Tissue Factor Pathway Inhibitor Binds to Platelet Thrombospondin-1*

Tissue factor pathway inhibitor (TFPI) is a Kunitz-type serine proteinase inhibitor that down-regulates tissue factor-initiated blood coagulation. The most biologically active pool of TFPI is associated with the vascular endothelium, however, the biochemical mechanisms responsible for its cellular binding are not entirely defined. Proposed cellular binding sites for TFPI include nonspecific association with cell surface glycosaminoglycans and binding to glycosyl phosphatidylinositol-anchored proteins. Here, we report that TFPI binds specifically and saturably to thrombospondin-1 (TSP-1) purified from platelet α-granules with an apparentKD of ∼7.5 nm. Binding is inhibited by polyclonal antibodies against TFPI and partially inhibited by the B-7 monoclonal anti-TSP-1 antibody. TFPI bound to immobilized TSP-1 remains an active proteinase inhibitor. Additionally, in solution phase assays measuring TFPI inhibition of factor VIIa/tissue factor catalytic activity, the rate of factor Xa generation was decreased 55% in the presence of TSP-1 compared with TFPI alone. Binding experiments done in the presence of heparin and with altered forms of TFPI suggest that the basic C-terminal region of TFPI is required for TSP-1 binding. The data provide a mechanism for the recruitment and localization of TFPI to extravascular surfaces within a bleeding wound, where it can efficiently down-regulate the procoagulant activity of tissue factor and allow subsequent aspects of platelet-mediated healing to proceed.

Blood clotting is initiated following a vascular injury when blood is exposed to tissue factor (TF) 1 present on the surface of perivascular smooth muscle cells and fibroblasts. TF is a 50 kDa membrane associated protein that binds to plasma factor VII⅐VIIa forming a catalytic complex that initiates the blood coagulation cascade through activation of factors IX and X, which lead to thrombin generation and fibrin formation. TF procoagulant activity is regulated, in part, by tissue factor pathway inhibitor (TFPI). TFPI is a trivalent, Kunitz-type serine proteinase inhibitor that inhibits the active site of factor Xa with the second Kunitz domain and the active site of the factor VIIa⅐TF catalytic complex with the first Kunitz domain. The third Kunitz domain does not have an identified function (1). Following the third Kunitz domain, TFPI has a highly basic C-terminal region that is required for rapid inhibition of factor Xa by the second Kunitz domain (2)(3)(4) and for its association with cell surfaces (5). Although antithrombin has also been shown to inhibit factor VIIa in vitro (6), TFPI is the only proteinase inhibitor that down-regulates TF procoagulant activity at physiologically relevant rates (7)(8)(9). When used as a therapeutic agent, TFPI has been shown to prevent disseminated intravascular coagulation and death from Escherichia coli sepsis in baboons (10) and to attenuate endotoxin-induced coagulation in humans (11).
It appears that the majority of TFPI is produced by endothelial cells and remains associated with the endothelial surface (12,13). This pool of TFPI contains an intact basic C-terminal region and is thought to be localized and oriented on the cell surface in a manner that allows it to simultaneously inhibit factor VIIa and factor Xa prior to dissociation of the newly activated factor X from the factor VIIa⅐TF catalytic complex (14). Thus, it is likely that TFPI is most effective as a surfacebound inhibitor of blood coagulation. However, the mechanisms responsible for TFPI binding to the endothelium are not entirely defined. Because heparin infusion results in a 2-to 10fold increase in the circulating TFPI concentration (15)(16)(17), nonspecific interactions with glycosaminoglycans are often cited as a primary mode of cell surface association. However, there is a growing body of evidence indicating that a portion of endothelial-associated TFPI is bound to glycosyl phosphatidylinositol (GPI)-anchored proteins in a manner that is not dependent on glycosaminoglycans or altered by heparin (13,18,19). We have previously shown that glypican-3, a GPI-anchored proteoglycan, binds specifically to TFPI and that the binding is likely mediated by its protein core (20).
In addition to being associated with the endothelium, TFPI is also present in circulating plasma and within platelets. The plasma form of TFPI is largely associated with lipoproteins and is variably C-terminally truncated (21). Because of the reduced anticoagulant activity of circulating TFPI, it is not thought to be an important in vivo inhibitor of TF initiated coagulation (21,22). The TFPI in platelets represents about 8% (8 ng/ml) of the total TFPI in whole blood and is released after platelet activation. Platelet TFPI has an intact basic C-terminal region and may account for the increasing TFPI concentrations found in blood samples obtained from the site of a template bleeding time wound (23).
Because TFPI is thought to be a surface-associated inhibitor of coagulation, we investigated the mechanisms through which TFPI may down-regulate factor VIIa⅐TF catalytic activity in the extravascular space after vascular injury. We examined the interaction of TFPI with platelet ␣-granule proteins and found that TFPI binds specifically and saturably to thrombospondin-1 (TSP-1). TSP-1 accounts for about 25% of the protein within platelet ␣-granules and is secreted when platelets are activated at sites of vascular injury (24). After secretion, TSP-1 is a transient component of the inflammatory extracellular matrix of healing wounds (25,26) and also binds to several cell surface integrins (27)(28)(29)(30), thereby acting as a "molecular bridge" between activated platelets and other cells within the wound (31). The binding interaction between TFPI and TSP-1 described here suggests that TSP-1 released from platelet ␣-granules also acts to localize TFPI to surfaces within the extravascular space, where it can efficiently down-regulate TF-initiated coagulation after vascular injury.

EXPERIMENTAL PROCEDURES
Proteins-Recombinant full-length human TFPI produced in Escherichia coli was a gift of the Chiron Corp. (Emeryville, CA) and the Searle Corp. (Skokie, IL). TFPI-160, an altered form of TFPI truncated after Gly-160, was produced in E. coli and purified as described previously (5 Platelet ␣-granule proteins were obtained by thrombin stimulation of 6-day-old (freshly outdated) apheresis platelets (LifeBlood, Memphis, TN) as described by Frazier and Santoro (32). In brief, the platelets were washed three times in phosphate-buffered saline containing 5.5 mM glucose at 23°C and resuspended in the same buffer at a concentration of 1 ϫ 10 10 platelets/ml. Thrombin was added to 0.5 unit/ml, and the platelets were rocked gently. Phenylmethylsulfonyl fluoride was added to 1 mM final concentration immediately after visual observation of platelet aggregation, which typically occurred 1-2 min after the addition of thrombin. Calcium chloride was added to 1 mM final concentration, and the secreted ␣-granule proteins were separated from the activated platelets by ultracentrifugation at 25,000 rpm in a SW40Ti rotor (56,000 ϫ g) for 1 h at 4°C. TSP-1 was purified from the ␣-granule preparations by gelatin agarose, heparin agarose, and gel filtration chromatography as described (32). In some experiments TSP-1 purchased from Hematologic Technologies was used.
SDS-PAGE-Proteins were analyzed using continuous 5-15% linear gradient gels in the 2-amino-2-methyl-1,3-propanediol/glycine/HCl buffer system described by Bury (33). Prior to electrophoresis, some samples were mixed with sample buffer containing 1% SDS but not boiled or reduced, whereas others were boiled for 3 min in the presence of 1% SDS and 50 mM dithiothreitol as indicated.
TFPI Ligand Blots-After SDS-PAGE, proteins were transferred to nitrocellulose (Schleicher & Schuell, Keene, NH) and incubated in 3% (w/v) non-fat milk reconstituted in phosphate-buffered saline for 1 h to block nonspecific protein binding sites. TFPI was added to a final concentration of 20 nM and incubated for 2 h at 23°C. The nitrocellulose was washed three times in the above buffer and then incubated for 2 h with a 1/1000 dilution of the polyclonal anti-TFPI antibody. After washing, the nitrocellulose was incubated for 1 h with anti-(rabbit-IgG)horseradish peroxidase conjugate (Sigma) and then reacted with hydrogen peroxide and diaminobenzidine (Sigma) to develop color.
Slot Blots-This assay has been described and characterized in detail previously (20). Samples of TSP-1 (100 l), prepared as described for SDS-PAGE, were blotted on nitrocellulose with a Minifold II slot-blot system (Schleicher & Schuell). After blotting, the nitrocellulose was incubated in 3% (w/v) non-fat milk and TFPI binding was detected as described above for the TFPI ligand blots.
Western Blots-Proteins were separated by SDS-PAGE, transferred to nitrocellulose, and immunostained with appropriate primary and secondary antibodies. Proteins were visualized using hydrogen peroxide and diaminobenzidine or ECL Western blotting detection reagents (Amersham Pharmacia Biotech, Buckinghamshire, UK).
Protein Iodination-TFPI, was iodinated using either Iodobeads or Iodogen according to instructions provided by the manufacturer (Pierce Chemical Co., Rockford, IL). The concentration of the 125 I-TFPI was determined by titration with factor Xa and comparison to a stock of unlabeled TFPI of known concentration.
Microtiter Plate Assays Measuring 125 I-TFPI Binding to Immobilized Proteins-A microtiter plate assay was developed to measure direct binding of TFPI to purified proteins and to further characterize the interaction between TFPI and TSP-1. The recombinant TFPI produced in E. coli used in these experiments binds nonspecifically to a wide variety of laboratory plastics. Therefore, this assay was extensively characterized to minimize nonspecific binding to the microtiter plate. Nonspecific binding to the plate was prevented by performing all binding assays in a non-tissue culture, polystyrene 96-well plate (Costar, Corning, NY) in the presence of 5% BSA. Under the conditions of the assay, 95% of the nonspecific 125 I-TFPI binding was blocked. This concentration of BSA was more effective than either 5% non-fat milk or 10% fetal calf serum, which blocked only 63% and 65% of the nonspecific binding, respectively. Fluid phase 125 I-TFPI bound to wells coated with passively adsorbed TSP-1 in amounts ϳ10-fold higher than the background binding observed in wells coated with 5% BSA. The background binding of 125 I-TFPI to 5% BSA was measured in each experiment and subtracted from the total bound to determine the specific TFPI bound in all data presented.
The interaction between TFPI and TSP-1 was characterized by adsorbing 0.2 ml of TSP-1 or human Type I collagen at 4 g/ml for 2 h at 37°C, to a non-tissue culture-treated, 96-well polystyrene plate in Hepes-buffered saline with 1 mM calcium chloride, pH 7.4 (HBS). In some experiments 0.2 ml of non-diluted plasma was adsorbed to the plate. The plate was washed three times with HBS and blocked for 1 h with HBS containing 5% BSA. 125 I-TFPI was added at the indicated concentrations for 2 h at 37°C in HBS buffer containing 5% BSA. In experiments using antibodies, the antibodies directed against TSP-1 and fibronectin were added to the reaction mixture to a final dilution of 1:10 of the material supplied by the manufacturer. The anti-TFPI polyclonal antibody was used at a 1:10 dilution of non-purified, immune rabbit serum. The anti-TFPI 2H8 monoclonal antibody was used at a final concentration of 800 g/ml. In some experiments 5 mM EDTA, various concentrations of heparin (Fujisawa USA, Deerfield, IL), unlabeled TFPI, TFPI-160, K1K2C, or TSP-1 were added to the reaction mixture.
Tissue Factor Inhibition Assay-TFPI inhibition of factor VIIa⅐TF activity was measured using a two-step assay similar to that described by others (14,34,35). TFPI activity was measured both after binding to immobilized TSP-1 in the microtiter plate assay and in solution in the presence and absence of soluble TSP-1. In both assays, recombinant tissue factor, prepared according to manufacturer's instructions and diluted 1:10,000 to make a working stock, 0.2 nM human factor VIIa, and 20 nM human factor X were mixed in the presence of 50 mM Hepes, 100 mM NaCl, 5 mM CaCl 2 , 0.1% BSA (pH 7.4) and allowed to generate factor Xa for 30 min. The reaction was quenched by the addition of 100 mM EDTA after 30 min. The amount of factor Xa generated from this reaction was measured by monitoring cleavage of 500 M Spectrozyme Xa (methoxycarbonyl-D-cyclohexylglycyl-glycyl-arginine-p-nitroanilide acetate; American Diagnostica, Greenwich, CT). When the activity of TFPI bound to immobilized TSP-1 was measured, the plate was prepared with non-radiolabeled TFPI (2 nM) bound as described for the microtiter plate assay. In the assays using soluble proteins, 5 nM TFPI and the indicated concentrations of TSP-1 were added.

Demonstration of TFPI Binding to TSP-1 in Ligand Blots after SDS-PAGE of ␣-Granule
Proteins-Platelet activation and the secretion of ␣-granule proteins is a key component of hemostasis and vascular wound healing. We hypothesized that one or more platelet ␣-granule proteins may function to localize TFPI to extravascular surfaces after vascular injury. TFPI ligand blots of platelet ␣-granule proteins separated by non-reducing, non-boiled SDS-PAGE demonstrated that TFPI bound to several high molecular weight proteins present in the platelet ␣-granule preparations (Fig. 1, lane 2). In a control experiment, Western analysis of the platelet ␣-granule preparations for TFPI did not demonstrate any bands, indicating that the proteins identified in the ligand blot did not represent platelet TFPI, which is also secreted from thrombin-stimulated platelets (23) (data not shown).
In an initial attempt to identify the protein(s) binding to TFPI, platelet ␣-granule proteins were applied to heparin agarose and eluted with a 0.15-2 M NaCl gradient. When the column fractions were analyzed, a single band, eluting at ϳ0.5 M NaCl and migrating with an apparent molecular mass of Ͼ200,000 Da, bound to TFPI in the ligand blot assay while the lower molecular weight proteins eluted with the column flowthrough (data not shown). Because TSP-1 is a high molecular weight ␣-granule protein that elutes from heparin agarose at ϳ0.5 M NaCl, Western analysis of the platelet ␣-granule proteins using a polyclonal antibody for TSP-1 was performed. This revealed a pattern of high molecular mass bands nearly identical to that seen in the TFPI ligand blot (compare lanes 2 and 3 in Fig. 1) and indicated that TFPI bound to full-length TSP-1 as well as fragments of TSP-1 that lack the N-terminal, heparin binding domain. Because there are multiple platelet ␣-granule proteins that did not bind to TFPI in the ligand blot assay (compare lanes 1 and 2 in Fig. 1), these data suggest that TSP-1 is the primary TFPI binding protein in the platelet ␣-granule preparations. It is possible that a second high molecular weight protein with a migration pattern similar to TSP-1 also bound to TFPI, however, no proteins other than TSP-1 were detected in TFPI ligand blot analysis of platelet ␣-granule proteins fractionated by either heparin agarose or ion exchange (MonoQ) chromatography (data not shown).
Demonstration of TFPI Binding to Purified TSP-1 in Ligand Blots after SDS-PAGE and in Slot Blots-When the TFPI ligand blot was repeated using purified TSP-1, binding of TFPI was again observed (Fig. 2A, lane 4). Purified TSP-1, under reducing and non-reducing conditions, is shown in lanes 1 and 2 of Fig. 2A to demonstrate that there are no detectable con-taminating proteins that migrate similar to non-reduced TSP-1 present in the purified material. Thus, the binding of TFPI to non-reduced TSP-1 observed in lane 4 is not due to a contaminating protein. When the ligand blot was performed under reducing conditions, TFPI binding was not reliably detected. To determine the effect of reduction and SDS treatment, TFPI binding to purified TSP-1 after various treatments was measured in a slot blot assay (Fig. 2B). In this assay, binding to reduced and reduced and boiled TSP-1 was consistently observed, whereas binding to reduced TSP-1 in 1% SDS was greatly reduced. Thus, it appeared that TFPI bound to the reduced subunits of TSP-1 but not in the presence of SDS.
Demonstration of 125 I-TFPI Binding to Immobilized TSP-1 in a Microtiter Plate Assay-When increasing amounts of 125 I-TFPI are added to wells containing immobilized TSP-1, binding appeared to saturate at ϳ40 nM (Fig. 3A). Analysis of these data with a double reciprocal plot yielded an estimated appar- non-boiled, non-reduced TSP-1; 3, non-boiled, reduced TSP-1; 4, boiled, non-reduced TSP-1; 5, boiled, reduced TSP-1; 6, non-boiled, reduced TSP-1 in 1% SDS; 7, boiled, reduced TSP-1 in 1% SDS. The increased binding observed to boiled, reduced TSP-1 (5) present in this slot blot was not a consistent finding. ent K D of 26.9 nM (data not shown). However, this assay tends to underestimate the apparent K D due to dissociation of TFPI during washing steps. Therefore, the binding of 5 nM 125 I-TFPI was measured in the presence of increasing amounts of unlabeled TFPI. A significant reduction in binding of 125 I-TFPI was observed in the presence of 10 nM unlabeled TFPI. The apparent K D , estimated from the point where binding of 125 I-TFPI was decreased by 50%, is ϳ7.5 nM (Fig. 3B).
Binding of 2.5 nM 125 I-TFPI to immobilized TSP-1 was compared with that of immobilized plasma and immobilized human Type I collagen. As shown in Fig. 4A, significantly more (5-to 10-fold) 125 I-TFPI binding occurs in the TSP-1-coated wells than to either collagen-or plasma-coated wells. Binding of TFPI to plasma proteins was further investigated by performing slot blot analysis of plasma fractionated by Superose 6 gel filtration chromatography and TFPI ligand blot analysis of plasma proteins separated by non-reducing SDS-PAGE. These assays did not demonstrate any proteins present in plasma that bound exogenously added TFPI. Endogenous plasma TFPI bound to plasma lipoproteins was detected in these assays (data not shown). These findings are consistent with the low binding of 125 I-TFPI to plasma observed in the microtiter plate assay and the very low concentration of TSP-1 present in circulating blood. When purified TSP-1 was added to plasma and the mixture separated by non-reducing SDS-PAGE, TFPI bound to TSP-1 in the ligand blot assay (data not shown). Binding of 125 I-TFPI was confirmed using TSP-1 purchased from Hematologic Technologies. No differences in binding were observed between TSP-1 purified in our laboratory and the commercially prepared material (data not shown).
The Effects of Antibodies, Soluble TSP-1 and Calcium on the Binding of 125 I-TFPI to Immobilized TSP-1-Polyclonal anti-TFPI antibodies and the monoclonal 2H8 anti-TFPI antibody blocked the binding of 5 nM 125 I-TFPI to TSP-1 in the microtiter plate assay by 71% and 57%, respectively. Polyclonal anti-TSP-1 antibodies did not block binding, however, the mono-clonal B7 anti-TSP-1 antibody reduced 125 I-TFPI binding by 33% (Fig. 4B). A 100-fold molar excess of soluble TSP-1 reduced the binding of 125 I-TFPI to immobilized TSP-1 by 51% (Fig. 4B). Because TSP-1 undergoes a conformational change upon binding to calcium, 1 mM calcium chloride was included in all buffers used in the microtiter plate assay. When the binding interaction was examined in the presence of 5 mM EDTA, 125 I-TFPI binding was not affected (data not shown).
Effect of Heparin on the Binding of 125 I-TFPI to Immobilized TSP-1-Because both TFPI and TSP-1 are heparin binding proteins, the binding of 125 I-TFPI to TSP-1 in the presence of varying heparin concentrations was measured. As demonstrated in Fig. 5, 125 I-TFPI binding in the presence of heparin concentrations ranging from 0.0001 to 0.1 unit/ml was not significantly different from that observed in the absence of heparin. However, in the presence of 1 unit/ml heparin, binding was reduced to 20%, and 10 units/ml heparin reduced binding to slightly below the background binding observed in the presence of 5% BSA with no TSP-1 adsorbed to the plate.
Effect of Altered Forms of TFPI on the Binding of 125 I-TFPI to Immobilized TSP-1-The ability of heparin to block binding suggests that the basic C-terminal region of TFPI is required for binding to TSP-1. Two altered forms of TFPI were used to further investigate the role of the C-terminal domain. TFPI-160 is truncated after Gly-160 and contains the first two Kunitz domains but lacks the third Kunitz domain and the C-terminal region, whereas K1K2C contains the first two Kunitz domains and the C-terminal region but lacks the third Kunitz domain. SDS-PAGE of full-length TFPI, TFPI-160, and K1K2C is shown in Fig. 6A. Western analysis, using an antibody that recognizes only the C-terminal region of TFPI, confirmed that the full-length TFPI and K1K2C contain the C-terminal region, whereas the TFPI-160 does not (Fig. 6A). The K1K2C has a larger predicted molecular weight than TFPI-160, but it migrates more rapidly in SDS-PAGE. The reason for this behavior is not known. A 100-fold molar excess of TFPI-160 had no effect on the binding of 125 I-TFPI to TSP-1, whereas a 100-fold molar excess of K1K2C decreased binding to 15% (Fig. 6B), demonstrating that the C-terminal region of TFPI has a critical role in the binding of TFPI to TSP-1.
TFPI Bound to Immobilized TSP-1 Remains an Active Inhibitor of Factor VIIa⅐TF Catalytic Activity-The microtiter plate assay was performed using unlabeled TFPI. In these experiments the relative amount of TFPI bound to either immobilized TSP-1 or BSA was measured using the TF inhibition assay. In wells coated with TSP-1, there was a 60% reduction in the amount of factor Xa generated compared with wells coated with 5% BSA (data not shown). These data demonstrate that the TFPI binding to TSP-1 observed in the microtiter plate assays is not an artifact induced by the radiolabeling of TFPI and, importantly, they indicate that TFPI bound to immobilized TSP-1 remains an active proteinase inhibitor.
Soluble TSP-1 Enhances the Inhibition of Factor VIIa⅐TF Catalytic Activity by TFPI-To determine the effect of soluble TSP-1 on TFPI inhibitory activity, rates of factor Xa generation by factor VIIa⅐TF were measured in the presence of 5 nM TFPI and a range of TSP-1 concentrations from 0 to 100 nM (Table I). When TSP-1 and TFPI were at equimolar (5 nM) concentration, there was no effect on the rate of factor Xa generation. However, when TSP-1 was present at 50 nM, the rate of factor Xa generation decreased by over 50%. This appeared to be a saturating amount of TSP-1, because the rate of factor Xa generation did not slow further in the presence of 100 nM TSP-1. DISCUSSION TSP-1 is a 450-kDa protein with affinity for cell surfaces and extracellular matrix proteins. It consists of three identical 150-kDa subunits linked by disulfide bonds. Each subunit is made up of a linear series of functional domains, including an ϳ30-kDa N-terminal heparin binding domain, regions homologous to procollagen, properdin, and epidermal growth factor, a calcium binding domain, and a C-terminal domain (36). Cultured endothelial cells (37), fibroblasts (38), and monocytes (39) synthesize and secrete TSP-1. In vivo, it is transiently expressed in skin wounds and is incorporated into the extracellular matrix of healing tissues (25,26). As a result of its complex structure and properties, multiple potential functions have been proposed for TSP-1.
We have demonstrated that TFPI binds to TSP-1 purified from platelet ␣-granules in three separate assays, a ligand blot after SDS-PAGE, a slot blot assay, and a microtiter plate assay. Additionally, TFPI bound to immobilized TSP-1 remains functionally active and soluble TSP-1 enhances the ability of TFPI to inhibit factor Xa generation by the factor VIIa⅐TF catalytic complex. Although we cannot absolutely rule out binding to another high molecular weight ␣-granule protein, the data suggest that TSP-1 is the only TFPI binding protein present in either platelet ␣-granules or plasma. In plasma, a C-terminally truncated form of TFPI circulates associated with lipoproteins. However, exogenously added, full-length TFPI does not bind to plasma lipoproteins with high affinity (40), and we could not identify any plasma protein to which TFPI binds in the ligand blot after SDS-PAGE, slot blot, or microtiter plate assays. Additionally, TFPI does not bind to human Type I collagen, a prominent protein in the extracellular matrix that could potentially compete with TSP-1 for binding TFPI. The affinity of TFPI for numerous other extracellular matrix proteins remains to be determined. The binding of TFPI to immobilized TSP-1 is saturable with an estimated apparent K D of ϳ7.5 nM. This K D indicates that TFPI binds to TSP-1 more avidly than it does to its cellular degradation receptor on hepatocytes, the low density lipoprotein receptor-related protein, (apparent K D ϳ30 nM) (41) and is within the physiological range of TFPI concentration that would be present at the site of a vascular wound.  The binding of 125 I-TFPI to immobilized TSP-1 is readily blocked by both polyclonal anti-TFPI antibodies and the monoclonal 2H8 anti-TFPI antibody. However, anti-TSP antibodies are much less effective at blocking binding. This suggests that TFPI may preferentially bind to surface-associated TSP-1, perhaps via a cryptic epitope of TSP-1 that is fully exposed after surface binding. This hypothesis is supported by the high concentration of soluble TSP-1 (500 nM) required to block 50% of the binding of 5 nM TFPI to immobilized TSP-1 (Fig. 4B) and the 50-fold excess of TSP-1 necessary for accelerated inhibition of factor Xa generation by TFPI in the solution phase TF inhibition assay (Table I).
Experiments were performed to define structural characteristics of TSP-1 and TFPI that are important for the binding interaction. Although TSP-1 undergoes a distinct conformational change upon binding calcium (42,43), 125 I-TFPI binding was not affected when the microtiter plate assay was performed in 5 mM EDTA. This is similar to the binding of plasminogen, fibrinogen, and fibronectin to TSP-1 which also are not dependent on calcium (44,45) but different from TSP-1 binding to cellular binding sites, which tend to be calcium-dependent (28). It appears that TFPI binds to the individual 150-kDa subunits of TSP-1, because TFPI bound to reduced TSP-1 in the slot blot assay, however, the binding domain on TSP-1 remains to be localized. Heparin, at concentrations above 0.1 unit/ml, greatly reduced the binding of 125 I-TFPI to TSP-1 in the microtiter plate assay. This is likely due to heparin blocking an interaction between the basic C-terminal region of TFPI and TSP-1 based on the following interpretation of the data. First, the experiments with the altered forms of TFPI strongly indicated that the C-terminal region of TFPI was required for binding to TSP-1. K1K2C, a form of TFPI that is missing the third Kunitz domain but has the C-terminal region, competed with 125 I-TFPI for binding to TSP-1 in the microtiter plate assay, whereas TFPI-160, a form of TFPI that is missing both the third Kunitz domain and the C-terminal region, did not (Fig. 6B). Second, it appeared that the heparin binding domain of TSP-1 was not required for binding to TFPI. In the TFPI ligand blot of platelet ␣-granule proteins, TFPI bound to partially degraded forms of TSP-1 (Fig. 1). However, none of the degraded forms of TSP-1 bound to heparin agarose, suggesting that the thrombin used to activate the platelets cleaved the N-terminal, heparin binding domain of these fragments (46,47). Therefore, a portion of TSP-1, other than the heparin binding domain, is likely involved in binding TFPI. Although it appears that the majority of endogenously bound TFPI on cultured endothelial cells is associated with a GPIanchored protein and is not released from the cell surface with heparin (13), TSP-1 is made by endothelial cells in culture and TFPI bound to endothelial TSP-1 may account for a portion of the TFPI released into the circulation after heparin infusion.
It is well established that TFPI is a key regulator of TFinduced coagulation in vivo. The in utero death of mice lacking the first Kunitz domain of TFPI due to disseminated intravascular coagulation demonstrates that TFPI has a critical role in maintaining the anticoagulant properties of the endothelium (48). The intravascular function of TFPI is likely down-regulation of factor VIIa⅐TF activity transiently present on endothelial cells or monocytes that have been stimulated by inflammatory cytokines. However, under normal conditions, TF is predominantly expressed in extravascular locations surrounding the blood vessels where TFPI is not typically located. Because plasma TFPI is largely truncated at the C terminus and a poor inhibitor of blood clotting (21,22), down-regulation of extravascular TF initiated coagulation by TFPI most likely requires the release of TFPI from activated platelets or the transfer of endothelial associated TFPI into the extravascular space.
Because TSP-1 accounts for approximately 25% of the platelet ␣-granule protein secreted at sites of vascular injury (24), it is likely that TSP-1 contributes to hemostasis within the wound site, but its exact role is not clear. TSP-1-deficient mice do not have bleeding diatheses, and their platelets aggregate normally in response to thrombin (49). However, TSP-1 is involved in the early organization of the extracellular matrix of healing wounds (25,26,50). We propose that TSP-1 secreted by platelets plays an important role in recruiting and localizing TFPI to surfaces within the extravascular matrix. Once localized, it can efficiently down-regulate the procoagulant activity of TF, which initiates blood clotting within the wound, and allow subsequent aspects of platelet-mediated healing to proceed. As mentioned above, TSP-1 is an adhesive, multifunctional protein with many proposed functional roles. The in vitro data presented here indicate that a binding interaction between TFPI and TSP-1 likely occurs at the site of a bleeding wound and that binding to TSP-1 enhances TFPI inhibitory activity. Further characterization of its in vivo importance is warranted.