min with D. biflorus lectin (50 g/ml), which is known to react specifically with ␣-D-GalNAc. The cells were then washed in PBS and analyzed by flow cytometry. These conditions were chosen (after testing various lectin concentrations) to demonstrate almost complete staining (Ͼ95% of positive cells) of K562 cells.
Cytotoxicity Assay-The NK cell activity of PBL from healthy donors was tested in a standard 4-h 51 Cr release assay against target cells labeled with 51 Cr as described previously (20). Briefly, various numbers of PBL (effector cells) were mixed in triplicate with 10 4 labeled target cells in microtiter plates. After 4 h at 37°C, 100 l of the supernatants were counted in a ␥-counter. The percentage of cell-mediated lysis was calculated as follows: % cell-mediated lysis ϭ ((cpm exp Ϫ cpm spont )/ (cpm max Ϫ cpm spont )) ϫ 100. cpm max was determined by counting an aliquot of labeled target cells, and cpm spont (spontaneous release) by counting the supernatant from wells without effectors. Cytotoxicity calculated from individual effector cell/target cell curves is expressed as lytic units (LU 25 )/10 6 effector cells. One LU 25 is defined as the number of lymphoid cells required to lyse 25% of 10 4 target cells under the assay conditions used.
Target Cell Binding Assay-PBL were depleted of CD3 ϩ cells by using anti-CD3 reactive magnetic microbeads (Immunotech). The percentage of effectors conjugated to target cells was determined using the procedure of Grimm and Bonavida (28). Briefly, 100 l (10 6 /ml) of cells from suspension of the effector and target cell populations in culture medium ϩ 10% fetal calf serum were mixed in centrifuge tubes. The tubes were placed in a water bath for 5 min at 30°C. The cells were then centrifuged for 5 min at 400 ϫ g at room temperature to promote conjugate formation. Then the pellet was resuspended 10 times with a micropipette. A small drop of this suspension was removed, and the percentage of the conjugates was determined in a hemocytometer.

Electroinsertion of Glycophorin A into the Membrane of K562
Cells-When K562 control cells were examined with our anti-GPA mAb, ϳ10% of the cells were weakly positive (ϳ1000 molecules of GPA/cell). Electroinsertion of GPA into the plasma membrane was mediated by submitting a GPA/K562 cell mixture to an electric field pulse (one pulse of 7-ms duration at 0.6 kV/cm). Electrical field application allowed us to insert the GPA, detected by anti-GPA monoclonal antibody, on Ͼ90% of the K562 cells. However, a relatively large percentage (44 Ϯ 5% for 89 M GPA; n ϭ 3) of fluorescent cells was detected even in the absence of electric field and for weak field intensities. In this cell population, the number of stained GPA molecules was always ϳ10 4 /cell. With an electrical pulse, the number of detectable glycophorin molecules/fluorescent cell rose with the increase in glycophorin concentration in medium to reach ϳ6 ϫ 10 4 molecules in the cell membrane when there was 89 M GPA in the pulsing medium (Fig. 1A).
We previously demonstrated using Chinese hamster ovary cells that electropermeabilization mediates a stable insertion of GPA into the cell membrane (24). To determine the stability of STRUCTURE 1. STRUCTURE 2. the interaction between GPA and pulsed or non-pulsed K562 cells, the GPA molecules were stained by mAb at 0, 24, and 48 h after pulsing. The percentage of fluorescent cells strongly decreased when the cells were not electropulsed and cultured for 24 and 48 h (27 Ϯ 4 and 6 Ϯ 1%, respectively, versus 41 Ϯ 5% at 0 h), whereas it was stable at 24 and 48 h after electroinsertion (83 Ϯ 6 and 79 Ϯ 6%, respectively, versus 89 Ϯ 5% directly after the pulse). In addition, the number of bound GPA molecules/cell decreased from 12,000 to 2500 after 1 day of non-pulsed cell culture and only decreased by a 2-fold factor every 24 h after electropulsation. Since the growth rate of K562 cells is one doubling/24 h, the above results indicate that (i) the GPA molecules bound to the cell membrane (termed electroinserted GPA) were shared between the daughter cells (Fig. 1B) and (ii) the GPA-cell membrane interaction is stable after pulsing. Consequently, the nature of the interaction with the membrane of electropulsed cells is likely to be different from that of the control cell membrane.
Electroinserted Glycophorin A Decreases K562 Cell Susceptibility to Attack by NK Cells- Fig. 2 (A1 and A2) shows a representative experiment of the effect of glycophorin insertion on susceptibility to NK cell-mediated lysis. The sensitivity of K562 cells to NK cell attack increased when the cells were pulsed without GPA (Fig. 2A1), whereas the presence of ϳ10 4 GPA molecules at the surface of 40% of the non-electropulsed cells did not alter susceptibility to lysis. Considering that the electropulsed cells were the effective target control, glycoph-orin had an inhibiting effect at all concentrations tested, and the resistance to NK cell lysis increased with the number of glycophorin molecules on the cell surface (Fig. 2A2).
NK cell-mediated cytotoxicity, calculated from individual effector cell/target cell curves, was cumulated and expressed as lytic units (LU 25 ) (Fig. 2B). The sensitivity of the K562 cells with electroassociated GPA (termed GPA ϩ cells) was reduced by a factor of ϳ2 as compared with the electropulsed K562 cells in the absence of GPA (P-K562; p Ͻ 0.001) and with the nonpulsed K562 cells with or without GPA in the pulsing medium (p Ͻ 0.001 and p Ͻ 0.001, respectively) (Fig. 2B). In addition, no effect was observed when GPA was directly added at different concentrations (from 8.9 to 89 M) to the NK cell cytotoxicity assay (data not shown). The spontaneous lysis of target cells induced by natural killer cell activity is accomplished in two main distinguishable steps: binding between target and effector cells and post-binding events leading to target cell destruction. To examine the possibility that the decreased lysis was due to a defect in the first step of binding, direct conjugate-forming cell assays were performed after CD3 ϩ cell depletion from PBL suspensions. Table I illustrates that the GPA ϩ K562 cells were as efficient in binding NK cells as the control samples.
Effect of the Glycophorin Insertion on the Expression of Surface Antigens-To determine whether electroinsertion of GPA into K562 cell membranes triggers a modulation of the expression of negative or positive signals for NK cells, CD71, CD15, and MHC class I antigens and certain adhesion molecules were stained by specific mAbs. No significant difference was observed between GPA ϩ cells and the control samples (Table II). Similar results were obtained when the cell positivity to mAbs was expressed as mean fluorescence intensity (data not shown). These results indicate that the resistance to NK cell attack induced by glycophorin insertion into the membrane depends neither on the triggering of MHC class I antigen expression nor on the modulation of the expression of other molecules. In addition, they demonstrate that insertion of GPA in the cell membrane does not interfere with accessibility of mAbs to natural epitopes of the cell surface.
Effect of the Glycosylated Structures on NK Cell-mediated Lysis-It has been reported that cell-surface sialic acid may contribute to the development of NK cell resistance directly or by masking the target structure(s) to NK cells (6,10). As GPA is a highly sialylated protein, we tested the hypothesis that the resistance induced by GPA insertion could be correlated to a simple contribution from the sialic acid. Asialoglycophorin was electroinserted into K562 cells. Fig. 3 shows that the same number of asialoglycophorin and GPA molecules induced the same resistance of target cells to NK cell-mediated cytotoxicity.
We were also interested in determining whether changes, other than sialic acid, in glycosylation of the inserted glycophorin would alter the resistance of GPA ϩ cells to NK cell-mediated lysis. After GPA electroinsertion into the cell membrane, N-and O-glycosylation were eliminated by enzymatic treatment as indicated under "Materials and Methods." A comparative study of lectin binding to enzyme-treated and -untreated K562 cells (GPA ϩ and K562) revealed a significant reduction of binding after treatment of GPA ϩ cells, suggesting a decrease of N-linked (Fig. 4A) or O-linked (Fig. 4B) oligosaccharides, whereas the enzyme treatment did not affect the glycosylation of control cells. The effect of enzyme treatments on the susceptibility of K562 cells to NK cell-mediated lysis was then determined while fluorescein-labeled lectins were used in parallel to control deglycosylation on the target cells. K562 cells treated with endoglycosidase F or O-glycanase were as sensitive to NK cell lysis as the control samples. GPA ϩ cells treated with the O-glycanase enzyme were lysed at the same level as the untreated GPA ϩ cells (Fig. 5A), whereas the elimination of Nlinked oligosaccharides led to the restoration of the K562 cell sensitivity to NK cell attack (Fig. 5B). DISCUSSION In this report, we have shown the direct involvement of glycophorin A in the NK cell-mediated lysis mechanism by electroinserting this molecule into the K562 cell membrane. In our system, ϳ6 ϫ 10 4 GPA molecules/cell were correctly oriented so as to be detected by an anti-GPA monoclonal antibody that reacts with an extracellular epitope of GPA. As previously demonstrated with Chinese hamster ovary cells (23,24), the inserted protein was stable in the membrane, was transferred from mother to daughter cells, and was able to diffuse freely across the surface of the cell membrane (data not shown). On the non-electropulsed cell membrane, it is likely that the GPA molecule was only adsorbed, as previously reported (24). On these control cells, the presence of ϳ10 4 GPA molecules did not induce resistance to lysis, suggesting that the putative inhibitive structure is not efficiently presented to NK cells when GPA was only adsorbed on the target cell surface.
Our present data show that the resistance to natural killer cells correlates with the increase in the number of GPA molecules on the K562 cell membrane. A significant resistance to

Effect of glycophorin insertion on the binding of NK cells to K562 cells
Values are the percentage of conjugate-forming cells using CD56 ϩenriched PBL (67% of CD56 positive cells in Experiment 1 and 51% of CD56 positive cells in Experiment 2) from two healthy donors. The results are expressed as the means Ϯ S.D. of triplicate samples. K562 represents original cells, P-K562 represents electropulsed cells (one pulse of 7-ms duration at 0.6 kV/cm in the absence of glycophorin), and GPA ϩ K562 represents cells with electrically inserted GPA (one pulse of 7-ms duration at 0.6 kV/cm in the presence of glycophorin (89 M)).

T ABLE II Effect of glycophorin insertion on the surface antigen expression
The expression of surface antigens was determined by fluorescence using specific monoclonal antibodies against each antigen. The levels of antigen expression on K562 cells with electrically bound GPA (GPA ϩ K562; 50,000 -60,000 GPA molecules/cell) were compared with those of the original K562 cells and electropulsed K562 cells (P-K562) (one pulse of 7-ms duration at 0.6 kV/cm). Results (fluorescent cell percentage after autofluorescence background deduction) are expressed as the means Ϯ S.D. of three separate experiments.  NK cell attack was observed with ϳ3 ϫ 10 4 electroinserted molecules/cell (i.e. 1 GPA molecule/2 ϫ 10 4 nm 2 of cell surface or 1 GPA molecule/2 ϫ 10 4 phospholipids). Consequently, the resistance to NK cell-mediated lysis can be effective when the GPA number is higher than a threshold value of between 10 4 and 3 ϫ 10 4 molecules. On the other hand, the target resistance induced by electroinserted GPA may be attributed, at least partly, to the stabilization of the membrane. Indeed, a GPA molecule incorporated into experimental bilayers interacts with ϳ500 -1000 phospholipids (29,30), which could alter the membrane stability. However, this possibility seems somewhat unlikely since (i) the interaction between GPA and phospholipids would involve only 5-10% of the membrane phospholipids; (ii) GPA electroinserted into Chinese hamster ovary cell membrane showed a free lateral diffusion with a diffusion coefficient in agreement with what would be expected for an intrinsic protein embedded in a viable cell membrane (23); and (iii) removal of N-linked oligosaccharide moieties suppresses the resistance. This relationship between the resistance to NK cell-mediated lysis and the inserted glycophorin A level is in accordance with previous results that showed a reduced sensitivity to NK cell lysis of (i) K562 cells differentiated in vitro by drugs that increased the levels of GPA on the cell surface (10,21) and (ii) a K562 cell clone expressing a very high number of GPA molecules (21). It is important to stress that the presence of ϳ6 ϫ 10 4 glycophorin molecules on the cell surface does not confer total resistance to NK cell lysis. This can be explained by the intervention of several NK cell subsets using different mechanisms to lyse K562 cells in chromium assay or by the choice of each effector cell between different recognition strat-egies. Moreover, the resistance to NK cells might only be partial because GPA ϩ cells maintain malignant features, i.e. positive signals for NK cells such as the absence of MHC class I antigens and/or the presence of ligands able to activate NK cells such as CD71 and CD15 antigens or some carbohydrate determinants (3)(4)(5)(6)(7)(8)(9)16).
In our system, the protective effect of GPA might be due to masking (i) of cellular adhesion molecules involved in NK cell mechanisms and/or (ii) of putative epitopes able to deliver an activating signal to NK cells. Both explanations are unlikely because we demonstrated that GPA electroinsertion did not alter the accessibility of cell adhesion molecules (CD18, CD29, CD54, CD56, and CD58), CD71, and CD15 by mAbs. In addition, no consistent reduction of target cell binding to NK cells was recorded in the experiments using GPA ϩ cells. However, it is well known that conjugate formation measures integrin binding and represents an early stage of binding that is followed by a cellular reorientation and firm adhesion that probably uses other molecules. Consequently, GPA might modulate a limiting step at the level of the membrane molecular interactions involved in post-conjugating mechanisms, such as in the stabilization of binding and/or in the lethal hit (e.g. it is quite possible that GPA blocks perforin adherence).
GPA is a sialoglycoprotein made up of 131 amino acid residues with no disulfide bonds and is composed of 60% carbohydrate by weight. The sugar chains contain ϳ45% sialic acid by weight (25). As mentioned above, several works have shown that the level of sialic acid on target cells may modulate their sensitivity to NK cell-mediated lysis, e.g. neuraminidase medi- ates an increase of the susceptibility to NK cell lysis (10). However, in agreement with previous data (20), we have demonstrated that sialic acid is not an essential residue in the inhibition of NK cell-mediated lysis due to inserted GPA.
On the contrary, this work indicates a role for the N-linked oligosaccharide of GPA in NK cell activity modulation because endoglycosidase F, which eliminates N-linked oligosaccharide structure, completely reverses the resistance to NK cells. On the other hand, the glycophorin O-linked oligosaccharide seems to have no effect on NK cell activity. However, glycosidase treatment of control K562 cells did not clearly alter their reactivity to lectins or their sensitivity to NK cells. This is probably due to the biosynthesis of new sugar chains for the membrane molecules, whereas these mechanisms cannot act on electroinserted GPA. The GPA molecule contains 15 O-linked tetrasaccharides and a single complex N-linked oligosaccharide (13 monosaccharide residues). Several works have emphasized the importance of carbohydrate molecules in target cell-effector cell interactions (17). According to the structures and the location of the extra sugar residues added to a pentasaccharide common core, all the N-linked sugar chains are classified into three subgroups (31): (i) complex-type sugar chains, (ii) high mannose-type sugar chains, and (iii) hybrid-type sugar chains. Recently, by using N-glycan processing inhibitors, it has been demonstrated that the presence of high mannose-type glycans on K562 cells correlates with increased binding of effectors and a greater susceptibility to lysis. The high mannose-type glycans can influence the NK cell-target cell interaction at the level of the adhesion molecules (32). However, the N-linked sugar chain of GPA belongs to the complex-type sugar chain, and its presence on the target cell surface decreases the susceptibility to NK cells without altering the conjugate formation. Thus, two types of sugar chain with strong differences in their structure can modulate the NK cell activity in opposite directions.
As mentioned above, NK cell receptors have been identified that bind MHC class I molecules and inhibit natural killer cell activation (14,15), and it is possible that carbohydrates of the MHC glycoprotein participate in the inhibition of NK cellmediated cytotoxicity (15). On the other hand, it has recently been shown in mice that members of the type II transmembrane lectin family are preferentially expressed on NK cells and can deliver either positive (NKR-P1 protein) (33) or negative (Ly49 protein) signals to the effector cell (34). In addition, NKR-P1 binds a diversity of oligosaccharides that activate NK cells and cytotoxicity (16). In humans, the same type of membrane proteins (named NKG2 proteins and with carbohydratebinding external domains) was also detected on NK cells (34). Their ligands and their effect on natural killing are unknown (34). It is possible that sugar residues of the N-linked sugar chain, shared by several glycoproteins, fulfill the role of ligand molecule for these putative NK cell receptors or for other unknown molecules characterized by lectin activity. From this point of view, sugar residues of the N-linked oligosaccharide of GPA could be a ligand for a putative receptor that delivers a negative signal to NK cells.