Binding of Vitronectin-Thrombin-Antithrombin I11 Complex to Human Endothelial Cells Is Mediated by the Heparin Binding Site of Vitronectin"

The interaction of vitronectin-thrombin-antithrom- bin I11 (VN*TAT) complex with endothelial cells (EC) was investigated. Binding was specific and time- and concentration-dependent. Kinetics revealed an appar- ent dissociation constant of 16 nM and 1.7 X 10' binding sites/endothelial cell. The binding determinant of the ternary complex was located on the VN moiety. Since the association of VN to TAT adds its specific properties to the VN-TAT complex, the involvement of the heparin binding domain and the cell attachment site of VN was investigated. Neither addition of RGD peptide nor blocking of the vitronectin receptor with a monoclonal antibody interfered with VN*TAT binding to EC. Addition of heparin, a VN-derived peptide comprising two heparin binding consensus sequences or a monoclonal antibody directed against the heparin binding domain on VN, completely inhibited VN-TAT binding to EC. These results indicate that the interac- tion is mediated through the heparin binding domain of VN. Digestion of heparan sulfate proteoglycans re- sulted in a decrease of VN-TAT binding to EC, indicating the involvement of heparin-like structures on the EC surface.

The function of the key enzyme of the blood clotting system, thrombin, as well as that of other coagulation proteases is controlled predominantly by direct interaction with serine protease inhibitors (serpins).' Antithrombin I11 (ATIII) is considered to be the most important direct inhibitor of thrombin, and the formation of a carboxylester bond between of ATIII and the active site serine of thrombin leads to an * This work was supported by Grant 88.001 from the Netherlands Thrombosis Foundation. 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. inactive, equimolar protease-inhibitor (TAT) complex. The low rate at which ATIII normally inhibits thrombin is accelerated in the presence of heparin (1). In human serum, most of the TAT complex is associated to the plasma protein, vitronectin, with which it forms a ternary complex (2, 3). Vitronectin is a glycoprotein with a molecular mass of 78,000 Da and a plasma concentration of 200-400 pg/ml (4). It is involved in several physiological processes, such as attachment of different cell types to surfaces (5), inhibition of the complement system (6), stabilization of the major inhibitor of fibrinolysis, plasminogen activator inhibitor (PAI-1) (7), and neutralization of heparin stimulation of ATIII (8). The association of VN to TAT labels the complex with the specific properties of VN, such as heparin affinity (9) and the ability to bind to cell surfaces via the RGD sequence (5). These newly generated properties of VN.TAT complex are speculated to promote a rapid disappearance of VN. TAT from the blood stream.
This prompted us to study the first event of clearance of VN. TAT, namely the interaction of the ternary VN.TAT complex with human endothelial cells, and to define the molecular characteristics of VN. TAT binding to the cell surface. Our present results demonstrate that this interaction is mediated via the heparin-binding domain of VN and heparan sulfate proteoglycan molecules on the luminal endothelial cell surface.

EXPERIMENTAL PROCEDURES
Materials-All chemicals obtained from commercial sources were of the highest grade available. The culture plastics were purchased from Costar (Cambridge, MA). RPMI-1640 medium and the antibiotics penicillin/streptomycin and fungizone were obtained from Gibco (Biocult, Paisley, UK). The Phast-System and the fast performance liquid chromatography (FPLC) equipment were obtained from Pharmacia (Sweden). Unfractionated heparin was obtained from Organon (Oss, The Netherlands). Heparinase, heparitinase, chondroitinase ABC, andp-nitrophenyl-@-D-xylopyranoside (8-D-xyloside) were purchased from Sigma.
Polyclonal antibodies against VN and prothrombin were raised in rabbits as described by Preissner et al. (2). The polyclonal antibody against ATIII was obtained from Behring Werke (Marburg, Federal Republic of Germany TAT complex was formed by incubating a-thrombin with ATIII in a molar ratio of 1:2 for 30 min at 37 "C. Purification of Proteins-VN. TAT was purified at room temperature from 100 ml of fresh human serum by affinity chromatography on heparin-Sepharose (50 ml, Pharmacia), equilibrated with Hepes buffer (10 mM Hepes, 137 mM NaCI, 0.02% sodium azide, pH 7.4). Unbound material was removed by extensive washing with Hepes buffer, and hound proteins were eluted with a salt gradient from 0.15 t o 1.00 M NaCl in a total volume of ROO ml of Hepes buffer. The different fractions were analyzed on 4-15?; sodium dodecyl sulfatepolyacrylamide gel electrophoresis (SDS-PAGE, Phast-System) and transferred to Immobilon-P membrane (Millipore, Bradford, USA) by diffuse blotting. The blots were blocked with 5% (w/v) milk powder (Protifar, Nutricia, Holland) and incuhated with polyclonal antibodies against VN, ATIII, and prothrombin. The blots were washed with Tris-buffered saline (TBS-buffer; 50 mM Trizma (Tris base), 150 mM NaCI. 0.5% Tween 20, pH 7.4). incubated with peroxidase-conjugated anti-rabbit IgG (Dako, Denmark), and stained with 5 mg of :I,:<'diaminobenzamidin (Sigma) in 50 ml of TBS-buffer. VN.TAT was further purified by ion-exchange chromatography on FPLC equipment. Therelore fractions containing VN .TAT were pooled, dialyzed, against 20 mM ethanolamine (Merck, Germany), pH 9.0, and loaded on a Mono Q column. VN.TAT was eluted by a salt gradient from 0 to 1.0 M NaCl a t 0.4 M NaCl in ethanolamine (20 mM, pH 9.0).
Purification of VN.TAT with a Tris buffer, pH 7.4, instead of ethanolamine, pH 9.0, did not change the characteristics of the complex (data not shown). The purified complex was kept frozen a t physiological pH and remained stable.
AT111 was purified according to the method described by de Swart et 01. (10). Denatured VN was purified according to the method of Yatohgo et al. (11). Purified proteins were >95% homogeneous as judged by SDS-PACE. Modified AT111 was prepared by incubating native ATIII with 2-fold molar excess of cr-thrombin for 1 h a t 37 "C, dialyzed overnight in 20 mM ethanolamine a t 4 "C, and the modified form of AT111 was separated from thrombin on Mono Q ( F P I X ) , in 'LO mM ethanolamine (pH 9.0) using a salt gradient from 0 to 1.0 M NaCI.
Binding of VN.TAT to HUVEC-HUVEC were washed three times with Hepes buffer (10 mM Hepes, 137 mM NaCI. 4 mM KCI, 2 mM CaCI,, 15 mM Glucose, 5 mg/ml BSA, pH 7.4) and incubated a t 4 'C with 20 ng of ""I-labeled VN-TAT, diluted in a total volume of 50 pl of Hepes buffer (2.5 nM). After a 2-h incubation the cells were washed five times with cold Hepes buffer and the cells were lysed in 0.1 M NH,OH for 15 min a t room temperature. The cell-associated label was determined by counting the lysate in a y-scintillation counter. Measurements were performed in triplicate unless stated otherwise on two different cell hatches. Representative results are shown, and binding of ""1-VN.TAT is expressed in nanograms/lO" cells. T o check for nonspecific VN.TAT binding to plastic, RSAcoated wells were assayed identically. Specific binding of iodinated ligand was defined as the difference in binding in the absence and presence of 100-fold molar excess of unlabeled ligand. The binding constant and number of binding sites were calculated using a computer-assisted iterative curve fit program (Ligand).
Competition assays were performed by incubating 2.5 nM radiolabeled VN.TAT in the presence of 100-fold molar excess of unlabeled compounds. The inhibitory potential of the different compounds was expressed in percentages of specific binding. Pretreatment of HUVEC-Preincubation a t 37 "C with the monoclonal antibodies anti-/& and VN27 (10 pg/ml) was performed 30 min prior to the binding assay. Non-bound antibody was washed away. I'roteoglycan synthesis was inhibited by adding 2.5 mM if-a-xyloside in the culture medium 2 days before the cells were used. Glycanases were incubated for 30 min prior to the binding experiment and washed away after the incubation period. The following glycanases were used: ABC (0.5 units/ml).

RESULTS
Purification of VN-TAT-Fresh human serum was used as starting material for purification of VN.TAT with heparin-Sepharose affinity chromatography. The bulk of proteins was eluted between 0.1 and 0.3 M NaCl from the heparin-Sepharose column, whereas VN .TAT eluted at 0.4-0.5 M NaCI. After further purification and concentration of VN-TAT using Mono Q ion-exchange chromatography, SDS-PACE revealed a major protein band at 160 kDa (Fig. 1, lane 3 ) . The protein, transferred to Immobilon-P membrane, was recognized by polyclonal antibodies against VN, ATIII, and prothrombin (lanes 4-6), although anti-prothrombin reacted poorly. The immunoblots also showed bands larger than 200 kDa reacting with all three antibodies used, indicating that multimeric forms of VN .TAT were present in the eluate from the heparin-Sepharose column. Lane I of Fig. 1 shows an autoradiograph of ""1-VN.TAT. The mobility of labeled and unlabeled VN .TAT was identical, indicating that the labeling procedure did not modify the VN .TAT molecule.
Binding of VN-TAT to HUVEC-A confluent monolayer of HUVEC was incubated with radiolabeled VN .TAT at 4 "C for the time periods indicated, in the presence or absence of 100-fold molar excess of unlabeled VN .TAT. Binding of VN .
TAT to HUVEC was specific and time-dependent and reached an equilibrium within 2 h of incubation (Fig. 2). No binding was observed to albumin-coated wells. HUVEC, incubated for 1 h, were lysed in SDS-containing sample buffer and run on a 4-15% SDS-PAGE. The autoradiograph (Fig. 1, lane 2 ) showed that VN . TAT bound to the HUVEC did not differ from the VN.TAT before incubation (Fig. 1, lane f ).To determine whether labeling of VN.TAT altered the binding affinity to endothelial cells, confluent monolayers of HUVEC were incubated a t 4 "C with labeled and nonlabeled VN. TAT a t varying ratios for 2 h keeping the total amount of VN. TAT added constant. After measuring the cell-associated radioactivity, a linear correlation between percentage of added labeled VN .TAT and percentage of bound labeled VN .TAT was found, indicating no difference in binding affinity of labeled and nonlabeled VN. TAT (data not shown). The number and affinity of VN . TAT binding sites on HUVEC was Binding of 20 ng/well (+ 100,000 cpm/well, 2.5 nM) of radiolabeled VN.TAT to HUVEC was performed at 4 "C for the time periods indicated. Total binding (0) was determined in the absence of unlabeled VN.TAT, nonspecific binding (V) in the presence of 100-fold molar excess of unlabeled VN. TAT. Specific binding (0) was calculated by subtracting nonspecific binding from total binding. As control, BSA-coated wells were assayed identically (0). Cell-associated binding was performed in duplicate and expressed in nanograms of VN .TAT bound/106 cells.
determined by binding of 1'51-VN-TAT a t 4 "C as a function of VN . TAT concentration (Fig. 3). Analysis of specific binding, using a computer-assisted iterative curve fit program (Ligand), yielded an apparent dissociation constant (Kd) of 16 nM and 1.7 X lo5 binding sites per EC. In all cases the one ligand/one binding site model fitted statistically better to the binding data than the two-binding site model. Characterization of V N . TAT Binding to HUVEC-To examine the domains of VN.TAT involved in the binding to HUVEC, competition assays were performed at 4 "C. A 100fold molar excess of urea-treated VN (VNurea) inhibited Iz5I-VN e TAT binding to HUVEC by 92% (Table I). Urea-treated VN was used to mimick the conformationally changed form of VN, in which it exists in the ternary complex (9). Addition of 10 ~L M RGDW peptide in the competition assay had no influence on the binding of VN'TAT to HUVEC. When HUVEC were preincubated with an antibody against the p3subunit of the VN receptor, no influence on "9-VN.TAT binding to HUVEC was found either. The RGDW peptide and the antibody against the p3-subunit were added at a concentration at which platelet aggregation, induced by ADP, was inhibited completely. Addition of 10 p~ RGDW peptide did not detach the HUVEC.
The influence of heparin on the binding of ''51-VN-TAT to HUVEC is depicted in Fig. 4. Half-maximal binding was observed in the presence of 10 units/ml (100 pg/ml) unfractionated heparin, while with 300 units/ml heparin, the binding of VN.TAT to HUVEC was completely blocked. When a monoclonal antibody directed against VN (8E6) was added together with T -V N . TAT, complete inhibition of VN. TAT binding to HUVEC was observed (Fig. 4). This antibody was found to recognize the carboxyl-terminal region of VN, containing the entire heparin binding domain.' Further insight into the VN binding domain involved was obtained by utilizing five synthetic peptides derived from the heparin binding region of VN (Fig. 5b). Addition of 10 pg/ml either peptide 4 or 5 inhibited the binding of l2'1-VN. TAT by 40%, 10 pg/ml peptide 3 inhibited binding by 60%, whereas peptide 1 or 2 completely inhibited VN.TAT binding to HUVEC (Fig. 5a). The concentration for peptide 2 to reach half-maximal binding was 10 times lower than for peptide 1  and 3 and more than 100-fold lower then for peptides 4 and 5.
T o determine whether domains on the serpin-and the protease-component of the ternary complex play an additional role in VN. TAT binding to HUVEC, unlabeled 11,-DIP or different forms of ATIII were added in the competition assay (Table I). IIa-DIP, native (ATnat) and thrombin-modified ATIII (ATmod) did not interfere with Iz5I-VN. TAT binding to HUVEC, whereas preformed binary complex TAT inhibited binding by about 50%. To explain the inhibitory effect of TAT on radiolabeled VN. TAT binding to HUVEC, the presence of cell surface-bound VN was taken into consideration. Preincubation of HUVEC with a monoclonal antibody, directed against VN, which was found to interfere with VN. TAT formation3 reduced the inhibitory effect of TAT on lZ5I-VN. TAT binding to HUVEC from 50 to 25%.   VEC-The observation that the interaction of VN .TAT to HUVEC was mediated by the heparin binding domain of VN suggested that proteoglycans present on the EC surface are involved in this process. When HUVEC were cultured in the presence of p-D-xyloside (2.5 mM), which leads to the synthesis of glycosaminoglycan-deficient proteoglycans (13,14), 50% inhibition of '251-VN.TAT binding was found (Fig. 6). Treatment of the cells with heparinase for 30 min at 37 "C (5 units/ ml) reduced '2sI-VN .TAT binding by 40%. Hardly any influ-a I r " FIG. 6. Treatment of HUVEC prior to VN .TAT binding.
HUVEC were either cultured in the presence of 2.5 mM P-D-xyloside for 2 days or confluent cells were preincubated with heparinase (5 units/ml), heparitinase (5 units/ml) or chondroitinase ABC (0.5 units/ml) for 30 min at 37 "C prior to the binding assay. The excess of agents was washed away, and the cells were incubated with 20 ng of VN.TAT/well (2.5 nM) for 2 h at 4 "C in the absence or presence of unlabeled VN.TAT (250 nM) or after pretreatment as indicated.
Cell-associated binding was determined in triplicate and expressed in nanograms/106 cells (mean -+ S.E.).
ence was found when HUVEC were treated with heparitinase (5 units/ml, 30 min) or chondroitinase ABC (0.5 units/ml, 30 min). Increasing concentrations or longer incubation times did not alter these data (not shown).

DISCUSSION
The ultimate end products of the clotting cascade are complexes between activated proteases and protease inhibitors. Thrombin, the final enzyme of the clotting cascade, is immediately inactivated by ATIII. In human serum, thrombinantithrombin 111 complexes are associated with VN (2, 3). In these ternary complexes different functional properties are combined, related to adhesive activity of VN (5) and possibly of thrombin (15) as well as potent heparin binding capacity residing in the VN component (4). Together with possible additional active site-independent mitogenic properties of thrombin (15) these complexes may provoke cell attachment, cell binding, and proliferation at the site of vessel wall injury. In the present study the interaction of ternary VN .TAT complexes with HUVEC was investigated, and the results presented here demonstrate that VN .TAT complexes bind to HUVEC in a time-and concentration-dependent manner. Binding kinetics revealed an apparent dissociation constant of 16 nM and 1.7 X lo5 binding sites/endothelial cell. The heparin binding domain accessible on vitronectin was responsible for the binding of VN.TAT to HUVEC, since binding was inhibited by unfractionated heparin and by a monoclonal antibody directed against the region containing the heparin binding domain of VN (8E6). Additional evidence was obtained utilizing five synthetic peptides overlapping the heparin binding domain of VN, which were able to differentially inhibit the binding of VN.TAT to HUVEC. The most efficient inhibition was found with peptide Lys"'-Arg36l, whose concentration needed to reach half-maximal inhibition of VN . TAT binding was 10 times lower than of peptide Ala341-Arg3" o r Arg3"7-Arf0. This indicates that the primary binding site in VN .TAT for HUVEC is located between amino acids 348 and 361 of VN, which comprises two consensus sequences for glycosaminoglycan recognition (X-B-E-X-B-X and X-B-B-B-X-X-B-B, where B represents the probability of a basic residue and X represents a non-basic residue) (16). Peptide Ala341-Arg3s5 entails the first of the two consensus sequences, whereas peptide Arg"1-Arg350 has a high homology with the second consensus sequence. The potential of these peptides to directly bind to heparin follows the same order as found for interference with VN. TAT binding to HUVEC. 4 NO participation was found of the cell attachment site of VN during binding of VN .TAT to HUVEC. This is in agreement with Lampugnani et al. (17), who described that the VN-receptor (c&) on endothelial cells is only expressed a t focal contact sites (in vitro and in viuo) and is thus not available for soluble ligands.
The binding site on the endothelial cell surface for VN. TAT was characterized as proteoglycans, based on experiments in which proteoglycan processing was altered by @-D-xyloside, which leads to synthesis of 50-60% glycosaminoglycan-deficient proteoglycans (14), or in which HUVEC were pretreated with heparinase. In both cases binding of VN. TAT was decreased by about 50%. Chondroitinase ABC was not effective to modify the cell surface binding site. From these experiments it is likely that heparan sulfate proteoglycans comprise the binding site on HUVEC for VN.TAT complexes. It is interesting to note that the Kd (10-40 nM) for binding of heparin to denatured VN (8) (18).
The two other components of the ternary complex were not directly involved in the binding of VN. TAT to HUVEC. No inhibition was found when unlabeled 11.-DIP and native or thrombin-modified ATIII was added as competitor. Preformed TAT complexes, however, partly inhibited VN . TAT binding to HUVECs, although competitive interaction with heparan sulfate proteoglycans is unlikely due to the lack in heparin affinity of TAT (19). A possible explanation for the partial inhibition by TAT is that VN present at the cell surface, originating from the culture medium, may harvest TAT complexes in situ. This introduces newly formed, nonlabeled VN .TAT into the binding assay. Support for this hypothesis stems from the finding that in the presence of a VN-specific monoclonal antibody VN27, which inhibits VN . TAT formation, the inhibitory effect of TAT could partly be prevented. These findings are in agreement with Knoller et al. (20), who found that a monoclonal antibody, recognizing the binding site on ATIII involved in the binding of TAT complex to VN, inhibited TAT binding to endothelial cells. Therefore, we conclude that epitopes present on ATIII, thrombin, or TAT are not primarily involved in binding of K. T. Preissner, unpublished observation. information is available a t present whether this receptor is also expressed on endothelial cells. Nevertheless, we added peptide 105Y, which comprises the recognition sequence of the SEC receptor, in the binding assay. Addition of peptide 105Y did not inhibit the binding of VN.TAT complex to endothelial cells (data not shown). Since we saw no residual binding of VN.TAT to HUVEC after blocking the heparin binding region of VN, we concluded that there is no participation of the SEC receptor. These findings indicate that besides a liver-dependent mechanism of TAT removal from the bloodstream, there is an endothelial cell-dependent clearance of locally formed TAT complex.
At present, ternary complexes are described for VN with three thrombin-serpin complexes, thrombin-AT111 (2, 3), thrombin-heparin cofactor I1 (22), and thrombin-protease nexin-1 (23). Whether these three complexes compete with each other for binding to heparan sulfate proteoglycans present on the endothelial cell surface or on other cells awaits further investigation. It remains to be established as well to which extent these nonintegrin VN-receptors participate in the subsequent metabolic processing of VN. TAT complexes and to which extent these proteoglycans are involved in VNmediated cell-matrix interactions.