High Affinity Binding of b 2 -Glycoprotein I to Human Endothelial Cells Is Mediated by Annexin II*

b 2 -Glycoprotein I ( b 2 GPI) is an abundant plasma phos- pholipid-binding protein and an autoantigen in the antiphospholipid antibody syndrome. Binding of b 2 GPI to endothelial cells targets them for activation by anti-b 2 GPI antibodies, which circulate and are associated with thrombosis in patients with the antiphospholipid antibody syndrome. However, the binding of b 2 GPI to endothelial cells has not been characterized and is assumed to result from association of b 2 GPI with mem- brane phospholipid. Here, we characterize the binding of b 2 GPI to endothelial cells and identify the b 2 GPI binding site. 125 I- b 2 GPI bound with high affinity ( K d ; 18 n M ) to human umbilical vein endothelial cells (HUVECs). Using affinity purification, we isolated b 2 GPI-binding proteins of ; 78 and ; 36 kDa from HUVECs and EAHY.926 cells. Amino acid sequences of tryptic peptides from each of these were identical to sequences within annexin II. A role for annexin II in binding of b 2 GPI to cells was confirmed by the observations that annexin II-transfected HEK 293 cells bound ; 10-fold more 125 I- b 2 GPI than control cells and that anti-annexin II antibodies inhibited the binding of 125 I- b 2 GPI This resulted in the immobilization of 1062 RU of b 2 GPI. The binding of increasing concentrations of annexin II, delivered at a flow rate of 30 m l/min, was then measured in real time. Binding data was analyzed by Global analysis using BiaEval 3.0 software (Biacore), in which the association and dissociation data for a series of annexin II concentrations is fit simultaneously (65). In parallel, association data was analyzed following linear transformation (66). The equation used in these studies was dRU/dt 5 k a [annexin II]RU max – RU( k a [annexin II] 1 k d ), where RU max 5 the maximal binding response. The use of this equation to derive the K d from real time surface plasmon resonance data has been described (66). This approach allows the definition of fast and slow components of association and thus reveals binding heterogeneity (66, 67).

␤ 2 -Glycoprotein I (␤ 2 GPI) is an abundant plasma phospholipid-binding protein and an autoantigen in the antiphospholipid antibody syndrome. Binding of ␤ 2 GPI to endothelial cells targets them for activation by anti-␤ 2 GPI antibodies, which circulate and are associated with thrombosis in patients with the antiphospholipid antibody syndrome. However, the binding of ␤ 2 GPI to endothelial cells has not been characterized and is assumed to result from association of ␤ 2 GPI with membrane phospholipid. Here, we characterize the binding of ␤ 2 GPI to endothelial cells and identify the ␤ 2 GPI binding site. 125 I-␤ 2 GPI bound with high affinity (K d ϳ18 nM) to human umbilical vein endothelial cells (HUVECs). Using affinity purification, we isolated ␤ 2 GPI-binding proteins of ϳ78 and ϳ36 kDa from HUVECs and EAHY.926 cells. Amino acid sequences of tryptic peptides from each of these were identical to sequences within annexin II. A role for annexin II in binding of ␤ 2 GPI to cells was confirmed by the observations that annexin II-transfected HEK 293 cells bound ϳ10-fold more 125 I-␤ 2 GPI than control cells and that anti-annexin II antibodies inhibited the binding of 125 I-␤ 2 GPI to HU-VECs by ϳ90%. Finally, surface plasmon resonance studies revealed high affinity binding between annexin II and ␤ 2 GPI. These results demonstrate that annexin II mediates the binding of ␤ 2 GPI to endothelial cells. ␤ 2 -Glycoprotein I (␤ 2 GPI) 1 is an abundant plasma glycoprotein that consists of five homologous domains of approximately 60 amino acids each (1,2). The first four of these are classical short consensus repeat domains, with extensive homology to those found in the complement-type repeats of Factor H (3). Each of these contains four conserved cysteines with a characteristic C 1-3 , C 2-4 disulfide bonding pattern, whereas domain 5 contains six cysteines, with C 1-4 , C 2-5 , and C 3-6 disulfide linkages (4,5). Domain 5 is also unique in its high content of lysine residues (5,6). In the recently solved crystal structure of ␤ 2 GPI, these have been shown to contribute to the formation of a positively charged phospholipid binding region in domain 5 (7,8).
The physiological function(s) of ␤ 2 GPI is uncertain. A role in lipid metabolism is suggested by the observations that 30% of plasma ␤ 2 GPI circulates in complex with lipoproteins (2,9) and that ␤ 2 GPI accelerates triglyceride clearance in mice (10). ␤ 2 GPI also binds with high affinity to the atherogenic lipoprotein, Lp(a) (11). A role for ␤ 2 GPI as a naturally occurring anticoagulant is suggested by reports that it inhibits the prothrombinase (12) and Factor X activating complexes (13), although the physiologic importance of these effects is uncertain. ␤ 2 GPI may also promote the clearance of senescent cells (14,15) and regulate the uptake of lipoproteins by macrophages (16).
Recently, ␤ 2 GPI has been found to be an important autoantigen in the antiphospholipid antibody syndrome (17)(18)(19), a disorder characterized by thrombosis and recurrent fetal loss in patients with circulating "antiphospholipid" antibodies (20,21). It is now generally accepted that most antiphospholipid antibodies associated with the antiphospholipid antibody syndrome recognize ␤ 2 GPI bound to the cardiolipin-coated microplates used in clinical "anticardiolipin" assays (19,(22)(23)(24)(25)(26)(27). Binding of ␤ 2 GPI to cardiolipin or another appropriate surface results in either a conformational change in the protein, exposing antigenic neoepitopes (28 -30), or concentration of ␤ 2 GPI to an antigenic density at which it is more avidly bound by low affinity anti-␤ 2 GPI antibodies (31). Antiphospholipid antibodies may also recognize epitopes occurring as a consequence of the formation of adducts between ␤ 2 GPI and oxidized cardiolipin (32,33).
It has been suggested that ␤ 2 GPI may contribute to the pathogenesis of aPS-associated thrombosis by binding to platelets or endothelial cells and targeting them for anti-␤ 2 GPI antibody-dependent activation. Binding of ␤ 2 GPI to these cells has been assumed to result from its interaction with membrane phospholipid. However, although ␤ 2 GPI binds with high affinity to purified anionic phospholipids (34,35), its affinity for phospholipid preparations with a composition resembling that of cell membranes is low (36,37). Consistent with this observation, ␤ 2 GPI binds with only micromolar affinity (38,39), if at all (40), to activated platelets. In addition, human anti-␤ 2 GPI * This work was supported by Grant HL50827, a grant from the American Diabetes Foundation, and an Established Investigator Award from the American Heart Association (to K. R. M. antibodies have not been convincingly shown to induce platelet activation (41).
In the nonactivated state, several anticoagulant moieties that play a central role in the maintenance of blood fluidity are expressed on the endothelial cell (42,43). These include, among others, heparan sulfate proteoglycans, which bind and activate antithrombin (42), and thrombomodulin, which redirects the proteolytic activity of thrombin toward the activation of the natural anticoagulant, protein C (44). However, a variety of stimuli may induce endothelial cell activation, a process in which numerous transcriptional and posttranscriptional events occur that lead to the expression of adhesion molecules and procoagulant activity on the endothelial surface (44 -46). These changes are associated with an increased risk of thrombosis in several clinical settings (46). That endothelial activation contributes to the pathogenesis of antiphospholipid antibody-associated thrombosis is suggested by the presence of increased levels of endothelial-derived proteins and microparticles in the plasma of patients with antiphospholipid antibodies (47)(48)(49). Furthermore, in contrast to platelets, anti-␤ 2 GPI antibodies have been convincingly shown to activate endothelial cells in a ␤ 2 GPI-dependent manner (50,51), although the mechanisms by which they do so remain undefined.
We hypothesized that endothelial cell activation mediated through this ␤ 2 GPI-dependent pathway would require a high affinity interaction between ␤ 2 GPI and a specific endothelial cell receptor, with receptor cross-linking subsequently induced indirectly through binding of anti-␤ 2 GPI antibodies to receptorbound ␤ 2 GPI. As an initial step in evaluating this hypothesis, we have characterized the binding of ␤ 2 GPI to endothelial cells. ␤ 2 GPI bound to endothelial cells through a high affinity interaction with annexin II, an endothelial cell receptor for tissuetype plasminogen activator (t-PA) (52-55) and plasminogen (52,55,56). Preliminary studies also suggest a potential role for annexin II in mediating anti-␤ 2 GPI antibody-mediated endothelial cell activation.
Cells-Human umbilical vein endothelial cells (HUVECs) were isolated as described and cultured in medium 199 (M199) containing 10% fetal bovine serum, 75 g/ml endothelial cell growth supplement, 2 mM glutamine, and penicillin-streptomycin (complete medium) (58). All cells used were of passage 3 or lower. EAHY.926 cells (a hybrid human cell line derived from fusion of HUVECs and A549 carcinoma cells (59)), HEK 293, and MDA-MB-231 cells were cultured in RPMI 1640 medium containing 5% fetal bovine serum and 2 mM glutamine.
Proteins-Fibronectin (60) and ␤ 2 GPI (9, 18) were isolated from human plasma as described previously (9,18,60). Purified, nonreduced ␤ 2 GPI migrated as a single ϳ50-kDa band when analyzed by SDS-PAGE, displayed the characteristic increase in M r to ϳ65,000 upon reduction, and was recognized on immunoblots by anti-␤ 2 GPI antibodies. ␤ 2 GPI was radiolabeled using Iodogen, per the manufacturer's protocol. Anionic phospholipid binding activity of the radiolabeled protein was confirmed by its ability to bind specifically to cardiolipincoated microplates (5).
Binding of 125 I-␤ 2 GPI to HUVECs-In initial studies, we observed that ␤ 2 GPI bound "specifically" to polystyrene tissue culture plates. We therefore developed an assay to minimize such interactions. Briefly, 96-well Immulon II plates were pretreated with Aquasil (0.5 mg/ml) and then coated with 50 l of a 20 g/ml solution of fibronectin. Endothelial cells were plated in individual wells and cultured in complete medium until confluent. Prior to binding assays, cells were washed and incubated for 6 h with serum-free medium 199 containing 0.1% fatty acid-free bovine serum albumin. Binding was measured at 4°C, by incubating quadruplicate wells for 2 h with increasing concentrations of 125 I-␤ 2 GPI in the absence (to determine total binding) or presence (to determine nonspecific binding) of a 100-fold molar excess of unlabeled ligand. After washing, bound 125 I-␤ 2 GPI was measured in cell lysates using a gamma counter (61). Specific binding was defined as the difference between total and nonspecific binding, and binding isotherms were analyzed by nonlinear curve fitting using the least squares method (Kaleidograph, Abelbeck Software, Reading PA), as well as by the method of Scatchard (62).
In selected experiments, the reversibility of 125 I-␤ 2 GPI binding was determined by incubating cells with 40 nM 125 I-␤ 2 GPI at 4°C for 2 h. The supernatant, containing radiolabeled ligand, was then removed and replaced with 100 l of fresh medium containing 4 M unlabeled ␤ 2 GPI. The radioactivity in cell supernatants at selected time points was then determined.
To assess the role of annexin II as an endothelial cell ␤ 2 GPI receptor, the ability of monoclonal and polyclonal anti-annexin II antibodies to inhibit the binding of 125 I-␤ 2 GPI to endothelial cells was measured. Cells were prepared for binding studies and then incubated for 2 h with 20 nM 125 I-␤ 2 GPI either alone, in the presence of 2 m unlabeled ␤ 2 GPI (to determine nonspecific binding), or with the specified concentration of antibody. After washing, cell-bound radioligand was determined.
Isolation of Endothelial Cell Surface ␤ 2 GPI-binding Proteins-To determine whether ␤ 2 GPI might bind to an endothelial cell surface protein, we determined whether endothelial cell-bound 125 I-␤ 2 GPI was incorporated into an SDS-stable complex following exposure of cells to the membrane-impermeable, bifunctional cross-linker, BS 3 . HUVECs were incubated with 40 nM 125 I-␤ 2 GPI for 2 h and then exposed to 4 mM BS 3 for 15 min. Detergent extracts were prepared and analyzed using 10% SDS-PAGE and autoradiography. The specificity of cross-linking was assessed by determining whether the amount of complex detected was reduced when studies were performed in the presence of a 100-fold molar excess of unlabeled ␤ 2 GPI and whether complexes of similar M r were detected when studies were performed using human MDA-MB-231 breast carcinoma cells as a control.
To further assess whether ␤ 2 GPI bound specifically to an endothelial cell protein, biotinylated endothelial cell surface proteins were affinitypurified using immobilized ␤ 2 GPI. Cells were biotinylated using sulfosuccinimidyl 6-(biotinamido) hexanoate (63), and detergent extracts were prepared in a buffer containing 0.1 M Tris-HCl, pH 8.1, 1.0% Triton X-100, 2.5 mM EDTA, 10 g/ml aprotinin, 50 g/ml phenylmethylsulfonyl fluoride, 12.5 g/ml leupeptin, 10 mM benzamidine and 10 g/ml soybean trypsin inhibitor. ␤ 2 GPI-binding proteins were isolated by affinity chromatography using a 1-ml column of ␤ 2 GPI-conjugated to Affi-Gel-HZ (10 mg of ␤ 2 GPI/ml of gel). After washing the column with Tris-buffered saline containing 0.8 M NaCl, bound proteins were eluted using 0.1 M glycine-HCl, pH 3.0. Fractions of 0.5 ml were collected, and proteins within each separated by 10% SDS-PAGE (nonreducing conditions) and transferred to polyvinylidene difluoride. ␤ 2 GPI-binding proteins were detected by incubating the membrane for 30 min with streptavidin-peroxidase, followed by development using chemiluminescence. The binding of proteins within detergent extracts to a control column containing Affi-Gel-immobilized bovine serum albumin was determined in parallel, to assess the specificity of the affinity purification procedure.
Definitive identification of endothelial cell ␤ 2 GPI-binding proteins was pursued through larger scale affinity purification studies. Briefly, extracts from 6 ϫ 10 9 EAHY.926 cells were subjected to affinity chromatography using a 5-ml column of ␤ 2 GPI-Affi-Gel HZ. After washing the column and eluting bound proteins, fractions were analyzed using 10% SDS-PAGE. After fixation, gels were stained using Coomassie Brilliant Blue, and protein bands were then excised, dehydrated in 100% acetonitrile, and stored at -80°C. Sequences of tryptic peptides from the ϳ78-kDa band were determined by LC-mass spectrometry at Harvard MicroChem (Cambridge, MA). Parallel studies with the ϳ36and ϳ60-kDa bands were performed at the William M. Keck Biomedical Mass Spectrometry Center, University of Virginia School of Medicine (Charlottesville, VA). Data base searches using nonredundant spectral information (SEQUEST) and partial peptide sequences (MS-Edman and BLAST) were performed to identify the tryptic peptides.
Effect of Transfection of HEK 293 Cells with Annexin II cDNA on ␤ 2 GPI Binding-Transient transfection of HEK 293 cells with annexin II cDNA was performed by incubating subconfluent cells either with 2 l of an empty plasmid (pCMV5) or with the same plasmid containing the annexin II coding sequence, in the presence of Lipofectin (pCMV5-AII) (52). Three days after transfection, cells were trypsinized and replated in 96-well plates. The total, nonspecific, and specific binding of 125 I-␤ 2 GPI to cells transfected with the empty and annexin II cDNA-containing vectors were then determined.
Measurement of the Binding of ␤ 2 GPI to Annexin II by Surface Plasmon Resonance-To determine whether ␤ 2 GPI bound to annexin II in a cell-free system, we assessed the interaction between these proteins by surface plasmon resonance using a Biacore 2000 (Biacore, Piscataway, NJ). ␤ 2 GPI was immobilized on a carboxymethyldextran (CM-5) biosensor chip using either amine or aldehyde coupling (64). For amine coupling, the chip was exposed to ␤ 2 GPI (3 g/ml) in 10 mM sodium acetate, pH 6.0, at a flow rate of 5 l/min for 2 min; this resulted in the immobilization of 198 response units (RU) of ␤ 2 GPI. For aldehyde coupling, carbohydrate residues on ␤ 2 GPI were oxidized by adding 20 l NaOI 4 to 1 ml of a 1 mg/ml solution of ␤ 2 GPI in 100 mM sodium acetate, pH 5.5. After 15 min, the solution was desalted on a NAP-5 column (Amersham Pharmacia Biotech). The oxidized protein was immobilized by exposure to a CM-5 chip at a flow rate of 5 l/min for 1 min, at which time the hydrazone bond was reduced by exposure to sodium cyanoborohydride. This resulted in the immobilization of 1062 RU of ␤ 2 GPI.
The binding of increasing concentrations of annexin II, delivered at a flow rate of 30 l/min, was then measured in real time. Binding data was analyzed by Global analysis using BiaEval 3.0 software (Biacore), in which the association and dissociation data for a series of annexin II concentrations is fit simultaneously (65). In parallel, association data was analyzed following linear transformation (66). The equation used in these studies was dRU/dt ϭ k a [annexin II]RU max -RU(k a [annexin II] ϩ k d ), where RU max ϭ the maximal binding response. The use of this equation to derive the K d from real time surface plasmon resonance data has been described (66). This approach allows the definition of fast and slow components of association and thus reveals binding heterogeneity (66,67).

RESULTS
Binding of 125 I-␤ 2 GPI to HUVECs-Preliminary attempts to measure the binding of 125 I-␤ 2 GPI to HUVECs plated in standard 96-well tissue culture plates were complicated by specific binding of the ligand to control wells that contained no cells. However, because 125 I-␤ 2 GPI bound specifically to fluid-phase endothelial cells, we focused on developing an assay in which its binding to endothelial cell monolayers could be assessed. Pretreatment of Immulon II plates with Aquasil abolished the binding of 125 I-␤ 2 GPI to the plates, which, however, could still be coated with sufficient fibronectin to support endothelial cell adhesion and growth. Using this system, we observed that the binding of 125 I-␤ 2 GPI to wells treated with Aquasil and fibronectin was only ϳ5% of that to identically prepared wells in which confluent monolayers of HUVECs were present (Table I). Therefore, this assay allowed us to selectively measure the specific binding of 125 I-␤ 2 GPI to endothelial cells. 125 I-␤ 2 GPI bound to HUVECs specifically and in a time-dependent manner (Fig. 1A). Binding was reversible in the presence of excess unlabeled ligand (Fig. 1B). Analysis of saturation isotherms revealed saturable, high affinity binding (K d ϳ18 nM, B max ϳ645,000 sites/cell) (Fig. 2). Binding was not inhibited by heparin (1-10 units/ml) or by pretreatment of cells with heparinase. However, although 125 I-␤ 2 GPI bound to endothelial cells with an affinity similar to that of annexin V (68), a phospholipid-binding protein that binds with high affinity (K d ϳ7 nM) to anionic phospholipid exposed on the surface of activated platelets (69), it bound to 14-fold fewer sites (68). These results suggested a fundamental difference in the interactions of ␤ 2 GPI and annexin V with endothelial cells. Furthermore, the high affinity binding of 125 I-␤ 2 GPI to HUVECs, when compared with its weak interaction with platelets (K d ϭ 0.5-1.0 M) (38 -40), suggested that the mechanisms of its binding to these two cell types differed.
Evidence for a ␤ 2 GPI-binding Protein(s) on the Endothelial Cell Surface-To assess whether a cell surface protein might mediate the binding of ␤ 2 GPI to endothelial cells, we determined whether endothelial cell-bound 125 I-␤ 2 GPI could be cross-linked to such a protein using the membrane impermeable, homobifunctional cross-linker, BS 3 . SDS-PAGE analysis of detergent extracts from cells incubated with 125 I-␤ 2 GPI followed by exposure to BS 3 revealed radioactive bands of ϳ50 and ϳ90 kDa, consistent with free 125 I-␤ 2 GPI, and 125 I-␤ 2 GPI cross-linked to a second protein of ϳ40 kDa (Fig. 3). The specificity of this interaction was demonstrated by the observations that formation of the ϳ90-kDa complex was inhibited by a 100-fold molar excess of unlabeled ␤ 2 GPI and that this complex a Data are expressed as mean of triplicate points Ϯ S.D.

FIG. 1. Time course and reversibility of ␤ 2 GPI binding to endothelial cells.
A, time course. HUVECs were prepared for binding assays as described under "Experimental Procedures" and then incubated at 4°C with 40 nM 125 I-␤ 2 GPI, in the absence or presence of 4 M unlabeled ligand. At various times thereafter, the amount of 125 I-␤ 2 GPI specifically bound to cells was determined. B, reversibility of binding. HUVECs were incubated with 40 nM 125 I-␤ 2 GPI for 2 h at 4°C. Supernatant containing the radiolabeled ligand was then replaced with cold PBS containing 4 M unlabeled ␤ 2 GPI (time ϭ 0 min). At various times thereafter, supernatants were removed, and the amount of 125 I-␤ 2 GPI that remained bound to the cells was measured. All points were determined in quadruplicate. These experiments are representative of three so performed.
was not detected when identical experiments were performed using MDA-MB 231 breast cancer cells (Fig. 3).
To further examine the interaction of ␤ 2 GPI with endothelial cell proteins, we affinity-purified biotinylated endothelial cell surface proteins using a column of ␤ 2 GPI conjugated to Affi-Gel HZ. Analysis of biotinylated proteins purified from detergent extracts of 4 ϫ 10 6 endothelial cells yielded primary bands of ϳ98 and ϳ78 kDa (Fig. 4, left lane). A less intense band of ϳ36 kDa was also observed, and a ϳ60-kDa band was present inconsistently. No endothelial cell proteins were isolated using control columns containing Affi-Gel-bovine serum albumin.
Identification of the Endothelial Cell ␤ 2 GPI-binding Protein-We next wished to identify the endothelial cell ␤ 2 GPIbinding proteins detected in the small-scale studies described above. However, due to the difficulty associated with culturing sufficient HUVECs to allow preparative scale isolation of such protein(s), we sought a transformed cell line that expressed similar proteins. No ␤ 2 GPI-binding proteins could be affinitypurified from THP-1, CHO, HEK 293, or MDA MB-231 cells (not shown). However, protein bands with a mobility identical to that of the proteins isolated from HUVECs were detected in detergent extracts of EAHY.926 cells, which also expressed an additional ␤ 2 GPI-binding protein of higher M r (Fig. 4, right  lane). These studies suggested that EAHY.926 would be suitable for isolation of endothelial ␤ 2 GPI-binding protein(s).
Definitive identification of endothelial cell ␤ 2 GPI-binding proteins was pursued by affinity purification of extracts from 6 ϫ 10 9 EAHY.926 cells, using a 5-ml ␤ 2 GPI-Affi-Gel HZ affinity column. Coomassie Blue-stained gels of fractions eluted from the affinity column revealed bands of ϳ78 and ϳ36 kDa (Fig. 5), corresponding to bands isolated from extracts of cell surface-biotinylated HUVECs. A faint band of ϳ60 kDa (not well visualized in Fig. 5) was also observed, although the ϳ98-kDa band isolated from the biotinylated cells was not. Mass spectrometric sequencing of two tryptic peptides from the ϳ78-kDa band and nine peptides from the ϳ36-kDa band revealed sequences corresponding to annexin II (Table II). These results suggested a binding interaction between annexin II and ␤ 2 GPI.
Evidence That Annexin II Serves as an Endothelial Cell Receptor for ␤ 2 GPI-To assess the role of annexin II as an endothelial cell ␤ 2 GPI receptor, we determined whether antiannexin II antibodies inhibited the binding of 125 I-␤ 2 GPI to endothelial cells. A 10-fold molar excess of a monoclonal antiannexin II antibody (mAb Z014) inhibited the binding of 20 nM 125 I-␤ 2 GPI to endothelial cells to a similar extent (90%) as a 100-fold excess of unlabeled ␤ 2 GPI. Binding was unaffected by a control, nonimmune murine IgG 1 (Fig. 6). Specificity of the monoclonal antibody was confirmed by the observation that it recognized only a single protein (ϳ36 kDa) in detergent extracts of endothelial cells when assessed in immunoblot studies. A polyclonal anti-annexin II antibody also inhibited the

FIG. 4. Affinity purification of ␤ 2 GPI-binding proteins from endothelial cells (EC) and EAHY.926 cells (EAHY).
HUVEC cell surface proteins were biotinylated using sulfosuccinimidyl 6-(biotinamide) hexanoate. Cells were then washed, and detergent extracts were prepared. Extracts were subjected to affinity chromatography on Affi-Gel HZ. Columns were washed with tris-buffered saline containing 0.8 M NaCl until the A 280 of the effluent reached 0, and bound proteins were then eluted using 0.1 M glycine-HCl, pH 3.0. Proteins within each fraction were separated by 10% SDS-PAGE under nonreducing conditions and transferred to polyvinylidene difluoride membranes. Biotinylated proteins were detected by incubating membranes with streptavidin-peroxidase and developing the membranes using chemiluminescence. HUVEC extracts yielded bands of ϳ98, ϳ78, ϳ60 and ϳ36 kDa, whereas extracts of EAHY.926 cells contained these proteins, as well as an additional protein of Ͼ200 kDa. This experiment is representative of six so performed.
binding of ␤ 2 GPI to HUVECs, although somewhat less potently (Fig. 6). The extent of inhibition caused by this antibody was similar to that with which it inhibited the binding of t-PA to endothelial cells in a prior report (52).
To further assess the role of annexin II in ␤ 2 GPI binding, we measured the binding of 125 I-␤ 2 GPI to HEK 293 cells transfected with either an empty vector (pCMV5) or the same vector containing an annexin II cDNA (pCMV5-AII). As previously reported, immunoblot analyses revealed only trace amounts of annexin II in detergent extracts of untransfected or pCMV5transfected HEK 293 cells (52,70). In contrast, cells transfected with pCMV5-AII expressed abundant annexin II (not shown). Furthermore, these cells bound 10-fold more 125 I-␤ 2 GPI than either untransfected cells or cells transfected with the empty vector (Fig. 7), with an affinity identical to that with which 125 I-␤ 2 GPI bound to HUVECs (K d ϳ13 nM). These observations, together with the effects of anti-annexin II antibodies on ␤ 2 GPI binding, support the conclusion that annexin II represents an important endothelial cell ␤ 2 GPI binding site.
Measurement of the Binding of Annexin II to ␤ 2 GPI by Surface Plasmon Resonance-The studies described above demonstrate that annexin II mediates the binding of ␤ 2 GPI to endothelial cells but do not address the issue of whether phospholipid, glycosaminoglycans, or other cell surface pro-teins are also required for this interaction. To address this issue, we determined whether ␤ 2 GPI and annexin II bound to one another in a cell-free system, by measuring the binding of fluid-phase annexin II to ␤ 2 GPI immobilized on a carboxymethyldextran biosensor chip. Regardless of whether ␤ 2 GPI was immobilized by amine or aldehyde coupling, rapid and concentrationdependent binding of annexin II was observed (Fig. 8). The dissociation constant (K d ) for binding was determined using the association and dissociation rate constants, calculated using the Global analysis feature of BiaEval 3.0 (65), as well as by linear transformation of the association data (66). The former method yielded a dissociation constant 2-4-fold higher than that obtained when measuring the equilibrium binding of FIG. 5. Large scale affinity purification of ␤ 2 GPI-binding proteins from EAHY.926 cells. Detergent extracts were prepared from 6 ϫ 10 9 EAHY.926 cells and subjected to affinity chromatography on ␤ 2 GPI-Affi-Gel HZ. The column was washed and eluted as described in the legend to Fig. 4. Proteins within each fraction were analyzed by 10% SDS-PAGE, and gels were stained using Coomassie Brilliant Blue. The majority of eluted protein was contained within fractions 4 -6, corresponding to the lanes shown here (left to right). The ϳ78and ϳ36-kDa bands (arrows 1 and 2, respectively) from each lane were excised for preparation of tryptic peptides, which were isolated and sequenced using LC-mass spectrometry.  I-␤ 2 GPI binding assays. The specific binding of 125 I-␤ 2 GPI to untransfected cells (q) or cells transfected with pCMV5 (E) or pCMV5-AII (Ⅺ) was determined as described. Significantly more 125 I-␤ 2 GPI bound specifically to cells transfected with pCMV5-AII, with a K d of ϳ13 nM. This experiment is representative of two so performed, with all points determined in duplicate. ␤ 2 GPI to HUVECs (Table III). The latter revealed fast and slow components of association, reflecting binding heterogeneity in this system (66,67). Use of the more relevant fast component for calculation of the K d , however, yielded values (5-12.75 nM) identical to those determined for the binding of ␤ 2 GPI to intact endothelial cells. Hence, although these results do not exclude the possibility that other cell surface components may stabilize or promote the ␤ 2 GPI-annexin II interaction, they demonstrate that such components are not essential for binding. DISCUSSION These studies demonstrate that annexin II mediates the binding of ␤ 2 GPI to endothelial cells. This conclusion is supported by 1) the affinity purification of annexin II from endothelial cells using ␤ 2 GPI-Affi-Gel, 2) the near-complete inhibition of ␤ 2 GPI binding to endothelial cells by anti-annexin II antibodies, 3) enhanced binding of ␤ 2 GPI to annexin II-transfected HEK 293 cells, and 4) the direct demonstration of high affinity binding of annexin II to ␤ 2 GPI using surface plasmon resonance.
Identification of annexin II (M r ϳ 36) as the primary ␤ 2 GPI binding site on unactivated endothelial cells is consistent with the pattern of protein bands affinity-purified using ␤ 2 GPI-Affi-Gel. Mass spectrometric sequencing of tryptic peptides from the ϳ36and ϳ78-kDa bands isolated in both the small and large-scale affinity purification procedures confirmed that each of these contained annexin II. The ϳ78-kDa band most likely represented an annexin II homodimer, because it migrated identically to a spontaneously forming homodimer present in preparations of recombinant annexin II. Mass spectrometric sequencing of tryptic peptides from the ϳ60-kDa band also yielded sequences identical to those within annexin II, as well as heat shock protein 27 (M r ϳ 22.3) (71) and human DNAbinding protein A (M r ϳ 38.6) (72). Hence, this band may represent a partially degraded annexin II homodimer or, more likely, a complex between annexin II and an additional, intracellular protein. Finally, the ϳ98-kDa band observed only in the small scale affinity purification procedure may represent an annexin II heterotetramer (AIIt) containing 2 molecules of annexin II and 2 molecules of p11 (11 kDa) (70,73), although the inability to detect this band in the large scale procedure precluded its definitive identification.
An additional high molecular mass ␤ 2 GPI-binding protein was observed only in extracts of cell surface biotinylated EAHY.926 cells and thus was most likely derived from A549 carcinoma cells (the non-HUVEC fusion partner used to create the EAHY.926 cell line). We speculate that this protein may be megalin, a low density lipoprotein receptor family member recently shown to bind ␤ 2 GPI (74), or perhaps another low density lipoprotein receptor family member. Taken together with the endothelial cell studies, however, these results suggest that ␤ 2 GPI may interact with unique cell surface proteins differentially expressed on specific cell types. However, we would emphasize that our studies do not exclude a biologically important interaction of ␤ 2 GPI with cellular phospholipids under some circumstances (75,76).
The annexins are a family of structurally related proteins, each of which consists of an N-terminal "tail" and C-terminal "core" domain (77). The core domains of different annexins share 40 -70% homology (78) and consist of a series of 70 amino acid endonexin repeats (78). In contrast, the length and amino acid composition of the tail domains are highly variable among different family members (78). Despite the lack of a hydrophobic signal peptide, the presence of annexin II on cell surfaces is well established (78), and approximately 4.3% of total endothelial annexin II is associated with the external plasma membrane (79). Annexin II mediates the binding of t-PA to endothelial cells through interactions with an LCKLSL sequence in the tail domain (53) and is also an endothelial cell receptor for Glu and Lys-plasminogen (54). Plasminogen binding may result from exposure of a C-terminal lysine (Lys 307 ) in the core domain following cleavage of the Lys 307 -Arg 308 bond (52,55,56), although others have presented compelling evidence for an important role of the N-terminal lysine residues of the p11 polypeptides in mediating the binding of plasminogen to the annexin II heterotetramer (70). Regardless, annexin II greatly enhances the catalytic efficiency of t-PA-mediated plasminogen activation on cell surfaces (54,56), with even more potent enhancement mediated by the heterotetramer (55,70).
Given the abundance of ␤ 2 GPI in plasma (plasma concentration, 2-4 M) and the affinity with which it binds to annexin II, we would expect most endothelial cell surface annexin II molecules to be occupied by ␤ 2 GPI under normal conditions. This hypothesis is supported by recent immunohistochemical studies demonstrating an association of ␤ 2 GPI with endothelial cells in vivo (81). Preliminary studies performed in our laboratory suggest that this interaction may be of particular importance in the presence of circulating anti-␤ 2 GPI antibodies. These antibodies, which are strongly associated with thrombo- FIG. 8. Binding of annexin II to immobilized ␤ 2 GPI-measurement by surface plasmon resonance. ␤ 2 GPI was immobilized on a carboxymethyldextran biosensor chip using amine (A) or aldehyde (B) coupling. The immobilized ligand was then exposed to annexin II, and binding was measured in real time. Following the binding of annexin II, dissociation was measured in a similar manner. K d values were determined from experimentally derived association and dissociation rate constants. The concentrations of annexin II used to obtain the binding curves are depicted on the right. sis, activate endothelial cells in vitro only in the presence of ␤ 2 GPI (50,51,82,83), and the elevated plasma levels of endothelial cell adhesion molecules (48) and von Willebrand factor (49,84) in patients with antiphospholipid antibodies/anti-␤ 2 GPI antibodies suggest that these antibodies may also disrupt endothelial function in vivo. We have observed that the same annexin II monoclonal antibodies that block ␤ 2 GPI binding to endothelial cells directly induce endothelial cell activation, as measured by the expression of endothelial adhesion molecules. 2 These results suggest that annexin II cross-linking induces signaling responses in endothelial cells that lead to cellular activation and the development of a proadhesive and procoagulant phenotype. Because annexin II does not span the cell membrane, this interaction may require an "adaptor" protein, the identity of which is under investigation. However, we hypothesize that "indirect" cross-linking of cell surface annexin II through ligation of bound ␤ 2 GPI by anti-␤ 2 GPI antibodies might induce a similar response, and hence our studies may provide an explanation for ␤ 2 GPI-dependent endothelial cell activation by anti-␤ 2 GPI antibodies. Alternatively, it has been suggested that annexin II might be capable of initiating signaling responses by mediating calcium channel activity (80). If so, the binding of ␤ 2 GPI and anti-␤ 2 GPI antibodies might induce a conformational change in annexin II that stimulates this effect. At present, however, this hypothesis remains speculative.
Although the ␤ 2 GPI binding site on annexin II has not yet been determined, the anti-annexin II mAb Z014, as well as previously described anti-annexin II polyclonal antibodies (52), blocked binding of both t-PA and annexin II to endothelial cells. Although these studies suggested that the ␤ 2 GPI binding site may reside within the LCKLSL t-PA binding sequence (amino acids 8 -13) in the tail domain of annexin II (53), ␤ 2 GPI itself did not inhibit the binding of 125 I-t-PA to endothelial cells (not shown). Hence, the t-PA and ␤ 2 GPI binding sites within annexin II are likely to be distinct, although spatially near one another in the tertiary structure of the molecule.
In summary, the demonstration that ␤ 2 GPI binds with high affinity to endothelial cell annexin II suggests that the paradigm in which the interaction of ␤ 2 GPI with endothelial cells is assumed to occur solely through binding to membrane phospholipid should be reconsidered. Further characterization of the ␤ 2 GPI-annexin II binding interaction, as well as definition of the role of annexin II in endothelial cell activation, may provide additional insight into the physiologic and pathophysiologic roles of these abundant proteins.