Thrombin-induced platelet aggregation is inhibited by the heptapeptide Leu271-Ala277 of domain 3 in the heavy chain of high molecular weight kininogen.

The ability of kininogens to modulate thrombin-induced aggregation of human platelets has been assigned to domain 3 (D3) in the common heavy chain coded for by exons 7, 8, and 9 of kininogen gene. We expressed each of the exons 7, 8, and 9, and various combinations as glutathione S-transferase fusion proteins in Escherichia coli. Each of the exon products 7 (Lys236-Gln292), 9 (Val293-Gly328), and 8 (Gln329-Met357), and their combinations were evaluated for the ability to inhibit thrombin induced platelet aggregation. Only products containing exon 7 inhibited platelet aggregation induced by thrombin with an IC50 of > 20 microM. A deletion mutant of exon 7 product, polypeptide 7A product (Lys236-Lys270) did not block thrombin-induced platelet aggregation, while 7B product (Thr255-Gln292) and 7C product (Leu271-Gln292) inhibited aggregation. These findings indicated that the inhibitory activity is localized to residues Leu271-Gln292. Peptides Phe279-Ile283 and Phe281-Gln292 did not block thrombin, and Asn275-Phe279 had only minimal inhibitory activity. A heptapeptide Leu271-Ala277 inhibited thrombin-induced aggregation of platelets with an IC50 of 65 microM. The effect is specific for the activation of platelets by thrombin but not ADP or collagen. No evidence for a thrombin-kininogen complex was found, and neither HK nor its derivatives directly inhibited thrombin activity. Knowledge of the critical sequence of kininogen should allow design of compounds that can modulate thrombin activation of platelets.

Platelets are critical cells in physiological hemostasis and pathological thrombosis. Platelet activation can be defined by distinct cellular events initiated by thrombin, as well as other agonists such as collagen, ADP, and thromboxane A 2 . Stimulation of platelets results in a orderly process consisting of shape change, followed by aggregation and/or secretion, finally resulting in the formation of a platelet plug (Holmsen, 1994).
Thrombin has been shown to bind to two receptors on the platelet surface: platelet glycoprotein Ib-IX complex (Takamatsu et al., 1986) and a seven-membered transmembrane receptor coupled to G proteins. The latter polypeptide is cleaved to expose a new NH 2 -terminal sequence SFLLRN (Vu et al., 1991), which can bind to distal sequences in the receptor to initiate signal transduction. Glycoprotein Ib-IX complex modulates thrombin action since, in its absence in Bernard-Soulier disease, 10 times as much thrombin is required for aggregation of platelets (Jamieson and Okumura, 1978).
The multifunctional protein high molecular weight kininogen (HK) 1 serves as the procofactor of the kallikrein-kinin system (also known as the contact system), a surface-mediated defense system (Colman, 1984). HK binds to platelets (Gustafson et al., 1986), neutrophils (Gustafson et al., 1989), and endothelial cells (Schmaier et al., 1988) and serves as a source of bradykinin after cleavage by a number of plasma proteases (Scott et al., 1984(Scott et al., , 1985. HK and low molecular weight kininogen (LK) are derived from the same gene by alternate splicing of the primary transcript and have identical heavy chains, which consist of three segments (domains 1, 2, and 3), each highly homologous to cystatin (Salvesen et al., 1986;Ishiguro et al., 1987). The sequence HK diverges from LK at a position on their light chains 12 residues COOH-terminal of the bradykinin moiety. Each molecule then exhibits a unique sequence . Kininogens can inhibit papain and cathepsins B, H and L, and are the most potent plasma inhibitors of the calcium-activated cellular protease, calpain . The concentration of the kininogens in plasma is 3.95 M (Kerbiriou-Nabias et al., 1984) comprising HK (0.67 M) and LK (3.28 M).
HK (Puri et al., 1991) and LK (Meloni and Schmaier, 1991) both inhibit the aggregation of human gel-filtered and washed platelets by thrombin. The primary determinants for thrombininduced platelet aggregation are present on the heavy chain, which is shared in both HK and LK. Unlike calpain, which exclusively requires domain 2 of kininogens (Bradford et al., 1990), and papain, which can be inhibited by ether domain 2 or 3, the inhibition of thrombin-induced platelet aggregation is uniquely found in domain 3 of kininogens (Jiang et al., 1992).
In the present study, we have expressed kininogen domain 3 and several fragments of D3, in E. coli as glutathione S-transferase (GST) fusion proteins, and then tested each polypeptide * This study was supported by a Specialized Center of Research (SCOR) in Thrombosis HL45486 (to R. W. C.) from the National Institutes of Health and a focused research grant from Johnson and Johnson (to R. W. C. and B. A. J.), a grant from the W. W. Smith Charitable Foundation, and a special investigator award (to S. P. K.) from American Heart Association Southeastern Pennsylvania affiliate. 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.
** To whom all correspondence should be addressed: Sol Sherry Thrombosis Research Center, Temple University School of Medicine, 3400 N. Broad St., Philadelphia, PA 19140. Tel.: 215-707-4665;Fax: 215-707-2783. for its ability to inhibit thrombin-induced platelet aggregation. Recombinant fragments that demonstrated inhibition were further dissected for critical regions by deletion mutagenesis. Smaller peptides were synthesized and tested for functional inhibition of platelet aggregation. Identification of the inhibitory sequences of kininogen should yield new insights into the structural requirements for the modulation of thrombin-induced platelet aggregation by kininogens.

Purified Proteins
Human fibrinogen (20.7 mg/ml) and HK (1 mg/ml) was purchased from Enzyme Research Laboratories, South Bend, IN. HK was supplied as a single chain and migrates as a single band on sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) of 120,000 daltons. Human ␣-thrombin was a generous gift of Dr. John W. Fenton II (Division of Laboratories and Research, New York State Department of Health, Albany, NY) and had an activity of 2393 NIH units/mg containing 99.24% ␣-thrombin. Hirudin (9230 units/mg) was purchased from Sigma. All other reagents were of the highest purity available.

Construction of Various D3 Deletion Mutant Expression Plasmids
The primary structure of D3 is illustrated (Fig. 1). The details of bacterial expression and purification of D3 (Gly 235 -Met 357 ) are described earlier (Wachtfogel et al., 1994). We have utilized similar methodology using LK cDNA (pHKG36) (Kitamura et al., 1994), kindly provided by Dr. S. Nakanishi, as template to express various D3 fragments in bacteria. The DNA coding for various fragments was separately amplified by polymerase chain reaction (PCR) using specific sets of primers and inserted in-frame into pGEX2T vector (Pharmacia Biotech Inc.) by the procedures described earlier (Wachtfogel et al., 1994;Kunapuli et al., 1993). This construct expresses the protein products as fusion proteins with GST as the NH 2 -terminal region (Fig. 2). All the sense primers contained a BamHI recognition site, and the antisense primers contained the EcoRI recognition site. The nucleotide sequences of the sense and antisense primers used for these constructions are given in Table I. Products of exons 7, 8, and 9 (Gly 235 -Gln 292 , Gln 292 -Gly 328 , and G1n 329 -Met 357 respectively) were constructed using SPK-82A/B, SPK-82C/D, and SPK-82E/F primers, respectively. G1y 235 -Gly 328 (exon products 7 and 8) and Gln 299 -Met 357 (exon products 8 and 9) were constructed using SPK-82A/D and SPK-82C/F, respectively. Thr 255 -Gln 292 (exon 7B product) was constructed using SPK-89 and SPK-82B primers.

Construction of ⌬311-346 D3 Expression Plasmid
Deletion of residues 293-328 in D3 was accomplished by construction of D3 expression plasmid lacking the coding region for Val 293 -Gly 328 by overlapping PCR strategy similar to the one described earlier for ⌬493-520 HK LC . The first PCR was carried out using SPK-82A as a sense primer and SPK-83B (complementary to nucleotides 912-930 and 1039 -1050) as an antisense primer. The second PCR was carried out using SPK-83A (corresponding to nucleotides 916 -930 and 1039 -1053) as a sense primer and SPK-82F as an antisense primer. The products of PCR 1 and 2 were used as a template in a third PCR with SPK-82A and SPK-82F as primers. The product of the third PCR was cleaved with BamHI and EcoRI and inserted into BamHI-and EcoRI-cut pGEX2T as described earlier . This plasmid is capable of expressing D3 lacking residues 293-328 (exon 8 product) and represents a fusion of exons 7 and 9 products.

Construction of Gly 235 -Leu 271 Fragment Expression Plasmid
The plasmid DNA capable of expressing G1y 235 -Gln 292 (exon 7A product) was digested with HindIII and EcoRI, filled in with Klenow fragment of DNA polymerase in the presence of four dNTPs, and religated. The unique HindIII site at nucleotide 862 thus eliminates the coding segment of Asn 272 -Gln 292 .

Construction of Lys 270 -Gln 292 Fragment Expression Plasmid
The expression plasmid capable of expressing Lys 270 -Gln 292 (exon 7C product) was constructed by phosphorylated synthetic oligonucleotides SPK-84 (A-F). These oligonucleotides were annealed to form a fragment with BamHI-and EcoRI-compatible sites, which was ligated into BamHI-and EcoRI-digested pGEX2T. All of these plasmids were checked for the authenticity of coding sequences and junction sequences by DNA sequence analysis (Sanger et al., 1977). Fig. 2 summarizes the protein sequences coded by the recombinant fragments of D3.

Purification of Recombinant Proteins
The recombinant proteins were purified by the procedure of Smith and Johnson (1988) as described earlier . Briefly, E. coli harboring the expression plasmid of interest were grown overnight in 100 ml of LB broth containing 100 g/ml ampicillin at 37°C. This culture was then added to 1 liter of fresh LB broth containing 100 g/ml ampicillin (1:10) and the culture grown for 2-5 h to an A 600 of 0.6. Then, 1 mM isopropyl-1-thio-␤-D-galactopyranoside was added and grown for an additional 2 h. Cells were harvested by centrifugation at 3000 ϫ g for 20 -30 min at 4°C. The cells were resuspended in 20 ml of PBS (0.1 M sodium phosphate, pH 7.4, 0.15 M NaCl) containing 1 mM p-mercuribenzenesulfonyl fluoride, 10 mM EDTA, and 0.5 M pepstatin. The cells were disrupted by sonication, and 0.1 volume of 10% Triton X-100 was added. The cell debris was removed by centrifugation at 12,000 ϫ g for 15 min at 4°C. The clear supernatant was then passed through a glutathione-Sepharose column (Pharmacia), pre-equilibrated in PBS containing 1% Triton X-100 and protease inhibitors. The breakthrough was passed through the column once more, and the column was washed thoroughly with PBS containing Triton X-100, followed by wash with PBS alone. The recombinant fusion protein was eluted with 5 mM reduced glutathione in 20 mM Tris-HCl, pH 8.0, and fractions (1 ml) were collected. The fractions containing the protein were identified by A 280 and by SDS-polyacrylamide gel, followed by Coomassie staining.

Cleavage and Purification of Fusion Products
The proteins were concentrated, exchanged into Tyrode's buffer, and incubated with thrombin at 1/100 of the molar concentration of the recombinant polypeptide exon product for 1 h at 37°C, D-Phe-Pro-Arg chloromethyl ketone was then added at a 10-fold molar excess to thrombin to inactivate any remaining active thrombin. After verification by SDS-PAGE that the fusion peptide had been cleaved, reverse phase HPLC using a linear gradient from 0 -80% acetonitrile in a trifluoroacetic acid buffer on a C-18 column was utilized to purify the expressed polypeptide from GST and thrombin. The recovered polypeptides were lyophilized, and their purity and molecular weights were verified using mass spectroscopy.

Electrophoresis
SDS-PAGE (10%) was used to monitor protein purification and cleavage was performed by the method of Laemmli (1970).

Peptide Synthesis
The peptides were synthesized using a modified standard t-butoxycarbonyl solid phase chemistry on an Applied Biosystems model 430 fully automated synthesizer and purified on HPLC as described previously . The average yield of peptide was 40 mg, and the purity by HPLC analysis was 90 -95%.

Platelet Preparation
Fresh human blood was collected into acid citrate dextrose and centrifuged for 18 min at 990 ϫ g to separate the red cells from the platelet rich plasma component used in the following step. Five ml of platelet rich plasma is applied onto a 50-ml packed Sepharose 2-B CL-300 column equilibrated in calcium-free Tyrode's solution buffered at pH 7.3 (Parker, 1962) containing 3 mg/ml bovine serum albumin (BSA). Platelet fractions were collected, pooled and counted on a model ZM Coulter counter, Coulter Scientific Instruments, Luton, Beds, United Kingdom, yielding an average platelet concentration of 250,000 cells/l.

Functional Platelet Assays
Platelet aggregation was monitored on a Chrono-Log model 440 dual aggregometer (Chrono-Log Corp., Havertown, PA). Aggregation was performed by using a 0.5-ml assay volume at 37°C containing 400 l of the gel-filtered platelets and the volume made up with buffer or test inhibitors. A base line was generated for 2 min, followed by the addition of 0.2 units/ml (2 nM) ␣-thrombin, ADP (10 M), or collagen (2 g/ml) to initiate the platelet aggregation response. The rate of aggregation in the presence of recombinant polypeptides or synthetic peptides was compared with the rate in the absence of any inhibitor. As described previously (Wachtfogel et al., 1994), the fusion proteins contain a thrombin-cleavable sequence between GST and the kininogen sequence to allow removal of the GST portion of the recombinant fusion protein.

Studies to Detect Thrombin-HK Complexes
Chromogenic Assays-Hydrolysis of the chromogenic substrate D-Phe-Pro-Arg-p-nitroaniline (S-2238) (Chromogenix, Franklin, OH) at 0.1 mM at 37°C using 0.04 units/ml ␣-thrombin was performed to assess the ability of kininogen and its derivatives (2-500 M) to inhibit the active site of thrombin. To determine whether thrombin binds to HK at a site other than the active site (without inhibiting enzymatic activity) we first coated HK (2 g/ml) by incubating it overnight in a borate buffer, pH 8.5, blocked it for 1 h with 1 mg/ml BSA, washed with PBS-Tween 20 and then reacted with thrombin, washed again, and finally detected with the chromogenic substrate for thrombin S-2238; this technique has been modified from that used to detect kallikrein complexes with HK (Page et al., 1994).
Fibrinogen Coagulation Assays-The binding of HK (2 M) to thrombin was also assessed by observing the ability of HK (2 M) to inhibit thrombin's activity to convert fibrinogen to fibrin. Clotting was initiated by addition of thrombin (50 units/ml, 500 nM) to fibrinogen (0.125 mg/ml) in a 20 mM Tris-Cl buffer, pH 7.4. After thrombin was added, the clotting time was monitored by a BBL Fibrometer (BBL, Cockeysville, MD) equipped with a 0.3-ml probe. The reciprocal of the clotting time is directly proportional to the thrombin concentration. We also tested whether HK could prevent the access of hirudin (25 units/ml, 330 nM) to the active site and/or exosite, both of which are occupied by the inhibitor. The concentration of hirudin was used at a value that resulted in a 50% inhibition of the reciprocal of the thrombin-fibrinogen clotting time. Kininogen or its derivatives were added to thrombin, followed by hirudin and fibrinogen.

FIG. 2. Recombinant fragments of kininogen domain 3.
The schematic representation of the 10 recombinant HK D3 GST-linked constructs. The corresponding amino acid sequence location for each construct is appropriately indicated. The resulting recombinant protein product identification is listed for each exon product construct. The position of the thrombin cleavage site located between GST and the NH 2 terminus of each construct is indicated with an arrow.
Ϫ nt 930-912 and nt 1039-1050 Ϫ nt 826-879 5Ј-ATTCTCTGCATTAAGCTTG-3Ј SPK-89 ϩ nt 817-834 5Ј-ATTAGGATCCACCAACAGCCCAGAGCTG-3Ј Thrombin was adjusted to pH 9.5 with 0.5 M sodium carbonate, after which 0.2 mg of FITC was added to 1 mg of thrombin and then incubated on ice for 2 h in total darkness. The unbound FITC was removed by gel filtration on a G-25 Sephadex column, equilibrated in PBS. The labeled thrombin present in the first colored band was collected from the column. The thrombin binding assay was performed by capturing HK on latex beads (Seradyne, Inc. (Division of Mitsubishi Chemical Corp.) Indianapolis, IN) with a monoclonal antibody, C1C11 (Schmaier et al., 1987), directed against the light chain of HK (domain 5), which was chosen to avoid interference with the D3 domain. Increasing amounts of FITC labeled thrombin (5-100 g/ml) were then added and incubated for 60 min. The beads were then collected on a porous membrane microplate and washed with PBS. The fluorescence was measured in a fluorescent concentration analyzer (Idexx Laboratories, Westbrook, ME). Cross-linking of HK with Thrombin-HK (3.8 M) and thrombin were incubated in PBS for 10 min at 25°C. Bis(sulfosuccinimidyl) suberate (Pierce), in Me 2 SO was added to yield a concentration of 250 M and incubated for 30 min, then quenched for 15 min with Tris-OH (5 mM). Samples were withdrawn and analyzed by SDS-PAGE under reduced (10 mM dithiothreitol) and nonreduced conditions. HK and thrombin were each treated separately for comparison with the test samples.

Complex Formation of HK with Radiolabeled Thrombin
Thrombin was labeled by the IODOGEN method with 125 I (Fraker and Speck, 1978). 125 I-thrombin (at 20 g/ml, 13,000 cpm/l) was incubated with HK at molar ratios of 0.5, 1.0, and 2.0. The mixture was electrophoresed on 10% PAGE (without SDS), and the gels were dried and autoradiographed on Cronex 4 film, DuPont Medical Products, Wilmington, DE.

Generation and Purification of Recombinant Kininogen Domain 3
Fragments-Using procedures described under "Methods and Materials," we have generated various recombinant fragments of kininogen domain 3, and the structures of these fusion proteins are shown in Fig. 2. All the recombinant fragments contain GST at their amino terminus, unless otherwise stated. These recombinant fragments were purified using glutathione-Sepharose as affinity matrix in a single step. The purified recombinant fragments were electrophoresed on SDSpolyacrylamide gels and stained with Coomassie Brilliant Blue R250. As can be seen in Fig. 3, the GST fusion proteins show mainly single components at the predicted molecular size. Some heterogeneity was observed in the recombinant fragments for exons 7A and 8, both of which contained a second band (27 kDa) consistent with GST that apparently had leached from the glutathione-Sepharose affinity column along with each of the resulting fusion proteins. The resulting purified fragment for exon 7 showed three bands comprising exon 7, GST, and an undetermined component migrating between GST and exon 7.
Inhibition of Thrombin-induced Platelet Aggregation by Recombinant Kininogen D3 Fragments-Various kininogen domain 3 recombinant fragments were evaluated for the ability to block the aggregation of gel-filtered platelets by (2 nM) thrombin. Initial screening of the seven initial recombinant fusion proteins showed that D3 (Gly 235 -Met 357 ), coded for by exons 7, 8, and 9; exon 7 product (Gly 235 -Gln 292 ); and products of exons 7 and 8 and exons 7 and 9, were capable of blocking thrombininduced platelet aggregation with an IC 50 less than or equal to 20 M (Table II). These results suggest that only those polypeptides containing the sequence coded by exon 7 had inhibitory activity. The IC 50 of the GST-free exon 7 product is 13.4 M, only 3 times more than D3. These results indicate that the primary determinants for inhibition of thrombin-induced aggregation reside in the polypeptide coded by exon 7. In order to further define the sequence responsible for inhibition, three additional recombinant fragments were made from exon 7, namely 7A (Gly 235 -Leu 270 ), 7B (Thr 255 -Gln 292 ), and 7C (Lys 270 -Gln 292 ). When these recombinant polypeptides were tested for their ability to inhibit thrombin-induced aggregation, 7B, containing the COOH-terminal half of exon 7 product, was the most effective inhibitor (25 M). Since GST (27 kDa) could potentially modify the inhibitory activity of the much smaller kininogen polypeptides, the recombinant polypeptides were cleaved from fusion proteins, purified to homogeneity, and retested in the aggregation assay (Table II). Several attempts were made to cleave and remove GST from exon 7A product, but this construct proved resistant to thrombin cleavage even at 10-fold higher thrombin concentrations and incubation times of up to 18 h; therefore, only the IC 50 value for the GST-linked product is reported. However, the IC 50 of the fusion protein exon product 7A was the least potent of the exon 7 products. When the cleaved polypeptides were repurified on HPLC (see "Materials and Methods" and Table II), D3 was the most potent (IC 50 ϭ 4 M, similar to native HK ϭ 2 M). Exon 7, 7B, and 7C products were highly active, only 3-5 times less potent than D3 (IC 50 13.4 -19 M) (Table II, Fig. 4). The shortest inhibitory sequence 7C was the 23-amino acid Lys 270 -Gln 292 (exon 7C).  at 275 M (Table III). In contrast, Leu 271 -Ala 277 showed an IC 50 of 65 M (about 15 times that of recombinant D3) (Fig. 4).

Molecular Model of Leu 271 -Ala 277 within D3-
We restudied a homology model of D3 constructed as described previously , based on the crystalline structure of egg white cystatin (Bode et al., 1988). The polypeptide L 271 NAENNA 277 appears as a subdomain on the surface (Fig.  5). An ␣ carbon trace (data not shown) indicates that it is in a helix-turn conformation.
Thrombin Binding to Kininogen-HK was evaluated to determine if the observed inhibition of platelet aggregation by HK could be due to direct complexing of thrombin to HK by four different methods. The ability of HK to inhibit the thrombin conversion of fibrinogen to fibrin was first evaluated and compared with hirudin. Hirudin could completely block thrombin clotting of fibrinogen, while HK showed no effect up to a 50-fold molar excess to thrombin. Second, when HK (2 g/ml) or a control BSA (2 g/ml) was coated onto an Immobilon plate overnight, washed, reacted with increasing amounts of thrombin, washed, and probed for bound thrombin with S-2238, modest amounts of thrombin activity could be detected bound to both HK and BSA, but no specific binding of thrombin to HK was observed when the amount of thrombin bound to the BSA control was subtracted. Similar results were observed when HK bound to a C1C11 monoclonal antibody (directed to HK domain 5) linked to latex beads (see "Materials and Methods") were used and reacted with increasing concentrations of FITClabeled thrombin. Fluorescence concentration analyzer detection of FITC-labeled thrombin bound to HK immobilized to latex beads yielded negative results. 125 I-Thrombin was reacted with kininogen at 1/0.5, 1/1, and 1/2 molar ratios and run on 10% native gels (no SDS), then autoradiographed to determined if a mobility shift could be observed or the presence of a new band could be seen. No difference could be found between the thrombin control and the lanes containing thrombin and kininogen mixtures (data not shown). These experiments provide further evidence against the formation of a thrombin-HK complex to explain the ability of HK to inhibit thrombin-induced platelet aggregation.
Kininogen Peptide Effects on the Amidolytic Activity of Thrombin-Selected synthetic peptide was evaluated for the ability to block the thrombin cleavage of the chromogenic substrate S-2238 (H-D-Phe-Pip-Arg-p-nitroanilide). The synthetic peptides Leu 271 -Ala 277 or the intact HK molecule did not inhibit the active site of thrombin at a concentrations below 500 M, while D-Phe-Pro-Arg-chloromethyl ketone completely abolished thrombin activity at concentration of 50 M or above. The addition of 1 unit of hirudin could inhibit 80% of thrombin amidolytic activity, and the addition of 2 M HK could not compete with hirudin for thrombin regardless of whether HK was added before or after hirudin addition (data not shown). Taken together, kininogens or their derivatives do not block the active site of thrombin.

DISCUSSION
The central role of thrombin in the pathogenesis of venous thrombosis has emphasized the actions of thrombin not only to clot fibrinogen and form the fibrin plug, but also to convert procofactors factor V and VIII to active cofactors factor Va (Colman et al., 1970) and VIIIa (Eaton et al., 1986), and thereby enhance the formation of thrombin. In contrast, on the arterial  side, coronary or cerebral vessels are occluded by thrombi composed of both platelets and fibrin. The importance of thrombin in re-occlusion (Puri and Colman, 1993) after angioplasty (Heras et al., 1989) or thrombolysis (Gash et al., 1986) has emphasized the role of this serine protease in activating platelets. Various strategies have been suggested to prevent this process, including preventing platelet aggregation by all agonists and by blocking the binding of fibrinogen to its platelet receptor glycoproteins IIb/IIIa (Bennett and Vilaire, 1979). Unfortunately, monoclonal antibodies to this platelet integrin, while effective in inhibiting platelet aggregation by thrombin, cause excessive bleeding (Hanson et al., 1988). This result is not unexpected, since the hereditary absence or functional impairment of glycoprotein IIb/IIIa results in a hemorrhagic disease, Glanzmann's thrombasthenia (Nurden and Caen, 1974;Phillips et al., 1975). A second strategy uses direct thrombin inhibitory polypeptides such as hirudin (Markwardt, 1991) or small organic inhibitors such as argatroban (Kikumoto et al., 1984). Since these compounds inhibit the active site of thrombin, they can prevent thrombin activation of platelets, which requires proteolytic activity. By inhibiting the actions of thrombin on the coagulation process, they, like heparin, also increase the possibility of hemorrhage.
Spurred by the cloning of a seven-transmembrane thrombin receptor (Coughlin et al., 1992) and discovery of a tethered peptide revealed after cleavage that initiated changes in intracellular calcium, investigators have attempted, thus far without remarkable success, to develop peptides that inhibit the binding of thrombin to this receptor. We found that the presence of physiological concentrations of HK, not only in purified systems but also in plasma, modulates thrombin-induced platelet aggregation by shifting the concentration dependence so that 10 times more thrombin is required to stimulate platelets than in washed platelets (Puri et al., 1991). Thus, we hypothesized that HK contained critical sequences that could modulate thrombin-induced platelet aggregation. Jiang et al. (1992) demonstrated that D3 of kininogen was able to prevent binding of thrombin to platelets. Using monoclonal antibodies, they found that they could distinguish two activities in D3, one of which blocked HK binding to platelets and the other thrombin binding. However, the location of their binding sites was unknown. Recently, Herwald et al. (1995) reported that a peptide Leu 331 -Met 357 present in the polypeptide coded for by exon 9 inhibited HK binding with an IC 50 ϭ 60 M. We have studied the similar oxidized peptide Cys 333 -Cys 352 for its ability to inhibit platelet aggregation induced by thrombin. At 84 M, no inhibition was observed (data not shown). It should be noted that the exon 7 peptides blocked thrombin-induced aggregation, while the exon 9 peptides blocked HK binding, and therefore the biological activities of the peptides were distinct.
Domain 3 is coded for by 3 exons designated exons 7, 8, and 9 . We reasoned that each exon might express distinct functions. Therefore, we expressed each exon product as a GST fusion protein and tested its ability to inhibit thrombin-induced aggregation of platelets from normal human donors, and compared the potency to HK and D3. Recombinant D3 was similar in potency to HK, with only a 2-fold difference in the concentration needed to inhibit the maximum aggrega- FIG. 5. Thrombin inhibitory sequence of HK. Figure is a three-dimensional homology model of D3 based on cystatin with the position of the peptide sequence LNAENNA highlighted and displayed as a hydrated region. The amino terminus and carboxyl terminus are indicated on the left and in the center of this model, respectively. The LNAENNA peptide sequence exhibits a helix-loop conformation (when viewed as an ␣ carbon trace) and is completely exposed on the exterior surface of the molecular model.
FIG. 6. Specificity of LNAENNA on platelet agonists. Panels 1-4 demonstrate the specificity of 26 M LNAENNA on three platelet agonists. Panel 1 shows the inhibition by LNAENNA of gel-filtered platelets stimulated by the addition of 2 nM thrombin. Panel 2 shows an apparent increase in inhibitory potency with the same concentration of inhibitor but using thrombin at 1 nM instead, suggesting that both inhibitor and agonists compete for the same site on the platelet surface. Panel 3 indicates that LNAENNA has no effect on ADP-stimulated platelets (10 mM). Panel 4 shows that the aggregation by collagen (2 mg/ml) was unaffected by this peptide sequence. tion by 50%. Only the recombinant polypeptide coded by exon 7, Gly 235 -Gln 292 , exhibited potency equivalent to HK or D3. In fact, exon 7 product when used as GST fusion protein showed an IC 50 ϭ 0.2 M, 10 times more potent than D3. This increase in potency is probably due to more favorable folding, since, after cleavage of the thrombin-sensitive linker and purification of free exon product 7, the potency decreased 67-fold and was one-third as potent as D3 (Table II). Cleavage of the fusion protein to yield the free recombinant fragment of kininogen did not always result in a less potent polypeptide. In fact, in the other cases of exon product 7 fragment, potency was increased (see below) by a factor of 2-3. Since two activities were distinguished within D3, we also made all three combinations of the exon product, 7ϩ8, 8ϩ9, and 7ϩ9, to test whether the difference between exon 7 product and D3 could be accounted for by another site on a different exon product. No increase in potency was found in products of exons 7ϩ8 or exons 7ϩ9, and the product of exons 8ϩ9 showed no activity.
Since exon 7 product comprises 56 amino acids, we sought to localize the responsible sequence more precisely. We generated three overlapping recombinant polypeptides. Exon product 7A, Gly 235 -Leu 271 , comprised the NH 2 -terminal half of the exon product 7, while 7C represented the COOH-terminal portion, Lys 270 -Gln 292 . In order to avoid splitting the active sequence, we also expressed 7B, Thr 255 -Gln 292 . Since 7B and 7C both inhibited thrombin-induced activation, but 7A did not, the functional region lay in the COOH-terminal 23-mer, Lys 270 -Gln 292 . Examination of the molecular model of D3 revealed that the proximal portion of the sequence of 7C was on the surface, but much of the rest was associated with the protein core. To test the hypothesis that the NH 2 -terminal portion of 7C contained all or part of the site responsible for inhibition of thrombin-induced platelet aggregation, we synthesized four peptides that subsumed the sequence Lys 270 -Gln 292 . Only the peptide Leu 271 -Ala 277 , which exists in a surface loop at the junction of a helix-turn configuration in the molecular model, inhibited platelet aggregation by thrombin. A search of the protein data base indicates no significant homologies except to domain 3 of rat and bovine kininogen. the latter shows complete identity for LNAENNA. The rat HK and LK contains an identical heptapeptide except for substitution of histidine for Ala 277 . Rat T-kininogen, which is a potent inhibitor of thrombin-induced platelet aggregation, 2 shows high homology with 5 identical amino acids. Gln and His are substituted, respectively, for Glu 274 and Asn 275 , which are extremely conservative substitutions.
We then addressed the mechanism of the action of exon product 7 and its contained peptide, LNAENNA, on thrombininduced activation. Several possibilities were considered. HK might form a complex with thrombin. HK is known to form complexes with prekallikrein and factor XI using portions of domain 6 (Tait and Fujikawa, 1987), with plasminogen via the light chain (Humphries et al., 1994), and with thrombospondin (DeLa Cadena et al., 1994b), involving both HK light and heavy chains. A peptide in D3 within exon 7 product is involved in HK binding to thrombospondin on the activated platelet surface . However, it is contained within exon product 7A, which did not inhibit thrombin-induced aggregation. We found that HK did not inhibit thrombin cleavage of fibrinogen or a peptide substrate (amidolytic activity). Moreover, thrombin failed to bind to immobilized HK or to HK in the fluid phase or to HK on native gel electrophoresis. While this study was in progress, Hasan and colleagues  also were unable to demonstrate a thrombin-HK complex using several techniques different from those used in this report. The second possibility is that the sequence LNAENNA in D3 directly blocks the binding of thrombin to the seven transmembrane receptor. Three pieces of evidence make this unlikely. First, HK does not inhibit the tethered peptide SFLLRN from aggregating platelets (Jiang et al., 1992) nor did LNAENNA in our study. Second, higher concentrations of thrombin can reverse the inhibition by HK (Puri et al., 1991). Finally, exon 7C product and LNAENNA fail to block shape change induced by thrombin. 2 Shape change is the earliest morphological change following ligand binding and is associated with an increase in intracellular Ca 2ϩ and with phosphorylation of myosin light chain (Daniel et al., 1984). The calcium increase has been shown to be a consequence of the interaction of SFLLRN or thrombin with the seven-transmembrane receptor. We cannot rule out an effect on this receptor at a site distant from the tethered peptide, but consider this unlikely. We are currently testing the hypothesis that HK may inhibit the binding of thrombin to glycoprotein Ib.
Finally, our studies show that the HK D3 and its constituents inhibit neither ADP and collagen-induced aggregation nor thrombin protease activity. This specificity would allow selective inhibition of thrombin-induced platelet aggregation, thus modulating thrombin-platelet interaction without either compromise of platelet function or thrombin effects on blood coagulation. These properties make it an attractive template for construction of peptidomimetic drugs to inhibit reocclusion and thrombosis.