Characterization of the H-kininogen-binding Site on Factor XI

Factor XI (FXI), the zymogen of the blood coagulation protease FXIa, and the structurally homologous protein plasma prekallikrein circulate in plasma in noncovalent complexes with H-kininogen (HK). HK binds to the heavy chains of FXI and of prekallikrein. Each chain contains four apple domains (F1–F4 for FXI and P1–P4 for prekallikrein). Previous studies indicated that the HK-binding site on FXI is located in F1, whereas the major HK-binding site on prekallikrein is in P2. To determine the contribution of each FXI apple domain to HK-FXI complex formation, we examined binding of recombinant single apple domain-tissue plasminogen activator fusion proteins to HK. The order of affinity from highest to lowest is F2 ≫ F4 > F1 ≫ F3. Monoclonal antibodies against F2 are superior to F4 or F1 antibodies as inhibitors of HK binding to FXI. Antibody αP2, raised against prekallikrein, cross-reacts with FXI F2 and inhibits FXI-HK binding with an IC50 of 8 nm. HK binding to a platelet-specific FXI variant lacking the N-terminal half of F2 is reduced > 5-fold compared with full-length FXI. A chimeric FXI molecule in which F2 is replaced by P2 is cleaved within P2 during activation by factor XIIa, resulting in greatly reduced HK binding capacity. In contrast, wild-type FXI is not cleaved within F2, and its binding capacity for HK is unaffected by factor XIIa. Our data show that HK binding to FXI involves multiple apple domains, with F2 being most important. The findings demonstrate a similarity in mechanism for FXI and prekallikrein binding to HK.

Coagulation factor XI (FXI) 1 is the zymogen of a plasma serine protease (FXIa) that contributes to blood coagulation by proteolytically activating factor IX (1)(2)(3)(4). Deficiency of FXI results in a mild to moderate bleeding disorder, whereas elevated FXI levels have been associated with increased risk for deep venous thrombosis (5,6). Human FXI is a 160-kDa dimeric glycoprotein consisting of two identical 80-kDa polypeptides connected by a single disulfide bond. The N-terminal "heavy chain" portion of each polypeptide is comprised of four 90 -91-amino acid repeats called apple domains, whereas the C-terminal "light chain" region is a typical trypsin-like serine protease domain (7,8). FXI circulates in plasma as a noncovalent complex with H-kininogen (HK), the high molecular mass precursor of kinin hormones (7,9). Plasma prekallikrein (PPK), another protease zymogen that circulates as a complex with HK (11), has a similar structural organization to the 80-kDa FXI polypeptide, with which it shares 58% amino acid identity (12). The interaction of HK with FXI or PPK facilitates binding of the latter two proteins to surfaces such as cell membranes (13). HK docks to cell surface heparan and chondroitin sulfatetype proteoglycans, indirectly anchoring FXI and PPK to the cells (14,15). In the case of PPK, this may be a critical step in localizing kinin production to the surface of vascular endothelial cells (14,16). Along similar lines, it has been shown that HK facilitates the binding of FXI to activated platelets (17,18). Once bound to the platelet surface, FXI is efficiently activated to FXIa by coagulation proteases such as thrombin and factor XIIa (FXIIa) (4,19,20).
The FXI-binding site of HK has been mapped to a contiguous sequence of 56 amino acids in the extreme C-terminal domain D6 H (21), to which FXI binds with high affinity (apparent K D ϭ 1.8 ϫ 10 Ϫ8 M). The FXI-binding site on HK overlaps with the PPK-binding site, and the two zymogens compete for binding to HK (21,22). Therefore, it appears that there are similarities in the mechanism by which HK binds to FXI and to PPK. A series of studies using recombinant whole molecules and individual apple domains, as well as monoclonal antibodies, have determined that PPK binding to HK requires the first, second, and fourth apple domains (P1, P2, an P4, respectively) but not the third apple domain (P3) (23-25). The P2 domain appears to be most important for this process (25). In contrast, studies of FXI binding to HK using conformationally constrained peptides suggest that HK is bound through a relatively small area between amino acids Phe 56 and Ser 86 within the first apple domain (F1). The second, third, and fourth FXI apple domains (F2, F3, and F4, respectively) appear to contribute little to the interaction with HK (26,27). The significant differences in the interactions of FXI and PPK with HK suggested by published studies are surprising considering the high degree of structural similarity between the proteins. To address this issue, we have conducted studies of FXI binding to HK using recombinant FXI apple domains, recombinant FXI/PPK chimeras, and monoclonal antibodies directed against specific FXI apple domains. The results indicate that the binding interaction between FXI and HK is similar to the interaction between PPK and HK, involving multiple apple domains, with the second apple domain being most important. We tested these findings with additional experiments on a recently identified FXI splice variant lacking the N-terminal half of F2 (28) that is present in platelets (29). This protein has greatly reduced HK binding compared with full-length FXI with potential (patho)physiological consequences.

EXPERIMENTAL PROCEDURES
Nomenclature-For clarity the following nomenclature was followed throughout the manuscript. FXI apple domains are designated by a capital F followed by the number of the domain, whereas PPK apple domains are designated by capital P followed by the domain number. Murine monoclonal antibodies are indicated by the symbol ␣ followed by F for antibodies raised against FXI and P for antibodies raised against PPK and then the number of the apple domain recognized by the antibody. For example, ␣P2 is an antibody raised against PPK that recognizes the P2 domain. The names of some antibodies used in previous studies were changed to comply with the new nomenclature system. Specifically, anti-PPK antibodies ␣P1, ␣P2, and ␣P4 were formerly designated PKH19, PHK6, and PKH1, respectively (25), and anti-FXI antibodies ␣F1, ␣F2, and ␣F4 were previously named XI-5, XI-3, and XI-1, respectively (30). For chimeric FXI/PPK proteins, the molecule that makes up the majority of the chimera is listed first, while the abbreviation for the apple domain substituted into the protein is listed second. For example, FXI in which the F2 domain has been replaced with the P2 domain from PPK is designated FXI/P2.

SDS-PAGE and Western
Blotting-Proteins were resolved by electrophoresis on 12.5% polyacrylamide gels containing 0.1% (w/v) SDS at 30 mA for 90 min. The proteins were visualized either by silver staining or staining with Coomassie Brilliant Blue. Alternatively, proteins were transferred to nitrocellulose at 100 mA for 30 min using a semi-dry technique for Western blotting. The membranes were blocked with PBS containing 5% (w/v) dry milk powder and 0.05% (v/v) Tween 20. Typically the primary detection antibody was diluted 1:1,000 in PBS/milk, and bound antibody was detected by a horseradish peroxidase-coupled secondary antibody against mouse immunoglobulin (DAKO, Hamburg, Germany), followed by chemiluminescence detection according to the manufacturer's instructions (Amersham Biosciences, Inc.).
Expression of FXI Apple Domain-tPA Fusion Proteins-Complementary DNAs for the four apple domains of FXI, F1 (Glu 1 -Ser 90 ), F2 (Ala 91 -Leu 180 ), F3 (Ala 181 -Val 271 ), and F4 (Phe 272 -Glu 361 ) were amplified by PCR with Taq polymerase using primers that introduced a BglII site and a XhoI site at the 5Ј-and 3Ј-ends of the PCR products, respec-tively (33). The primers also introduced an additional six amino acids (Pro-Arg-Ile-Lys-Gly-Gly) between the apple domains and the tPA molecule (33) to allow for maximum flexibility. The cDNAs encoding single apple domains were excised by BglII and XhoI digestion and ligated into the corresponding sites in the modified tPA expression vector ZpL7(S478A) as described previously (33,34). The sequences of the PCR products were verified by dideoxy chain termination sequencing. The constructs encode fusion proteins that contain the signal peptide sequence of tPA, followed by a single FXI apple domain (F1, F2, F3, or F4), a spacer sequence of 6 amino acids, the kringle 1 and kringle 2 domains of tPA, and the tPA catalytic chain containing an alanine substitution for the active site serine (S478A). Expression plasmids were stably transfected into baby hamster kidney cells grown in the presence of 1 M methotrexate, and fusion proteins were purified from serum-free medium (Opti-MEM; Invitrogen) by affinity chromatography using a monoclonal antibody to tPA as described previously (24). Protein concentrations were determined with an ELISA system for tPA (33).
Expression of Full-length FXI, PPK, and FXI/PPK Chimeric Constructs-Two different methods were used to prepare recombinant proteins. The preparation of mammalian tissue culture expression constructs in vector pJVCMV for wild-type FXI, wild-type PPK, and chimera FXI/P2 has been described previously (35). The proteins were expressed in stably transfected HEK293 cells in serum-free medium (Cellgro Complete; Mediatech, Herndon, VA), and purified on an antihuman FXI monoclonal antibody affinity column (35). Alternatively, chimeric constructs were prepared using overlap extension with PCR in vector pZEM (36). The sequence of all constructs was verified by dideoxy sequencing. Chimeric cDNAs were excised from pZEM with EcoRI and cloned into the EcoRI site of mammalian expression vector pcDNA3(ϩ) (Invitrogen). Constructs were transiently transfected into HEK293 cells using LipofectAMINE (Invitrogen) according to the manufacturer's instructions. Transfection efficiency was Ն40% as monitored by co-transfection with a vector encoding green fluorescent protein. Recombinant proteins were collected in serum-free medium (Dulbecco's modified Eagle's medium).
Expression and Purification of FXI Truncation Protein FXI/⌬F2N-The cDNA for the FXI splice variant FXI/⌬F2N (28) was a generous gift from Dr. Peter Walsh (Temple University School of Medicine, Philadelphia, PA). The cDNA was subcloned into the EcoRI site of the pcDNA3(ϩ) vector, and the HEK293 cells were transiently transfected using LipofectAMINE. The transfected cells were resuspended in 2ϫ PBS supplemented with 10 g/ml each of soybean trypsin inhibitor, benzamidine, leupeptin (Sigma), and 0.1 mM Pefabloc SC (Roth, Karlsruhe, Germany), placed on ice, and lysed by repetitive application of weak ultrasonic pulses for 3 min. Following centrifugation at 20,000 ϫ g for 10 min at 4°C, the cell lysates were applied to an affinity chromatography column containing polyclonal anti-FXI antibodies (affinity-purified from AS 199 antiserum) immobilized on AffiGel-10 matrix (Bio-Rad). After extensive washing with PBS, bound FXI/⌬F2N was eluted with 1 M NaSCN in PBS, dialyzed against PBS followed by gel filtration high pressure liquid chromatography on a Sephadex 200 column (Amersham Biosciences, Inc.) equilibrated in PBS. Proteincontaining fractions were concentrated using an Amicon concentrator, and protein concentration was determined by ELISA (see below).
Determination of Protein Concentration-Protein concentrations were determined by ELISA. Polyclonal anti-FXI antiserum (AS199) at a 1:1,000 dilution was used to coat the wells of microtiter plates (Nunc, Wiesbaden, Germany). After blocking with 1% BSA in PBS, culture supernatants or purified proteins were added in serial 1:2 dilutions. Bound protein was detected using monoclonal antibodies directed against the PPK or FXI light chain (25,32). Purified FXI and PPK or supernatants from mock-transfected cells supplemented with known concentrations of FXI or PPK were used as standards. Protein concentration was determined by the Bradford assay (Bio-Rad) or by biospecific interaction analysis using surface plasmon resonance spectroscopy (BIAcore, Freiburg, Germany). CM5 sensor chips were coated with ␣-FXI (AS199) or ␣-PPK (AS176) antibodies using the amine coupling kit provided by the manufacturer. Serial 1:2 dilutions of the supernatants were applied at a continuous flow rate of 20 l/min, and association was followed for 90 s. Dissociation of the complex induced by applying PBS alone was monitored for 3 min. For calibration, supernatants from cells transfected with control vector supplemented with purified FXI or PPK were used. The chip was reconstituted by briefly washing with 30 mM HCl. The relative concentrations of proteins were determined from the measured response units with the BIAevaluation 2.1 program (BIAcore). The concentrations calculated by ELISA or biospecific interaction analysis differed by Ͻ10%.
Immunoprecipitation of Recombinant Proteins-HEK293 cells were transiently transfected with pcDNA3 vectors containing wild-type FXI, FXI/P2, or FXI/⌬F2N using the LipofectAMINE method. After 60 h the cells were washed with Cys/Met-free Dulbecco's modified Eagle's medium and incubated with the same medium for 45 min at 37°C. The cells were labeled with 100 Ci/ml [ 35 S]Cys/Met (Trans-label, ICN, Eschwege, Germany) for 12 h at 37°C. The supernatants were collected, centrifuged at 2,500 ϫ g for 10 min, and used directly for immunoprecipitation. Adherent cells were washed four times with PBS and lysed in 150 mM NaCl, 50 mM Tris-HCl, pH 8.0, 0.1% SDS, 0.5% deoxycholic acid, 1% Nonidet P-40, 10 g/ml phenylmethylsulfonyl fluoride, 10 g/ml benzamidine-HCl (RIPA buffer) under rotation for 1 h at 4°C. The cell lysates were centrifuged at 14,000 ϫ g for 15 min at 4°C to separate insoluble components. The cell lysates and supernatants were incubated for 60 min at 4°C with 20 l of polyclonal anti-FXI antibodies (AS199) bound to staphylococcus A (Pansorbin; Calbiochem, La Jolla, CA) cells. The mixtures were pelleted at 8,000 ϫ g for 2 min at 4°C and washed four times with RIPA buffer. Immunoprecipitates were dissolved in SDS sample buffer and size-fractionated by SDS-PAGE. The gels were fixed with sodium acetate, incubated with 15% (m/v) sodium salicylate for 30 min at room temperature, dried for 2 h at 55°C, and exposed to Fuji x-ray film for 6 or 24 h at Ϫ80°C (14,25). Alternatively, immunoprecipitated proteins from unlabeled cells were transferred to nitrocellulose and analyzed by Western blot analysis using monoclonal antibody ␣P2 or ␣F4. HK Binding Assays-Maxisorp TM microtiter plates (used throughout; from Nunc) were coated overnight with 25 nM (3 g/ml) HK in 100 mM NaCl, pH 5.5 ("coating solution"). Coated plates were washed six times with PBS and blocked with 1% (w/v) BSA in PBS and incubated with serial 1:2 dilutions of individual FXI or PPK apple domains fused to tPA starting at 500 nM (25 g/ml PBS/BSA). After washing six times with PBS, bound apple domains were probed by 13.3 nM (2 g/ml) of ␣-tPA (from rabbit) specific for the tPA portion shared by all constructs, followed by a horseradish peroxidase-coupled secondary antibody to rabbit immunoglobulin F t fragment and the substrate solution, i.e.
After 30 min the absorbance at 405 nm was monitored by an ELISA plate reader (Dynatech, Deppendorf, Germany). Alternatively, a sandwich ELISA was employed. Microtiter plates were coated with 20 nM (3 g/ml) ␣-tPA in coating buffer, washed, blocked with 1% BSA, and incubated with serial 1:2 dilutions of apple-tPA fusion proteins (starting concentration, 50 nM in PBS/BSA). Plates were washed, and 8.3 nM (1 g/ml) biot-HK in PBS/BSA was applied. Bound biot-HK was probed by streptavidin-peroxidase (1 g/ml; Roche Molecular Biochemicals) and the substrate solution, as above.
For competitive ELISAs, the microtiter plates were coated with 45.4 nM (8 g/ml) of FXI. Serial 1:2 dilutions of monoclonal antibodies (starting concentration, 1.2 M ϭ 180 g/ml), FXI apple-tPA constructs (starting concentration, 4 M ϭ 240 g/ml), or full-length recombinant proteins FXI or PPK (starting concentration, 2 M) were prepared and made 8.3 nM in biot-HK (final concentration, 1 g/ml), and the resultant mixtures were applied to the coated plates. Bound biot-HK was probed by streptavidin-peroxidase (1 g/ml) and the substrate solution. To follow binding of HK to immobilized target proteins, 5 nM of full-length FXI and PPK, FXI/PPK chimeras, or proteolytic cleavage products thereof were coated on microtiter plates, followed by incubation with serial 1:2 dilutions (starting at 2 M ϭ 240 g/ml) of biot-HK in PBS/ BSA. Bound biot-HK was measured as above.
To determine apparent dissociation constants (K D ) and maximum binding (B max ) for HK binding to the individual constructs, nonspecific binding of biot-HK to recombinant full-length constructs was determined in the presence of 100-fold molar excess of unlabeled HK. Binding data were calculated using the Prism 2.0a software (GraphPad Software, San Diego, CA). Relative B max values for the recombinant constructs PPK, PPK/F2, FXI/P2, and FXI/⌬2N differed by Ͻ15% from the B max value of HK binding to FXI, which was arbitrary set to 100%.
To measure binding of FXI constructs to immobilized HK, 4 nM (4.8 g/ml) HK was coated on microtiter plates, and serial 1:2 dilutions of FXI or FXI/⌬2N (starting concentration 300 nM) were applied, followed by 6.7 nM (1 g/ml) of ␣-FXI (AS 199), an horseradish peroxidasecoupled secondary antibody to rabbit F c , and the substrate solution. If not otherwise stated, incubation steps were at 37°C for 45 min, except for the coating step, which was done at 4°C overnight.

RESULTS
Binding of HK to Single FXI Apple Domains-Initially, we investigated the capacity of HK to bind to individual FXI apple domains. FXI apple domains were expressed as fusion proteins with tPA in baby hamster kidney cells and purified using antibodies against the tPA portion of the fusion protein (Fig.  1A). The assay measures binding of the fusion proteins to HK immobilized on a microtiter plate. A rank order of F2 Ͼ Ͼ F4 Ͼ F1 Ͼ Ͼ F3 was found for the binding of individual apple domains to HK (Fig. 1B). The binding of F3 to HK was not significantly different from that of the control (tPA). An identical rank order was found when fusion proteins were bound to microtiter plates via antibodies to their common tPA portion and probed by increasing concentrations of biot-HK (data not shown). Interestingly, these results are very similar to previously published data examining binding of HK to individual PPK apple domains, using identical techniques: P2 Ͼ Ͼ P4 Ͼ P1 Ͼ Ͼ P3 Ϸ control (25). These results indicate that the F2 domain is crucial for FXI binding to HK and emphasize a similarity in the interactions between HK and FXI and between HK and PPK.
Anti-FXI and PPK Monoclonal Antibodies Interfere with FXI-HK Complex Formation-Previously, we developed a panel of monoclonal antibodies against epitopes on the FXI heavy chain (30). Using the recombinant apple domain-tPA fusion proteins, we identified several antibodies from this panel that interact with individual apple domains on Western blots ( Fig. 2A). Because of the structural similarity of PPK to FXI, we tested the monoclonal anti-FXI antibodies and a similar panel of monoclonal anti-PPK antibodies for cross-reactivity using full-length FXI and PPK. The anti-PPK antibody ␣P2 (25, 32) readily detected FXI, whereas all other antibodies tested selectively detected their target protein only (Fig. 2B). ␣P2 cross-reacted selectively with recombinant F2 domain (Fig.  2C), suggesting that the epitope for ␣P2 is conserved in the apple 2 domains of FXI and PPK. To investigate the role of individual apple domains in HK binding in the context of a complete FXI molecule, we tested the capacities of monoclonal anti-FXI and anti-PPK antibodies and individual apple domains to interfere with FXI binding to HK in a competitive ELISA. ␣P2 specifically blocked HK binding to FXI with an apparent IC 50 of 8 nM (Fig. 3A). This anti-PPK antibody was even more effective than its anti-FXI counterpart ␣F2 (apparent IC 50 ϭ 35 nM). Antibodies ␣F4 and ␣F1 inhibited binding with IC 50 values of 1 M and 2 M, respectively, whereas ␣P4 and ␣P1 failed to interfere with HK-FXI complex formation (IC 50 Ͼ 5 M). We obtained similar results using individual apple domains of FXI and PPK fused to tPA. Apple domains F2 and P2 most efficiently inhibited HK-FXI complex formation with apparent IC 50 values of 35 and 50 nM, respectively, whereas domains F4 and F1 (IC 50 ϭ 1.3 and 4 M, respectively) and P4 and P1 (IC 50 Ͼ 10 M) were much less effective in blocking biot-HK binding to FXI (Fig. 3B). These results again point to the F2 domain as playing a key role in formation of the FXI-HK complex formation. Furthermore the cross-reactivity studies highlight the similarity in the ␣P2 epitopes of FXI and PPK and establish ␣P2 as a generic probe for studying HK binding to apple 2 domains.
Cleavage within the Apple 2 Domain Impairs HK Binding-Activation of PPK by FXIIa initially produces a two-chain protease called ␣-kallikrein (␣-PKa) that binds to HK and antibody ␣P2 in a manner similar to zymogen PPK (25). With continued incubation, a second proteolytic cleavage occurs in the P2 domain creating ␤-kallikrein (␤-PKa) (31). We have previously demonstrated that this cleavage in P2 greatly reduces binding of both HK and ␣P2 (25). To further characterize the importance of the apple 2 domain for HK binding to FXI and PPK, we constructed chimeric molecules containing homologous exchanges of the apple 2 domain between PPK and FXI. HEK293 cells were stably transfected with expression constructs for FXI/P2, PPK/F2, wild-type FXI, or PPK (schematics shown in Fig. 4), and recombinant proteins were affinity-purified from culture supernatants. The proteins were activated by incubation with FXIIa, and the reaction mixtures were ana-lyzed under reducing conditions by SDS-PAGE (Fig. 5A). To determine the identity of the bands, identical gels underwent Western blot analysis using domain specific antibodies. Zymogen PPK (migrating at 88 kDa) was cleaved into the three fragments of ␤-kallikrein: the light chain (37 kDa), fragment N (25 kDa), and fragment C (31 kDa). In contrast, activated FXI (FXIa) has two fragments: the light chain migrating at 33 kDa and the heavy chain (49 kDa). FXI/P2 activation results in a  three-chain form (light chain, 33 kDa; fragment N, 28 kDa; and fragment C, 31 kDa), demonstrating that introduction of the P2 domain into FXI also introduces the ␤-kallikrein cleavage site. PPK/F2, in contrast, is activated to a two-chain form (light and heavy chains; data not shown) consistent with a loss of the ␤-kallikrein cleavage site.
Next we investigated the consequence of cleavage within the apple 2 domain for binding of ␣P2, the monoclonal antibody most effective in blocking HK binding to FXI (Fig. 5B). ␣P2 detected the zymogen forms of PPK, FXI, and FXI/P2 (lanes 1,  3, and 5 of Fig. 5B) and the heavy chain of FXIa (lane 4 of Fig.  5B). However, ␣P2 failed to detect any fragment of the cleaved heavy chains of PPK or FXI/P2 (lanes 2 and 6 of Fig. 5B). This clearly indicates that the epitope for ␣P2 is critically dependent on the conformation of the apple 2 domain and that proper conformation is lost upon proteolytic cleavage of the domain. Based on the premise that the epitope of ␣P2 overlaps with the HK-binding site, we tested HK binding to the zymogen and activated proteins in a direct binding assay (Fig. 5C). Biotinylated HK bound to the zymogens of recombinant PPK (apparent K D ϭ 10 Ϯ 3 nM) and the FXI/P2 chimera (apparent K D ϭ 11 Ϯ 3 nM) with almost identical affinities. Cleavage of these zymogens within the apple 2 domain decreases the apparent HK binding affinity more than 25-fold to K D ϭ 270 Ϯ 37 nM for cleaved PPK and K D ϭ 284 Ϯ 41 nM for cleaved FXI/P2. In contrast the binding of HK to FXI was not affected by FXI activation (K D ϭ 19 Ϯ 5 nM for FXI and K D ϭ 20 Ϯ 4 nM for FXIa), indicating that HK binding is not changed by a cleavage separating the light and heavy chain portions of FXI. Together these data indicate that loss of HK binding during activation is caused by a cleavage in the apple 2 domain of PPK and FXI/P2 and support the hypothesis that the integrity of the apple 2 domain is important for HK binding to both FXI and PPK.
Characterization of FXI Lacking the N-terminal Portion of F2 (FXI/⌬F2N)-Recently a splice variant of FXI mRNA lacking exon 5 was identified in a megakaryocytic leukemia cell line (28). It has been proposed that this message, coding for a factor XI protein lacking the N-terminal half of the F2 domain (amino acids Ala 91 -Arg 144 ) represents a nonsecreted platelet specific form of factor XI (FXI/⌬F2N; Fig. 4) (29). FXI/⌬F2N offers an opportunity to test the hypothesis that an intact F2 domain is critical for HK binding in a system of possible physiologic relevance. HEK293 cells were transiently transfected with cDNAs coding for FXI/⌬F2N, wild-type FXI, or FXI/P2. After 72 h of incubation in medium containing [ 35 S]Cys/Met, factor XI proteins were immunoprecipitated from culture supernatant and whole cell lysates using polyclonal anti-FXI antibody (AS199) and size fractionated on SDS-polyacrylamide gels. Bands for wild-type FXI and FXI/P2 migrating at 80 kDa were identified in precipitates of culture supernatants (Fig. 6A); however, FXI/⌬F2N was not detected in the supernatant. A band representing FXI/⌬F2N (ϳ73 kDa) was identified in immunoprecipitates from whole cells (Fig. 6A). A faint band representing intracellular wild-type FXI can also be seen (Fig. 6A). Similar results were obtained when COS-7 cells were transfected with the cDNAs (data not shown). On Western blots of nonlabeled immunoprecipitates, antibody ␣P2 detected FXI and FXI/P2 but failed to recognize FXI/⌬F2N (Fig. 6B). However, cellular FXI/⌬F2N was detected by ␣F4, an antibody that also recognizes FXI and FXI/P2 (Fig. 6B). These results indicate that the deletion in F2, while destroying the epitope recognized by ␣P2, does not alter the conformation of the F4 domain sufficiently to prevent ␣F4 binding.
FXI/⌬F2N Binding to HK-Because failure to bind the ␣P2 antibody is associated with poor HK binding to other recombinant proteins, we examined HK binding to FXI/⌬F2N. HEK293 cells were transiently transfected with expression vectors for wild-type FXI or PPK, PPK/F2, FXI/P2, FXI/⌬F2N, or mock vector (control), and proteins were affinity-purified using columns with immobilized polyclonal anti-PPK or anti-FXI antibodies. Protein concentrations were determined by ELISA and by biospecific plasmon-resonance spectroscopy (data not shown). We employed Western blots using antibodies ␣P2 and ␣F4 to demonstrate expression of proteins of proper molecular size (Fig. 7A). Direct binding of HK binding to the various proteins was examined in a microtiter plate assay (Fig. 7B). HK bound to immobilized PPK and the FXI construct where the F2 domain had been replaced by P2, with highest affinities (ap- FIG. 5. Cleavage in the apple 2 domain impairs ␣P2 and HK binding. Noncleaved (Ϫ) zymogen forms of PPK, FXI, and FXI/P2 were activated with FXIIa for 72 h at 37°C. Zymogens (Ϫ) and activated products (ϩ) were subjected to electrophoresis on a 10% polyacrylamide-SDS gel under reducing conditions. A, Coomassie Brilliant Bluestained gel. The positions of the molecular mass standards in kDa are indicated along the left edge of the figure. B, Western blot of a gel identical to the one in A with 1 pmol of protein/lane transferred to nitrocellulose using the anti-PPK antibody ␣P2 (1 g/ml) as primary antibody. C, HK binding assay. Microtiter plates were coated with 5 nM (0.9 g/ml) of zymogen (Ϫ) or activated (ϩ) recombinant proteins. Following incubation with biot-HK (serial 1:2 serial dilution; starting concentration, 2 M), bound HK was detected with the biotin-avidinperoxidase method. A schematic diagram of the assay is shown at the top of C; the symbols are identical to those in Fig. 3. A representative result of a series of five independent experiments is shown. parent K D ϭ 10 Ϯ 3 and 11 Ϯ 3 nM, respectively). Replacement of P2 of PPK by F2 of FXI lowered the affinity for the PPK/F2 construct (K D ϭ 18 Ϯ 5 nM) almost to that of native FXI (K D ϭ 19 Ϯ 5 nM). Deletion of the N-terminal portion of F2 (FXI/ ⌬F2N) further reduced HK binding affinity to an apparent K D of 107 Ϯ 19 nM. However, FXI/⌬F2N still bound HK significantly over background (Fig. 7B), indicating that other portions of the FXI molecule contribute to HK binding, although to a minor extent. This conclusion was confirmed by an assay system using a "reverse" set-up, where microtiter plate-bound HK was probed by FXI and FXI/⌬F2N (Fig. 7B, inset).
To rule out the possibility that immobilization of the various constructs to microtiter plates may induce subtle conformational changes and thus contribute to differential binding affinities, we employed a competitive ELISA. Constructs such as PPK and FXI/P2 holding apple domain P2 competed with biot-HK binding to immobilized FXI with apparent IC 50 values of 9 and 10 nM, respectively. Replacement of P2 by F2 in constructs PPK/F2 and FXI lowered the apparent IC 50 values to 27 and 29 nM, respectively, whereas truncation of F2 in FXI/⌬F2N increases the apparent IC 50 value to 312 nM (Fig.  7C). Taken together, these results underline the importance of the apple 2 domain in HK binding to FXI and PPK. Furthermore, the data indicate that FXI/⌬F2N binds poorly to HK. DISCUSSION HK circulates in plasma in binary complexes with FXI or PPK (9, 11). The protease zymogens remain bound to HK upon attachment of the kinin precursor to its acceptor structures on cell membranes such as heparan and chondroitin sulfate proteoglycans (13,14). In this way HK works as an adapter that links the proenzymes indirectly to the cell. For PPK the physiological consequence of the association between the proenzyme and prohormone (HK) on the cell surface is obvious. In the course of contact phase reactions, PPK is converted to the active enzyme ␣-kallikrein (␣-PKa) that proteolytically cleaves HK, liberating the vasoactive peptide hormone bradykinin (38,39). Because bradykinin has a half-life of Ͻ15 s in plasma, the generation of the hormone from its precursor in proximity to its cellular receptors is a prerequisite for an effective hormone response (14,16,40). In contrast the physiological consequences of the interaction between FXI and HK are less clear. HK is required for proper FXI activation during contact activation-initiated coagulation in vitro. However, although congenital FXI deficiency is associated with a bleeding disorder, deficiency of HK is not (39,41,42). Several recent studies have shed light on this conundrum. It has been shown that HK facilitates FXI binding to the surface of activated platelets, where it is rapidly activated to FXIa (18,20,43). Prothrombin may serve as a substitute for HK in this role, providing a plausible explanation as to why HK-deficient humans do not bleed (19).
Despite their different biological functions, PPK and FXI share 58% amino acid identity and identical domain organizations (7,12,46). This homology is reflected in the organization of the PPK and FXI genes, both containing 15 exons and identical intron-exon boundaries (47,48). In fact, the two genes are separated by Ͻ10 kilobases on the distal end of the long arm of chromosome 4, indicating they are the products of a duplication event involving a common ancestral gene. 2 Despite the many similarities between FXI and PPK, previously published data suggest that the mechanisms by which these proteins interact with HK are quite different. A series of experiments using conformationally constrained peptides representing various portions of the FXI apple domains have assigned binding sites for several macromolecules on FXI (49 -51). Sequence segments important for the interaction between HK and FXI have been localized to amino acids Phe 56 -Ser 86 of the F1 domain, based on the capacity of peptides from this area to competitively inhibit HK binding to FXI. Peptides representing F2, F3, and F4 were poor inhibitors of HK binding (26,27). In contrast, an approach using recombinant individual apple domains, monoclonal antibody interference, PPK deletion mutants, and PPK/ FXI chimeras indicates that the HK-binding site on PPK is discontinuous, involving three apple domains (P1, P2, and P4) with P2 being most important (25). Using a similar strategy to analyze the HK interaction with FXI we have found that the apple 2 domain of FXI is most important for HK binding, with minor contributions coming from the F1 and F4 domains. The results are strikingly similar to our previous results with HK binding to PPK and suggest that the interactions between HK and FXI and between HK and PPK are similar.
The reasons for the disparity between our results and those of the previously published peptide inhibition studies are not clear. Both approaches determined the F1 domain to be important for HK binding. It is possible that the peptides for F2 and F4 used in the previous study (26) were not in the proper conformation to inhibit FXI-HK complex formation. Indeed, early studies examining HK binding to PPK, using peptides based on the PPK sequence, identified P1 and P4 as important for HK binding but failed to identify P2 as a critical element of the binding site (23, 52). This may be indicative of a general problem with using peptides to model complex protein structures. Each apple domain in FXI and PPK contains three or four disulfide bonds, and the resulting structures may be difficult to reproduce with a short linear peptide. Supporting this notion, mapping studies of the factor IX-binding site on FXI using peptides localized the binding site to F2 (53), whereas studies using FXI/PPK chimeras and alanine scanning mutagenesis make a strong case for F3 (35). Our approach using fusion proteins of tPA and individual apple domains may have some limitations as well. For example, the bulky tPA portion of ϳ50 kDa could sterically hinder the access of HK to binding site(s) exposed by the smaller apple domain of ϳ10 kDa. To minimize this possibility, we have inserted spacer sequences of 6 amino acids, each separating the apple domain from the flanking regions of tPA. Although we cannot entirely exclude an underestimation of the HK binding capacity of individual apple domains by the strategy employing fusion proteins, our alternative approaches using FXI deletion mutants or domaindirected antibodies are consistent with the conclusions drawn from individual apple domain experiments.
Factor XIIa cleaves PPK to form the active two-chain enzyme, ␣-PKa. Prolonged incubation with FXIIa results in a second cleavage in the P2 domain, producing the three-chain form ␤-PKa (31). Although the enzymatic activities of ␤-PKa and ␣-PKa are similar (31), the HK binding capacity of ␤-PKa is decreased Ͼ20-fold compared with ␣-PKa or PPK (24,25). Because kallikrein docking to cell surfaces is mediated through HK, the dramatic loss in affinity for HK will promote dissociation of ␤-PKa from cells. It is possible, therefore, that FXIIa proteolysis of PPK serves two purposes: the initiation and the termination of kinin release on endothelial cells. The present study clearly demonstrates that human FXI is not susceptible to secondary cleavage in its heavy chain by FXIIa. Thus the FXIa-HK complex may persist on cell surfaces such as platelet membranes for extended periods of time. The differences in the sequences of P2 and F2 may well reflect different requirements for continuous cell-associated activity between ␣-PKa and FXIa, respectively.
Recently, a splice variant of FXI mRNA lacking the sequence encoded by exon 5 of the FXI gene (coding for amino acids Ala 91 -Arg 144 in F2) has been described (28). Flow cytometry studies indicate that this variant is expressed on platelet membranes independent of plasma FXI expression and thus may compensate partially for plasma FXI deficiency (29). Although the (patho)physiological relevance of this variant is debated (10), the protein offers an opportunity to test our model for HK binding with a molecule of possible physiologic importance. Initial studies predicted that FXI/⌬F2N is a tetramer comprised of identical 50 -55-kDa subunits that migrates at 220 kDa (37,44). In contrast, the protein we precipitated from transfected cells has an apparent molecular mass of 73 kDa. This is consistent with the loss of the 53 amino acids encoded by exon 5 from the 80-kDa FXI polypeptide. We note that FXI/⌬F2N is expressed poorly compared with FXI but was recoverable using affinity purification with polyclonal anti-FXI antibodies. Furthermore, a monoclonal antibody (␣F4) raised against full-length FXI recognizes FXI/⌬F2N on Western blot. This indicates that FXI/⌬F2N may share a similar conformation with FXI outside of the F2 domain. Most importantly for the present study, FXI/⌬F2N bound poorly to HK compared with FXI, supporting an important role for F2 in HK binding.
We note limitations in our experiment analyzing antibody interference with FXI-HK complex formation. We did not have monoclonal antibodies directed against F3 (or P3) to make a complete study. However, there are several lines of evidence to indicate that F3, as in the case of P3, is not involved in HK binding. We demonstrated that the recombinant F3 domain bound HK poorly. Furthermore, a previous study involving saturation mutagenesis of the F3 domain failed to identify abnormalities in HK binding using surface plasmon resonance techniques (18). These findings, in conjunction with the observation that P3 does not contribute to HK binding to PPK (25), support the premise that F3 is not involved in HK binding to FXI. Because F3 appears to be important for binding to other components of the coagulation mechanism, such as factor IX (35,45), platelet membranes (51), and heparin (52), it makes sense that it would not be involved in the interaction with HK (43). Taken together, our data are consistent with a model of a discontinuous HK-binding site jointly formed by F2, F1, and F4 of FXI, where domain F2 harbors the core binding sequence and F1 and F4 contribute directly and/or indirectly to this high affinity binding site. The generation of constructs comprising doublets or triplets of distinct apple domains and of recombinant proteins holding duplicated or multiple individual apple domains should allow us to address the issue of cooperativity among the apple domains in more detail.