Mapping of a putative surface-binding site of human coagulation factor XII.

We have localized the binding epitope(s) of two murine monoclonal antibodies (B7C9 and P5-2-1) that were shown previously to inhibit the activation of human coagulation factor XII by negatively charged surfaces. A factor XII cDNA expression library in lambda gt11 was screened with antibody B7C9, and 16 immunoreactive bacteriophage were isolated. Fusion proteins from each of the recombinant phage were reactive with both monoclonal antibodies. Two of the phage cDNA inserts were found to code for amino acid residues -6-+31 and +1-+47 of factor XII, respectively, thereby defining the limits of the antigenic peptide to amino acids +1-+31. Each of the remaining 14 recombinant phage contained longer factor XII cDNA inserts that included sequences coding for the amino-terminal 31 amino acid residues. These results were confirmed by direct binding of antibody B7C9 to synthetic peptides containing amino acids 1-14 and 1-28 of factor XII. Further experiments with a set of nested peptides also indicated that amino acid residues 1-4 were essential but not sufficient for binding of B7C9 to the peptides. Hydrophobicity analysis of the amino-terminal region of plasma factor XII revealed a highly hydrophilic region between amino acid residues 5 and 15 that contained positively charged lysine residues at positions 8, 11, and 13. We conclude that a major epitope(s) recognized by monoclonal antibodies B7C9 and P5-2-1 is present in the amino-terminal 28 amino acids of factor XII. It is proposed that binding of these antibodies to factor XII blocks interaction of the positively charged region between residues 5 and 15 with negatively charged surfaces, thereby inhibiting activation.

We have localized the binding epitope(s) of two murine monoclonal antibodies (B7C9 and P5-2-1) that were shown previously to inhibit the activation of human coagulation factor XI1 by negatively charged surfaces. A factor XI1 cDNA expression library in Xgtll was screened with antibody B7C9, and 16 immunoreactive bacteriophage were isolated. Fusion proteins from each of the recombinant phage were reactive with both monoclonal antibodies. Two of the phage cDNA inserts were found to code for amino acid residues -6-+31 and +1-+47 of factor XII, respectively, thereby defining the limits of the antigenic peptide to amino acids +1-+31. Each of the remaining 14 recombinant phage contained longer factor XI1 cDNA inserts that included sequences coding for the amino-terminal 31 amino acid residues. These results were confirmed by direct binding of antibody B7C9 to synthetic peptides containing amino acids 1-14 and 1-28 of factor XII. Further experiments with a set of nested peptides also indicated that amino acid residues 1-4 were essential but not sufficient for binding of B7C9 to the peptides. Hydrophobicity analysis of the amino-terminal region of plasma factor XI1 revealed a highly hydrophilic region between amino acid residues 5 and 15 that contained positively charged lysine residues at positions 8, 11, and 13. We conclude that a major epitope(s) recognized by monoclonal antibodies B7C9 and P5-2-1 is present in the amino-terminal 28 amino acids of factor XII. It is proposed that binding of these antibodies to factor XI1 blocks interaction of the positively charged region between residues 5 and 15 with negatively charged surfaces, thereby inhibiting activation.
Human Hageman factor, or coagulation factor XII, circulates in plasma as a single glycopolypeptide chain of M, 80 (1). In uitro, the enzymatically inert zymogen form of factor XI1 is activated following contact with a negatively charged surface such as glass in the presence of the plasma constituents high molecular weight kininogen, kallikrein, and factor XI1 to form an active serine protease, thus initiating the intrinsic coagulation pathway and blood clot formation (2). Recently, the organization of the human factor XI1 gene has been determined (3), confirming the previously published cDNA sequence (4-6) and the amino acid sequence of plasma factor XI1 (7, 8). Several notable similarities to other mammalian proteins occur within the factor XI1 polypeptide. These include areas resembling both the type I and type I1 homologies found in fibronectin, two epidermal growth factorlike regions, a kringle homology found also in prothrombin, plasminogen, and the plasminogen activators, a unique proline-rich region, and the carboxyl-terminal catalytic portion of factor XI1 that is homologous to other serine proteases (4-8). Other than the catalytic region, the functional significance of these sequences has not yet been determined.
Despite our knowledge of the DNA and protein structure of factor XII, its physiological role remains unclear. Those rare individuals who are deficient in factor XI1 do not have prolonged bleeding times, thereby raising doubts as to the significance of factor XI1 in normal hemostasis (9). Alternative theories for the function of this abundant plasma protein include roles in inflammation and neutrophil activation (9, 10). Progress in these areas of factor XI1 biology has been hindered by the absence of a plausible physiological activator of factor XI1 (2). As an initial step towards the elucidation of this rate-limiting function, we sought to assign the surfacebinding site to one of the structural domains in factor XII. Recently, two murine monoclonal antibodies have been shown to inhibit the activation of factor XI1 zymogen by negatively charged surfaces (11, 12). We have used these antibodies as probes for the identification of the corresponding factor XI1 epitope(s) by expressing regions of the factor XI1 molecule as fusion proteins in Escherichia coli (13,14). The localization of the epitope was confirmed by using synthetic peptides. MATERIALS

RESULTS
Screening the Factor XII cDNA Library with the B7C9 Morwclonal Antibody-Approximately 1 X 10' bacteriophage from the factor XII 200-300-base pair Xgtll cDNA library were plated on E. coli lOgOr-, and fusion proteins were induced with isopropyl-1-thio-@-D-galactopyranoside (16). After transfer to nitrocellulose, the plaques were incubated with the monoclonal antibody B7C9 followed by incubation with alkaline phosphatase conjugated to goat anti-mouse immunoglobulin and a calorimetric substrate. 20 strongly positive and 26 weakly positive B7C9 immunoreactive plaques were identified on eight separate filters. Only the strongly positive plaques were examined further. Each of the 20 positive phage populations were rescreened until a single immunoreactive phage population was obtained. DNA was then isolated from each of the positive phage.
Characterization of the Immunoreactiue Phage-The factor XII cDNA insert was released from the vector DNA by digestion with EcoRI followed by purification of the insert DNA by electrophoresis on a 5% polyacrylamide gel. After subcloning into the EcoRI site of bacteriophage M13mp18, the factor XII cDNA sequences were determined by the chain termination method (20). The cDNA sequences of 16 of the 20 positive phage were determined.
Each insert contained DNA coding for a region of factor XII ( Table I). Many of the factor XII fragments began at a common point (amino acid -6 of the hydrophobic leader sequence) resulting from shearing within the poly(G.C) region from the cloning procedure (4). However, only clones 6 and 17 of the 16 factor XII insert DNAs isolated were identical, attesting to the random nature of the factor XII plasmid DNA fragments generated by sonication.
The two shortest factor XII inserts, contained in phage clones 9 (amino acids -6-+31) and 16 (amino acids +l-+47), limit the B7C9 epitope to the amino-terminal 31 amino acids of factor XII (summarized in Fig. 1). Each of the other 14 factor XI1 insert DNAs also contained DNA coding for this peptide (Table I). The DNA sequences of the four remaining phage inserts were not analyzed because their longer lengths made it unlikely that new information would result. Because the factor XI1 plasmid pcHXII501 was originally isolated from a human liver cDNA library in the plasmid pKT218 (24) generated by using the G:C tailing method, seven of the 16 factor XI1 insert DNAs contained a poly(G. C) tail at one end (Table I). In some instances, this created a problem in reading the DNA sequence due to stalling of the Klenow fragment of DNA polymerase I during the chain termination reactions. This technical difficulty was overcome by determining the DNA sequence of the complementary strand of insert DNA to that originally chosen (25).
All of the data generated using X phage clones selected with antibody B7C9 were consistent in identifying the region containing amino acids 1-31 as the B7C9 immunoreactive region of factor XI1 (Fig. 1). A second monoclonal antibody P5-2-1 (12) has also been described that inhibits the surface activation of factor XII. To determine if the two antibodies recognized the same region in factor XII, we tested 14 of the 16 phage clones for which we had determined the complete factor XI1 insert DNA sequence by immunoscreening of the phage with monoclonal antibody P5-2-1 under conditions identical to those used for B7C9. In every phage tested, the plaques were strongly reactive with P5-2-1 (Table 1). We conclude that the independently derived anti-factor XI1 monoclonal antibodies B7C9 and P5-2-1 recognize identical or contiguous antigenic determinants in the amino-terminal 31 amino acids of factor XII. Expression of the Factor X I I Fusion Protein in E. coli-To characterize the factor XI1 fusion proteins further and to allow their expression in E. coli in biochemically useful amounts, the recombinant Xgtll clone B2-2B (containing the 83 amino-terminal amino acids of factor XI1 (Table I)  Hydropathy Plot of the Amino-terminal Region of Factor XZZ-In view of the predicted antigenicity of hydrophilic regions of proteins (27), we were interested in determining the hydrophilicity of the region of factor XI1 that appeared to contain the epitope for the two monoclonal antibodies B7C9 and P5-2-1. Fig. 3 shows a computer-generated hydropathy plot (28) of the amino-terminal region of factor XII. Scanning of the entire factor XI1 polypeptide indicated that the region between amino acids 5 and 15 included one of the most hydrophilic areas of the molecule. Particularly notable was the presence of three positively charged lysine residues at Dositions 8. 11. and 13 that could Dotentially interact Binding of Anti-factor XII Monoclonal Antibodies to Synthetic Peptides-To establish the reactivity of the B7C9 monoclonal antibody with the amino-terminal region of factor XI1 and to define the immunoreactive epitope(s) implicated in the surface-mediated binding of factor XII, a set of four peptides was synthesized that included amino acids 1-28, 5-28,9-28, and 14-28 of factor XII. Acontrol peptide containing residues 1-17 of interleukin-3 (22) was also synthesized. The binding of the B7C9 antibody to the peptides was first tested by slot blot analysis of the individual peptides in 10-pg amounts on nitrocellulose paper followed by incubation with B7C9 and the alkaline phosphatase-conjugated second antibody as described for the phage screens. The B7C9 antibody reacted only with the peptide containing amino acids 1-28 of factor XI1 (Fig. 4). To test the specificity of this binding, a A -

D-
FIG. 4. Binding of B7C9 antibody to factor XI1 peptides. Row A , control peptide (from human interleukin-3); row E , factor XI1 peptide 14-28; row C, factor XI1 peptide 9-28; row D, factor XI1 peptide 5-28 row E, factor XI1 peptide 1-28. In each case, 10 pg of peptide was spotted onto the nitrocellulose filter. duplicate blot of peptides was incubated with an anti-factor XI1 monoclonal antibody (KOK-5 obtained from Dr. Erik Hack, Central Red Cross Laboratory, Amsterdam) that does not inhibit the surface-mediated activation of factor XII. This antibody bound to none of the peptides (data not shown). Together, these results confirmed the previous data that the B7C9 epitope(s) resides in the first 31 amino acids of factor XI1 but suggested that amino acids 1-4 are critical for binding of the antibody. Because different peptides may not bind quantitatively to nitrocellulose, an ELISA was established to test the ability of B7C9 antibody to bind to peptides 1-28,5-28, 9-28, and 14-28 that had been bound to the microtiter dish wells. Again, only peptide 1-28 bound to the antibody (data not shown), confirming the results of the nitrocellulose slot blot assay.
To map the epitope more precisely, a competitive ELISA was established to test the ability of various peptides to block the binding of B7C9 antibody to 33 nmol of peptide 1-28 that had been immobilized on the microtiter dish wells. When peptides 1-28,5-28,9-28, and 14-28 were tested, only peptide 1-28 competed for the antibody binding (Fig. 5), confirming the previous ELISA results. In addition, peptide 1-14 competed with peptide 1-28, although a 10-fold higher concentration of peptide 1-14 was required to obtain the same degree of competition as peptide 1-28 competing for binding with itself (Fig. 5). Interestingly, peptide 4-14 did not compete at all (Fig. 5).
Because these results suggested that the region from 1 to 4 was important for binding of the B7C9 antibody, a series of short peptides (1-4,l-5,l-6,2-7,343, and 4-9) were synthesized and tested in the competitive ELISA, but none of them affected the binding (Fig. 5). Taken together, these results suggest that residues 1-4 of factor XI1 are essential but not sufficient for binding of B7C9 antibody. Moreover, although peptide 1-14 was sufficient to compete for binding, it was not as effective as peptide 1-28, suggesting that the epitope may also involve some as yet undefined secondary structure (29). These results are summarized in Fig. 1.

DISCUSSION
In this study, screening of a factor XI1 cDNA expression library with two murine monoclonal antibodies known to inhibit the activation of factor XI1 by negatively charged surfaces (11, 12) resulted in the isolation of 16 independent bacteriophage clones, all of which contained DNA encoding the first 31 amino acids of factor XI1 (Table I, Fig. 1). Hydropathic analysis of this region of the factor XI1 polypeptide (Fig. 3) revealed a highly hydrophilic cluster between amino acids 5 and 15 with positively charged lysine residues at positions 8,11, and 13. The marked hydrophilic nature and positive electrostatic charge of this area fulfill the predicted criteria for epitope antigenicity in polypeptides (27) and provide a possible basis for involvement of this peptide in binding to a negatively charged surface. However, definitive proof that this area makes up part of the binding site requires demonstration that these peptides interfere with the binding of factor XI1 to a negatively charged surface.
A recent paper has suggested that the surface-binding site of factor XI1 lies between amino acids 134 and 153 on the M, 50,000 heavy chain (11). This result was based on the kallikrein-mediated cleavage of purified factor XI1 followed by purification of the peptide fragments by anti-factor XI1 affinity chromatography using the B7C9 murine monoclonal antibody. None of the three factor XI1 peptides isolated by these authors contained sequences covering the region from 1 to 14 of factor XII. To support our results using recombinant DNA techniques, we synthesized a set of nested peptides between amino acids 1 and 28 of factor XI1 and tested their reactivity with the B7C9 antibody used in both studies. These experiments strongly implicated that the first 14 amino acids of factor XI1 in antibody binding (Fig. 41, as peptides 1-28 and 1-14, are easily detectable by the B7C9 antibody at peptide concentrations two orders of magnitude below those employed for peptide 134-153 (11). However, it is possible that the B7C9 antibody recognizes two noncontiguous sequences of factor XI1 as has been reported for human von Willebrand's factor (30) where the binding of a monoclonal antibody to the protein was blocked by two peptides that were separated by 220 amino acid residues of linear sequence. We plan to investigate this further by functional analysis of deletion mutants of recombinant factor XI1 expressed in eukaryotic cells.
The first 19 amino acids of the amino-terminal region of factor XI1 are encoded by the second exon in factor XI1 genomic DNA (3). In contrast to the sequence homologies between factor XI1 and other plasma proteins ( Fig. 1; Ref. 3), the exon %encoded region of factor XI1 is unique within the blood-clotting factors (4). A similar general organizational structure of the first and second exons of factor XII, tissuetype plasminogen activator and urokinase-type plasminogen activator, has been noted (3), but there is no sequence identity among these three proteins in exon 2. Interestingly, four out of 15 amino acids encoded by the second exon of tissue-type plasminogen activator are arginine residues in close proximity to one another. This constitutes a positively charged region reminiscent of the lysine cluster between amino acids 5 and 15 in factor XII. More interesting, however, is the presence of a common tetrapeptide in factor XI1 and bovine (31) and human (32) high molecular weight kininogen. These positively charged peptide fragments have been implicated in the binding of high molecular weight kininogens to negatively charged surfaces (2). Bovine and human high molecular weight kininogen contain two and three copies, respectively, of the sequence His-Lys-X-Lys where X represents either asparagine, histidine, or phenylalanine in the kininogen sequences or tyrosine in the factor XI1 protein sequence between amino acids 10 and 13. Thus, the highly positively charged protein sequence His-Lys-X-Lys may be involved in the binding of both coagulation factor XI1 and high molecular weight kininogen to negatively charged surfaces.