Interaction of thrombospondin with resting and stimulated human platelets.

The interaction of isolated and radioiodinated thrombospondin with washed human platelets has been characterized. The ligand bound to nonstimulated and thrombin-stimulated platelets in a time-dependent manner, and apparent steady state was reached within 25 min. Binding was not due to iodination of the ligand and was inhibited by nonlabeled thrombospondin but not by unrelated proteins, and bound ligand was identical with thrombospondin in terms of subunit structure. Nonlinear curve-fitting analyses of binding to resting platelets suggested the presence of a single class of sites which bound 3,100 +/- 1,000 molecules/platelet with an apparent Kd of 50 +/- 20 nM. This interaction was not attributable to contaminating cells or inadvertant platelet activation. Binding to thrombin-stimulated platelets had a lower apparent affinity (Kd = 250 +/- 100 nM) and higher apparent capacity (35,600 +/- 9,600 molecules/platelet). Thrombin-enhanced binding was dependent upon agonist dose and platelet stimulation. Fibrinogen, a monoclonal antibody to GPIIb-IIIa, temperature, and divalent ions had differential effects upon thrombospondin binding to resting and stimulated platelets, suggesting the presence of two distinct mechanisms of thrombospondin binding to platelets. While thrombospondin binding to thrombin-stimulated platelets occurs with characteristics similar to those observed for fibrinogen, fibronectin, and von Willebrand Factor, its high affinity interaction with resting platelets is unique to this adhesive glycoprotein.

During the hemostatic response, platelets adhere and spread on the subendothelial matrix and aggregate with one another. These platelet reactions can be directly mediated or influenced by specific proteins. Fibrinogen, von Willebrand Factor, and fibronectin may influence platelet adhesion and spreading on subendothelial matrices and artificial surfaces (1-3). Fibrinogen is the major regulator of platelet aggregation (4, 5 ) , and von Willebrand Factor and fibronectin may also influence this cell-cell interaction (6,7). These three proteins share the common properties of being large, glycosylated, and multimeric, suggesting that they may function to bridge platelets to one another or to substrata. All three glycoproteins are also present within platelet a! granules and are secreted from stimulated platelets (reviewed in Ref. 8), as well as being * This work was supported by National Institutes of Health Grants HL 28235 and HL 16411. This is publication number 3747-1" from the Research Institute of Scripps Clinic. 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.
$ Recipient of Research Career Development Award AM 00720. To whom correspondence and requests for reprints should be addressed. plasma proteins. Each of these proteins binds specifically to the surface of activated platelets (9-12), providing a mechanism for their participation in bridgmg functions. Similarities in the binding of these proteins to platelets, include: (a) requirements for platelet stimulation, (b) divalent ion dependence, (c) reduced binding to platelets from patients with Glanzmann's thrombasthenia, and ( d ) ADP dependence (9-15). These similarities may be explained in part by the existence of shared binding sites as suggested by the capacities of the same sets of monoclonal antibodies (16) and fibrinogen y chain peptides (17) to inhibit their binding. Nevertheless, important differences in the number of sites and induction requirements suggest unique features for the interaction of each protein with the platelet.
Thrombospondin (TSP') has many features in common with the three adhesive glycoproteins. It is a large glycoprotein (Mr -450,000) comprised of three subunits of similar size (18). Although present at only very low levels in normal plasma (19), TSP is a major constituent of platelet a! granules (20). When secreted from platelets, a portion of released TSP becomes associated with the cell surface in a calcium-dependent interaction raising the possibility of a divalent ion receptor-mediated interaction (21). Gartner and Dockter (22), however, have recently reported that the binding of TSP to platelet membranes is divalent ion-independent, suggesting a mechanism by which George et al. (23) observed increased surface expression of TSP on thrombin-stimulated platelets in the presence of 5 mM EDTA. Several studies have implicated TSP in platelet function. It is now clear that TSP is at least in part responsible for a lectin-like activity of platelets (24-27), and antibodies to TSP can inhibit thrombin-induced platelet aggregation (28). To define a basis for the role of TSP in platelet function and to further the analogy between TSP and the other three platelet-adhesive glycoproteins, we have examined in detail the interactions of TSP with stimulated and nonstimulated human platelets.
Purified human a-thrombin ws the generous gift of Dr. John Fenton, New York State Department of Health, Albany, NY. This material was diluted to 100 units/ml in modified Tyrode's buffer, stored in aliguots at -70 "C, and thawed once prior to use. Purified The abbreviations used are: TSP, thrombospondin; BSA, bovine serum albumin; SDS, sodium dodecyl sulfate; PG, prostaglandin; EGTA, ethylenebis(oxyethylenenitri1o)tetraacetic acid. human fibrinogen (29) and fibronectin (30) were prepared as previously described. The 10E5 monoclonal was a generous gift of Dr. Barry Coller, SUNY at Stonybrook. This monoclonal antibody immunoprecipitates platelet membrane glycoprotein IIb-IIIa in the presence of calcium (31).
TSP Purification and Radioiodination-TSP was isolated by the procedure of Lawler et al. (18) with the modifications previously described (32). On 5% polyacrylamide gels in SDS (33), the purified protein yielded a single Coomassie Blue staining band of apparent M, = 170,000, under reducing conditions and was greater than 90% homogeneous as judged by scanning densitometry (32). For selected studies, TSP was isolated from the releasate of thrombin-stimulated platelets maintaining a concentration of 2 mM calcium throughout purification.
TSP was radioiodinated by a modified chloramine-T procedure. To 500 pl of a 240 pg/ml TSP solution, 20 p1 of chloramine-T at 2 mg/ml and 0.5 mCi of Na' "1 was added. After 5 min at 22 "C, 40 pg of sodium metabisulfite and 200 pg of KI were added, followed by 50 mg of BSA (pretreated with 2 mM phenylmethylsulfonyl fluoride).
The sample was then diluted with 800 pl of Hz0 to a final volume of 1.42 ml. Free 12SI-TSP was 0.5-1.0 pCi/pg. The radioiodinated TSP exhibited a mobility identical with that of nonlabeled TSP on SDSpolyacrylamide gel electrophoresis under reducing or nonreducing conditions, and 95% of the radioactivity migrated as a single peak based on densitometric scanning of autoradiograms. In addition, 295% of the radioactivity was precipitated by 10% trichloracetic acid or by monospecific polyclonal or monoclonal antibodies to TSP.
Cell Isolation-Platelets were isolated from acid/citrate/dextrose anticoagulated fresh human blood by differential centrifugation and gel filtration on Sepharose 2B as previously described (11). Briefly, platelet pellets, obtained by centrifugation of platelet-rich plasma, were resuspended in 2 ml of modified Tyrodb's buffer, pH 6.5, containing 2 mM MgClz and 0.1% BSA and gel filtered on a column (2.5 X 10 cm) of Sepharose 2B equilibrated in modified Tyrode's buffer at pH 7.4. In some experiments, platelets were prelabeled with ["C] serotonin by addition of 2 pCi/ml platelet-rich plasma as previously described (34).
Red and mononuclear cells were prepared as previously described (35). Briefly, the cell pellet, obtained by an initial low speed centrifugation (800 X g for 15 min) to obtain a platelet-rich plasma, was resuspended in platelet-poor plasma and layered onto Ficoll-Hypaque (d = 1.074 ml/g). The mononuclear cells were removed from the interface after a 15-min centrifugation at 2260 X g and washed twice by centrifugation. Red cells pelleting through the Ficoll-Hypaque were separated from white cells by sedimentation in 3% dextran T250 and were subsequently washed by centrifugation.
Binding Assays-In a typical binding assay, platelets were diluted in modified Tyrode's buffer containing 0.1% BSA, pH 7.4, to 8 X loB/ ml. To 50 ~1 of the platelet suspension, 110 pl of buffer or unlabeled ligand, 30 pl of 1.9 p M (350 pg/ml) lZ5I-TSP (precentrifuged at 11,750 rpm for 5 min in a Beckman microfuge) were added, followed immediately by 10 +l of stimulus or buffer. Unless otherwise indicated, incubations were performed at 37 "C. At selected time points, triplicate 50-pl aliquots were layered onto 300 pl of 20% sucrose in modified Tyrode's buffer, pH 7.4, and centrifuged for 3 min in the Beckman microfuge. Tips were amputated and counted, aqd platelet-bound TSP was calculated from the specific activity of the ligand, utilizing a molecular weight of 450,000 (18). Unless otherwise indicated, the TSP recovered in the centrifuge tip in the absence of platelets was subtracted as background and was always less than 0.1% of the total added ligand. SDS-Polyacrylamide Gel Ekctrophores&"he polyacrylamide gel system of Laemmli (33) was utilized in a vertical gel apparatus. To analyze lZ5I-TSP bound to platelets, all pellets were extracted with 10% SDS, 20% 8-mercaptoethanol, and 6 M urea (32) for 5 min at 22 "C. Platelet extracts were then heated to 100 "C for 5 min prior to loading onto a 3% stacking gel. Electrophoresis was performed at 30 mA for 1 h, and the gels were fiied with 10% acetic acid in methanol and dried. Autoradiograms were developed using Kodak RP film and a Cronex enhancing screen at -70 "C.
Statistical Analysis-Binding data were fitted to models for N ligands binding to M sites utilizing the Ligand program (36) modified for the Apple I1 by T. Jackson, Department of Nuclear Medicine, Middlesex Hospital Medical School, London, England. This program permits estimation of the nonsaturable binding, N1, as a fitted parameter with the assumption that non-saturable binding is a constant fraction of the free ligand. Correlation coefficients were calculated on a Texas Instrument TI55 calculator.

Binding of TSP to Unstimulated and Thrombin-stimulated
Platelets-When lZ5I-TSP (6 nM) was incubated with unstimulated platelets at 37 "C, time-dependent binding was observed ( Fig. 1). An apparent steady state was attained at 25 min as the extent of binding did not change with an additional 20 min of incubation. Platelets, stimulated with 1.0 unit/ml thrombin, bound TSP with a similar time course. At the TSP input concentration utilized (6 nM), binding was augmented 4-fold by thrombin stimulation. Subsequent studies of 12'1-TSP interaction with either thrombin-stimulated or resting platelets were evaluated at 30 min to ensure that steady state binding has been attained. Approximately 70% of the TSP bound to the resting or thrombin-activated platelets was initially reversibly bound. As has been observed with fibrinogen binding (5), bound TSP was subsequently stabilized, becoming irreversibly bound by 30 min of incubation ( Table  I).
To show that the binding of lZ51-TSP to platelets was not due to its modification by radioiodination, varying proportions of 1251-TSP and unlabeled TSP were added to platelets, maintaining a constant total TSP concentration. As shown in Fig. 2.4, a plot of the "'I-TSP bound to the resting platelets uersus the per cent of 1251-TSP added was linear (correlation coefficient ( r ) = 0.99). A similar plot was obtained for 1251-TSP binding to thrombin-stimulated platelets (Fig. 2B). The correlation coefficient (r) was 0.98. These observations suggested that labeling of TSP did not alter its affinity for either   shown. Left lune, thrombin-stimulated platelets; center lune, resting platelets; and right l a n e , the starting '"I-TSP. Scanning densitometry revealed >95% of radioactivity was in the major bands in each case. Note the smearing in the lower portions of the gels of the plateletbound material due to the presence of 1 mg/ml BSA.
The "'I-TSP bound to both the resting and thrombinstimulated platelets was characterized by polyacrylamide gel electrophoresis. Autoradiograms of the gels are shown in Fig.  3. Under reducing conditions, the starting '"I-TSP yielded a single major band. The estimated molecular weight of 170,000 is consistent with that reported for the subunits of TSP (18).
The ligands extracted from both resting platelet and thrombin-stimulated platelets had mobilities identical with that of the starting "'1-TSP.
To evaluate the specificity of TSP binding to resting and thrombin-stimulated platelets, the ability of four proteins (fibronectin, ovalbumin, transferrin, and BSA) to inhibit "'I-TSP binding was analyzed (Table 11). For both resting and the thrombin-activated cells, the four proteins had a minimal effect on lzSI-TSP binding; inhibition of binding did not exceed 7%. In contrast, unlabeled TSP (500 pg/ml) produced 64% inhibition of lZ5I-TSP binding to thrombin-activated platelets and 54% inhibition of binding to resting cells.
The capacity of nonlabeled TSP to inhibit "'I-TSP binding indicates that its interaction with both resting and thrombinstimulated platelets is mediated by a limited number of binding sites. To estimate the number of these sites and their affinity, varying concentrations of TSP, containing lZ5I-TSP as a tracer, were added to unstimulated platelets. The T S P molecules bound per platelet were calculated from the specific activity of the '"I-TSP and was plotted as a function of the t.otal TSP added. The binding of TSP to resting platelets appeared to be saturable (Fig. 4A). These data could be fit to a single class of TSP binding sites by nonlinear curve-fitting analysis in the "Ligand" computer program. In the sample shown, a good fit was obtained (mean square error = 28.7) when the nonspecific binding (N1) was estimated to be 7. To derive the binding parameters indicated above, TSP had been isolated in the presence of EDTA, removal of calcium may alter the conformation of T S P (37), thereby changing its platelet binding functions. Therefore, TSP was isolated maintaining 2 mM calcium throughout its purification. In analyses such as those shown in Fig. 4    Effect of inhibitors on TSP binding PGE, and theophylline were used at final concentrations of 1 pg/ ml and 1 mM, respectively. EDTA was at a final concentration of 5 mM. Other conditions were the same as in Table 11 PGEl-theophylline mixtures were used as inhibitors of platelet activation. As shown in Table 111, this combination of antagonists did not reduce TSP binding to resting platelets (990 uersus 1020 molecules/platelet). To verify the activity of the inhibitors, their effect on TSP binding to thrombinactivated cells was assessed in parallel. At the doses of PGEl and theophyIline used, TSP binding to thrombin-activated platelets was inhibited by 73%. EDTA (5 mM) inhibited binding to stimulated cells to a similar degree and the com-bination of the two inhibitors was no more effective than either alone. Thus, TSP binding to unstimulated platelets, is unlikely to arise from low level platelet activation during cell preparation.
To assess the contribution of contaminating cells to TSP binding, the interaction of the ligand with red and mononuclear cell preparations was examined. lZ5I-TSP binding to these cells was minimal, and thrombin had no effect on the limited binding observed. A platelet preparation from the same donor contained red cells and mononuclear cells at levels of 0.3 and 0.03%, respectively, relative to platelets. At these levels of contamination, the red and mononuclear cells contributed less than 5 molecules to the total of 767 molecules bound per cell to the nonstimulated platelet preparation. Stimulus Requirements for Enhanced TSP Binding-Experiments were performed to establish that thrombin augmented binding of TSP to platelets was due to thrombin stimulation of the platelets rather than to proteolysis of the ligand (Table IV). Thrombin was preincubated with platelets prior to the addition of hirudin and lZ5I-TSP. Under this condition, binding was virtually identical with that observed in the absence of the thrombin inhibitor. In contrast, incubation of the TSP with thrombin, followed by the addition of hirudin and platelets, resulted in TSP binding similar to that seen in the absence of stimulus (Table IV). This result, coupled with the near complete inhibition of serotonin release, indicates that the hirudin was an effective thrombin antagonist and suggests that the critical interaction of thrombin is with the platelet rather than with the ligand. The ability of ADP, a stimulus for fibrinogen binding, to induce TSP binding was also assessed. Under conditions where thrombin-stimulated platelets bound 10,150 f 60 molecules/platelet, ADP (5 pM)-stimulated cells bound 1,070 f 140 molecules/platelet and nonstimulated cells bound 850 f 90 molecules/platelet. In contrast, the same ADP-stimulated cells bound 12 times more fibrinogen than resting platelets (9,300 f 500 molecules/platelet uersus 720 k 60 molecules/ platelet), confirming the activity of the ADP. ADP doses as high as 40 PM produced no increment in TSP binding. The thrombin requirement for augmented TSP binding was eval-  1'

TSP Binding to Platelets
uated. The thrombin effect was maximal at 0.25 units of thrombin/ml (Fig. 5).
Comparison of Requirements for TSP Binding to Unstimuluted and Thrombin-stimulated Platelets-The divalent ion requirements for the binding of lZI-TSP to resting and thrombin-activated platelets has been further evaluated in Table V. Platelets were prepared in divalent ion-free Tyrode's buffer (using Chelex 100) to reduce calcium and magnesium to cO.1 p~. TSP binding to thrombin-stimulated platelets was similar in the presence of added calcium or magnesium and were enhanced relative to the absence of added divalent cation. When EGTA was added to reduce extracellular calcium, TSP binding was diminished by &fold. Addition of EDTA to chelate M$+, as well, reduced binding an additional factor of 2 (Table V). In contrast, TSP binding to resting platelets was much less sensitive to alterations in the divalent cation composition of the medium.
The effects of temperature also distinguished TSP binding to resting and thrombin-activated platelets. As shown in Table VI, the extent of 1251-TSP binding to unstimulated platelets was temperature-independent in the 4-27 "C range. In contrast, binding of the ligand to thrombin-activated platelets was temperature-sensitive. Although binding to the thrombin-stimulated cells was similar at 22 and 37 "C, the interaction was reduced approximately 75% at 4 "C. The level of residual binding at 4 "C is consistent with the extent of TSP binding to nonstimulated platelets at the lmI-TSP input concentration.
As shown in Fig. 6B, fibrinogen had no effect on the binding of TSP to resting platelets. In contrast, fibrinogen inhibited    6. A, effect of fibrinogen on TSP binding to thrombin-stimulated platelets. Platelets at a final concentration of 2 X 108/ml, in modified Tyrode's buffer containing 2 mM CaC12, were stimulated for 5 min with 1 unit/ml a-thrombin. Hirudin was added at 10 units/ml; and, after 2 min, varying concentrations of fibrinogen and 200 nM lZ6I-TSP were added. Binding was measured after an additional 30 min. All incubations were at 37 "C. B, the effect of fibrinogen on TSP binding to resting platelets. Conditions are the same as in A with the omission of thrombin and hirudin, and lZ5I-TSP was present at a final concentration of 50 nM. Fg, fibrinogen.
TSP binding to thrombin-stimulated cells (Fig. 6A). This inhibition was dose-dependent, and 50% inhibition occurred at 40 nM fibrinogen. The maximal inhibition of TSP binding by fibrinogen was 82% which corresponded to 3600 TSP molecules bound per platelet. This level was consistent with the binding of TSP to unstimulated platelets at the TSP input concentration utilized.
Monoclonal antibody 10E5 immunopurifies glycoprotein GPIIb-IIIa in the presence of calcium (31). Dilutions of 10E5 inhibited TSP binding to thrombin-stimulated platelets in a dose-dependent fashion, and, at high concentrations of the antibody, 94% inhibition was achieved (Fig. 7). These same doses of 10E5 lowered lmI-TSP binding to the resting cells by less than 11%.

DISCUSSION
Both unstimulated and thrombin-activated platelets bound lZ5I-TSP. These interactions were time-dependent, not due to modification of the ligand during radioiodination and not inhibited by unrelated proteins. Authentic TSP was bound to both stimulated and nonstimulated platelets as assessed by polyacrylamide gel electrophoresis of the bound ligand. The inhibition of '251-TSP binding by nonlabeled TSP and binding isotherms, such as those shown in Fig. 4, provided clear evidence for a saturable component in the binding of TSP to both unstimulated and thrombin-activated platelets with comparable estimates for nonspecific binding in both cases (less than 0.17% of input TSP).
The interaction of TSP with resting platelets had an apparent dissociation constant, K d , of 50 nM, with 3100 k 1000 TSP molecules maximally bound per platelet. This binding was not due to the presence of a small number of activated platelets or to other contaminating cells within platelet preparations. The presence of a specific binding site for TSP on unstimulated platelets distinguishes this ligand from fibrinogen, fibronectin, and von Willebrand Factor as saturable binding of these other three adhesive proteins to unstimulated platelets has not been demonstrated. Normal plasma contains less than 25 ng/ml (50 PM) TSP (19), and this level would not result in significant occupancy of the TSP binding site on unstimulated cells. Thus, circumstances which result in secretion of TSP from platelets or other cells (38)(39)(40)(41) or exposes TSP within subcellular matrices (41) should permit occupancy of the TSP binding site on resting platelets.
With thrombin-stimulated platelets, TSP bound with an apparent dissociation constant of 250 nM and 35,600 k 9,000 molecules were maximally bound per platelet. Although the fit of the experimental data to a single site model was quite good, the release of endogenous platelet TSP (4 nM released under the conditions of the binding anlyses (42)), as well as fibrinogen, which inhibits TSP binding, and the probable presence of the nonstimulated TSP site on thrombin-stimulated platelets (see below) indicate that these parameters must be forwarded with considerable caution.
The capacity of thrombin to enhance TSP binding was due to the effect of this agonist on platelets rather than to its proteolytic modification of the ligand. This conclusion is based upon the experiments using a thrombin inhibitor hirudin, as well as on the analysis of the ligand bound to thrombin-stimulated platelets. The doses of thrombin required for optimal augmentation of TSP binding are compatible with those required for induction of platelet secretion and fibronectin receptor expression (11). ADP failed to augment TSP binding, although it did induce fibrinogen receptors on the same platelet preparation. The capacity of thrombin but not ADP to support specific bindig also occurs with fibronectin (11) and distinguishes TSP binding from that of fibrinogen (9) and von Willebrand Factor (43) which bind with high affinity to ADP-stimulated cells.
In addition to the differences in estimated binding parameters, several lines of evidence indicate that the TSP binding sites on unstimulated and thrombin-activated platelets are distinct. The differential effects of inhibitors of platelet activation (PGE, +theophylline), temperature, fibrinogen, monoclonal antibody 10E5, and divalent ions all point to the nonidentity between the TSP binding sites. Phillips et al. (21) reported that the association of endogenous TSP with the surface of thrombin-stimulated platelets was calcium dependent. In contrast, Gartner and Dockter (22) measured TSP binding to platelet surfaces and found it to be divalent ionindependent. This apparent dichotomy can now be explained by the identification of two independent TSP binding sites, one divalent ion-dependent and one divalent ion-independent. It seems most likely that both the binding sites co-exist on the surface of thrombin-stimulated platelets. Inhibitors of platelet activation, 4 "C and fibrinogen reduced lZ5I-TSP binding to thrombin-stimulated platelets by 75-85%. The level of residual TSP binding was compatible with the occupancy of the unstimulated binding sites at the TSP input concentration utilized. Monoclonal antibody 10E5 had a somewhat greater effect on TSP binding to thrombin-stimulated platelets, inhibiting the interaction by 94% rather than by the predicted 75-85%. The antibody, however, also had a slight effect on TSP binding to nonstimulated platelets (11% inhibition).
The effects of fibrinogen and fibronectin on TSP binding to thrombin-stimulated platelets merit comment. TSP has been reported to interact with fibrinogen in the solid phase with dissociation constants of 3.4 nM (44). While the affinity of TSP for thrombin-stimulated platelets appears to be 2 orders of magnitude lower, the possibility that platelet fibrinogen or fibrin, expressed on the cell surface as a result of thrombin stimulation (45), serves as a TSP binding site must be considered. This possibility provides one explanation for the inhibition of TSP binding by fibrinogen. Alternative explanations for the inhibitory effect of fibrinogen may be competition for a shared or sterically related binding site, or interaction of the two molecules in solution preventing TSP binding to the platelets. Lahav et al. (46) have shown that platelet TSP comes in close proximity to surface-bound fibronectin during spreading. The presence of less than 4000 molecules of fibronectin per platelet (30) makes it unlikely that fibronectin serves as a univalent TSP receptor for thrombin-activated platelets.
In sum, evidence has been provided for two classes of TSP binding sites on platelets. One class of sites is thrombininducible, and this interaction is calcium ion-dependent. The identification of this inducible site extends the analogy between thrombospondin and the other three adhesive proteins, fibrinogen, fibronectin, and von Willebrand Factor. The second class of sites is expressed by unstimulated platelets, and TSP binding to the site is divalent ion-independent. The existence of this site distinguishes the interaction of TSP with platelets from those of fibrinogen, fibronectin, and von Willebrand Factor, and establishes a potentially unique mechanism for thrombospondin to influence platelet function.