The interaction of factor XIa with activated platelets but not endothelial cells promotes the activation of factor IX in the consolidation phase of blood coagulation.

We have previously shown that the zymogen factor XI (FXI) binds to activated platelets but not to human umbilical vein endothelial cells (HUVEC), a conclusion that is in conflict with previous reports stating that FXI binds to 2.7-13 x 10(6) high affinity sites per HUVEC (Berrettini, M., Schleef, R. R., Heeb, M. J., Hopmeier, P., and Griffin, J. H. (1992) J. Biol. Chem. 267, 19833-19839; Shariat-Madar, Z., Mahdi, F., and Schmaier, A. H. (2001) Thromb. Haemostasis 85, 544-551). It has also been reported that activated FXI (FXIa) binds to 1.5 x 10(6) sites per HUVEC and promotes the activation of factor IX by cell bound FXIa (Berrettini, M., Schleef, R. R., Heeb, M. J., Hopmeier, P., and Griffin, J. H. (1992) J. Biol. Chem. 267, 19833-19839). Therefore, the binding of FXIa to activated platelets was compared with FXIa binding to HUVEC and HEK293 cells immobilized on microcarrier beads. Specific and saturable zinc-dependent FXIa binding was demonstrated to 250 +/- 48 sites per activated platelet (K(D) = 1.7 +/- 0.78 nm) and 6.5 +/- 0.4 x 10(4) sites per HUVEC (K(D) = 2.4 +/- 0.5 nm), whereas no binding to HEK293 cells was detected. A titration with high molecular weight kininogen had no effect on FXIa binding to platelets, but revealed a concentration-dependent decrease in the amount of FXIa bound to HUVEC. The rate of factor IXa generation catalyzed by FXIa was unaffected by the presence of surfaces; however only the activated platelet surface protected FXIa from inhibition by protease nexin 2. The results presented here confirm the conclusion that activated platelets are procoagulant while unstimulated endothelial cells are not.

Human factor XI (FXI) 1 is a homodimeric protein that is essential for initiating the consolidation phase of blood coagulation. FXI is activated to factor XIa (FXIa) by factor XIIa (FXIIa), by thrombin, or by autocatalysis (1)(2)(3). The activation of FXI to FXIa requires the presence of a negatively charged surface, which under physiological conditions is provided by activated platelets (4). The binding of FXI to platelets (and also to heparin) is mediated by the Apple 3 domain of FXI (5)(6)(7), whereas the binding of FXIa to heparin occurs solely through a cysteine constrained loop within the catalytic domain (8). The FXIa domain that mediates its binding to activated platelets has not been identified. The natural substrate for FXIa is factor IX (FIX), which is activated to FIXa, the enzyme component of the FX activating complex on activated platelets (9,10).
The homodimeric nature of FXIa is thought to be necessary for the simultaneous binding of the enzyme to the activated platelet surface, mediated through one subunit and the binding of FIX by the second subunit (11). Alternatively, the activation of FIX, through proteolytic cleavage resulting in the release of the activation peptide by FXIa on activated platelets, may require the active sites of both subunits because monomer FXIa can activate FIX only in solution, not in the presence of activated platelets (11). Interestingly, FIX is activated by hydrolysis of two scissile bonds without release of an intermediate (12). The substrate binding site for FIX within FXIa has been postulated to reside within the Apple 2 domain from studies utilizing small peptide competition, whereas studies with chimeric proteins suggest that this site resides within the Apple 3 domain (13)(14)(15)(16)(17).
FXIa binds with high affinity (K D ϳ3.25 nM) to ϳ225 sites on the activated platelet surface (18,19). Binding of FXIa to activated platelets was shown to require the presence of high molecular weight kininogen (HK), but the rates of FIX activation in the presence and absence of activated platelets were similar (18,19). This suggests either that binding of FXIa (18) and FIX (10) to the platelet surface do not increase rates of FIX activation or alternatively that an actual rate enhancement is obscured by simultaneous inhibition of FXIa by protease nexin 2 (PN2) secreted from platelets (20 -22). It has also been reported that FXIa binds with high affinity (K D ϳ1.5 nM) to sites (ϳ1.5 ϫ 10 6 per cell) exposed on unstimulated human umbilical vein endothelial cells (HUVEC) (23). Binding of FXIa to unactivated HUVEC was reported to require the presence of both HK and zinc ions, and HUVEC-bound FXIa was reported to activate FIX at reduced rates compared with solution phase FIX activation (23).
Previously (24), we have addressed the question of whether the initiation of the consolidation phase of blood coagulation occurs on platelets or on endothelium, since it had previously been shown that zymogen FXI binds both to activated platelets (ϳ1500 sites/platelets, K D ϳ10 nM) (4,25) and to unactivated HUVEC (2.7-13 ϫ 10 6 sites/cell; K D ϭ 4.5-6.9 nM) (23,26). We demonstrated (24) that the reported interaction of FXI with resting HUVEC represented an artifact of the binding assay carried out in microtiter wells (23,26). During the course of these previous studies we developed an assay for examining the binding of FXI to HUVEC cultured on collagen-coated microcarrier beads, permitting direct comparison with FXI binding to activated platelets (24). In the present study, we have utilized this assay (24) to examine the binding of FXIa to both platelets and HUVEC. We have confirmed that FXIa binds specifically and reversibly with high affinity to sites that are distinct from FXI on both platelets and HUVEC. The binding of FXIa to HUVEC is unlikely to occur physiologically, since plasma concentrations of HK abrogate FXIa binding to HU-VEC. Furthermore, the rate of FIX activation by FXIa is not enhanced by physiological surfaces; however, the activated platelet surface protects the catalytic activity of FXIa from inhibition by PN2. This report supports the hypothesis that the activated platelet surface is procoagulant and the endothelium is non-thrombogenic. Preparation of 125 I-FXIa-125 I-FXIa was prepared by the activation of radiolabeled FXI. FXI was radiolabeled with 125 I using the IODO-GEN method (27). Approximately 100 g of protein was incubated with 1 mCi of carrier-free Na 125 I (Amersham Biosciences) for 20 min in an IODO-GEN-coated vial and centrifuged through a 1-cc Sephadex G-50 packed spin column at 200 ϫ g for 5 min to separate 125 I-bound protein from free 125 I. 125 I-FXI was incubated with FXIIa (20:1 molar ratio) for 4 h at 37°C. FXIIa was removed by incubation with ␣-FXIIa linked to agarose beads for 1 h at 37°C. The antibody-linked agarose beads were generated by incubation with CNBr-activated Sepharose (1 g, Sigma) in 0.001 N HCl. The antibody ␣-FXIIa was diluted to 1 mg/ml in sodium bicarbonate buffer (pH ϭ 8.3) and an equal volume of beads was added overnight at 4°C with constant stirring. Ethanolamine (1 M final concentration) was added to the reaction mixture for 2 h with constant stirring. The resin was then washed three times in alternating sodium acetate and sodium bicarbonate. Optical density was measured for each eluate wash for the absorbance at wavelength 280 nm to detect noncross-linked antibody. The resin was stored at 4°C in sodium bicarbonate buffer with 0.03% sodium azide. After final purification and removal of FXIIa 125 I-FXIa was visualized using SDS-PAGE analysis for the appearance of the 30-and 50-kDa enzyme bands under reducing conditions. Specific radioactivity (1.1 ϫ 10 8 cpm/mol) and biological activity (246 units/mg) were determined as reported previously (24).

Proteins
Measurement of Amidolytic Activity-The substrate pyro-Glu-Pro-Arg-pNA (S-2366, Chromogenix, Môlndal, Sweden) was used to determine FXIa activity. The rate of cleavage of S-2366 (250 M) was determined by the absorbance at 405 nm and compared with a standard curve generated from FXIa.
Cell Culture-Normal pooled HUVEC were purchased from Cascade Biologicals (Portland, OR) and cultured in low serum growth supplement (2% fetal bovine serum) in a humidified atmosphere of 5% CO 2 at 37°C. Medium was changed after 24 h and then on alternate days until the cells reached confluence. Cells were passaged and no longer used after the fifth passage. Cells (1 ϫ 10 6 passage 1) were frozen in 1 ml of Me 2 SO Freeze buffer and stored in liquid nitrogen. HEK293 cells and Chinese hamster ovary cells were cultured similarly in Dulbecco's modified Eagle's medium medium supplemented with 10% fetal bovine serum and penicillin/streptomycin. HUVEC or HEK293 cells (1 ϫ 10 6 cells/ml) were added to 5 ml (1.5 ϫ 10 5 beads) of swelled Cytodex-3 collagen-coated microcarrier beads (Amersham Biosciences) that were washed in complete endothelial cell media. After 30 min at room temperature the beads and cells were transferred to a spinner flask (Wheaton Scientific, Millville, NJ) and incubated under constant stirring (60 rpm) in a humidified atmosphere of 5% CO 2 at 37°C. Microcarrier bead cultures (0.5 ml) were mixed with 0.5 ml of trypan blue (Invitrogen) to determine the extent of confluence, since the microcarrier beads, but not cell-coated beads, retained trypan blue staining (24).
Briefly, platelets were obtained from platelet rich plasma after centrifugation of whole blood collected into acid-citrate-dextrose. Platelet-rich plasma was centrifuged through a bovine serum albumin (BSA) gradient and gel-filtered through a 50-ml Sepharose 2B column. Platelet eluates were counted electronically using a particle counter (Coulter Electronics, Hialeah, FL) and then washed in 10 mM Hepes containing 5 mM EDTA once, and three times in 10 mM Hepes buffer without EDTA. Activation of platelets was measured by shape change, aggregation, and release of ADP using a Lumi-Aggregometer (Chrono-Log Corp., Havertown, PA). Equilibrium Binding Experiments Using Cells in Suspension-To eliminate the effect of FXIa binding to an artificial surface, HUVEC were cultured in suspension on microcarrier beads, and binding was measured using a procedure similar to that previously described for platelets (29). HUVEC or HEK293 cells grown to confluence on microcarrier beads were washed and resuspended in incubation buffer (0.05 M Tris-HCl, 0.138 M NaCl, 100 M ZnCl 2 , 2 mM CaCl 2 , 0.5% BSA (pH ϭ 7.35)) to a density of 10 6 cells/ml. Platelets (10 8 /ml) were incubated with 25 M TRAP in incubation buffer. Silicone oil (15 l, DC500:DC200, 4:1) was centrifuged to the bottom of a 200-l microcentrifuge tube with a narrow bore tip (Sarstedt, Inc., Princeton, NJ). The cell suspension was added to the radiolabeled protein mixture (50 l each) and incubated at 37°C for 1 h with shaking by hand every 15 min. To separate the bound from free protein, tubes were centrifuged at 12,000 ϫ g for 5 min in a vertical centrifuge (model B, Beckman Instruments, Inc., Cedar Grove, NJ). The tubes were placed in dry ice/ethanol slurry for 5 min, and the tips of the microtubes were amputated. Both the pellets (cell-bound proteins) and 10 l of the supernatant (free protein) were measured for ␥-emission. Membranes were blocked with 3% BSA in PBST overnight. Goat antihuman FIX polyclonal antibody (Affinity Biologicals, Hamilton, Ontario, Canada) was diluted 1:2000 in PBST containing 1% BSA and added to the membranes for 1.5 h. After three washes in PBST, a 1:20,000 dilution of the horseradish peroxidase-conjugated rabbit antigoat secondary antibody in PBST with 1% BSA was added to the membranes for 30 min. The FIX migration pattern was detected on the membranes using the Supersignal West Picochemiluminesent Substrate (Pierce) according to the manufacturer's suggestions and visualized by autoradiography (Kodak Eastman). Densitometry was performed using Un-Scan-It software (Silk Scientific, Inc., Orem, UT).

FXIa Binding to Platelets and HUVEC-Either
HUVEC grown to confluence on microcarrier beads or platelets isolated from fresh whole blood and activated with the thrombin receptor (PAR1) activation peptide (SFLLRN-amide or TRAP) were incubated with increasing concentrations of 125 I-FXIa. After centrifugation through silicone oil and separation of cell-bound from free protein, binding curves were generated as shown in Fig. 1. Specific and saturable binding was detected to both platelets (Fig. 1A) and HUVEC, but not HEK293 cells (Fig. 1B). Specific binding was calculated by subtracting nonspecific binding (i.e. the amount of 125 I-FXIa not displaced by a 100-fold molar excess of non-labeled FXIa) from the total amount of 125 I-FXIa bound. FXIa binds to 250 Ϯ 48 high affinity sites per platelet (K D ϭ 1.7 Ϯ 0.78 nM) and to 6.5 Ϯ 0.4 ϫ 10 4 high affinity sites per HUVEC (K D ϭ 2.4 Ϯ 0.5 nM). Both studies were carried out in the presence of 2 mM calcium ions and 25 M zinc ions (divalent metal concentrations determined to be op-timal for FXI binding to platelets (4, 25) but in the absence of HK). No binding sites were detected on either HUVEC or platelets when experiments were performed in the absence of zinc ions (data not shown).
Effect of HK-HK had previously been reported to be essential in the binding of FXIa to both HUVEC (23) and activated platelets (18,19); however, we were able to detect binding sites on both activated platelets and HUVEC in the absence of HK (Fig. 1). To examine the role of HK in the binding of FXIa to both activated platelets and HUVEC, HK was titrated into the binding reaction. A saturating concentration of FXIa (5 nM) was incubated with increasing concentrations of HK in the presence of zinc ions (25 M) and either TRAP-activated platelets ( Fig.  2A) or HUVEC (Fig. 2B). FXIa binding to activated platelets was unaffected by the presence of HK. In contrast, increasing concentrations of HK decreased the amount of FXIa bound to HUVEC, suggesting that under physiological conditions, FXIa would not bind to HUVEC.
Specificity of FXIa Binding-To determine the specificity of FXIa binding to both the activated platelets and to the HUVEC surface, we attempted to compete FXIa binding sites from the surface of platelets and HUVEC using different proteins at 100ϫ molar excess (Fig. 3). The FXIa receptor on platelets is distinct from that of the FXI receptor, because FXI was incapable of displacing FXIa. Furthermore, FXI did not displace FXIa binding to HUVEC, most likely because there are no FXI binding sites on HUVEC (24). The proteins, prekallikrein, FIX, and the IgG ␣-FXII, were also incapable of displacing FXIa binding to either HUVEC or platelets, as were the enzymes, kallikrein, FIXa, and FXa. The presence of HK did not have any effect on the binding of FXIa to platelets, but did reduce the binding of FXIa to HUVEC 83%, confirming the results in Fig.  2. We conclude from these experiments that FXIa binds specifically to both platelets and HUVEC and that FXIa binding to HUVEC is almost completely abrogated in the presence of HK at physiological concentrations.
Activation of FIX by FXIa-The activation of FIX by FXIa is the only reaction within the intrinsic or consolidation phase of the coagulation cascade where the rate appears to be unaffected by the presence of the activated platelet surface (9, 11, 30 -33). However, evidence exists that suggests activated platelets protect the activity of FXIa from inactivation by protease inhibitors (34) including PN2, which is secreted from activated platelets (22). To elucidate the role of both platelets and HU-VEC in the activation of FIX by FXIa, studies were performed in the presence and absence of PN2 (Fig. 4). In the presence of platelets (Fig. 4, B and C), HUVEC (Fig. 4, D and E) or a buffer control (Fig. 4A), no change was observed in rate of FIXa formation in the absence of PN2. The activation of FIX by FXIa was virtually completely inhibited by PN2 in solution or in the presence of unstimulated platelets and HUVEC (Fig. 4, A and  B-E). However, in the presence of activated platelets, the activation of FIX by FXIa was preserved in the presence of exogenously added PN2 (Fig. 4B). Thus the activated platelet surface had little or no direct effect on the rate of FIX activation by FXIa and the reaction proceeded at uninhibited rates when PN2 was present. In contrast, FXIa-catalyzed FIX activation was completely inhibited by PN2 in solution or in the presence of HUVEC.

DISCUSSION
The central biological question addressed by the present studies relates to the identification of the cellular locus of coagulation reactions involved in the initiation of the consolidation phase of blood coagulation. Cellular surfaces have previously been suggested to participate in this important physiological process by binding coagulation FXI, promoting its proteolytic activation, binding the enzyme, FXIa, and facilitating the activation of FIX to FIXa and subsequent coagulation reactions leading to generation of sufficient amounts of thrombin to effect normal hemostasis. Two of these cellular surfaces are the surface membranes of activated platelets (35) and resting endothelial cells (23,26). We have previously compared these two cellular surfaces for their capacity to bind FXI and promote its activation to FXIa (24). This comparison required the development of a novel method for immobilizing endothelial cells (HUVEC) on microcarrier beads because of the very high level of nonspecific binding of HK to microtiter plates that resulted in the artifactual appearance of a very large number of FXI binding sites originally, which were erroneously interpreted as FXI receptors on resting endothelial cells (23,24,36). Our studies demonstrated that activated platelets exposed a discrete number of saturable, specific, high affinity FXI binding sites and promoted the activation of FXI preferentially by thrombin (24,37) and also by FXIIa and FXIa (37). In contrast HUVEC, either in the resting state or after activation with thrombin or the thrombin receptor (PAR1) activation peptide (SFLLRN-amide), were unable to bind FXI (24).
FXIa binding to activated platelets did not require the presence of added HK (Figs. 1A and 2A). This result differs from a previous report of studies carried out in the absence of added zinc ions that demonstrated a requirement for added HK to promote FXIa binding to activated platelets (19). HK that is bound to platelets either from plasma or released from platelets may explain this discrepancy. HK did, however, abrogate the binding of FXIa to HUVEC (but not activated platelets, Fig.  2A) in a concentration-dependent manner (Fig. 2B). FXIa and HK may compete for the same receptor on HUVEC; proposed HK receptors include the uPAR, C1qR, cytokeratin-1 receptor complex (38 -40), the glycoprotein Ib-IX-V receptor complex (41), and heparan sulfate glycosaminoglycans located at cell junctions (42). Another potential explanation is that HK does not compete with FXIa for the same binding site, but rather the FXIa/HK complex remains sequestered in solution and does not associate with the HUVEC surface. It can be concluded from these observations that under physiological conditions in vivo, where the concentration of HK in blood plasma is ϳ650 nM, FXIa would fail to bind to HUVEC, but would bind to the surface of activated platelets.
The presence of the activated platelet surface in both the thrombin-generating and FXa-generating complexes increases the overall catalytic efficiency by many orders of magnitude. In contrast, the rate of FIX activation by FXIa is unaffected by the presence of activated platelets or phospholipids (9,11,30,31,33) even though both enzyme (FXIa) and substrate (FIX) bind to activated platelets with high affinity (10,19). The fact that activated platelets release protease inhibitors such as PN2, which potently inhibits FXIa (K i ϳ 500 pM) (20,43) suggests that the activated platelet surface could theoretically increase the rate of FIX activation and simultaneously, through the release of PN2, inhibit FXIa, thereby limiting the detection of an increased rate of FIXa generation.
Studies by Scandura et al. (22) using a small florescent peptide substrate (BOC-EAR-AMC) determined that activated platelets release 1-1.5 nM PN2/10 8 platelets and that in the presence of HK the surface of activated platelets protects FXIa from inactivation. Platelets activated by TRAP secreted enough PN2 to significantly inhibit low concentrations of FXIa (25 pM) (22); however, this does not represent FXIa bound to activated platelets, because FXIa was used at a concentration 70-fold lower than the K D (ϳ1.7 nM). We tested the ability of both platelets and HUVEC to promote the activation of the physiological substrate for FXIa, FIX, in the presence and absence of PN2 utilizing a FXIa concentration (2 nM) approximating the equilibrium dissociation constant that allows for the maximum number of FXIa molecules to be bound to the surface, compared with the number of FXIa molecules free in solution (i.e. ϳ50%).
Since there are no substrates available that are specific for FIXa and not FXIa, we detected FIXa generation using Western blot analysis (Fig. 4). Using this method we were unable to detect an increase in rate of FIXa formation over that found in solution, regardless of the presence of surface. Exogenously added PN2 inhibited the formation of FIXa unless platelets pretreated with TRAP were added to the reaction.
TRAP-activated platelets (3 ϫ 10 8 /ml) have been shown to secrete 3-5 nM PN2 into solution (22) and possibly as much as 30 nM PN2 (20). Nonetheless, the rate of FIX activation by FXIa was virtually identical in solution (Fig. 4A) and in the presence of activated platelets (Fig. 4C) in our experiments. Since FXIa contains two active sites a 2-fold molar excess of PN2 is required for complete inactivation of FXIa (22). However, when a 10-fold molar excess of PN2 to FXIa was present, TRAP-activated platelets supported FXIa-catalyzed FIX activation at rates identical to those observed in the absence of PN2 (Fig.  4C). In contrast, in the presence of HUVEC, PN2 completely abrogated FXIa-catalyzed FIX activation (Fig. 4, D and E). We conclude that activated platelets, but not endothelial cells, protect FXIa from inactivation by PN2. Thus whereas neither the platelet nor the endothelial cell membrane increases the rate of FIX activation, the activated platelet promotes the initiation of the consolidation phase of blood coagulation by providing a protective nidus for FXIa-catalyzed FIX activation. By secreting PN2, activated platelets promote the inactivation of FXIa in solution and localize subsequent coagulation reactions to the platelet hemostatic thrombus. These observations are consistent with the concept that the primary phenotype of the endothelium is anticoagulant and non-thrombogenic, whereas the major function of activated (but not resting) platelets is to participate in the initiation of the consolidation pathway of blood coagulation leading to the local explosive generation of thrombin at sites of vascular injury.