Identification and Characterization of a Binding Site for Factor XIIa in the Apple 4 Domain of Coagulation Factor XI*

E, K, K,S/K,,, where E, equals the total enzyme concentration and S equals the substrate concentration. Thus, the S, K, (the inhibitor constant) can be determined. In the case of noncompetitive inhibition, Is0 is independent of the substrate concentration and related to K, as follows: IsO E, K,. K, be evaluated determining enzyme


M.
These results, interpreted in the context of the model, suggest that the sequence of amino acids from Ala317 through Gly350 of the heavy chain of the A4 domain of factor XI contains three peptide structures, possibly consisting of three antiparallel &strands that together comprise a contact surface for interacting with factor XIIa.
Human plasma factor XI is a 160,000-dalton homodimer consisting of two identical disulfide-linked polypeptide chains each of which can be cleaved a t a single peptide bond (Arg369-Ile370) by factor XIIa to give rise to factor XIa (1,2). This *This study was supported by Research Grants HL46213, HL45486, and HL25661 from the National Institutes of Health and from the W. W. Smith Charitable Trust. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
YTo whom reprint requests should be addressed: Sol Sherry Thrombosis Research Center, Temple University School of Medicine, 3400 N. Broad St., Philadelphia, PA 19140. Tel.: 215-221-4375;Fax: 215-221-2783. plasma glycoprotein participates in the contact phase of blood coagulation and circulates in plasma as a complex with a nonenzymatic cofactor, high M , kininogen (3, 4). The importance of factor XI in normal hemostasis is supported by the fact that its deficiency in plasma can give rise to excessive bleeding (5, 6). We and others have been interested in the complex intermolecular interactions involved in the activation of factor XI (7-14) and in the expression of the enzymatic activity of factor [15][16][17]. It is well known that the activation of factor XI by factor XIIa occurs on negatively charged surfaces to which factor XI is bound by high Mr kininogen (4, 9). From the sequence of a cDNA insert coding for factor XI, the primary structure has been elucidated (2). Four repeat sequences (Al, A2, A3, and A4 domains) or Apple domains are present in the heavy chain(s), and these four tandem repeat sequences apparently comprise four separate domains, each encoded by two exons (18). Studies in our laboratory have characterized a binding site for high M , kininogen in the first (Al) of the four Apple domains (13, 14), whereas a substrate-binding site for factor IX was localized to the A2 domain (17). The present study was undertaken to determine whether domains in the heavy chain region of factor XI are important for activation by factor XIIa. Evidence is presented to support the conclusion that the A4 domain of factor XI is capable of interacting with factor XIIa.

MATERIALS AND METHODS
Purification of Coagulation Proteins-Factor XI, purified from human plasma by immunoaffinity chromatography using a monoclonal antibody to factor XI (lo), had a specific activity of 250 units/ Peptide Synthesis-Peptides were synthesized on an Applied Biosystems 430A peptide synthesizer by a modification of the procedure described by  as previously described (21). The sequences of the synthetic peptides utilized in this study are given in Table I. Refolding and Reduction and Alkylation of Peptides-In order to refold peptides containing cysteine residues, the peptide was dissolved in deionized water as a 0.1 mg/ml solution in a flask containing a stir bar. The pH was adjusted to 8.5 with NH,OH, and the solution was allowed to stir at 5 "C for at least 3 days. The resulting solution was lyophilized. Alternatively, peptides were reduced with dithiothreitol and alkylated with iodoacetamide as previously described (10).
High Performance Liquid Chromatography (HPLC)-The HPLC system employed was from Waters (Waters 600 Gradient Module, model 740 Data Module, model 46K Universal Injector, and Lambda-Max model 481 Detector). The reverse-phase chromatography described here was performed using a Waters C8 pBondapak Column, whereas gel filtration was carried out using a Waters Protein-Pak 60 Column as previously described (14,17,21).
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 refolded peptides were examined by HPLC (both reverse phase and gel filtration), single homogeneous peaks with identical retention times to the original mixtures were observed. This demonstrates the presence of a single homogeneous mixture of refolded peptides and not a mixed population of diverse polymers. The results were the same after reduction and alkylation of these same peptides, i.e. single peaks with identical retention times to the original mixtures were observed by both reverse-phase and gel filtration HPLC. In addition, all reduced and alkylated or refolded peptides were examined for free -SH groups using the Ellman reagent, 5,5'dithiobis(2-nitrobenzoic acid). It was determined (22) that there was less than 0.02 mol of free -SH/mole of peptide, which further verifies that these refolded peptides were homogeneous preparations consisting of intramolecular disulfide-bonded peptide.
Fast Atom Bombardment (FAB) Mass Spectrometry-FAB mass spectrometry was utilized to assess homogeneity and molecular mass of each synthesized peptide. FAB was performed as previously described (17,21). The spectra of synthesized peptides revealed masses almost identical to calculated values (i.e. within 0.1%) with a single ion species detected in each case (data not shown).
Computer Modeling-A structural model of the A4 domain was constructed using the computational chemistry package supplied by Molecular Simulations, Inc., Pasadena, CA and a Silicon Graphics 4D 280 Parallel Processing Supercomputer. A description of the modeling package and methods has been previously published (21, 23). Information concerning cysteine disulfide constraints (24) was used to initiate model building after which extended energy minimization calculations were carried out according to previously published methods (14,17,21). Subdomains were selected for synthesis directly from our molecular model. Conformational restraints were built into selected peptides using artificially introduced cysteine pairs. Coagulation Assays-Factor XI(a) was assayed by minor modifications (25) of the kaolin-activated partial thromboplastin time (26) using factor XI congenitally deficient substrate plasma, results of which were quantitated on double logarithmic plots of clotting times versu.s concentrations of pooled normal plasma.
Preparation and Characterization of Factor XZa-Purified factor XI was activated by incubation at 37 'C with human factor XIIa as previously described (10) to use as a reference standard in assays of factor XIa generation. Gel electrophoresis in the presence of Na-DodSO4 of this preparation under reducing conditions showed two major bands of M, 50,000 and 30,000.
Amidolytic Assays-Factor XIIa was assayed using the chromogenic substrate s-2302 according to the procedure previously described (27). The amidolytic assay of factor XIa was carried out by a modification of the method of Scott et al. (25). Factor XIa generation in the experiments reported here was quantitated from a standard curve prepared using purified factor Xla. Thrombin, trypsin (obtained from Sigma), and activated protein C at a concentration of 2 nM were assayed using 1 mM of the chromogenic substrate S-2238 (H-D-Phepip-Arg-pNA, Kabi). Plasma kallikrein (2 nM) was assayed using 1 mM of the chromogenic substrate S-2302 (H-D-Pro-Phe-Arg pNA, Kabi).
Assay of Factor XI Activation-The rate of factor XI activation by factor XIIa was determined as previously reported (11). Human factor XI (0.17 p M ) was incubated with factor XIIa (0.017 p~) for 15 min at 37 "C, and the rate of factor XIa generation was measured using the substrate S-2366 at 405 nm.
Effect of Peptides on the Rate of Activation of Factor X I by Factor Xlla-The assay procedure was the same as described above except factor XIIa was incubated for 3 min at 37 'C with either peptide or buffer solution before the addition of other components of the assay mixture.
Kinetics of Activation of Factor X I by Factor XIZa-The assay was the same as described above except that the initial rates of activation were determined over a wide range of substrate concentrations. For determination of the effects of peptides on the kinetics of the reaction, the mixture was incubated with various concentrations of peptide for 3 min at 37 "C before the other components of the assay were added. The initial rates of release of factor XI activation were linear and were determined under conditions where less than 20% of the substrate had been consumed. Values for the Michaelis constant (K,) and the maximum velocity ( VmaX) were obtained by the Lineweaver-Burk method (28) and were calculated using least-squares fit as previously described (16).
Analysis of Kinetic Data for the Quantitation of the Inhibitor Constants-The 150 method of c h a (29) was used to determine the inhibitor constants as previously described (16). In the case of classical competitive inhibition, I,, (total inhibitor concentration at which the enzyme reaction velocity is 50% of the uninhibited reaction) is related to the substrate concentration as follows: I,,, = 1/2 E, + K, + K,S/K,,, where E, equals the total enzyme concentration and S equals the substrate concentration. Thus, from the plot of 150 versus S, K, (the inhibitor constant) can be determined. In the case of noncompetitive inhibition, Is0 is independent of the substrate concentration and related to K, as follows: IsO = 1/2 E, + K,. K, can be evaluated by determining I,, at several enzyme concentrations.

Effects of Heavy Chain-derived Peptides on the Activation of Factor XI by Factor XIIa-Our previous
studies of the interactions of factor XI (XIa) with high M , kininogen (13, 14) and factor IX (17) have delineated binding sites for these two proteins within the carboxyl-terminal half of the first two Apple domains of factor XI. Preliminary computer modeling studies with all four Apple domains suggested that a similar pattern of folding might be present in each domain, Al-A4, each characterized by three stem-loop structures that might interact with one another to form a solvent-accessible contact surface (17). Such a surface, comprising residues ValSg-Lysa3, is apparently the high M , kininogen-binding site in the A1 domain of factor XI (10, 11), whereas a similar surface, comprising residues Ala134-Leu1T2, forms a substrate (factor IX) binding site in the A2 domain of factor XIa (12). Because of the structural similarities predicted from computer modeling and the 23-34% sequence identity among these four tandem repeat domains (a), we prepared a panel of heavy chain-derived peptides similar to those representing factor IX and high M , kininogen-binding sites ( Table I) to determine their effects on factor XI activation by factor XIIa. Peptide G l~~'~-L y s~'~ from the A4 domain inhibited factor XI activation by factor XIIa 50% at a concentration of 5 PM (Fig. 1). In contrast, the concentration of the peptide derived from the A1 domain (Phes6-SerR6) required to produce 50% inhibition of factor XI activation was 7.5 mM (>l,OOO-fold higher concentration). Moreover, peptides from the A2 domain Ala17') and the A3 ( A~n '~~-A r g '~~) were not effective even at a 10,000-fold higher concentration.
Effect of A4 Peptide (Gly326-Lys3B7) on Factor XI Coagulant Activity-In order to determine if the A4-derived peptide Gly326-Lys357 inhibits factor XI coagulant activity, coagulant assays were carried out (see "Materials and Methods") in the presence of various concentrations of peptide. The concentration of Gly'v2fi-Lys3s7 required to inhibit factor XI coagulant activity 50% was 7.5 X M, i.e. almost the same as the concentration required for 50% inhibition of factor XI activation by factor XIIa (data not shown).
Kinetics of Factor XI Activation by Factor XIIa in the Presence of A4 Peptide Gly326-Lys3"7-The Lineweaver-Burk plots of the activation of factor XI by factor XIIa in the presence of various concentrations of the heavy chain synthetic peptide Gly""-LyP7 are shown in Fig. 2 A . The double-

F T C V L K D S V T E T L P R V -N R T A A I S G Y S F K Q C S A1 Phe56-SerRG (C)* N I C L L K H T Q T G T P T R I T K L D K V V S G F S L K S C
The numbers indicate the numbering of residues in each Apple domain as reported by Fujikawa et al. (2). Gaps were inserted in the A1 and A4 domains a t residue 72 and 55, respectively, to maintain maximal alignment of residues.
Designates a peptide in which cysteine replaced one or more amino acid(s) in the normal factor XI sequence. Designates a residue where cysteine was substituted for an amino acid in the native factor XI sequence. Designates a residue where alanine was substituted for a cysteine in the native factor XI sequence. A~n '~~-A l a '~~ (0), and Asn235-Arg266 (A). Factor XIIa (0.017 p~) was incubated for 3 min at 37 "C either with the designated peptide at the concentration shown or with buffer solution. Then human factor XI (0.17 PM) was added, the mixture was incubated a t 37 "C for 15 min, and the rate of factor XIa formation was determined as described under "Materials and Methods." reciprocal plots yielded patterns consistent with classical noncompetitive inhibition. It is clear from the plots that the A4 peptide reduces the V,,, and increases the slopes of the double-reciprocal plot (i.e. has no effect on Kmapp). This indicates that the catalytic site is blocked thereby reducing the effective enzyme concentration.
The Effect of A4 Peptide on the Kinetics of the Amidolytic Activity of Factor XIIa Using the Substrate S-2302-Because the catalytic site of factor XIIa is apparently blocked by the peptide Gly"2G-Lys""7 we asked whether the amidolytic activity might also be affected. The Lineweaver-Burk plots of the amidolytic activity of factor XIIa using the substrate S-2302 in the presence of various concentrations of heavy chain peptide Gly32G-Lys"57 are shown in Fig. 2B. The double-reciprocal plots yielded patterns consistent with classical competitive inhibition; that is, the V,,, was unaffected by the peptide while progressively higher concentrations of peptide yielded commensurately increased slopes (i.e. higher values of Kmapp).
Thus, the binding of the peptide to factor XIIa reduces the effective concentration of binding sites within the enzyme for its substrate S-2302.
In order to determine whether the A4 peptide (Gly326-Lys:357) has inhibitory activity against serine proteases other than factor XIIa, the peptide was incubated with trypsin, kallikrein, a-thrombin, and activated protein C. Chromogenic assays of enzymatic activity demonstrated no significant inhibition of cr-thrombin, trypsin, kallikrein, or activated protein C using up to 1 mM A4 peptide (Gly3"-Lys"s7) (data not shown). Thus, this provides evidence of a high degree of specificity of the A4 peptide (Gly"6-Lys357) as an inhibitor of factor XIIa.
Calculation of Binding Constants from Kinetic Parameters-The inhibitor constant Ki (the dissociation constant of the enzyme-inhibitor complex) was determined from the doublereciprocal plots as described by Cha (29). In the case of noncompetitive inhibition, I,, is independent of substrate concentration and related to enzyme concentration according to the equation under "Materials and Methods." The inset of Fig. 2A shows that the IhO, which is independent of substrate concentration, is 3.75 pM. In the case of competitive inhibition, IsO is related to the substrate concentration according to the equation under "Materials and Methods." It can be seen from this equation that a plot of 1," versus substrate concentration should be a straight line, from which Ki can be determined. The inset of Fig. 2B represents such a plot with the factor XI heavy chain peptide (Gly"6-Lyss57) used as the inhibitor. The calculated value of Ki from the slope is 3.8 p~.
Computer Modeling of the A4 Peptide Gly"G-Lys3'"-In order to gain more information about the possible structure of the factor XIIa-binding site on factor XI, we used computer modeling of the A4 domain to generate a testable threedimensional structure (Fig. 3), using the primary amino acid sequence and the disulfide linkages (24) within the A4 domain (Phe271-Lys357). This model may or may not accurately reflect the solution structure of the factor XI A4 domain. However, it does predict a possible contact surface for interaction with factor XIIa. It is important to emphasize that the model is not presented to depict the structure of the A4 domain of factor XI, but rather as a hypothetical paradigm to be used to design functionally active synthetic peptides. The calculated structure shows three antiparallel P-strands connected by pturns defined by amino acid residues Ala317-Gly"G, Lys3"-

Effects of Stem-loop Synthetic Peptides on the Activation of Factor X I by Factor
XIIa-In order to determine whether the three stem-loops assume a conformation that comprises a contact site that interacts with factor XIIa, we prepared synthetic peptides, rationally designed on the basis of the model (Fig. 3). In some of the peptides, cysteines were introduced so that the resulting disulfide bond might stabilize this loop structure in a conformation likely to resemble the native Lys340, and Gly344-GlY:jSO.

FIG. 2. Double-reciprocal plots of the activation of factor XI by factor XIIa (panel A ) and the amidolytic activity of factor XIIa (panel B ) in the presence of various concentrations of A 4 domain peptide Gly326-Lys3s7. Panel A , factor XIIa
(0.017 p~) was incubated with various concentrations of the peptide for 3 min at 37 "C. Then human factor XI was added at the concentrations indicated on the abscissa (0.8-2.5 phi) and the rate of factor XIa formation was determined as described under "Materials and Methods." The results shown are double-reciprocal plots in the absence (0) and in the presence of peptide Gly326-Lys357 at the following concentrations: 2 (A), 4 (O), 8 (V), and 10 (0) p~. The inset is the secondary plot of I,, peptide concentration uersw substrate concentration as described by Cha (27). Panel B , factor XIIa (0.017 p M ) was incubated with various concentrations of the peptide for 3 min at 37 "C. Then S-2302 was added at the concentrations (40-320 phi) shown on the abseissa, and the amidolytic assay was carried out according to Scott et al. binding surface. These peptides were examined for their capacity to inhibit the activation of factor XI by factor XIIa after refolding or reduction and alkylation (see "Materials and Methods"). The peptide designated Ala"17-Gly"26, in which cysteines were substituted, inhibited 50% the activation of factor XI by factor XIIa at a concentration of 8 X M when subjected to a folding procedure (see "Materials and Methods") and lost significant inhibitor activity when reduced and alkylated ( Fig. 4 and Table 11), suggesting that the correct conformation of this peptide is required for optimal binding. This supports the suggestion that this first loop constitutes a part of the surface of the factor XIIa-binding site.
Since synthetic peptide Gly326-Lys:i57 inhibits the activation of factor XI by factor XIIa (Fig. 1) and since the computer model (Fig. 3) suggests the possibility that the stem-loop structures Ly~""'-Lys'~" and Gly"44-Gly:'so form a portion of the factor XIIa-binding site, we synthesized conformationally unconstrained Ly~:~"'-Lys:'~" and a conformationally constrained Gly'344-"y''50 (C) ( Table I). The folded conformation of Gly:'44-Glyn5" (C) was more than 100-fold more effective as an inhibitor of factor XI activation by factor XIIa than the reduced and alkylated peptide ( Fig. 4 and Table 11). However, the unconstrained peptide Lys""'-Lys"(' representing the second stem loop did not inhibit the activation of factor XI by factor XIIa when used at concentrations up to M. Because our computer model (Fig. 3) predicted that the distance between a-carbons of amino acids within the antiparallel @ ; strands (stem portion of stem-loop 2 ) was greater than 10 A we were reluctant to constrain this region by introduction of cysteines, since the distance between g-carbons of disulfidebonded cysteines is approximately 6 A. Thus, the failure of the unconstrained Lys""-Lys""" peptide to inhibit factor XIIamediated factor XI activation could be attributed to the failure of the peptide to assume the correct conformation. This possibility was borne out by additional experiments described below. Another peptide representing a sequence of amino acids from the amino-terminal half of the A4 domain (Asp278-Gly288) (C), not implicated by the computer model to contribute to the putative binding surface, failed to affect factor XIIa-catalyzed factor XI activation at concentrations as high as 1 mM (Table 11).
Effects of Peptide Combinations on the Activation of Factor XZ by Factor XZIa-The results presented in Fig. 4 together with the model (Fig. 3) suggest that Ala:'1'-Gly:'2fi and G~Y : '~~-Gly:i"l form stem-loop structures that contribute to the formation of a binding site for factor XIIa. It was, therefore, important to examine the inhibitory effects of these peptides when added in combination a t equimolar concentrations. Fig.   4B indicates that using equimolar mixtures of folded Ala'"'-Gly"26 (C) plus folded Gly:'44-Gly"50 (C) the inhibitory effects observed were strictly additive, compared with the effects of either peptide examined alone. Similarly when all three peptides, i.e. folded Ala""-Gly"' (C) plus folded Gly'44"Gly~'"1 (C) plus Lys':11-LyS:'40, were added at equimolar concentrations the inhibitory effects were similar to those obtained with the longer peptide encompassing two of the putative stem loops (Gly"26-Lys"'). Therefore, an additional experiment was performed to determine whether Ly~"~-Lys"~' is active in inhibiting the activation of factor XI by factor XIIa when folded 5 shows that when the concentration of LyP'-Lys'"' is increased from lo-' M to lo-'' M while the combined concentration of folded Ala:"7-Gly:32ci (C) and folded Gl~:'~~-Gly:'"" (C) is increased from 0 to M, evidence of inhibitory effects of Ly~""'-Lys:'~" is disclosed. Thus, it is apparent that concentrations of M to lo-:' M Ly~:'"'-Lys"~", which by themselves have no inhibitory effects, enhance the inhibitory effects of Ala""-Gly"'6 (C) plus Gly344-Gly350 (C), e.g. 10% inhibition at lo-' M combined peptides becomes 60% inhibition and 50% inhibition at M combined peptides becomes 100% in the presence of Ly~:':''-Lys~'~~~, which by itself has no inhibitory effect. Therefore, it would appear that the three peptides interact in a complex way either with each other or with factor XIIa or both to potentiate the inhibition of factor XIIacatalyzed factor XI activation. As a final test of this hypoth-~1~:317-~1~326 ((7) and folded (C) are present.

I I B 348
Factor X I Activation by Factor X I I a u 6 3 u )

FIG. 3. Computer model of the A4 domain of the heavy chain of factor XI.
A plausible three-dimensional structure was calculated using the primary structure of the A4 domain (2), known disulfide linkages (24), Biograf software (Molecular Simulations, Inc.), and a Silicon Graphics 4D 280 Parallel Processing Computer (see "Materials and Methods"). A stereographic pair of a-carbon tracings of the protein backbone is shown with the positions of amino acids numbered according to their sequence in the mature protein.
esis, a peptide comprising all three putative antiparallel @strands, i e . Ala"7-Gly"') (C), was examined for its capacity to inhibit factor XIIa-catalyzed factor XI activation (Fig. 4B). The concentration of this peptide required for 50% inhibition was 5 X M, compared with 5 X lo-' M for the shorter peptide G l~~'~-L y s "~~ (C) (see Table 11), indicating that the sequence Ala"7-Gly326 does in fact contribute to the binding energy. Data not shown indicated that upon reduction and alkylation the Ala"7-Gly"" peptide lost about one order of magnitude activity indicating that its conformation affects its binding activity.

DISCUSSION
The participation of factor XI in the contact phase of blood coagulation involves intermolecular interactions with factor XIIa, high M , kininogen, and factor IX. Our most recent experiments have resulted in the identification and characterization of conformationally specific solvent-accessible contact surfaces within the first two Apple domains of the factor XI heavy chain that form binding sites for high M , kininogen and factor IX (13,14,17). Previous studies in our laboratory using monoclonal antibodies have revealed that domains in both the heavy chain and light chain are essential for the efficient activation of factor XI by factor XIIa (13). In order to gain information about the mechanisms of activation of factor XI by factor XIIa and to identify and characterize the domains and amino acid sequences within factor XI that interact with factor XIIa, the present study was undertaken.
The results of our experiments support the hypothesis that a sequence of amino acids (Ala317-Gly3s") in the A4 domain of the heavy chain may contain three antiparallel @-strands connected by @-turns, which comprise a continuous surface that interacts with factor XIIa. The evidence supporting this possibility is listed as follows: 1) an A4-derived peptide (Gly32G-Lys"s7) inhibits factor XI activation by factor XIIa noncompetitively with a K: of 3.75 FM (Figs. 1 and 2). This same peptide is a competitive inhibitor ( K , = 3.8 p~) of factor XIIa amidolytic activity (Fig. 2) suggesting that it binds near the active site of factor XIIa. 2) A model of the A4 domain derived from energy minimization computations (Fig. 3) predicts the presence of @-stranded loop structures (Ala"7-Gly326, Lys"'-Ly~~~", and GlyB44-Gly3s") that fold together to form a solvent-accessible surface that might comprise a binding site for factor XIIa. 3) Based on this model, two conformationally constrained peptides and one peptide without conformational constraints were synthesized. Both of the constrained peptides (Ala317-Gly326 and Gly344-Gly352) inhibited the activation of factor XI by factor XIIa when refolded and lost inhibitory activity when reduced and alkylated (Fig. 4A). 4) The unconstrained peptide Lys""'-Ly~"~ (inactive by itself) acted to potentiate the inhibitory effects of the constrained peptides (Ala"17-Gly326 and Gly"44-Gly352) on the activation of factor XI by factor XIIa whereas when the two constrained peptides were added at equimolar concentrations simple additive inhibitory effects were demonstrated (Figs. 4B and 5 ) . 5 ) A conformationally constricted peptide (Ala317-Gly"") containing all three putative @-stranded loop structures was the most effective inhibitor of factor XI activation by factor XIIa (Fig.  4B and Table 11). These experiments demonstrate that the optimal function of the factor XIIa site requires the presence and the proper conformation of all three putative @-stranded loop structures.
In our approach to this problem, in the absence of any crystallographic or solution structural information about the Apple domains of factor XI, we have used computer modeling to predict the secondary and tertiary structure of the A4 domain of factor XI that appears to form a contact surface with factor XIIa. The molecular dynamic, energy minimization calculations suggested that this domain may have a "N/E, no effect of factor XI activation by factor XIIa (highest concentration tested given in parentheses).
R + A designates a reduced and alkylated peptide.
structural motif consisting of three antiparallel @-strands connected by @-turns. This construct was used to design peptides restricted with disulfide bonds, and we have shown that their activity ( i e . their ability to inhibit factor XI activation by factor XIIa) is dependent on their conformation (refolded uersus reduced and alkylated peptides). Thus, the model has been used as a tool to generate hypotheses, about the possible secondary and tertiary structure of the A4 domain, that were tested in functional studies of factor XI activation. Although our results are consistent with the pos- sibility that the predicted structural motif has some validity, the model is not presented as evidence of the three-dimensional structure of the A4 domain, which can be obtained only using physical techniques such as x-ray crystallography or nuclear magnetic resonance. However, it will be interesting to determine whether the model makes correct structural predictions.
We have investigated refolded peptides and reduced and alkylated preparations and have found they have identical retention times (by gel filtration and reverse-phase HPLC) thus suggesting they are homogeneous species and that polymer formation does not account for their inhibitory effects on factor XI activation. Also, all refolded peptides and all reduced and alkylated peptides contained no free thiols, thus, free -SH groups were confirmed to be either oxidized to disulfides or reduced and alkylated. Thus, from the refolding experiments we conclude that protein conformation is an important determinant of the structure of the surface that binds factor XIIa.
We have carried out similar experiments that identified binding sites for high M , kininogen (13, 14) within domain A1 and for factor IX (17) in domain A2. The binding site for factor IX in the A2 domain of the heavy chain also contains three putative antiparallel @-strands connected by @-turns which comprise a continuous surface. The A2 peptide comprising the sequence of amino acids, Ala1:'4-Le~'i2, was found to be a competitive inhibitor of factor IX activation by factor XIa and was therefore inferred to represent a substratebinding site for factor IX. In contrast, the A4 peptide Gly"jG-Lys'"' is a noncompetitive inhibitor of factor XI activation by factor XIIa (Fig. 2 A ) . Furthermore, this same peptide is a competitive inhibitor of factor XIIa amidolytic activity suggesting that it binds near the active site of factor XIIa (Fig.  2B). The demonstration that the A4 peptide Gly:'26-Lys35i has no inhibitory effects on the amidolytic activities of a variety of serine proteases other than factor XIIa (including thrombin, activated protein C, trypsin, and kallikrein) indicates that it binds specifically to factor XIIa and has no nonspecific inhibitory effects. Thus, it is clear from our experiments that the A4 domain of factor XI contains amino acid sequences that are important in binding factor XIIa and play a role in the activation of factor XI by factor XIIa.
,41a:317-Gly:%0 A recent report by Meijers et ul. (30) provides definitive evidence that the A4 domain of factor XI mediates dimer formation. Previously it was known that factor XI is present in plasma as a homodimer (1, 2) and is secreted as a dimer linked by a disulfide bridge between C Y S~~' of each A4 domain (24). Interestingly, whereas recombinant factor XI-Cys 321 Ser migrated as a monomer by SDS-PAGE, gel filtration studies indicated that the protein exists as a dimer under native conditions (30). Furthermore, when the A4 domain was inserted into tissue plasminogen activat,or, dimer formation under native conditions was also observed, whereas neither the A3 domain of factor XI nor the A4 domain of prekallikrein promoted dimer formation. Our present studies indicate that in addition to mediating dimer formation the A4 domain also participates in factor XI activation by factor XIIa.
The mechanisms by which the A4 domain peptides inhibit the enzymatic activities of factor XIIa are somewhat conjectural. However, several inferences concerning the mechanisms of factor XIIa-catalyzed factor XI activation can be drawn from our studies within the context of the general theory of enzyme activation and kinetics. Enzyme-catalyzed reactions are characterized by the phenomenon of saturation of reaction rates as a function of substrate concentration, a feature not generally observed in nonenzymatic reactions. From this saturation effect it has been deduced that the enzyme first forms a reversible complex with the substrate through a substratebinding site on the enzyme. Only when the substrate is anchored to this binding site can the active site efficiently catalyze the conversion of the substrate to product, which subsequently dissociates from the enzyme. Since the study of enzyme inhibitors can be a useful tool in elucidating mechanisms of catalysis, we prepared a variety of factor XI-derived synthetic peptides. The A4 peptide Gly"'-Ly~~~~ proved to be a competitive inhibitor of factor XIIa-catalyzed amidolysis of the chromogenic substrate, S-2302, which indicates that the peptide binds to a site in factor XIIa utilized for binding the tripeptide substrate and positioning it for efficient cleavage to generate p-nitroanilide and the tripeptide o-Pro-Phe-Arg. It can, therefore, be inferred that a small peptide substratebinding exosite exists near the catalytic site of factor XIIa that is recognized by the A4 peptide, G l~~*~-L y s~~~, which binds to it with a Ki of 3.8 PM. The fact that the same A4 peptide, Gly326-Lys35T, inhibits factor XI activation by factor XIIa with a nearly identical Ki (3.75 PM) implies that the peptide exerts its inhibitory effects through the same binding event. However, the kinetic effect is different since the A4 peptide is a noncompetitive inhibitor of macromolecular substrate (factor XI) hydrolysis but a competitive inhibitor of small peptide (S2302) hydrolysis. Therefore, the putative small peptide substrate-binding exosite does not appear to function as a macromolecular substrate-binding site, which may very well exist elsewhere in factor XIIa, with a complementary factor XIIa-binding site in factor XI. The suggested mechanism by which the A4 peptide, Gly"'-Ly~~~~, inhibits factor XIIa-catalyzed factor XI activation is by binding to the putative small peptide substrat~-binding site causing steric or conformational changes in the molecule and thus preventing access of the active site of factor XIIa to the cleavage site in factor XI, i.e. Arg169-Ile37". Kinetically this binding event would manifest itself as a progressive reduction in active site concentration as a function of progressive increases in peptide concentration. Thus. the concentration of substrate (factor XI) required to saturate the enzyme (factor XIIa) would remain unaltered (ix. the Kmapp would be unaffected) whereas the V,,, would progressively decrease with increased inhibitor concentration, thereby disclosing noncompetitive inhibition. The remaining question relates to the normal function of the putative small peptide substrate-binding exosite presumably near the active site of factor XIIa, apparently recognized by the chromogenic substrate S-2302 and by the factor XIderived A4 peptides. The answer to this question is somewhat conjectural, but it seems reasonable to suggest that it represents a secondary substrate-binding site that serves to position the factor XI cleavage site (Arg369-Ile"o) near the active site of factor XIIa once the two proteins are already docked through binding of factor XI to a macromolecular substrate binding site in factor XIIa.
A e k n o~l e~~e n~-W e are grateful to Frances S. Seaman for technical assistance; to Drs. Charles A. Grubmeyer, Edward W. Kirby,