Hirudisins HIRUDIN-DERIVED THROMBIN INHIBITORS WITH DISINTEGRIN ACTIVITY*

Recombinant hirudin variants have been designed which inhibit a-thrombin by the hirudin mechanism and which in addition exhibit disintegrin activity.

The thrombin-specific inhibitor hirudin is a polypeptide of 65 amino acid residues isolated from the leech Hirudo rnedicinalis (Markwardt, 1970;Bagdy et al., 1976;Dodt et al., 1986a;Harvey et al., 1986;Tripier, 1988). On the basis of its structure and mechanism of thrombin inhibition this protein appears t o be unique among serine proteinase inhibitors. From solution studies on thrombin-hirudin complex formation (Stone and Hofsteenge, 1986;Braun et al., 1988;Dennis et al., 1990;Dodt et al., 1990;Ni et al., 1990) as well as from the crystallographic structure of the complex (Grutter et al., 1990;Rydel et al., 1990) it became evident that hirudin is basically a bivalent inhibitor comprising two binding domains. These are located in the disulfide-bonded NH2-terminal part (residues 1-49) and the COOH-terminal region (residues 54-65). Ionic interactions of negatively charged residues in the COOHterminal part of hirudin with a positively charged surface groove on thrombin have been shown to play an important role in the first step of complex formation (Braun et al., 1988; *This work was supported by a grant from the Deutsche Forschungsgemeinschaft. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisernent" in accordance with 18 1J.S.C. Section 1734 solely to indicate this fact. $ TO whom correspondence should be addressed. Stone et aL, 1989;Betz et al., 1991). In the second step, interactions of the NHs-terminal core and especially of the NH2-terminal 3 amino acid residues (Wallace et al., 1989;Lazar et al., 1991) with the active site cleft inhibit the enzyme's hydrolytic activities. Hirudin-derived COOH-terminal peptides with a minimal length of 12 amino acid residues exhibit anticoagulant activities (Mao et al., 1988;Maraganore et al., 1989) but only slightly modulate the enzyme's hydrolytic activity towards peptide-p-nitroanilides (Naski et al., 1990;Schmitz et al., 1991;Liu et al., 1991). Thus, based on the desired therapeutical application hirudin and derivatives are potent potential therapeutic anticoagulants.
Adhesion of platelets to vessel walls, their activation, and aggregation is a central primary event in blood coagulation. On activated platelets the glycoprotein IIb/IIIa (GP IIb-IIIa),* a member of the integrin family, functions as a receptor for fibrinogen, fibronectin, von Willebrand factor, vitronectin, and thrombospondin (Hynes, 1987). Binding of GP IIb-IIIa to fibrinogen or von Willebrand factor mediates platelet aggregation, whereas binding to the other proteins may additionally allow platelet adhesion and spreading. The capability of GP IIb-IIIa to bind to adhesive proteins is due to its ability to recognize RGD motifs within their sequences (Ruoslathi and Pierschbacher, 1987;Phillips et a/., 1988). Peptides containing the RGD sequence inhibit the binding of fibrinogen or von Willebrand factor to GP IIb-IIIa to platelets, thus inhibiting platelet aggregation (Kiefer and Phillips, 1990). However, not only the primary structure around the RGD motif but also its conformation, maintained by the appropriate cysteine pairing, is necessary for full expression of disintegrin activity (Dennis et al., 1990). Pierschbacher and Ruoslathi (1984) have already suggested that the RGD motif of adhesive proteins may be located at the tip of a 0-turn. Recently, this proposal was established by 'H NMR studies on the platelet aggregation inhibitors kistrin and echistatin from snake venoms (Adler et al., 1991;Cooke et al., 1991;Dalvit et al., 1991;Saudek et al., 1991). The RGD binding motif of these proteins is located at the end of a long arm consisting of two antiparallel strands which are connected by a 0-turn with high flexibility.
Recently, a chimeric peptide has been described consisting of the cell adhesion sequence RGDS followed by the COOHterminal region of hirudin (residues 53-64) (Church et al., 1991). This peptide was found to exhibit both disintegrin and antithrombin activities. It displays the corresponding activities to the same extend as its individual constituents. From the crystal structure of the thrombin-hirudin complex it is known that a finger-like structure points outwards from the The abbreviations used are: GP, glycoprotein; AMC, 7-(4methyl)-coumarylamide; r-, recombinant; cm-, carboxymethylated; Tos, p-toluenesulfonyl-; ELISA, enzyme-linked immunosorbent assay; PRP, platelet-rich plasma. complex into solution (Rydel et al., 1990;Griitter et al., 1990). The finger is formed by an antiparallel p-sheet comprising residues 27-31 and 36-40 which are connected by a highly flexible p-turn (residues 32-35). Since hirudin obviously shares this structural element with the snake venom proteins, we have decided to replace residues 32-35 of hirudin by the fibronectin-derived adhesion sequence RGDS. Introducing this sequence into an appropriate environment should enhance its disintegrin activity relative to that of the linear RGD peptides, yielding anticoagulants on the basis of hirudin which exhibit both antithrombin and platelet aggregation inhibitory activity. In this paper we provide data that the chimeric hirudin, called hirudisin, is a thrombin inhibitor comparable to r-hirudin as well as an inhibitor of platelet aggregation by direct interaction with the integrin GP IIb-IIIa. Thus, these proteins are of potential use as integrintargeted antithrombotic agents.
Oligonucleotide-directed Mutagenesis-Site-directed mutagenesis was performed according to Kunkel et al. (1987). A DNA fragment coding for the tac promoter, the alkaline phosphatase signal sequence, and the hirudin gene was inserted into the phasmid pUC 118 (Vieira and Messing, 1987). Uracil-containing template single-stranded DNA was prepared using the Escherichia coli strain RZ1032 (American Type Culture Collection No. 39737). Oligonucleotides of 32 bases were used as mutagenic primers. I n vitro DNA synthesis was performed with T 7 DNA polymerase and T4 DNA ligase. E. coli BMH71-18 mutS (Kramer et al., 1984) was transformed with the mutagenized phasmid double-stranded DNA. Mutants were identified by sequencing single-stranded DNA according to Sanger et al. (1977).
Protein Purification-Human prothrombin was isolated according to Mann (1976) and converted to a-thrombin using the venom of Oryuranus scutellatus (Owen and Jackson, 1973). a-Thrombin concentrations were determined by active-site titrations (Jameson et al., 1973), the purity was >98%. Recombinant hirudisins were isolated from the periplasm of E. coli by anion-exchange Chromatography on DEAE Sephadex A-25 and reverse-phase high performance liquid chromatography on a Shandon ODS Hypersil column (Dodt et al., 1986b). Protein sequence analyses (Hunkapiller et al., 1983) of 0.1-0.8 nmol of the recombinant proteins or of tryptic peptides of the reduced and carboxymethylated inhibitors were performed in a gasphase sequenator (Applied Biosystems Model 470A connected to a Model 120A analyzer) and confirmed the correct sequence for hirudisins. G P IIb-IIIa was prepared by the method of Fitzgerald et al. (1985). After removing salts and reagents by chromatography on a PDlO column, reduced and carboxymethylated hirudisin (cm-hirudisin) was isolated by reverse-phase high performance liquid chromatography for inhibition studies.
FibrinogenlGP Ilb-IIIa ELISA-The ELISA was performed according to Dennis et al. (1990). Microtiter plates were coated with human fibrinogen (1 mg/ml) in 0.1 M sodium carbonate buffer, pH 9.0, overnight and then incubated with 3% Blotto (5%, w/v, nonfat dry milk) in phosphate-buffered saline. After washing with TBST (50 mM Tris-HC1, pH 7.5, containing 0.15 M NaCI, 10 mM CaC12, 0.05% Tween 20) the samples to be evaluated and then G P IIb-IIIa (1 mg/ ml in TBST) were added. After 1 h of incubation the plate was washed, and bound G P IIh-IIIa was labeled by the biotinylated monoclonal antibody (1 pg/ml). Bound antibodies were detected with a streptavidin-peroxidase complex (50 milliunits/ml).
Platelet Aggregation Assays-Platelet aggregation assays were performed in human platelet-rich plasma (PRP). P R P was prepared as described by Dennis et al. (1990). For the assays, P R P was diluted to 250,000 platelets/pl with platelet-poor plasma. P R P (200 pl) plus 25 p1 of sample in 0.9% NaCl or plus 25 pl of 0.9% NaCl alone was incubated in an aggregometer at 37 'C for 5 min. The aggregating agent (25 pl of a-thrombin (1 NIH unit) or 25 p1 of 100 mM ADP) was added and the light transmission recorded. Transmission was set at 100% for platelet-poor plasma and 0% for PRP.
Amidolytic Assays of Thrombin Activity-Assays were performed in polyacrylate cuvettes at 25 "C in 0.1 M Tris-HCI, pH 8.3, containing 0.2 M sodium chloride and 0.05% Triton X-100 (Dodt et al., 1988). Tos-Gly-Pro-Arg-AMC was used as substrate at a concentration of 50 p~. Under the specified conditions, the K , of the substrate was 5.0 f 0.2 p~. In tight-binding inhibition experiments, 50 pM athrombin was preincubated with 10-200 p~ inhibitor and the steadystate velocity was measured. In slow-binding inhibition experiments, 200-1000 PM inhibitor was incubated with 50 pM substrate and the reaction was started by the addition of 20 PM enzyme. Assays were performed with a Perkin Elmer LS50 spectrofluorimeter (Aex = 370 nm; X,, = 450 nm), and fluorescence intensities were calibrated with 7-amino-4-methylcoumanin (100 nM).
Datu Analysis-Data analysis was performed with the nonlinear regression program GraFit (Leatherbarrow, 1990). In slow-binding inhibition experiments for each of the inhibitors a set of progress curves a t several inhibitor concentrations was obtained and the data were fitted to Equation 1 by nonlinear regression analysis (Morrison, 1982;Morrison and Stone, 1985).
The symbols u,, us, and he,, represent the initial velocity, steady-state velocity, and an apparent first-order rate constant; d is a displacement term to account for the case that at t = 0 the fluorescence is not known accurately. Values for k,,, were plotted uersus [I] and fitted to Equation 2 to obtain k,,, (Morrison, 1982).
Equilibrium dissociation constants K, were obtained from tight-binding inhibition experiments by fitting the data to Equation 4 (Cha, 1975;Williams and Morrison, 1979), with the following relationship.
IC50 values of inhibitors in platelet aggregation assays as well as fibrinogen/GP IIb-IIIa ELISA were determined by numerical analysis of the data according to Equation 5 (Leatherbarrow, 1990).
Where E is the response variable, a the maximum range of y values, d the background y value, c the slope factor, and [I] represents the inhibitor concentration.

RESULTS
Interaction of Hirudisin with Human a-Thrombin-Amino acid residues Ser-Asp-Gly-Glu at position 32-35 of hirudin were replaced by the cell adhesion motif Arg-Gly-Asp-Ser of fibronectin to generate hirudisin. In order to determine the effect of the mutation at a site which should not be involved in the thrombin-hirudin interaction (Rydel et al., 1990;Grutter et al., 1990), we studied the inhibitory properties of hirudisin in slow-binding as well as tight-binding inhibition experiments. From tight-binding experiments a 2-fold decrease in the dissociation constant Ki for the chimeric protein (Ki = 0.16 f 0.06 PM) was obtained which corresponds well to the dissociation constant K; = 0.27 k 0.03 PM of r-hirudin (Table  I). In slow-binding inhibition experiments a set of progress curves was obtained at a fixed enzyme concentration (20 PM) and several inhibitor concentrations (200-800 PM). Curves were analyzed according to Equation 1 to yield the apparent first-order rate constants kapp. From the slope of a plot of k,,,  M" s" has been determined for hirudisin. As the intercept with the y axis was too close to zero we estimated the dissociation rate constant k,ff = 1.54 f 0.3 s" using the Ki from tight-binding and the kOn from slow-binding experiments (k,ff = k,, x Ki). Hirudisin-1, a variant of hirudisin with Arg3' replaced by Lys, has also been analyzed for thrombin inhibition properties and was found to display similar kinetics as determined for r-hirudin. Even deletion of the Ser-Asp-Gly-Glu sequence, which forms a flexible p-turn connecting strand I1 (residues 27-31) and strand 11' (residues 36-40) of the second antiparallel p-sheet of hirudin (nomenclature according to Folkers et al., 1989), does not interfere with thrombininhibitor interaction (Table I). Thus, modification of the amino acid sequence at the tip of the finger-like structure of hirudin can be performed without any loss of thrombin inhibitory activity.

Summary of GPIIb-IIIa antagonist activity
ICso value is the concentration necessary to inhibit total platelet aggregation or binding of G P IIb-IIIa to fibrinogen in the ELISA assay to 50% of the control. ADP-and thrombin-induced platelet aggregation assays were performed in P R P as described under "Experimental Procedures." Numbers in parentheses refer to maximal concentrations in the assays. NE, no effect. the linear unfolded NH2-terminal domain, inhibited platelet aggregation slightly better than hirudin(45-65). The 9-fold lower ICso appears to be due to additional contacts with thrombin. The GRGDS peptide did not inhibit thrombininduced platelet aggregation up to 10 p~. However, at higher concentrations the peptide showed the same dose-dependent aggregation inhibition as in ADP-stimulated assays. This observation has to be assigned to the RGD-mediated route.
Inhibition of ADP-induced Platelet Aggregation-GP IIb-IIIa antagonist activity was measured in vitro by dose-dependent inhibition of ADP-stimulated platelet aggregation in human platelet-rich plasma (Fig. 2). Disintegrin activity is most effectively exhibited by hirudisin (IC50 = 65 p~) . The cell adhesion sequence of this protein is about %fold more potent than that of cm-hirudisin and the linear GRGDS peptide (IC50 = 180 p~) as well as about 6-fold more potent than the RGDS peptide (IC50 = 330 p~) .
The GKGDS sequence of hirudisin-1 exhibited much weaker G P IIb-IIIa antagonist activity (Table 11). Under assay conditions only slight inhibition of ADP-induced platelet aggregation was observed with 120 p~ hirudisin-1.
Inhibition of GP IIb-IIIa-Fibrinogen Interaction-Using a solid-phase integrin binding assay (Dennis et al., 1990) we determined the relative potency of hirudisin, r-hirudin, and RGDS peptide, to inhibit the binding of G P IIb-IIIa to immobilized fibrinogen (Fig. 3). Hirudisin binds 9-fold weaker to GP IIb-IIIa in solution (ICso = 13.6 p~) than has been observed for the linear RGDS peptide (ICso = 1.5 p~) .
Up to 35 p~ r-hirudin did not inhibit the receptor-ligand interaction. It is noteworthy that the binding potency of the RGDS peptide is 200-fold enhanced in the ELISA as compared to the platelet aggregation assay, whereas the binding potency of hirudisin is only 4-fold increased (Table 11). The different results are probably caused by differences in the affinity of purified GP IIb-IIIa for the disintegrins relative to the affinitiy in intact platelets.

DISCUSSION
We have generated hirudin-based thrombin-specific inhibitors with RGD-mediated disintegrin activity. The major findings of this study are summarized as follows. First, modifications of the @-turn (residues 32-35) in the finger-like structure of hirudin appear not to be crucial to the inhibitor's interaction with a-thrombin. Second, replacing the SDGE sequence (residues 32-35) by the cell adhesion sequence RGDS results in the aquisition of disintegrin function. Third, enhanced inhibition of ADP-induced platelet aggregation by native hirudisin relative to cm-hirudisin illustrates a considerable contribution of the tertiary structure to integrin binding. Fourth, thrombin-stimulated platelet aggregation shows that combining antithrombin and disintegrin activity does not result in a cooperative effect for hirudisins available a t present.
Recently, the solution structures of the GP IIb-IIIa antagonists kistrin (Adler et al., 1991) and echistatin (Cooke et al., 1991;Dalvit et al., 1991;Saudek et al., 1991) from snake venoms illustrated tertiary structure requirements of functional RGD adhesion sequences. The recognition sequence of echistatin lies in a loop of 11 residues joining the two strands of an antiparallel @-sheet and protrudes from the tightly packed core of the molecule. The RGD adhesion site of kistrin is located at the end of a long arm (residues 49-51) connecting two antiparallel strands (residues 41-48 and 52-59). Since hirudin displays a related structural element in the loop (residues 32-35) linking two antiparallel @-strands (residues 27-31 and 36-40) (Folkers et al., 1989;Grutter et al., 1990;Haruyama and Wuthrich, 1989;Rydel at el., 1990) FIG. 4. Alignment of RGD containing loops of adhesive proteins. The primary structure of the RGD containing loop of hirudisin is compared with that of fibronectin and several snake venom disintegrins. Positively charged (0) amino acid residues are found at the NH,-terminal of the RGD motif, and negatively charged (0) amino acid residues accumulate at the COOH-terminal side. Church et al. (1991) it was already known that chimeric molecules could be constructed combining the RGDS sequence and hirudin-derived COOH-terminal peptides which exhibit both antithrombotic and anti-adhesive activities. Our results demonstrate that hirudisin (ICs0 = 65 ptM) is a 3-to 5fold better inhibitor of ADP-induced platelet aggregation in PRP than RGDS or GRGDS peptides, whereas by careful interpretation of our present data, maximum inhibitory potency of cm-hirudisin is the same as that of the RGD peptides (Fig. 2). However, the 200-to 2000-fold poorer activity of the cell adhesion motif in hirudisin relative to that in disintegrins from snake venoms like echistatin (ICso = 32 nM; Garsky et al. (1989)), flavoridin (ICso = 40 nM; Musial et al. (1990)), kistrin (ICso = 130 nM; Dennis et al. (1990)), albolabrin (IC50 = 220 nM; Calvete et al. (1991)), and barbourin (ICso = 300 nM; Scarborough et a1. (1991)) may have two reasons: (i) the RGD sequence is not in the proper conformation for optimal binding to GP IIb-IIIa, or (ii) the linear sequence or the environment around the adhesion sequence does not satisfy all requirements. Comparison of amino acid sequences of the RGD containing loops of snake venom proteins (Fig. 4) reveals some conserved properties. There are positively charged amino acid residues on the NH2-terminal side of the RGD motif and an accumulation of negatively charged amino acid residues on the COOH-terminal side of the binding sequence. In the hirudisin molecule a positively charged lysyl residue follows on the COOH-terminal side of the adhesion motif. Thus, the charge distribution around the RGD motif of hirudisin may be responsible for lower disintegrin activity with respect to the snake venom proteins. However, some conformational contribution to the integrin-directed cell recognition is indicated by the 3-to 5-fold better inhibition of ADPinduced platelet aggregation of hirudisin compared to cmhirudisin and RGD peptides.
Since RGD-containing peptides have been shown to block binding of fibrinogen to GP IIb-IIIa and to prevent formation of platelet thrombin, these peptides appear to be promising candidates as antiplatelet agents (Kieffer and Phillips, 1990;Phillips et al., 1991). However, a possible therapeutic application of RGD-containing peptides requires distinct specificity because a large number of integrins are known to bind RGD peptides (Kieffer and Phillips, 1990) which may increase the risk of side effects. The snake venom disintegrin barbourin has been found to exhibit a G P IIb-IIIa-directed specificity due to a KGD motif instead of the RGD sequence (Scarborough et al., 1991). Substitution of Lys for Arg in the disintegrin eristociphin converts a nonselective into a G P IIb-IIIa-specific ligand and demonstrates that specificity is achieved merely by a single amino acid replacement (Scarborough et al., 1991). In order to gain such specificity we replaced Arg"2 by Lys to generate hirudisin-1. However, careful interpretation of our results (maximum concentration of hirudisin-1 in ADP-induced platelet aggregation assay was 120 PM) shows that this inhibitor is an approximately &fold less active antagonist of GP IIb-IIIa than hirudisin. This finding is in agreement with earlier studies that have shown that peptides containing a KGD sequence are much weaker antagonists of integrins (Ginsberg et al., 1985;Scarborough et al., 1991).
One purpose of this study was the generation of a targeted hirudin. Hirudisins are designed to bind to platelet surfaces via their cell adhesion sequence, simultaneously acting as disintegrins and as inhibitors of thrombin functions just at the site where thrombin is generated. These proteins are considered to be potential anticoagulants with enhanced pharmacological properties. However, to avoid bleeding complications due to hirudisin's strong thrombin inhibitory activity, its RGD motif should exhibit disintegrin function in the same therapeutic window of 10-100 nM in human plasma as has been proposed for hirudin (Markwardt, 1989). In order to increase hirudisin's disintegrin activity and specificity we will focus on engineering the charge distribution around the RGD adhesion sequence as well as on the complete replacement of hirudisin's finger-like structure by disintegrin-related loops.