A Binding Site for the Kringle II Domain of Prothrombin in the Apple 1 Domain of Factor XI*

Previously we defined binding sites for high molecular weight kininogen (HK) and thrombin in the Apple 1 (A1) domain of factor XI (FXI). Since prothrombin (and Ca2+) can bind FXI and can substitute for HK (and Zn2+) as a cofactor for FXI binding to platelets, we have attempted to identify a prothrombin-binding site in FXI. The recombinant A1 domain (rA1, Glu1–Ser90) inhibited the saturable, specific and reversible binding of prothrombin to FXI, whereas neither the rA2 domain (Ser90–Ala181), rA3 domain (Ala181–Val271), nor rA4 domain (Phe272–Glu361) inhibited prothrombin binding to FXI. Kinetic binding studies using surface plasmon resonance showed binding of FXI (K d ∼71 nm) and the rA1 domain (K d ∼239 nm) but not rA2, rA3, or rA4 to immobilized prothrombin. Reciprocal binding studies revealed that synthetic peptides (encompassing residues Ala45–Ser86) containing both HK- and thrombin-binding sites, inhibit 125I-rA1 (Glu1–Ser90) binding to prothrombin,125I-prothrombin binding to FXI, and125I-prothrombin fragment 2 (Ser156–Arg271) binding to FXI. However, homologous prekallikrein-derived peptides (encompassing Pro45–Gly86) did not inhibit FXI rA1 binding to prothrombin. The peptides Ala45–Arg54, Phe56–Val71, and Asp72–Ser86, derived from sequences of the A1 domain of FXI, acted synergistically to inhibit 125I-rA1 binding to prothrombin. Mutant rA1 peptides (V64A and I77A), which did not inhibit FXI binding to HK, retained full capacity to inhibit rA1 domain binding to prothrombin, and mutant rA1 peptides Ala45–Ala54 (D51A) and Val59–Arg70 (E66A), which did not inhibit FXI binding to thrombin, retained full capacity to inhibit rA1 domain binding to prothrombin. Thus, these experiments demonstrate that a prothrombin binding site exists in the A1 domain of FXI spanning residues Ala45–Ser86 that is contiguous with but separate and distinct from the HK- and thrombin-binding sites and that this interaction occurs through the kringle II domain of prothrombin.

Factor XI (FXI), 1 a protein that participates in the intrinsic pathway of blood coagulation, consists of two identical polypep-tide chains that can be activated by FXIIa, thrombin, or factor XIa to give rise to factor XIa (1)(2)(3)(4)(5)(6)(7)(8)(9). FXI is present in plasma in its precursor or zymogen form noncovalently complexed to high molecular weight kininogen (HK) (10,11). Although the site at which FXI is cleaved by thrombin is identical to that cleaved by FXIIa (8,9), the concentrations of thrombin required are unphysiologically high and the rates of FXI activation are insignificant in solution (8,9,12,13). Moreover, although the kinetics of thrombin-catalyzed FXI activation are greatly enhanced in the presence of dextran sulfate and other unphysiologic surfaces (8,9), the fact that physiological concentrations of HK and fibrinogen potently inhibit F-XI activation by thrombin (8,9,12,13) has been used as an argument that the reaction may not occur in normal plasma (12,13). However, recently we have shown that the presence of physiological concentrations of prothrombin (ϳ1.5 M) can reverse the inhibitory effects of HK in the presence of activated platelets, which provide a surface for potentiation of F-XI activation by thrombin (14). Since prothrombin can also act as a cofactor for binding FXI to platelets (in place of HK), the activation of FXI by thrombin may be physiologically important on the platelet surface (14).
The amino-terminal portion of the FXI monomer consists of four tandem repeat sequences of 90 -91 residues, each with 6 -7 cysteine residues in a characteristic disulfide linkage termed Apple domains (4,5). Specific ligand-binding functions have been ascribed to each of the four FXI Apple domains (15)(16)(17)(18)(19)(20), including the Apple 1 (A1) domain, which has been shown to bind both thrombin (15) and HK (18,20). Our experiments also support the notion that prothrombin as well as thrombin bind FXI at a site encompassing residues Phe 56 -Ser 86 that also mediates HK binding to the A1 domain (14). The fact that prothrombin fragment 1.2 (PF1.2) but not PF1 displaced HK from its binding site on the A1 domain of FXI suggests that the kringle II domain of prothrombin binds to a site contiguous with the HK binding site in the A1 domain (14). Therefore, one objective of the present study is to identify the amino acid sequences in FXI that interact with prothrombin and PF2 (containing kringle II) in order to ascertain the relevant physiological mechanisms that promote activation of FXI by thrombin on the platelet surface. The conclusion drawn from these studies is that prothrombin kringle II can bind to the FXI A1 domain at a site contiguous with but separate and distinct from the A1 domain site utilized by thrombin to bind to and activate FXI.

EXPERIMENTAL PROCEDURES
Purification of Proteins-FXI (250 units/mg of protein) was purified from human plasma by immunoaffinity chromatography (21). FXI was assayed by a minor modification (22) of the kaolin-activated partial thromboplastin time. The recombinant Apple 1 (rA1) domain (Glu 1 -Ser 90 ) was prepared and characterized as described previously (15,18). Human prothrombin, PF1 and PF1.2, were purchased from Hematologic Technologies, Inc. (Essex Junction, VT). HK (specific activity, 15 units/mg) was purified by the method of Kerbiriou and Griffin (23). Human ␣-thrombin (2,800 NIH units/mg) was purchased from Enzyme Research Laboratories (South Bend, IN). The potent thrombin inhibitor D-Phe-Pro-Arg chloromethyl ketone (PPACK) was purchased from Calbiochem (Indianapolis, IN). Active site-inhibited thrombin was prepared by incubation of a 10-fold excess of PPACK with ␣-thrombin for 1 h at 37°C, and this mixture was dialyzed in Spectrophor tubing (M r 3,500 cut-off, Spectrum Medical Industries, Los Angeles, CA) overnight in phosphate-buffered saline at 5°C. All purified proteins appeared homogeneous by SDS-polyacrylamide gel electrophoresis.
Isolation and Purification of Prothrombin Fragment 2 (PF2) from Prothrombin-PF2 was isolated from prothrombin activation material depleted of thrombin by SP chromatography (the source of this material was from Enzyme Research Laboratories, South Bend, IN). This preparation contained 0.4 mg/ml protein in 50 mM sodium phosphate and 1 M benzamidine, pH 6.5. To 50 ml of this material, PPACK (from Calbiochem, Indianapolis, IN) was added to a concentration of 10 M. PF2 was isolated utilizing a fast protein liquid chromatography system (Amersham Pharmacia Biotech) at 23°C using a MonoQ (HR 5/5) column according to the procedure of Stevens and Nesheim (24). Preparative isolation of the material on this column results in PF1 and PF2 in two separate pools, with PF2 migrating at 14 kDa as a doublet on SDS-gel electrophoresis. From 50 ml (20 mg) of protein loaded on the column, approximately 7 mg of PF2 was recovered. An additional preparation of human PF2 was a generous gift from Dr. S. Krishnaswamy (University of Pennsylvania, Philadelphia, PA) and gave identical results in all experiments reported here.
Protein Analysis-Protein concentrations were determined by the Bio-Rad dye-binding assay according to the instructions provided by the manufacturer.
Peptide Synthesis-Peptides were synthesized on an Applied Biosystems 430A peptide synthesizer by a modification of the procedure described by Kent and Clark-Lewis (25) and purified to Ͼ99% purity using reverse phase high performance liquid chromatography (HPLC). The sequences of the synthetic peptides utilized in this study have been published previously (16 -20). All the peptides utilized in this work were rationally designed, conformationally constrained synthetic peptides based upon previously published (16 -20) molecular models for A1, A2, A3, and A4 domains of FXI. A previously published method (26) was used to oxidize the two cysteine residues in each peptide to form a disulfide bond and to conformationally constrain the peptide.
HPLC-The HPLC system employed was from Waters (600 gradient module, model 740 data module, model 46K universal injector, and Lambda-Max model 481 detector; Waters, Milford, MA). Reverse phase chromatography was performed using a Waters C8 Bondapak column, whereas gel filtration was carried out using a Waters Protein-Pak 60 column as described previously (26).
Characterization of Synthetic Peptides-All the peptides utilized in this study were examined by HPLC (both reverse phase and gel filtration), and all demonstrated a single homogeneous peak (data not shown). When the peptides were examined by HPLC (both reverse phase and gel filtration), single homogeneous peaks with identical retention times to the original mixtures were observed, demonstrating the presence of a single homogenous mixture of refolded peptides. All peptides were examined for free SH groups using the Ellman reagent (5,5Ј-dithiobis(2-nitrobenzoic acid)). It was determined (27) that there was less than 0.02 mol of free SH/mol of peptide, which further verifies that these peptides were homogeneous preparations consisting of intramolecular disulfide-bonded peptide.
Binding of 125 I-Labeled Prothrombin, Prothrombin Fragment 2, or Prothrombin Fragment 1.2 to Factor XI-The binding of prothrombin, PF2, or PF1.2 to FXI was studied using polyvinyl chloride microtiter plates, the wells of which were coated with FXI by incubation with 100 l of protein (100 g/ml) overnight at 4°C. After residual binding sites in the wells were blocked with bovine serum albumin for 2 h at 25°C, 100 l of 125 I-prothrombin, 125 I-PF2, or 125 I-PF1.2 (50 -3,000 nM) was added to the wells and incubated for 3-4 h at room temperature. The wells were thoroughly washed with phosphate-buffered saline/bovine serum albumin and dried and counted in a 1470 Wallac Wizard ␥ counter (Wallac Inc., Gaithersburg, MD).
Binding of 125 I-Recombinant Apple 1 (Glu 1 -Ser 90 ) Domain to Prothrombin-The binding of the rA1 domain of FXI to prothrombin was studied using polyvinyl chloride microtiter plates, the wells of which were coated with prothrombin by incubation with 100 l of protein (100 g/ml) overnight at 4°C. After residual binding sites in the wells were blocked with bovine serum albumin for 2 h at 25°C, 100 l of 125 I-rA1 (Glu 1 -Ser 90 ) (50 -3,000 nM) was added to the wells and incubated for 3-4 h at room temperature. The wells were thoroughly washed with phosphate-buffered saline/bovine serum albumin and dried and counted in a 1470 Wallac Wizard ␥ counter.
Binding of Factor XI and Recombinant Apple 1 Domain to Prothrombin on Surface Plasmon Resonance-Binding studies were performed on a Biacore 2000 Flow Biosensor (Biacore, Inc., Uppsala, Sweden). Prothrombin was immobilized on a carboxymethyl dextran (CM5) flow cell surface using either aldehyde linkage of oxidized carbohydrate (on prothrombin) or amine-coupling chemistry. For aldehyde linkage, prothrombin was diluted to 1 mg/ml in 100 mM sodium acetate buffer, pH 5.5, and placed on ice. After addition of sodium metaperiodate (1 mM), the prothrombin was incubated on ice in the dark for 20 min. The protein was desalted on a PD10 column that had been pre-equilibrated in 10 mM sodium acetate, pH 4.0. The CM5 cell surface was activated with a 3-min pulse of 1.15 mg of N-hydroxysuccinimide mixed with 7.5 mg of NЉ- (3-dimethylaminopropyl)carbodiimide hydrochloride at 5 l/ min. The surface was then hydrazide-modified with a 7-min pulse of 5 mM carbohydrazide in water. Non-reacted sites were blocked with a 7-min pulse of 1 M ethanolamine, pH 8.5. The oxidized prothrombin, ϳ10 g/ml in sodium acetate, pH 4.0, was injected to a response level of ϳ150 response units. The flow rate was lowered to 2 l/min. Cyanoborohydride was injected for 20 min, followed by three 1-min injections of 0.1 M glycine. For amine coupling, a 7-min pulse of N-hydroxysuccinimide/NЉ-(3-dimethylaminopropyl)carbodiimide hydrochloride was followed by injection of prothrombin, ϳ10 g/ml in sodium acetate buffer, pH 4.5, to a response level of ϳ500 response units. Any remaining derivatized carboxymethyl groups were blocked by a 7-min injection of 1 M ethanolamine. Nonspecific binding was determined by protein binding to a derivatized and blocked flow cell with an unrelated antibody bound. Wild type FXI was studied using both the aldehyde-linked prothrombin and amine-coupled prothrombin. Recombinant Apple 1 domain was studied on the amine-coupled prothrombin flow cell alone. Serial dilutions of wild type FXI and recombinant Apple domains in Hepes-buffered saline (HBS; 10 mM Hepes, 150 mM NaCl), 0.005% surfactant P20, 5 mM Ca 2ϩ , were injected with a 6-min association time and 5-min dissociation time. After subtraction of nonspecific binding curves, the association and dissociation rate constants were determined using a global fit to a one to one Langmuir association model on Biaevaluation software (Biacore, Inc.). The best fit was determined by a 2 value of less than 10 or less than 5% of the equilibrium response unit value for the highest concentration. The 2 value is the square of the differences between the theoretical ideal curve and the actual curve and was calculated according to Equation 1. r f indicates the fitted value at a given point, r x indicates the experimental value at that point, n indicates the number of data points, and p indicates the fitted parameters.

RESULTS
Binding of 125 I-Prothrombin to Factor XI-Since contiguous but distinct binding sites for HK (Val 59 -Lys 83 ) and thrombin (Ala 45 -Arg 70 ) reside within the A1 domain of FXI (18,20), we previously examined the effects of prothrombin and its fragments on the binding of FXI to HK (14). Our published studies (14) show that active site-inhibited thrombin, prothrombin, and PF1.2 all inhibit the binding of FXI to HK, whereas PF1 had no such effect. We reported that prothrombin as well as thrombin bind FXI at or near the HK binding site in the A1 domain (Phe 56 -Ser 86 ) (14). These experiments suggest that the kringle II domain of prothrombin may interact with the A1 domain of FXI at a site close to but distinct from the thrombinbinding site. We therefore investigated whether prothrombin can interact with the Apple domains of FXI by examining whether peptides derived from the amino acid sequences of the heavy chain of FXI could affect the binding of 125 I-prothrombin to FXI. The rA1 domain (Glu 1 -Ser 90 ) inhibited the binding of prothrombin to FXI with an IC 50 of 4 ϫ 10 Ϫ7 M (Fig. 1A, Table  I). In contrast, neither the rA2 domain (Ser 90 -Ala 181 ) containing a substrate binding site for FIX (19), nor the rA3 domain (Ala 181 -Val 271 ) containing both platelet and heparin binding sites (16,28), nor the rA4 domain (Phe 272 -Glu 361 ) containing the FXIIa binding site (17) were able to inhibit prothrombin binding to FXI.
To confirm these results and define the affinity of FXI binding to prothrombin, we also examined the direct binding of FXI and rA1 domains to prothrombin using surface plasmon resonance with prothrombin immobilized. The affinity of FXI binding to prothrombin was determined as shown in Fig. 2A and Table II (K d ϭ 53.9 Ϯ 5.8 nM). The K d values obtained by aldehyde linkage of prothrombin (50.7 nM) and amine coupling (58.9 nM) were essentially identical. No FXI binding was observed to biosensor chips underivatized with prothrombin. These studies also demonstrate that prothrombin interacts only with the rA1 domain (Glu 1 -Ser 90 ) with a calculated K d ϭ 239 Ϯ 83 nM (Fig. 2B, Table II). In contrast neither the rA2 domain (Ser 90 -Ala 181 ), nor the rA3 domain (Ala 181 -Val 271 ), nor the rA4 domain (Phe 272 -Glu 361 ) was able to bind prothrombin using surface plasmon resonance (Table II). Thus, these experiments reveal that a binding site for prothrombin exists in the A1 domain of FXI.
We next examined the effects of active site-inhibited (PPACK-treated) thrombin, PF1, and PF1.2 on the binding of prothrombin to FXI. The experiment presented in Fig. 1B and Table I demonstrates that PF1.2 inhibits the binding of prothrombin to FXI (with an IC 50 ϭ 1.0 ϫ 10 Ϫ6 M). PPACK-treated thrombin also displaced FXI from prothrombin (IC 50 ϭ 5.0 ϫ 10 Ϫ6 M), whereas PF1 had no effect. This experiment suggests that both prothrombin (utilizing the kringle II domain) as well as thrombin bind to the A1 domain of FXI.
The Binding of Factor XI Apple 1 Domain to Prothrombin-Since our experiments demonstrate that prothrombin can bind FXI and the rA1 (Glu 1 -Ser 90 ) inhibits this interaction (Fig. 1A), we therefore determined whether the rA1 (Glu 1 -Ser 90 ) binds prothrombin directly. Therefore, we performed experiments in which prothrombin was bound to the wells of a microtiter plate and determined whether 125 I-rA1 (Glu 1 -Ser 90 ) could bind to this protein. Our results indicate that 125 I-rA1 (Glu 1 -Ser 90 ) binds prothrombin in a saturable manner with a K d(app) of 471 Ϯ 175 nM ( Fig. 3 and Table I).
The Effect of Apple 1-derived Peptides on the Binding of Recombinant Apple 1 (Glu 1 -Ser 90 ) to Prothrombin-The true structure of the FXI A1 domain is not known, but we have constructed a molecular model of this region (18). Using this model we have made predictions about the structure of the HK and thrombin binding sites (15,18,29). Since prothrombin, thrombin, and PF1.2 inhibit the binding of FXI to HK (14) and rA1 (Glu 1 -Ser 90 ) and PF1.2 inhibit the binding of prothrombin to FXI (Fig. 1, A and B), we proposed to define the prothrombinbinding site within the A1 domain. Based upon an examination of a molecular model of the A1 domain that predicts the presence of three stem-loop structures (antiparallel ␤-strands connected by ␤-turns) defined by amino acid residues, Ala 45 -Arg 54 , Val 59 -Arg 70 , and Asn 72 -Lys 83 (15,18,29), we prepared conformationally constrained cyclic peptides comprising these three peptide loop structures to determine whether they might assume a conformation that comprises a prothrombin binding site. These peptides were identical to those tested to delineate the HK and thrombin binding sites (15,18,29). The peptides designated Ala 45 -Arg 54 (thrombin binding site) and Phe 56 -Ser 86 (HK binding site) both inhibited the binding of 125 I-rA1 (Glu 1 -Ser 90 ) to prothrombin with IC 50 values of 2.0 ϫ 10 Ϫ4 M and 7.0 ϫ 10 Ϫ5 M, respectively (Fig. 4A, Table I), suggesting that both amino acid sequences common to thrombin and HK binding sites may bind prothrombin. When both peptides were added together at equimolar concentrations, they were 1 order of magnitude more effective (demonstrating synergism) than either one alone in inhibiting FXI rA1 binding to prothrombin (Table I). Moreover, the three peptides, Ala 45 -Arg 54 , Asn 72 -Ser 86 , and Phe 56 -Val 71 , were effective inhibitors of rA1 binding to prothrombin with similar IC 50 values of ϳ2 ϫ 10 Ϫ4 M (Fig.  4B, Table I). When combinations of equimolar concentrations of these peptides were added to this assay, an even greater effect was observed (IC 50 ϭ 2 ϫ 10 Ϫ6 M), demonstrating synergism in the effects of the peptides since when the peptides were added together their effect was greater than a simple additive effect (i.e. the mixture of peptides was 2 orders of magnitude more effective than either one alone in inhibiting 125 I-rA1 domain binding to prothrombin). However, the three individual peptides, Ala 45 -Arg 54 , Asn 72 -Ser 86 , and Phe 56 -Val 71 , added to- gether were less effective (IC 50 ϭ 2 ϫ 10 Ϫ6 M) in inhibiting 125 I-rA1 binding to prothrombin than the rA1 domain (IC 50 ϭ 4 ϫ 10 Ϫ7 M) (Fig. 4 (A and B) and Table I). This may be due to a loss in conformational structure of the individual peptides. Thus, these experiments suggest that the binding site for prothrombin in the A1 domain of FXI is defined by the molecular model of the A1 domain that predicts the presence of three stem-loop structures defined by residues Ala 45 -Lys 83 (15,18,29).
Prekallikrein (PK) is a protein that shares a 58% sequence identity with FXI (4). It also binds HK in the A1 domain within the homologous amino acids Phe 56 -Gly 86 (29), but does not bind thrombin (15,30). Therefore, we tested the PK (Pro 45 -Lys 54 ) and PK (Phe 56 -Gly 86 ) peptides for their ability to inhibit FXI rA1 binding to prothrombin. Unlike the corresponding FXI peptides, the PK peptides did not inhibit FXI rA1 binding to prothrombin (Fig. 4A, Table I). Thus, the amino acid sequences involved in PK interactions with HK are not involved in binding prothrombin.
Mapping of the Prothrombin Binding Site in Factor XI-Previously, in order to gain information about the HK and thrombin binding sites in the A1 domain of FXI, we prepared synthetic peptides with amino acid substitutions, determined by examining a molecular model for residues that project their side chains into a predicted contact surface (15,29). We have identified specific amino acid residues within the A1 domain involved in binding thrombin (15). We have determined that Glu 66 and Asp 51 of the A1 domain of FXI are important components of the thrombin binding site in FXI (15). However, the altered peptides Val 59 -Arg 70 (E66A) and Ala 45 -Arg 54 (D51A) inhibited the binding of the rA1 domain to prothrombin (IC 50 ϭ 4 ϫ 10 Ϫ4 M and 1 ϫ 10 Ϫ4 M, respectively) with IC 50 values similar to that of the wild-type peptides (Fig. 5 (A and B), Table  I). Thus, prothrombin does not bind the A1 domain of FXI through amino acid sequences found to be important for thrombin binding to the A1 domain (15). By comparison, substitutions at position 50 of an alanine for glutamic acid in peptide Ala 45 -Arg 54 and at position 73 of an alanine for arginine in peptide Asn 72 -Lys 83 did not decrease the inhibitory potency (Fig. 5 (A and B), Table I). Utilizing mutational analysis, we NE, no effect up to 10 Ϫ4 M.
have determined that the binding of FXI to HK is mediated at least in part by Val 64 and Ile 77 in the A1 domain of FXI (29). Therefore, we examined the effects of mutations of these two residues on the capacity of the rA1 domain to inhibit rA1 binding to prothrombin. We found that the mutant rA1 domain constructs (V64A and I77A), which have lost the capacity to inhibit FXI binding to HK (29), retain the full capacity of the rA1 domain (Glu 1 -Ser 90 ) to inhibit rA1 binding to prothrombin (Fig. 5C, Table I). Therefore, the binding sites for HK and prothrombin, although contiguous, are apparently separate and distinct. The Binding of Prothrombin Fragment 1.2 or Prothrombin Fragment 2 to Factor XI-We have reported that PF1.2 inhibits the binding of FXI to HK, whereas PF1 has no effect (14). Our experiments suggest that prothrombin binds to FXI at a site spatially contiguous with the HK binding site and that PF1.2 (presumably using the kringle II domain) may also bind FXI near this site displacing HK from the A1 domain. In order to test this, we determined whether PF1.2 or PF2 binds directly to FXI bound to the wells of a microtiter plate. Our results indicate that 125 I-PF2 binds FXI in a saturable manner with a K d(app) of 417 Ϯ 190 nM (Fig. 6A), while 125 I-PF1.2 binds FXI in a saturable manner with a K d(app) of 449 Ϯ 172 nM (Fig. 6B).
The Effect of Apple 1-derived Peptides on the Binding of Prothrombin Fragment 1.2 or Prothrombin Fragment 2 to Factor XI- Fig. 1 and Table I show that the rA1 domain inhibits the binding of prothrombin to FXI. Furthermore, the rA1 domain binds to prothrombin and A1-derived synthetic peptides inhibit this interaction (Figs. 3 and 4, Table I). In order to extend and confirm these observations, we have determined that both PF1.2 and PF2 bind FXI in a specific and saturable manner (Fig. 6, A and B). Data presented in Fig. 7A and Table  I

DISCUSSION
Previously, we identified a sequence of amino acids in FXI that binds thrombin and is juxtaposed to the HK binding site (15). We also investigated the role of PF1 and PF1.2 and prothrombin in their interaction with FXI since prothrombin is the precursor of thrombin (14). We recently reported that prothrombin as well as thrombin binds FXI at or near the HK binding site in the A1 domain (Phe 56 -Ser 86 ) (14). These studies suggest that the kringle II domain of prothrombin binds at or near the HK binding site in the A1 domain (14). Additionally, PF1.2 abrogates the inhibitory effect of HK on thrombin-catalyzed F-XI activation in the presence of dextran sulfate or activated platelets, indicating that it cannot bind to the same site in FXI as thrombin (14). To understand the physiological importance of the FXI-prothrombin interaction, we have attempted to identify the amino acid sequences in FXI that interact with prothrombin.
Our experiments support the conclusion that a sequence of amino acids (Ala 45 -Ser 86 ) in the A1 domain of FXI that contains three antiparallel ␤-strands connected by ␤-turns comprises a surface that interacts with the kringle II domain of prothrombin. The evidence supporting this conclusion is as follows. 1) The rA1 domain inhibits the binding of prothrombin to FXI with an IC 50 of 4 ϫ 10 Ϫ7 M (Fig. 1A, Table I). 2) According to surface plasmon resonance studies, prothrombin interacts with FXI (K d(app) ϭ 71.5 Ϯ 14 nM) and with the rA1 domain of FXI (K d(app) ϭ 239 Ϯ 83 nM) ( Fig. 2 and Table II). 3) PF1.2 inhibits the binding of prothrombin to FXI (with a IC 50 of 5 ϫ 10 -7 M), whereas PF1 has no effect (Fig. 1B, Table I). 4) The rA1 domain (Glu 1 -Ser 90 ) binds prothrombin in a saturable manner with a K d(app) value of 471 Ϯ 175 nM (Fig. 3). 5) PF2 or PF1.2 binds FXI in a saturable manner with K d(app) values of 417 Ϯ 190 and 449 Ϯ 172 nM, respectively (Fig. 6, A and B). 6) Based on a molecular model of the Al domain, conformationally constrained peptides were synthesized, which act synergistically to inhibit rA1 binding to prothrombin (Fig. 4, A and B) or PF2 binding to FXI (Fig. 6 (A and B), Table I). We conclude from these studies that the binding of prothrombin to FXI is mediated by amino acid residues (Ala 45 -Ser 86 ) exposed on the surface of the A1 domain (Glu 1 -Ser 90 ) that interact with complementary residues within the kringle II domain of prothrombin.
Our subsequent studies were focused on the identification and mapping of the site within FXI that mediates the binding of prothrombin employing a variety of solid phase binding assays and a kinetic method (utilizing surface plasmon resonance) in which prothrombin was bound to a carboxymethyl dextran (CM5) flow cell using amine coupling chemistry, for examining FXI binding to prothrombin. Although these two methods gave similar estimates for the affinity of FXI binding to prothrombin (K d(app) ϳ250 nM utilizing the microtiter plate assay (Table I); K d(app) ϳ70 nM with the SPR method (Table II)), the latter kinetic method gives a more reliable estimate of affinity since the microtiter plate assay is likely to represent non-equilibrium binding measurements of a semiquantitative nature. Nonetheless, the microtiter plate assays yielded internally consistent results (comparing assays with either FXI or prothrombin immobilized and prothrombin, PF2, FXI, or rA1 domain as ligand) and provided a useful method for characterizing the prothrombin-kringle II domain binding site in the FXI A1 domain. Moreover, the K d(app) (250 Ϯ 50 nM) in the equilibrium binding experiments were almost identical to the K d (app) in the surface plasmon resonance experiments in which prothrombin was bound to the rA1 (Glu 1 -Ser 90 ) (239 Ϯ 83 nM). Therefore, we conclude that prothrombin binds FXI entirely through the A1 domain. Furthermore, we conclude that FXI binds in a saturable and reversible manner to immobilized prothrombin and that the fidelity of FXI as a probe for quantitative assessment of this interaction is not affected by the radiolabeling procedure.
In order to identify specific amino acid residues within the A1 domain of FXI involved in binding either HK or thrombin, conformationally constrained peptides and rA1 domains were synthesized containing conservative amino acid substitutions suspected on the basis of molecular modeling of the A1 domain to contain side chains involved in these interactions (15,29). Since the prothrombin binding site was localized to the A1 domain and found to contain amino acid sequences overlapping the thrombin and HK binding sites, we aimed to determine whether specific amino acid residues in the A1 domain involved in binding HK or thrombin might also bind prothrombin or kringle II. Our results are consistent with the following conclusions. 1) Val 64 and Ile 77 , which are important as contact sites for HK (but not for thrombin), do not participate in the interaction of FXI with prothrombin or kringle II (Figs. 4 and 6,   Table I). 2) Glu 66 and Asp 51 which are important contact sites for the interaction of the A1 domain with thrombin (but not for HK), are not involved in the interaction of the A1 domain with prothrombin or kringle II (Fig. 5A, Table I).
We have utilized four different, complementary assays to investigate prothrombin interaction with FXI (Table I): 1) pro-thrombin binding to FXI, 2) PF1.2 binding to FXI, 3) rA1 binding to prothrombin, and 4) PF2 binding to FXI. We have previously reported that prothrombin binds FXI in a saturable, specific manner with a K d(app) ϳ250 nM (14) and now demonstrate that PF1.2 binds FXI in a saturable manner with a K d(app) 449 Ϯ 172 nM (Fig. 6B). In order to examine direct interactions between the kringle II domain of prothrombin and FXI A1 domain, we devised two additional assays: PF2 binding to FXI and rA1 binding to prothrombin; both binding assays reveal essentially identical K d(app) values of 417 Ϯ 190 nM and 471 Ϯ 175 nM (Figs. 3 and 6A). The close correspondence of estimated affinities using these four different binding assays provides a basis for confidence in the validity of the results and the assumption that the conformation of FXI or prothrombin was not altered significantly when the proteins were bound to the wells of microtiter plates.
The plasma protein, PK, which shares a high degree of sequence identity (58%) with FXI, was also examined to determine whether homologous amino acid sequences can inhibit rA1 domain binding to prothrombin. Unlike the FXI peptides, the PK peptides, Pro 45 -Lys 54 and Phe 56 -Gly 86 , did not inhibit rA1 binding to prothrombin (Fig. 4A, Table I). Therefore, PK, which binds HK in the Phe 56 -Gly 86 region (31) but does not bind thrombin (Pro 45 -Lys 54 ) (15), also does not compete with FXI for binding sites on prothrombin.
In order to investigate the three stem-loop structures defined by the molecular model, the three peptides had cysteine residues introduced at the amino terminus and the carboxyl terminus of each peptide so that the resulting disulfide bond might stabilize the looplike structure (15,18,29). The individual loop-structures were tested in assays of rA1 binding to prothrombin to assess the contributions of three conformationally constrained loop structures, Ala 45 -Arg 54 , Asn 72 -Ser 86 , and Phe 56 -Val 71 , in constituting the prothrombin (kringle II) binding site in the A1 domain. The results (Fig. 4B, Table I) suggest that the three peptide loops interacting together might comprise a binding site for prothrombin. This conclusion is strongly supported by the observation that the peptide loop structures in equimolar combination (Ala 45 -Arg 54 plus Phe 56 -Ser 86 ; Phe 56 -Val 71 plus Asn 72 -Ser 86 ; Ala 45 -Arg 54 plus Phe 56 -Val 71 ; Ala 45 -Arg 54 plus Asn 72 -Ser 86 ; and Ala 45 -Arg 54 plus Phe 56 -Val 71 plus Asn 72 -Ser 86 ) inhibited the binding of the rA1 do-main to prothrombin with significantly greater potency than each of the stem loop peptides individually indicating synergistic interactions between the three conformationally constrained loop structures in comprising a binding site within the FXI A1 domain for the kringle II domain of prothrombin.
Since prothrombin (and PF1.2 but not PF1) binds to FXI through the A1 domain and displaces HK, it obviates the inhibitory effect of HK on thrombin-catalyzed F-XI activation on the platelet surfaces (14). We were therefore led to postulate that complex formation between FXI and prothrombin promotes the binding of FXI to the platelet surface (14). Our studies demonstrate that prothrombin (in the presence of Ca 2ϩ ions) can substitute for HK (and Zn 2ϩ ) as a cofactor for FXI binding to activated platelets (14). Therefore, it was important to examine the interaction of FXI and prothrombin. Our results suggest that prothrombin contains two separate and distinct domains, one in the kringle II domain of prothrombin, the other in the catalytic domain of thrombin, both of which bind to the A1 domain of FXI thereby displacing HK without displacing each other (14). We have mapped the thrombin binding site to spanning residues Ala 45 -Arg 70 (15) and the HK binding site to residues Phe 56 -Ser 86 (8,29), whereas the prothrombin (kringle II domain) binding site was found in the present study to reside within amino acid residues Ala 45 -Ser 86 . Thus, the data in this paper and previously published studies support the conclusion that prothrombin (kringle II), thrombin, and HK binding sites are all contained within the A1 of FXI and, although contiguous, are separate and distinct. However, these three binding sites overlap since they share a common sequence of amino acids. Our results are consistent with the conclusion that prothrombin binding to the A1 domain facilitates FXI binding to the activated platelet surface, thereby favoring thrombin-mediated activation of FXI (14). FXI binding to prothrombin is also predicted to block HK binding and to prevent FXIIa-catalyzed activation of FXI. Evidence to confirm this prediction has recently been reported (32). We have suggested that the interaction of FXI with prothrombin results in a conformational transition in FXI resulting in the exposure of a binding site within the A3 domain (Asn 235 -Arg 266 ) of FXI that is important in mediating a direct, high affinity interaction with the platelet surface (14). Thereby, platelet-bound FXI can be activated efficiently by thrombin (14). The presence of prothrombin at physiological concentration would favor this mechanism supporting the hypothesis that thrombin can activate the intrinsic pathway in a revised model of platelet-dependent blood coagulation (8,9,14).