Characterization of Phosphatidylserine-dependent β2-Glycoprotein I Macrophage Interactions

The binding and uptake of phosphatidylserine (PS)-expressing cells appears to involve multiple receptor-mediated systems that recognize the lipid either directly or indirectly through intermediate proteins that form a molecular bridge between the cells. Here we show that β2-glycoprotein I (β2GPI), a 50-kDa serum glycoprotein, binds PS-containing vesicles and serves as an intermediate for the interaction of these vesicles with macrophages. Chemical modification of lysines and cysteines abolished β2GPI-dependent PS uptake by inhibiting the binding of PS to β2GPI and the binding of PS·β2GPI complex to macrophages, respectively. Recognition was mediated by β2GPI and not by the lipid because antibodies to β2GPI inhibited binding of the complex to macrophages. These results indicate that human (THP-1-derived) macrophages bind β2GPI only after it is bound to its lipid ligand. Competition experiments with monosaccharides that inhibit lectin-dependent interactions, and PS·β2GPI binding experiments using deglycosylated β2GPI, suggested that carbohydrate residues were not required for macrophage recognition of the complex. Antibodies to putative macrophage PS receptors (CD36, CD68, and CD14) did not inhibit uptake of the complex. These data suggest that β2GPI can bind cells that fail to maintain membrane lipid asymmetry and generate a specific bridging moiety that is recognized for clearance by a phagocyte receptor that is distinct from CD36, CD68, and CD14.

The emergence of phosphatidylserine (PS) 1 in the cells outer leaflet results in the expression of altered cell surface properties that regulates their recognition by phagocytes (1)(2)(3). Although PS recognition might include binding to specific PS receptors (1,4,5), class B scavenger receptors (6 -8), the lipopolysaccharide receptor (2, 9 -11) or thrombospondindependent vitronectin receptors (12), recent evidence suggests that ␤ 2 GPI, a relatively abundant plasma protein (13), binds PS (14,15) and mediates its uptake by phagocytic cells. Indeed, the binding of ␤ 2 GPI to PS-containing liposomes (16), PS expressing apoptotic thymocytes (17,18) and symmetric red blood cell ghosts (18) has been shown to influence their clearance and phagocytosis. These observations raise the possibility that, in addition to regulating thrombosis (19), ␤ 2 GPI plays an important physiologic role by mediating the recognition of cells that fail to preserve membrane lipid asymmetry.
Although the binding of ␤ 2 GPI to PS-containing surfaces has been well characterized (15,20,21), little is known about its interaction with phagocytes. Here we report on the nature of ␤ 2 GPI/phagocyte interactions using human THP-1-derived macrophages as a model system. We show that human ␤ 2 GPI binds macrophages only after it is complexed to its lipid ligand. Our results suggest that PS binding by ␤ 2 GPI induces a lipiddependent conformational change that exposes a specific epitope that is recognized by a macrophage receptor that is distinct from other previously described lipid receptors.
Binding of ␤ 2 GPI to PS-The binding of ␤ 2 GPI to phospholipids was monitored by lipoblotting and by gel diffusion using 125 I-labeled small unilamellar vesicle. For the lipoblot, unmodified and chemically modified ␤ 2 GPI were electrophoresed on 10% SDS-polyacrylamide gels and transferred to polyvinylidene difluoride membrane. The membrane was blocked with 1% ovalbumin in TBS and incubated for 60 min at 20°C with 125 I-labeled PS/PC (1/1) small unilamellar vesicles at 0.5 mg of lipid/ml in TBS. Unbound lipid was removed by washing in TBS. Binding of the labeled vesicles to ␤ 2 GPI was determined by autoradiography. For the gel diffusion, microscope slides were coated with 0.9% agarose in 10 mM Tris-HCl, pH 7.4 with or without the indicated inhibitors. Two holes (20-l loading volume) were punched 1 cm apart and filled with ␤ 2 GPI (300 g/ml) and 125 I-labeled PS/PC (1/1) small unilamellar vesicles. The plates were developed for 24 h, and unbound protein and lipid was removed by washing for 24 h in the same buffer. The gels were then dried, and precipitates were detected by autoradiography (18).
Macrophage Uptake-Adherent macrophages (ϳ10 6 cells) in 24-well Costar plates were washed with TBS and resuspended in 250 l of RPMI 1640 medium (without serum) containing multilamellar vesicle-␤ 2 GPI complexes (200 g of lipid/100 g of ␤ 2 GPI/ml). The cells were incubated at 37°C for the indicated time, washed, and solubilized in 0.2% SDS. Lipid uptake was determined by scintillation counting.

RESULTS
␤ 2 GPI-dependent PS Uptake by Macrophages-The effect of ␤ 2 GPI and PS on the uptake of 125 I-labeled PS-containing vesicles was determined. Fig. 1A shows the uptake of PS/PC (1/1) vesicles by macrophages as a function of ␤ 2 GPI concentration. Maximal uptake occurred at ϳ2 M ␤ 2 GPI, which is approximately half the concentration found in normal human serum (13). Fig. 1B shows the effect of PS concentration on vesicle uptake. The inclusion of ␤ 2 GPI enhanced the uptake of 50 mol % PS Ͼ4-fold during the 1-h time course of the experiment. Interestingly, the degree of ␤ 2 GPI-dependent enhance-ment of uptake diminished at PS concentration Ͼ70 mol %, suggesting the existence of both ␤ 2 GPI-dependent and -independent uptake mechanisms. Increasing the incubation time resulted in ϳ7-fold enhancement in ␤ 2 GPI-dependent uptake (Fig. 1C).
Ligand Dependence of ␤ 2 GPI Binding to Macrophages-The influence of ␤ 2 GPI on the uptake of PS vesicles by macrophages can be explained by at least two mechanisms. 1) ␤ 2 GPI binds independent of its lipid ligand (PS) to macrophages, and 2) the binding of ␤ 2 GPI to macrophages is dependent on its prior binding to PS. To differentiate between these two possibilities, macrophages were preincubated with ␤ 2 GPI, washed, and then assessed for lipid uptake. Fig. 2A shows that macrophages incubated with ␤ 2 GPI in this manner did not exhibit enhanced PS uptake. On the other hand, uptake was similar to control levels when PS vesicles were added to macrophages that were not washed free of ␤ 2 GPI, suggesting that ␤ 2 GPI binding to macrophages has a lipid dependence. To verify that binding of ␤ 2 GPI was ligand-dependent, macrophages were incubated for 20 min with ␤ 2 GPI alone, or in complex with PS, and assessed for cell-bound ␤ 2 GPI by staining with fluorescent ␤ 2 GPI antibodies. No fluorescence was seen on cells incubated with ␤ 2 GPI alone confirming that ␤ 2 GPI binding is ligand dependent ( Fig.  2A, inset). Indeed, acylation of lysines critical to ␤ 2 GPI lipid binding (28) with diketene abrogated the uptake of PS vesicles (Fig. 2B).
While ␤ 2 GPI is clearly required for the efficient uptake of PS, it is possible that the binding of the PS⅐␤ 2 GPI complex to the macrophage membrane triggers a generalized phagocytic pathway that engulfs particles independent of ␤ 2 GPI. To test this, mixed PS/PC and PC vesicle experiments in which the 125 Ilabel was incorporated into either population was carried out. Fig. 3 shows that cells bound 125 I-PS/PC vesicles but not to 125 I-PC vesicles. Moreover, when mixed PS/PC and PC vesicles in which the radiolabel was incorporated in the PC population was added to the macrophages, uptake of the radiolabel was similar to that of 125 I-PC vesicles alone, indicating selective uptake of the ␤ 2 GPI bound population only.
Effect of Chemical Modification of ␤ 2 GPI on PS⅐␤ 2 GPI Uptake-To assess the amino acid moieties critical to macrophage binding of PS⅐␤ 2 GPI complexes, ␤ 2 GPI was treated with the indicated reagents, purified, complexed to 125 I-PS/PC vesicles, and assessed for macrophage uptake. As shown in Fig. 4A, modification of histidines (phenacylbromide) and arginines (cyclohexanedione and phenylglyoxal) did not inhibit uptake, 2 whereas blocking of lysines (diketene) and cysteines (NEM) did. Although inhibition in the case of lysine modification was indirect (due to inhibition in ligand binding; see Fig. 2B), acylation of cysteines with NEM directly inhibited the binding of PS⅐␤ 2 GPI to macrophages. This was concluded from the results of lipoblot (Fig. 4C) and gel diffusion (Fig. 4D) experiments which showed that the NEM-treated protein, unlike diketenetreated protein, still bound 125 I-PS/PC vesicles, and by the finding that the inhibition was bypassed by the addition of ␤ 2 GPI antibodies (Fig. 4B). It should be noted that circumvention of inhibition by ␤ 2 GPI antibodies was due to the binding of the PS⅐␤ 2 GPI⅐IgG complex to the Fc receptor of the cells. Similar experiments carried out with F(abЈ) 2 fragments resulted in inhibition of PS⅐␤ 2 GPI uptake, suggesting that the antibodies bind to a ␤ 2 GPI site that is critical to macrophage recognition (Fig. 5).
Role of Carbohydrate on Macrophage Uptake-To determine the role carbohydrates might play in the binding of PS⅐␤ 2 GPI to macrophages, PS⅐␤ 2 GPI was incubated with macrophages in the presence of various monosaccharides. The data presented in Fig. 6A show that these monosaccharrides did not inhibit uptake, suggesting that macrophage PS⅐␤ 2 GPI interactions do not involve lectin-like carbohydrate binding moieties. Interestingly, deglycosylation of ␤ 2 GPI enhanced uptake ϳ3-fold (Fig.  6B), possibly due to decreased charge repulsion by removal of sialic acid residues.
Effect of Charge on PS⅐␤ 2 GPI Interaction with Macrophages-To determine the nature of PS⅐␤ 2 GPI-macrophage interactions, uptake was assessed in the presence of amino acids and negatively charged groups. Table I shows that lysine and arginine inhibited uptake, whereas serine, leucine, and valine did not. Interestingly, while phosphoserine, phosphate, succinate, and butyrate inhibited uptake, aspartate and glutamate were without effect. Analysis of PS⅐␤ 2 GPI interaction by gel diffusion (18) in the presence of the inhibitors (data not shown) indicated that inhibition of binding to the macrophage surface was not because of dissociation of the PS⅐␤ 2 GPI complex. It should be noted that these inhibitors were specific for PS⅐␤ 2 GPI complexes because they did not significantly influence macrophage uptake of PS alone ( Table I).
Effect of Antibodies to Phagocyte Receptors on PS⅐␤ 2 GPI Uptake-To determine whether putative PS recognition pathways are involved in the specific uptake of PS⅐␤ 2 GPI complexes by macrophages, uptake was determined in the presence of antibodies to CD36 (6 -8), CD68 (4), and CD14 (2,9,11). As determined by fluorescence microscopy using fluorescein-conjugated secondary antibodies, all the monoclonals, with the exception of CD36, bound to the macrophage membrane (not shown). Unlike the significant inhibition (ϳ50%) obtained with apoptotic cells (9,11) and symmetric red blood cell ghosts (2), 61D3 antibodies (anti-CD14) did not significantly inhibit the uptake of PS⅐␤ 2 GPI (Table II). Antibodies against other macrophage surface antigens (CD11b, CD36, and CD68) and against other CD14 epitopes (monoclonals TUK4 and UCHM-1) were also without significant effect. Surprisingly, all of the monoclonals with the exception of 61D3 significantly enhanced the uptake of PS/PC vesicles in the absence of ␤ 2 GPI. The reasons for this observation are not clear.

DISCUSSION
␤ 2 GPI is a well characterized plasma glycoprotein that binds negatively charged phospholipids. This property is responsible for regulating thrombosis by competing with clotting factors for PS expressed on the surface of activated platelets (19). Similar PS binding activities have also been shown to occur with synthetic negatively charged phospholipid vesicles (15,16) and apoptotic thymocytes (17,18). Formation of these PS⅐␤ 2 GPI complexes is associated with rapid clearance of the PS-expressing particle from the peripheral circulation (16) and phagocytosis of PS-expressing cells in vitro (18). Interestingly, the binding of ␤ 2 GPI to PS has been shown to result in a major change in the proteins conformation (29,30) that might result in the expression of a new epitope (31), which, in certain individuals, generates an autoimmune response. While the mechanism for the generation of these immune responses is not known, this new epitope could be responsible for the recognition and removal of PS-expressing apoptotic and senescent cells from the host. 2 The reason for the ϳ7-fold increase in PS⅐␤ 2 GPI uptake in the case of phenylglyoxal-modified ␤ 2 GPI is not known. Using an in vitro model for macrophage binding, we showed that the uptake of PS liposomes by macrophages was greatly enhanced by the addition of ␤ 2 GPI that saturated at ϳ2 M. Since the average concentration of ␤ 2 GPI in normal human plasma is about 4 M (13), these data raise the possibility that this property could be the principal function of ␤ 2 GPI in vivo. Interestingly, as shown in Fig. 2, ligand-free ␤ 2 GPI did not bind macrophages. However, once ␤ 2 GPI bound PS, the complex became rapidly associated with the macrophage membrane. The requirement for ligand binding was further shown by the inability of macrophages to bind diketene-treated (lysine-blocked) ␤ 2 GPI even in the presence of PS.
PS⅐␤ 2 GPI complex binding to macrophages was shown to be inhibited by lysine and arginine as well as negatively charged FIG. 4. Effect of chemical modification on ␤ 2 GPI-dependent uptake of PS/PC vesicles. Chemically modified ␤ 2 GPI was preincubated with 125 I-PS/PC vesicles, and macrophage uptake was determined, A, in the absence and B, presence of ␤ 2 GPI antibodies. Lipid binding of chemically modified ␤ 2 GPI was determined by C, lipoblot and D, gel diffusion analysis. ␤ 2 GPI treatments: 1, control (not treated); 2, diketene; 3, NEM; 4, phenacylbromide; 5, cyclohexanedione; 6, phenylglyoxal; and the negative controls; 7, PS/PC vesicles alone; and 8, PC vesicles alone. D, clockwise left, control ␤ 2 GPI, phenacylbromide, NEM, cyclohexanedione, and diketene. Right, control, cyclohexanedione, and phenylglyoxal. ions including organic acids and phosphate (Table I). These compounds did not affect ␤ 2 GPI binding to its ligand since they did not inhibit precipitation of the complex in a gel diffusion assay. This suggests that the binding of the complex to the macrophage membrane requires specific electrostatic interactions that are unlike those involved in the binding of ␤ 2 GPI to its ligand.
Amino acid analysis showed that ␤ 2 GPI consists of five characteristic sushi domains (32) with domain V being principally responsible for the lipid binding properties of the protein (28). While the lysines present on domain V are known to be critical for lipid binding (28), the experiments described here cannot exclude the possibility that these or other lysines are also required for ligand-dependent macrophage binding. Acylation of sulfhydryls with NEM, on the other hand, inhibited PS-dependent macrophage binding but not PS binding to ␤ 2 GPI (Fig.  4). Because human ␤ 2 GPI might contain a free sulfhydryl (Cys 102 and/or Cys 169 ) (33,34), it is possible that acylation of one or both these moieties inhibits the putative conformational change required to promote macrophage binding of the PS⅐␤ 2 GPI complex. This raises the possibility that domains II and/or III, harbor a hidden moiety critical to macrophage binding. Additional support for the involvement of these domains in macrophage binding can be obtained from results which showed that removal of the carbohydrate moieties (33) resulted in Ͼ3-fold enhancement in ␤ 2 GPI-dependent uptake (Fig. 6).
Several studies have suggested that the redistribution of PS from the cells inner to outer leaflet signals for removal of these cells by the reticuloendothelial system (1)(2)(3)16). Although several mechanisms might be responsible for phagocyte recognition of PS expressing apoptotic cells, Price et al. (17) proposed that the interaction of circulating ␤ 2 GPI with redistributed anionic phospholipid may, by itself, generate a novel ligand by which apoptotic cells are recognized. Indeed, other studies have indicated that ␤ 2 GPI could play a central role in this recognition process (16). Combined with these previous studies, the data presented here provide evidence for the existence of a receptor on the macrophage membrane that specifically binds ␤ 2 GPI in a ligand-dependent manner. Although the motif on the protein responsible for the macrophage membrane interaction is not known, it could result from a lipid-dependent conformational change (29,30) that is specifically bound to a cell surface receptor, a process that can be inhibited with anti-␤ 2 GPI F(abЈ) 2 (Fig. 5). Several reports have suggested the involvement of macrosialin (CD68) (4), scavenger receptor (CD36) (6 -8), and the lipopolysaccharide receptor (CD14) (2,9,11) in the recognition of PS on apoptotic cells. Antibodies directed against these cell surface moieties, however, did not significantly inhibit ␤ 2 GPI-dependent PS uptake. The inability to obtain more than 20% inhibition raises the possibility that either more than one cell surface component is involved in ␤ 2 GPI-dependent recognition or that recognition is inhibited because of steric hindrance by antibody bound to an unrelated proximal site. Although further studies will be required to identify the putative PS⅐␤ 2 GPI-dependent macrophage receptor, the data presented here argue for the existence of such a receptor. Because of the relative abundance of ␤ 2 GPI in plasma, it could play an important physiologic role by bridging PS-expressing cells to phagocytes for their ultimate disposal.  a The antibody concentrations used were: ␤ 2 GPI F(abЈ) 2 , 80 g/ml; CD11b, 37 g/ml; CD36, 100 g/ml; CD68, 67 g/ml; CD14 (TUK4), 12.5 g/ml; CD14 (UCHM-1), 44 g/ml; CD14 (61D3), 67 g/ml.
b Results are expressed as counts/min uptake with percent uptake in parentheses.
c Fluorescent staining failed to detect CD36 antigen on the cell surface.