Thrombin Activation of Factor XI on Activated Platelets Requires the Interaction of Factor XI and Platelet Glycoprotein Ibα with Thrombin Anion-binding Exosites I and II, Respectively*

Activation of factor XI (FXI) by thrombin on stimulated platelets plays a physiological role in hemostasis, providing additional thrombin generation required in cases of severe hemostatic challenge. Using a collection of 53 thrombin mutants, we identified 16 mutants with <50% of the wild-type thrombin FXI-activating activity in the presence of dextran sulfate. These mutants mapped to anion-binding exosite (ABE) I, ABE-II, the Na+-binding site, and the 50-insertion loop. Only the ABE-II mutants showed reduced binding to dextran sulfate-linked agarose. Selected thrombin mutants in ABE-I (R68A, R70A, and R73A), ABE-II (R98A, R245A, and K248A), the 50-insertion loop (W50A), and the Na+-binding site (E229A and R233A) with <10% of the wild-type activity also showed a markedly reduced ability to activate FXI in the presence of stimulated platelets. The ABE-I, 50-insertion loop, and Na+-binding site mutants had impaired binding to FXI, but normal binding to glycocalicin, the soluble form of glycoprotein Ibα (GPIbα). In contrast, the ABE-II mutants were defective in binding to glycocalicin, but displayed normal binding to FXI. Our data support a quaternary complex model of thrombin activation of FXI on stimulated platelets. Thrombin bound to one GPIbα molecule, via ABE-II on its posterior surface, is properly oriented for its activation of FXI bound to a neighboring GPIα molecule, via ABE-I on its anterior surface. GPIbα plays a critical role in the co-localization of thrombin and FXI and the resultant efficient activation of FXI.

Human factor XI (FXI), 1 unique among the coagulation proteases in the clotting cascade, circulates in blood in zymogen form as a disulfide-linked homodimer (1)(2)(3) in a noncovalent complex with high molecular weight kininogen (HMWK) (4). Each FXI monomer is composed of an N-terminal heavy chain and a C-terminal light chain containing the serine protease catalytic triad (1,2). The heavy chain has four homologous subunits called Apple domains (designated A1, A2, A3, and A4, from the N terminus), which are not found in other coagulation proteases, with the exception of prekallikrein (3). Additionally, FXI is the only coagulation protease (with the exception of FXII and prekallikrein) that does not contain "Gla domains," sequences that are rich in ␥-carboxylated glutamic acids required for binding to phospholipids (5). However, despite the absence of Gla domains and their inability to interact with phospholipids, FXI and FXIa do bind to activated platelets in a specific and saturable manner (6,7). Because of these distinctive structural features, it is not surprising that the mode of activation of FXI is quite different from that of the other coagulation enzymes.
There are three biologically relevant proteases that can activate FXI, i.e. FXIIa, FXIa, and thrombin (8,9), all of which cleave FXI at the Arg 369 -Ile 370 bond and expose the active-site catalytic triad. FXIIa, a component of the contact phase of blood coagulation, is unlikely to be a physiologically relevant activator because FXII deficiency is not associated with clinical bleeding (10,11). Recent data suggest that thrombin is the physiological activator of FXI on the activated platelet (12)(13)(14). Normal hemostasis is initiated by the exposure of tissue factor at sites of vascular injury, followed by the formation of the FVII-tissue factor complex and the subsequent activation of FX. The resultant initial burst of thrombin generation is limited by the rapid and potent inhibition of the FVIIa-tissue factor-FXa complex by the tissue factor pathway inhibitor, but is sufficient to activate platelets, FVIII, FV, and FXI. The subsequent initiation of the consolidation (or intrinsic) pathway of coagulation by thrombin-mediated FXI activation on the activated platelet membrane leads to the sequential activation of FIX, FX, and prothrombin, generating a large secondary burst of thrombin. Thus, FXI activation is required for normal hemostasis only in cases of significant hemostatic challenge, which is consistent with clinical observations that patients with congenital FXI deficiency generally have much milder bleeding than those with FVIII deficiency (hemophilia A) (15,16). Consistent with these observations is the demonstration that thrombin is more efficient than FXIIa or FXIa in activating FXI in vitro (13).
Binding of HMWK or prothrombin appears to be required for the exposure of amino acids on the A3 domain of FXI, which binds to the glycoprotein Ib (GPIb)-IX-V complex on the activated platelet surface (12,22). The FXI homodimer binds to GPIb␣ in the GPIb-IX-V complex through one of its monomers (12,14,23). Thrombin then binds to the A1 domain and activates FXI (20,24). The binding site on thrombin for FXI has yet to be defined.
The active site of thrombin lies within a deep cleft (25,26). Access to the active site is restricted by two surface loops, the 50-insertion loop (Leu 45 -Asn 57 , thrombin B chain numbering system) and the autolysis loop (Leu 144 -Gly 155 ), that are situated on the northern and southern rims of the active-site cleft, respectively. Enzyme specificity is further defined by two ligand-binding sites (exosites) that are characterized by a high density of solvent-exposed basic residues. Many thrombin substrates, receptors, and inhibitors gain access to the active site by binding to either of the two exosites. For example, anionbinding exosite (ABE) I is important for binding to fibrinogen (27), fibrin (28), heparin cofactor II (29), PAR1 (30), thrombomodulin (31), and hirudin (32,33), whereas ABE-II is involved in binding to platelet GPIb␣ (34), protease nexin I (35), and glycosaminoglycans such as heparin (14).
Utilizing a collection of thrombin mutants generated by alanine scanning site-directed mutagenesis, we previously mapped the interactions of thrombin with some of its key substrates, including protein C, thrombin-activable fibrinolysis inhibitor, and thrombomodulin (36), fibrinogen (37), FV (38), and FVIII (39). In this study, we used this collection of thrombin mutants to identify the key residues on thrombin required for FXI activation. Our data suggest a model of thrombin activation of FXI that requires the engagement of thrombin ABE-I and ABE-II by FXI and the platelet GPIb-IX-V complex, respectively.

EXPERIMENTAL PROCEDURES
Materials-Purified human FXI and HMWK were purchased from Hematologic Technologies (Essex Junction, VT). The expression and purification of wild-type (WT) and alanine-substituted mutant thrombins from stably transfected Chinese hamster ovary AA8 cells have been described in detail previously (36,37). The concentration of active thrombin molecules was determined by titration with D-phenylalanylprolylarginyl chloromethyl ketone (PPACK; Calbiochem) using the chromogenic substrate H-D-Val-Leu-Arg-p-nitroanilide (S-2266, Chromogenix). A comparison of the catalytic activities of the purified WT and mutant thrombins toward the chromogenic substrate H-D-Phe-Pip-Arg-p-nitroanilide (S-2388, Chromogenix) has been described in detail (36,38,39). Dextran sulfate (average M r 500,000) and cyanogen bromide-activated Sepharose 4B-agarose were purchased from Sigma. Dimethylmethylene blue was obtained from Serva Fine Chemicals (Long Island, NY). Active site-inhibited thrombin was prepared by incubation of a 10-fold excess of PPACK with thrombin for 1 h at 37°C. This mixture was then dialyzed in Spectrapor tubing (M r cutoff of 3500; Spectrum Medical Industries, Inc., Los Angeles, CA) overnight in phosphate-buffered saline at 4°C. The thrombin receptor activation peptide (SFLLRN-amide) was synthesized on an Applied Biosystems 430A synthesizer and purified to Ͼ99% homogeneity by reverse-phase high pressure liquid chromatography.
Radiolabeling of Factor XI and PPACK-Thrombin-Purified FXI and PPACK-thrombin were radiolabeled with 125 I by a minor modification (20) of the IODO-GEN method to specific activities of 5 ϫ 10 6 and 2 ϫ 10 6 cpm/g, respectively. Radiolabeled FXI retained Ͼ98% of its biological activity.
Preparation of Glycocalicin-Glycocalicin was extracted from human platelets and purified as described previously (40).
Factor XI Activation in the Presence of Dextran Sulfate-Mutant thrombins were screened for their ability to activate FXI using a twostage chromogenic assay. Initially, 1 nM thrombin was incubated with 60 nM FXI for 10 min at 37°C in 100 l of assay buffer (50 mM Tris (pH 7.3), 150 mM NaCl, 0.1% bovine serum albumin (BSA), and 1 g/ml dextran sulfate) to activate FXI. Thrombin activity was inhibited with 25 units/ml hirudin. FXIa activity was measured by combining 20 l of the reaction mixture with 80 l of 600 M S-2366 and observing the change in absorbance at 405 nm on a Spectramax Plus microplate reader (Molecular Devices, Sunnyvale, CA). For 10 mutants demonstrating diminished (Ͻ50% compared with WT thrombin) FXI activation, dose-response curves were constructed over a range of thrombin concentrations using the above protocol. Linear initial rates of FXIa generation with respect to time and enzyme (thrombin) concentration were observed in the presence of dextran sulfate as previously reported (8,9,13). EC 50 values were calculated using SigmaPlot software.
Assay of Factor XI Activation in the Presence of Activated Platelets-Activation of FXI (60 nM) by thrombin (and thrombin mutants) was measured by chromogenic assay. Incubations were carried out at 37°C in 200 l of 50 mM Tris and 150 mM NaCl (pH 7.3) with 1% BSA and gel-filtered platelets (activated by incubation at 37°C for 1 min with the thrombin receptor activation peptide SFLLRN-amide at 25 M) in the presence of 42 nM HMWK, 2 mM CaCl 2 , and 25 M ZnCl 2 . After dilution to a final volume of 1 ml with 50 mM Tris and 150 mM NaCl (pH 7.3) with 1% BSA containing 600 M S-2366, the amount of free p-nitroanilide was determined by measuring the changes in absorbance at 405 nm. Hirudin (25 units/ml) was added at the end of the reaction to inhibit thrombin activity. Low levels of enzymatic activity after 5-fold dilution of the incubation mixture containing no added FXI were subtracted from all results to assure detection and measurement of only FXIa generated during experimental incubations. The amount of FXIa generated was assayed by reference to a standard curve constructed using purified FXIa. Linear initial rates of FXIa generation with respect to time and enzyme (thrombin) concentration were observed in the presence of activated platelets as previously reported (13). Dose-response curves were constructed over a range of thrombin concentrations to quantify rates of FXI activation by WT thrombin and by the 10 thrombin mutants identified as demonstrating diminished FXI activation during screening.
Thrombin-Dextran Sulfate Binding Assay-Dextran sulfate (average M r 500,000) was coupled to cyanogen bromide-activated Sepharose 4B-agarose according to the manufacturer's instructions. 3 g of Washed resin was combined with 30 ml of 10 mg/ml dextran sulfate in coupling buffer at room temperature for 2.5 h. Bound dextran sulfate was quantified by dimethylmethylene blue assay (35) and determined to be 3.2 mg of dextran/mg of resin. WT and mutant thrombins were screened for their ability to bind to dextran sulfate cross-linked to Sepharose 4Bagarose. 500 l of 10 nM WT or mutant thrombin solution was incubated at room temperature with 50 l of a 1:2 slurry of dextran sulfate-crosslinked agarose. After 10 min, the mixtures were filtered through spin columns (2000 ϫ g), and the resultant flow-through fractions were assayed for thrombin activity using the chromogenic substrate S-2238. 100 l of flow-through fractions was incubated for 10 min with 100 l of 0.4 mM S-2238, and the absorbance at 405 nm was measured at that point. Absorbance readings were determined to be linear within the prescribed thrombin concentration ranges. To quantify the amount of bound thrombin, these measurements were subtracted from the A 450 nm of the flow-through fractions from an incubation using non-cross-linked Sepharose 4B. This value was then divided by the A 405 nm of the flowthrough fractions from the non-cross-linked Sepharose 4B incubation to obtain a percentage of thrombin bound specifically to dextran sulfate.
Solid-phase Binding of 125 I-PPACK-Thrombin to Glycocalicin-We utilized a modified method of De Cristofaro et al. (41) to examine the binding of 125 I-PPACK-thrombin to plate-bound glycocalicin. Wheat germ lectin (10 g/ml) was coated on the wells of 96-well polystyrene trays (Immulon high protein capacity binding) and incubated overnight at 4°C in 50 mM carbonate buffer (pH 9.50). The remaining binding sites in the sample wells were blocked by incubation with 1% BSA in Hepes-buffered saline for 2 h. After aspiration of the BSA solution, purified glycocalicin was added to the wells at a concentration of 20 g/ml and incubated at 4°C for 1 h. After aspiration, 125 I-PPACKthrombin was applied to the wells and incubated for 1 h at 37°C. Each sample and blank well were washed with Hepes-buffered saline seven times for 1 min each, dried, and counted in a Wallac Wizard 1470 ␥-counter.
Solid-phase Binding of 125 I-PPACK-Thrombin to Factor XI-We utilized a modified method of De Cristofaro et al. (41) to examine the binding of 125 I-PPACK-thrombin to FXI. Wheat germ lectin was coated on the wells of 96-well polystyrene trays (Immulon high protein capacity binding) and incubated overnight at 4°C in 50 mM carbonate buffer (pH 9.50). The remaining binding sites of the sample wells were blocked for 2 h with 1% BSA in Hepes-buffered saline. After aspiration of the BSA solution, purified FXI was added to the wells at a concentration of 20 g/ml and incubated at 4°C for 1 h. After aspiration, 125 I-PPACKthrombin and the thrombin mutants were applied to the wells and incubated for 1 h at 37°C. Each sample and blank well were washed with Hepes-buffered saline seven times for 1 min each, dried, and counted in a Wallac Wizard 1470 ␥-counter. We utilized the Kaleidagraph program (Synergy Software, Reading, PA) to determine the K d of the thrombin-FXI interaction using the equation y ϭ (M 0 ϫ M 1 )/(M 2 ϩ M 0 ), where M 0 is the total ligand concentrations, M 1 is B max , and M 2 is K d . In competition experiments, the concentration of competitor that displaced 50% of the bound 125 I-thrombin (IC 50 ) was determined by plotting the amount of 125 I-thrombin bound to FXI or glycocalicin versus the amount of competitor ligand added.

Screening for Thrombin Mutants Defective in Dextran
Sulfate-mediated Factor XI Activation-A site-directed mutagenesis strategy was used to determine residues on thrombin important for activating FXI in the presence of dextran sulfate. A collection of 53 thrombin mutants in which solvent-exposed polar and charged residues were mutated to alanine was screened for FXI-activating ability in the presence of dextran sulfate (Fig. 1). A total of 16 mutants mapped to ABE-I, ABE-II, the Na ϩ -binding site, and the 50-insertion loop demonstrated Ͻ50% of the WT thrombin FXI-activating ability. Among these are seven ABE-I mutants, H66A, R68A, R70A, Y71A, R73A, K107A, and the double mutant K106A/K107A, with relative activating abilities ranging from 28 to 48% of that of WT thrombin. Five ABE-II mutants, R89A/R93A/E94A, R98A, R245A, K248A, and R245A/K248A/Q251A, showed greater deficiency in FXI activation, as their relative abilities were all Ͻ27%. Interestingly, the effects of the R245A and K248A mutations seem to be additive because the triple mutant R245A/ K248A/Q251A displayed an activating ability of 3%, well less than that of either the R245A or K248A mutant (27 and 11%, respectively). Mutation of Gln 251 does not contribute to the deficiency of the triple mutant because the single mutation did not cause significant loss of FXI-activating ability. The ABE-I and ABE-II mutants all have normal k cat /K m values for the substrate S-2238 (36,38,39), suggesting that their deficiency in activating FXI results from reduced binding to FXI and/or dextran sulfate rather than a deficiency in catalysis. The 50insertion loop (Leu 45 -Asn 57 loop) mutant W50A and the Na ϩbinding site mutants E229A and R233A also showed greatly diminished FXI-activating ability.
Binding of Thrombin to Dextran Sulfate-Dextran sulfate has been postulated to mediate FXI activation by acting as a template for the formation of a ternary complex along with thrombin (8). To identify region(s) of thrombin responsible for binding to dextran sulfate, thrombin mutants were screened for their ability to bind to dextran sulfate-cross-linked agarose (Fig. 2). Only four mutants bound to dextran sulfate-agarose with 50% or less efficiency. These four mutants all mapped to ABE-II: R89A/R93A/E94A (45.6 Ϯ 3.6%), R98A (34 Ϯ 1.3%), K248A (50.6 Ϯ 1.1%), and R245A/K248A/Q251A (23.7 Ϯ 2.0%). All of the other mutants bound to dextran sulfate with percentages ranging from 55 to 80%, with WT thrombin binding at 76.7 Ϯ 1.0%. As in the FXI activation screen, the ABE-II triple mutant R245A/K248A/Q251 demonstrated the largest deficiency compared with WT thrombin, and this deficiency was a result of the cumulative effects of the mutations at Arg 245 and Lys 248 , but not at Gln 251 . These data clearly implicate ABE-II as the region responsible for binding to dextran sulfate and suggest that the impairment in FXI-activating ability found among the ABE-II mutants is due solely to inefficient binding to dextran sulfate.
Thrombin Activation of Factor XI in the Presence of Dextran Sulfate-Dose-response curves were generated for 10 of the 16 thrombin mutants found deficient in activating FXI in the initial screen. The selection of these 10 single mutants is representative of the regions identified to be important in FXI activation from the initial activation screen. The EC 50 values of the mutants corroborate the findings of the initial screen, with all of the mutants demonstrating Ͻ10% activating ability over WT thrombin (0.47 Ϯ 0.05 nM). Among these, the ABE-I mu- Factor XI Activation in the Presence of Activated Platelets-To identify thrombin residues important for FXI activation under physiological conditions, dose-response curves and EC 50 values for the same 10 deficient thrombin mutants were generated using activated platelets (in the presence of 42 nM HMWK, 2 mM CaCl 2 , and 25 M ZnCl 2 ) instead of dextran sulfate in the assay (Table I). EC 50 values obtained in the presence of activated platelets were remarkably similar to those obtained from studies done in the presence of dextran sulfate. A notable exception was the ϳ5-fold discrepancy between the EC 50 values for WT thrombin in the presence of activated platelets (0.1 Ϯ 0.2 nM) compared with dextran sulfate (0.47 Ϯ 0.05 nM). This difference most likely reflects the higher affinity of WT thrombin for activated platelets than for dextran sulfate.
Binding of Thrombin Mutants to Glycocalicin-Glycocalicin, the extracellular domain of GPIb␣, was recently discovered to accelerate thrombin activation of FXI to a degree similar to that observed with activated platelets (12). Because GPIb␣ appears to be the platelet receptor for FXI co-localization with thrombin, the same 10 single thrombin mutants used in the dose-response studies were tested for their ability to competitively inhibit the binding of 125 I-PPACK-thrombin to glycocalicin in a solid-phase assay as a measure of their ability to bind to GPIb␣. Of these thrombin mutants tested, only the three ABE-II mutants R98A, R245A, and K248A demonstrated a significant reduction in their ability to bind to glycocalicin as demonstrated by their IC 50 values (Fig. 3 and Table II). All three ABE-II mutants showed at least a 100-fold decrease in binding affinity, whereas the other mutants displayed comparable or enhanced binding to glycocalicin compared with WT thrombin. These data clearly corroborate the findings of the dextran sulfate binding screen and indicate the functional similarity of dextran sulfate and glycocalicin in the activation of FXI.
Binding of Thrombin to Factor XI-To establish an experimental basis for determination of thrombin exosites that me-diate binding to FXI, thrombin was inactivated with PPACK and radiolabeled to determine direct binding to FXI in a solidphase assay. The results indicate that 125 I-PPACK-thrombin bound FXI in a saturable manner, with a K d(app) of 41.9 Ϯ 4.9 nM (Fig. 4).
Binding of Thrombin Mutants to Factor XI-The same 10 single thrombin mutants identified in the screening assay were tested for their ability to bind to FXI in a solid-phase competition assay. All of the ABE-I mutants displayed reduced binding affinity for FXI (increase in IC 50 ) compared with WT thrombin, with R73A showing a 20-fold reduction ( Fig. 5 and Table II). The three ABE-II mutants tested showed enhancement of or negligible reduction in binding ability compared with WT thrombin. The Na ϩ -binding site mutants E229A and R233A and the 50-insertion loop mutant W50A were also found to be deficient in FXI binding. Taken together, these results indicate the involvement of ABE-I, the Leu 45 -Asn 57 loop, and the Na ϩbinding site in direct interactions with FXI. DISCUSSION Thrombin activation of FXI plays an important role in amplifying thrombin generation at sites of vascular injury. In this study, utilizing a collection of thrombin mutants, we have identified the key thrombin residues important in this process (Fig. 6). We used two assay methods for FXI activation by thrombin. The initial screening was performed in the presence of dextran sulfate because it is technically more straightforward (8). The key residues identified were further characterized by dose-response studies, and their importance was confirmed by FXI activation in the presence of activated platelets. Although dextran sulfate is commonly employed in studies of FXI activation by thrombin (8,9,20), it should be noted that activated platelets and dextran sulfate may promote thrombincatalyzed FXI activation by similar but non-identical mechanisms because the binding site for the activated platelet surface has been localized to the A3 domain of FXI (14,23), whereas binding to dextran sulfate is mediated through the A1 domain (24).
Despite these differences, the key thrombin residues identified to be important in FXI activation in the presence of dextran sulfate were confirmed by the subsequent activated plate- FIG. 2. Comparison of WT and mutant thrombin abilities to bind to dextran sulfate. WT and thrombin mutants were incubated with either noncross-linked or dextran sulfate-crosslinked agarose, and the residual free thrombin activities were determined by S-2238 hydrolysis as described under "Experimental Procedures." Results are expressed as percent thrombin-bound (means Ϯ S.D. of at least three separate experiments, each done in duplicate). Mutants that bound with Յ50% efficiency are indicated by black bars, and WT thrombin is indicated by the hatched bar.
let assay system (Table I). All four major functional domains of thrombin were found to be involved in the process, including ABE-I, ABE-II, the 50-insertion loop, and the Na ϩ -binding site (Fig. 6). The specific roles of the exosites in the FXI interaction were delineated by solid-phase binding experiments using mutants that were found to be defective in activating FXI in the initial activation screen. Four ABE-I single mutants with impaired FXI-activating ability (R68A, R70A, Y71A, and R73A) were chosen for further characterization and found to be deficient in the activated platelet assay system (Table I) and in binding to FXI (Fig. 5 and Table II), but demonstrated normal binding to glycocalicin (Fig. 3 and Table II). These single mu-tations within ABE-I could directly disrupt specific contacts or indirectly alter the local topology of ABE-I for the interaction of thrombin and FXI. Future x-ray crystallographic studies on the thrombin-FXI complex will help to resolve this issue.
In contrast to the ABE-I mutants, three ABE-II single mutants with impaired FXI-activating ability (R98A, R245A, and K248A) displayed no significant impairment in binding to FXI (Fig. 5 and Table II), whereas they showed major impairment in binding to glycocalicin (Fig. 3 and Table II). These results implicate ABE-I in binding to FXI and ABE-II in binding to glycocalicin. Recently, De Cristofaro et al. (34) found the following ABE-II residues to be involved in binding to GPIb␣: Arg 89 , Arg 94 , Arg 98 , Arg 245 , and Arg 248 . These residues were also implicated in our dextran sulfate binding screen, suggesting similarity to the role of GPIb␣ in FXI activation. Furthermore, unlike the other ABE-II residues tested, substitution of the ABE-II residue Gln 251 with alanine does not result in reduced binding to dextran sulfate or to GPIb␣ (34). This confirms a role for electrostatic effects in ABE-II binding to GPIb␣, first suggested by its salt dependence (42). Whether binding to GPIb␣ allosterically modifies the structure of thrombin is still in question because conflicting results have been published indicating no effect on PAR1 peptide hydrolysis (42), but inhibitory effects on fibrinopeptide A release and D-Phe-Pro-Arg-p-nitroanilide hydrolysis (43) and, more recently, FVIII activation (44).
The present data indicate that Trp 50 in the 50-insertion loop is important in mediating FXI activation in the presence of dextran sulfate or activated platelets (Table I). Substitution of alanine for Trp 50 had no effect on thrombin binding to glycocalicin ( Fig. 3 and Table II), but resulted in severe impairment of thrombin binding to FXI (Fig. 5 and Table II). Trp 50 plays an important role in defining the apolar S2 subsite and has been shown to make substantial contacts in the thrombin complex with direct thrombin inhibitors such as hirudin (32), hemadin (45), and PPACK (26) as well as with fibrinopeptide A (46). Alanine substitution of Trp 50 significantly perturbs thrombin activity with all of the substrates tested thus far, disrupting fibrinogen clotting and the activation of protein C, thrombinactivatable fibrinolysis inhibitor, FV, and FVIII (36 -39, 47). However, it does not affect the ability of thrombin to bind the DNA thrombin aptamer or fibrin, both of which bind to a defined site in ABE-I (37, 47), suggesting that the insertion loop and the ABE-I site are functionally independent. The significant impairment of FXI activation by the W50A mutant  is consistent with either loss of direct contact with FXI or disruption of the S2 subsite.
Consistent with all of our previous thrombin-substrate mapping studies (36 -39) is the demonstration in this study of the importance of Glu 229 and Arg 233 at the Na ϩ -binding site in FXI activation (Table I) and in thrombin binding to FXI (Fig. 5 and Table II), whereas in contrast, mutations of the Na ϩ site residues (E229A and R233A) led to normal or enhanced binding to glycocalicin ( Fig. 3 and Table II). The Na ϩ -binding site is adjacent to the autolysis loop that forms the southern rim of the active-site cleft (Fig. 6). Na ϩ -bound thrombin ("fast" form) has a predominantly procoagulant activity, whereas Na ϩ -free thrombin ("slow" form) has a predominantly anticoagulant property (48 -50). Based on empiric screening using this collection of thrombin mutants, we previously identified the thrombin mutants E229A and E229K as essentially anticoagulant forms of the protease with markedly diminished procoagulant properties, but with substantial activity in activating protein C and thrombin-activatable fibrinolysis inhibitor (35,36,51,52). Recently, a new crystal structure of thrombin in which the active site was not occupied was obtained (53). It reveals a  I-PPACK-thrombin to uncoated wells was Ͻ0.1% of the total binding, and the maximum variation of counts/min bound for each experimental observation was Ͻ2% of the total counts/min bound. The amount of 125 I-PPACK-thrombin bound to wells coated with BSA (negative control) was subtracted from the total binding. Results shown represent the means Ϯ S.E. of three experiments, each done in duplicate.
previously unobserved closed conformation of thrombin, which satisfies the conditions of the slow form of thrombin, supporting the thesis that thrombin can switch between these two conformational states. The substantial loss of FXI activation by the E229A and R233A mutants is consistent with these previous findings. Alanine substitution at the Glu 229 or Arg 233 site may disrupt Na ϩ binding and significantly alter the active-site conformation, leading to the closed form of thrombin.
Thus, our current data suggest a quaternary complex model of thrombin activation of FXI on the activated platelet surface. Thrombin binds to the GPIb-IX-V complex via ABE-II on the posterior surface of thrombin and to FXI via ABE-I on its anterior face (Fig. 6). At least one thrombin-binding site on GPIb␣ has been localized to residues 269 -287 (54) and is mediated by electrostatic interactions and perhaps direct contacts. FXI binding to GPIb␣ through its A3 domain is achieved when one of its homodimers is first complexed with HMWK or prothrombin (14,23,24). Once bound to the GPIb-IX-V complex by one of its monomers, FXI can interact with an adjacent thrombin molecule through the A1 domain of its other, free monomer (12). Thus, thrombin bound to one GPIb␣ molecule, via ABE-II on its posterior surface, is properly oriented for its activation of FXI bound to a neighboring GPIb␣ molecule, via ABE-I on its anterior surface. In this model, GPIb␣ plays a major role in co-localizing thrombin and FXI to form the quaternary complex. It would account for the inability of either fibrinogen or HMWK to inhibit this activation in the presence of activated platelets (13).
From our recent thrombin-substrate mapping studies, we found that the binding of thrombin to the GPIb-IX-V complex on the platelet surface is also important in thrombin cleavage of platelet GPV and PAR1. 2 Thus, engagement of thrombin on GPIb-IX-V via ABE-II may facilitate the interaction of thrombin with its physiological substrates on the platelet surface, which may represent the first step in the thrombin-platelet interaction process. The GPIb␣-bound thrombin is then in a position to cleave PAR1, GPV, and FXI or to bind another GPI␣ molecule, all mediated by ABE-I. The specific outcome of this thrombin-substrate interaction will depend on the relative accessibility of these substrates or ligands and other factors that are not yet defined. One notable difference in the activation of FXI by thrombin is that it preferentially occurs only on the activated platelet surface (12), whereas the cleavage of GPV or PAR1 takes place before platelet activation. Because platelet activation also results in a substantial reduction in the number of GPIb molecules from the platelet surface (55) and redistribution to the surface canalicular system (56), this would imply that the binding site for FXI is mediated by a new conformational state of GPIb␣ that is formed only after platelet activation. One possibility is that thrombin activation leads to redistribution and co-localization of a subpopulation of thrombin, GPIb-IX-V, and FXI in lipid rafts on activated platelets, which is responsible for the bulk of FXI activation (22). A prediction from this model is that part of the platelet-bound thrombin should be found in lipid rafts, along with the other components of this enzyme-cofactor-substrate complex. Recently, cleavage of GPV by thrombin has been shown to play a signaling role in platelet activation (57). Whether the cleavage of GPV also facilitates the exposure of the FXI-binding site in the GPIb-IX-V complex remains to be tested. FIG. 6. Space-filling model of thrombin residues critical in activating FXI. The Rasmol Version 2.6 software package was used to depict thrombin (EC 3.4.21.5; Protein Data Bank code 1PPB) as a space-filling model with solvent removed. The 10 residues identified to be essential for efficient FXI activation are labeled using the single amino acid code. A, the classic frontal view of thrombin with the activesite cleft running horizontally, left to right, with the 50-insertion loop on the upper rim of the cleft occluding the active-site Ser 205 (orange). The active site is shown with the bound active-site inhibitor PPACK as a brown stick model. Critical Na ϩ -binding site residues are colored separately in shades of green, whereas the 50-insertion loop residue Trp 50 is colored yellow. ABE-I residues are colored in shades of blue and are located to the right of the active-site cleft, whereas ABE-II is located to the left of the cleft. B, the same molecule rotated 90°to the right with critical ABE-II residues now visible and colored in shades of red.