A strong thrombin-inhibitory prourokinase ... - Semantic Scholar

6 downloads 0 Views 357KB Size Report
(1994) Deutsches Patenamt, Patentschrift DE4323754C1 to a decreased binding of M35 ... Patentamt, Offenlegungsschrift DE4442665. specificity which may be ...
peng$$0109

Protein Engineering vol.10 no.2 pp.169–173, 1997

A strong thrombin-inhibitory prourokinase derivative with sequence elements from hirudin and the human thrombin receptor

Stephan Wnendt1, Elke Janocha, Gerd J.Steffens2 and Wolfgang Strassburger3 Department of Molecular Pharmacology and 3Department of Theoretical Chemistry, Gruenenthal Centre of Research, Zieglerstr. 6, D-52078 Aachen and 2Marienhof Laboratorien, Moenchengladbach, Germany 1To whom correspondence should be addressed

In order to design plasminogen activators with improved thrombolytic properties, bifunctional proteins with both plasminogen-activating and anticoagulative activity were constructed by fusing a thrombin-inhibitory moiety to the carboxy-terminus of a prourokinase derivative lacking the growth-factor domain. The thrombin-inhibitory moiety itself comprises four elements: linker 1, a motif directed to thrombin’s active site, linker 2 and a fragment of hirudin which binds to the fibrinogen-recognition site of thrombin. In order to improve further the anticoagulative activity, the thrombin-inhibitory domain was modified by substituting linker 2. Introduction of a linker (FLLRNP) from the human thrombin receptor conferred about a 10-fold increase in anticoagulative activity in protein M37 compared with the parent molecule M23 carrying an aliphatic linker. The increase in anticoagulative activity was also reflected in the shift of the Ki value from 159 K 20 nM for M23 to 2.0 K 0.5 nM for M37. The increased thrombininhibitory activity of M37 may be due to the presence of an arginine in the linker from the thrombin receptor which may interact with one of two glutamic acid residues located at the exit of the thrombin substrate binding pocket. This explanation is supported by the observation that another chimera (M35) carrying a linker sequence with two acidic residues has relatively weak thrombin-inhibitory activity. The thrombin-inhibitory activity of M37 may be strong enough to substitute anticoagulative co-medication during fibrinolytic treatment. Keywords: acute myocardial infarction/fibrinolysis/prourokinase/stroke

Introduction The treatment of occlusive disorders, in particular acute myocardial infarction, has been significantly improved by the introduction of streptokinase, urokinase-type (uPA) and tissuetype (tPA) plasminogen activators (for recent reviews, see Bachmann, in press). Both uPA and tPA convert plasminogen into plasmin by a specific cleavage between Arg560 and Val561. Plasmin then induces clot lysis by proteolytic degradation of fibrin. However, several lines of experimental and clinical evidence indicate that the thrombogenic activity of thrombin must be inhibited by anticoagulative co-medication during thrombolytic treatment in order to achieve acceptable potency rates and to avoid early reocclusion (Schneider, 1991; Klement et al., 1992; Yao et al., 1992). The most common © Oxford University Press

anticoagulative agent is heparin; however, more potent compounds such as recombinant hirudin or a synthetic thrombin-inhibitory peptide named Hirulog® are under clinical investigation. In order to improve the therapeutic profile of recombinant single-chain urokinase-type plasminogen activator (rscu-PA), we designed chimeric rscu-PA derivatives with intrinsic thrombin inhibitory activity that showed a higher clot specificity than prourokinase itself, probably through an increased affinity to the thrombus (Schneider et al., 1996; Wnendt et al., 1996). These bifunctional chimeras comprise the kringle and protease domain of human rscu-PA fused via the carboxyterminus end to a thrombin inhibitory sequence (Figure 1). The thrombin inhibitory element is built up from the aminoterminus to the carboxy-terminus by four sequence elements: a pentameric amino acid sequence called linker 1, a sequence directed to the thrombin active site, a penta- or hexameric sequence designated linker 2 and a sequence derived from the carboxy-terminal region of hirudin (Dodt et al., 1985; Rydel et al., 1990) or from the human thrombin receptor (Vu et al., 1991a), respectively, which both bind to the fibrinogen recognition site (FRS) of thrombin (Figure 2). In a previous report, it was shown that three of the four building blocks contribute to the anticogulative and thrombin-inhibitory activity of the chimeric rscu-PA derivatives: the active site-directed sequence, the FRS-binding sequence and, to a surprisingly high extent, also linker 1 (Wnendt et al., 1996). The role of linker 2 appeared to be less important. The current model of the complex between thrombin and synthetic inhibitory peptides like Hirulog® which contain an active site-

Fig. 1. View of a three-dimensional model of M37. The conformation of the kringle (pink) was taken from an NMR structure analysis of the aminoterminal fragment of prourokinase (Hansen et al., 1994). Coordinates from an X-ray structure analysis were used for the protease domain (yellow) (Spraggon et al., 1995). The structure of the thrombin inhibitory domain (purple) has been modelled in an arbitrary conformation using WHATIF.

169

S.Wnendt et al.

ing the desired amino acid sequences. The modified BamHI/ HindIII fragments were then inserted back into pSJ41 or pSJ95, yielding the respective expression plasmid. All modifications and the flanking regions were verified by DNA sequencing. A detailed description of the plasmid construction was given by Wnendt et al. (1996a). Recombinant expression, refolding of material from inclusion bodies and purification of the chimeric proteins were performed as described previously (Wnendt et al., 1996b). Protein concentrations were determined by using the BCA assay (Pierce, Oud Beijerland, The Netherlands). All proteins were analysed by SDS–PAGE (10% polyacrylamide) under reducing conditions.

Fig. 2. Alignment of amino acid sequences from the C-terminus of the protease domain (Lys358 to Leu365) and the thrombin-inhibitory domains of the chimeric rscu-PA derivatives or peptides. Gaps are marked by (–).

directed sequence fused via a four to six residues long linker to the FRS-binding sequence from hirudin suggests that the linker is not interacting with thrombin (DiMaio et al., 1990; Maraganore et al., 1990). The function of this linker sequence is thought to be merely the proper positioning of the active site directed-sequence and the FRS-binding sequence (Yue et al., 1992). Indeed, based on an X-ray structural analysis of a complex between Hirulog® and thrombin, the electron density for the linker is very weak, indicating no defined interactions between the pentaglycinyl linker and thrombin (SkrzypczakJankun et al., 1991). However, it should be taken into account that glycine has no side chain prone to molecular interactions. A similar model to that for the interaction between Hirulog® and thrombin was suggested for the binding of thrombin to the human thrombin receptor which contains a cleavage site for thrombin separated by a linker from an FRS-binding sequence (Vu et al., 1991b). The chimeric rscu-PA derivative M32 (Figure 2) which carries the linker and FRS-binding sequence from the human thrombin receptor displayed a relatively good anticoagulative activity compared with other chimeric rscu-PA derivatives containing an artificial spacer and the FRS-binding sequence from hirudin (Wnendt et al., 1996). This observation led to the assumption that the amino acid sequence linking the cleavage site and the FRS-binding sequence of the thrombin receptor may contribute to the interaction between thrombin and the chimeric rscu-PA derivative M32. In order to assess the role of linker 2 and to improve further the thrombin-inhibitory activity of the chimeric rscuPA derivative M23, we constructed three new chimeras in which linker 2 and also the FRS-binding sequence were modified by using sequences from the thrombin receptor and hirudin, respectively. Materials and methods Cloning and expression of chimeric rscu-PA derivatives All expression vectors were constructed by modifying pSJ41 and pSJ95 (Steffens et al., 1994). The BamHI/HindIII fragments of pSJ41 and pSJ95, respectively, were subcloned into pUC18 (Pharmacia, Freiburg, Germany) and modified by insertion of synthetic double-stranded oligonucleotides encod170

Preparation of the synthetic peptides Peptides M23P and M37P were synthesized and purified by NeoSystems (Strasbourg, France). Both preparations were .95% homogeneous based on HPLC analysis. The correct amino acid sequence was verified by automatic sequence analysis using an ABI sequencer (ABI, Weiterstadt, Germany). Determination of specific urokinase and anticoagulative activity The specific urokinase activity was assayed using the chromogenic substrate S-2444 (L-pyroglutamyl-L-phenylalanyl-Llysine-p-nitroanilide; Chromogenix, Antwerp, Belgium) as descibed previously (Wnendt et al., 1996b). The anticoagulative activity of the chimeric rscu-PA derivatives was determined by using a standard thrombin time assay: human citrate-plasma was diluted 10-fold with veronal buffer (Boehringer-Mannheim, Mannheim, Germany) and kept at 37°C for at least 5 min. A 200 µl volume of this dilution was mixed with 50 µl of solution containing the rscu-PA sample diluted in veronal buffer. The reaction was started by adding 50 µl of solution containing 0.2 NIH units of human thrombin (BoehringerMannheim) in veronal buffer and the coagulation time was measured automatically by a coagulometer (Sarstedt, Sarstedt, Germany). Each thrombin time determination was performed in duplicate. Determination of Ki values for inhibition of thrombinmediated cleavage of chromogenic substrates The Ki values were determined using S-2238 (D-phenylalanylL-pipecolyl-L-arginine-p-nitroanilide; Chromogenix) or L-1150 (benzoyl-L-phenylalanyl-L-valinyl-L-arginine-p-nitroanilide; Bachem, Heidelberg, Germany) at concentrations of 20, 30, 40, 60, 80 and 200 µM with 0.04 IU thrombin in a 200 µl assay volume. The chimeric rscu-PA derivatives were tested at concentrations between 1.2 and 870 nM. Hirudin (American Diagnostics, Loxo, Dossenheim, Germany) was tested at concentrations between 7.5 and 750 pM. The buffer solution consisted of 90 mM NaCl, 45 mM Tris–HCl, containing 0.02% Triton X-100, pH 8. Each determination was performed in duplicate. The initial rate of the reaction was determined using an SLT microplate reader (405 nm) and evaluated with EasyKin-software (SLT, Crailsheim, Germany). Based on the molar absorption coefficient of p-nitroanilide at 405 nm of 8800 M–1 cm–1 (Erlanger et al., 1965), the enzyme velocities were calculated and further analysed by the method of Lineweaver and Burk (Segel, 1975). Inhibition constants (Ki values) and inhibition types were determined from Dixonplots of the kinetic data (Segel, 1975).

New thrombin-inhibitory prourokinase derivatives

Fig. 4. SDS–PAGE of M23, M32, M35, M37 and M38. Approximately 2 µg of each protein were separated on a 10% polyacrylamide gel. As a molecular size marker a calibration kit from Pharmacia was used (phosphorylase b, 94 kDa; bovine serum albumin, 67 kDa; ovalbumin, 43 kDa; carbonic anhydrase, 30 kDa).

Fig. 3. Primary structure of M37. Amino acids Ser1 to Leu365 (corresponding to Ser47–Leu411 of rscu-PA) are linked via a 15 amino acid linker sequence to amino acids Gly381 to Gln392 (corresponding to Gly54– Gln65 of hirudin). Within the 15 amino acid linker amino acids Phe275– Pro280 correspond to Phe43–Pro48 of the human thrombin receptor. The amino acids are represented by their single letter symbols and black bars indicate disulphide bonds. The active site residues His158, Asp209 and Ser310 (corresponding to His204, Asp255 and Ser356 in rscu-PA) are indicated with an asterisk and the arrow indicates the cleavage site for conversion into the two-chain form.

Molecular structure modelling Pictures were generated using the program package RASMOL v. 2.5 (R.Sayle, Glaxo, Greenford, UK) and WHATIF (Vriend, 1990). Results Three new chimeric rscu-PA derivatives (M37, M38 and M35) were constructed based on M23 and M32 (Figure 2). In the case of protein M37 the sequence FLLRNP which corresponds to amino acids Phe43 to Pro48 of the human thrombin receptor (Vu et al., 1991a) was used as linker 2. The entire primary structure of M37 is shown in Figure 3. In protein M38, linker 2 and part of the FRS-binding sequence of hirudin were substituted by the corresponding amino acid sequence from the human thrombin receptor. Thus, M38 carries a chimeric FRS-binding sequence comprising a fragment from the thrombin receptor and from hirudin. Finally, in protein M35 the synthetic LGGGG linker was substituted by the sequence ESHND which is derived from the sequence located between amino acids 49 and 54, N-terminal to the FRS-binding sequence, of hirudin rHV2-Lys47 (ESHNN; Rydel et al., 1990) and native hirudin (QSHND; Dodt et al., 1985). The three chimeric rscu-PA derivatives were constructed by modification of Escherichia coli expression plasmids pSJ95 and pSJ41, encoding M23 and another chimeric rscu-PA derivative, M12 (Steffens et al., 1994; Wnendt et al., 1996b), using conventional cloning techniques. The recombinant pro-

Table I. Specific activities of the chimeric rscu-PA derivatives determined with S-2444 after conversion into the two-chain form by digestion with plasmin rscu-PA derivative

Specific activity (Ploug units/mg protein)

M23 M35 M37 M38

123 128 159 135

300 500 000 000

teins were obtained as at least .90% homogeneous singlechain preparations with an Mr(app) of ~43 kDa (Figure 4). The specific amidolytic activities of the chimeric u-PA derivatives against the chromogenic substrate S-2444, which mimics the cleavage site of plasminogen recognized by u-PA, were found to be in the range 123 000–159 000 Ploug units/ mg (see Table I). These values were also expected based on previous work (Wnendt et al., 1996b). The anticoagulative activity was determined using the thrombin-stimulated coagulation of human plasma as test system. The prolongation of the time necessary to induce coagulation (i.e. thrombin time) is a measure for the anticoagulative activity. Figure 5A shows the effect of the three new chimeric rscu-PA derivatives and also their parent molecules M23 and M32 on thrombin time. Clearly there are considerable differences in the degree of anticoagulation between the five proteins: M35 was weaker in its anticoagulative activity than any of the other proteins tested and M38 showed an activity which was only slightly stronger than that of its parent M32. Thus, by introducing the sequence ESHND from hirudin as linker 2 in M35 or by constructing a composite FRS-binding sequence with elements from hirudin and the thrombin receptor in M38, the anticoagulative activity could not be increased above the level reached by M23 and M32. However, a surprisingly strong anticoagulative effect was observed with M37. The concentration of M37 that induces a half-maximal prolongation of thrombin time (i.e. 150 s) is 171

S.Wnendt et al.

Fig. 5. Effect of chimeric rscu-PA derivatives and synthetic peptides on thrombin-stimulated plasma coagulation. The final concentration of the chimeras is given on the abscissa. Panel A: (u) M23; (s) M32; (j) M35; (d) M37; (m) M38. Panel B: (u) M23P; (d) M37P.

Table II. Ki values (means 1 s.d.) for chimeric rscu derivatives and recombinant hirudin determined with human thrombin and two different substrates Ki (nM) S-2238

L-1150 Hirudin M23 M32 M37

0.04 159 441 2.0

6 6 6 6

0.01 20 101 0.5

0.2 6 0.1 .2500 .2500 180 6 20

about 10-fold lower than that necessary for M23 (Figure 5A). These data suggest that the amino acid sequence between the cleavage site and the FRS-binding sequence of the human thrombin receptor used as linker 2 may contribute substantially to the thrombin-inhibitory activity of the chimeric rscu-PA derivative M37. The strong inhibitory activity of M37 was further confirmed in a direct assay of thrombin inhibition in which a Ki value of 2.0 1 0.5 nM for thrombin-mediated cleavage of the chromogenic substrate L-1150 was determined. This value is approximately two orders of magnitude lower than the Ki values found for the parent chimeras M23 and M32 and only 50 times higher than the inhibition constant of hirudin (Table II). Based on Dixon plot analysis it was found that M37 inhibits thrombin in a competitive fashion (not shown). The significant improvement in thrombin inhibitory activity was also demonstrated by the fact that M37, in contrast to M23, is able to inhibit the thrombin-mediated cleavage of S-2238 (Table II). S-2238 is another thrombin-specific chromogenic substrate has a much higher affinity for thrombin than L-1150, since the Km value of thrombin for S-2238 is 7 µM compared with 160 µM for L-1150 (supplier’s datasheet; Svendsen et al., 1972). Consequently, the Ki values of M37 and of hirudin obtained with S-2238 are higher; indicating that higher concentrations of the respective competitive inhibitor are necessary to displace the substrate from the active centre of thrombin. The inhibition of thrombin-mediated cleavage of S-2238 by M37 indicates that this chimeric rscu-PA derivitate has a substantially increased affinity to thrombin compared with its parent compound M23. In order to investigate further the unforeseen high thrombininhibitory activity of M37, two synthetic peptides were designed which carry the active-site-directed sequence, linker 2 and the FRS-binding sequence of either M23 or M37. In the case of the Hirulogs® it was found that a phenylalanine residue with D-configuration at the amino terminus is associated with a substantially higher thrombin inhibitory activity than L172

phenylalanine (Maraganore et al., 1990). Therefore, a D-Phe residue was also used as amino-terminal amino acid for the synthesis of the peptides M23P and M37P (Figure 2). Both peptides were tested with respect to their effect on thrombinstimulated plasma coagulation. M37P showed a stronger anticoagulative activity than M23P (Figure 5B); however, the difference between the two peptides was not so dramatic as observed between the corresponding proteins, M23 and M37. Comparing the molar concentrations of the corresponding peptides and proteins necessary to achieve a half-maximal prolongation of thrombin time (i.e. 150 s), the concentration was ~270 nM in case of the peptide M37P and only ~21 nM for the rscu-PA derivative M37 (0.9 µg/ml, corresponding to 21 nM). In the case of M23 and M23P the bias was towards the peptide M23P since it induced about a twofold stronger anticoagulation than the parent protein M23. A thrombin time of 150 s was reached with about 310 nM of the peptide M23P, while about 580 nM (25 µg/ml) of the protein M23 were necessary to induce the same degree of inhibition. These relations between peptide and protein may indicate that the very high thrombin-inhibitory activity of M37 is not only a question of an optimal sequence of the thrombin-inhibitory domain but is also due to a very good ‘presentation’ of this inhibitory domain by the larger rscu-PA moiety. Discussion The present study was carried out in order to evaluate whether specific ‘bridging’ sequences taken from hirudin or the human thrombin receptor may enhance the thrombin-inhibitory activity of the chimeric rscu-PA derivative M23 (Wnendt et al., 1996b). While no increase in inhibitory activity was observed when the sequence between the active site directed sequence and the FRS-binding sequence was substituted by a pentameric sequence derived from hirudin, a very significant and pronounced enhancement was conferred by introduction of the sequence FLLRNP as linker 2 in rscu-PA derivative M37. In contrast to M23, the new chimera is able to inhibit competitively the thrombin-mediated cleavage of S-2238, which has a relatively high affinity to thrombin. However, the enhancement of thrombin-inhibitory activity was less dramatic when the sequence from the human thrombin receptor was introduced in a chimeric bifunctional peptide with thrombin inhibitory activity. This may be explained by the introduction of the NH2-terminal phenylalanine with D-configuration, which confers a very high thrombin inhibitory activity to bifunctional peptides such as M23P or M37P (Maraganore et al., 1990) and, therefore, may superimpose the effect of the substitution of linker 2. Also, it may be conceivable that the protease domain supports a proper ‘presentation’ of the thrombin inhibitory moiety. The enhancing effect of the introduction of the sequence from the human thrombin receptor may be explained by a direct interaction of linker 2 with thrombin, which is in contrast to current models of interaction between thrombin and its receptor (Vu et al., 1991b). Based on the X-ray structure of thrombin, it is known that there are two glutamic acid residues (Glu39 and Glu192) positioned at the exit of the substrate binding pocket in the vicinity of the fibrinogen-recognition site (Stubbs and Bode, 1995; Figure 6). The basic arginine residue (Arg378) present in linker 2 of M37 may interact with one of these acidic residues and therefore facilitate the interaction of the inhibitory domain with thrombin. In addition, Glu39 and Glu192 may also account for the decreased thrombin

New thrombin-inhibitory prourokinase derivatives

or hirudin, which would simplify the treatment of patients with acute myocardial infarct. References

Fig. 6. View towards the exit of the east side of the substrate binding pocket of thrombin (Stubbs and Bode, 1995). Glu39 (right) and Glu (left) are highlighted [coordinates were taken from the X-ray structure of thrombin complexed with PPACK (Bode et al., 1989); chymotrypsinogen numbering is used].

inhibitory activity of M35, because this chimera carries two negatively charged amino acids (Glu375 and Asp379) in its linker 2 sequence. These acidic residues may therefore lead to a decreased binding of M35 to thrombin compared with M23 and M37 due to electrostatic repulsion forces. A more precise explanation of the interaction of M37 and M35 with thrombin may be possible after the establishment of an X-ray structure analysis of the complex between thrombin and the amino-terminal extracellular domain of the thrombin receptor. The rscu-PA derivative M23 was tested in two animal models for occlusive diseases and showed a remarkable thrombus specificity which may be explained by the high affinity of M23 for clot-bound thrombin (Schneider et al., 1996; Schneider et al., in press). However, these in vivo studies also revealed that the thrombin-inhibitory activity of M23 is not strong enough to substitute anticoagulative co-medication during thrombolysis. The increased thrombin-inhibitory activity achieved with M37 may be sufficient to fulfil this goal. Using the thrombin time assay it was demonstrated previously that the anticoagulative activity of M23 is ~20-fold lower than that of hirudin (Lijnen et al., 1995). However, a half-maximal prolongation of thrombin-time is achieved at a concentration of ~21 nM for M37 (see Results) and ~5 nM for hirudin (data not shown), indicating that the new chimeric rscu-PA derivative is only about fourfold weaker than hirudin. Since the plasminogen activating activity of M37 is very similar to that of Saruplase (about 150 000 Ploug units/mg) or M23 (see Table II), it may be assumed that the therapeutic dose for fibrinolytic therapy will be similar to that of Saruplase, which is about 1 mg/kg (Diefenbach et al., 1988). Recent data from clinical trials with hirudin as an adjunct during fibrinolytic therapy indicate that the initial dosis administered together with the plasminogen activator will be 0.1–0.2 mg/kg or even lower (Zeymer and Neuhaus, 1995). Therefore, when M37 is applied at a dose of 1 mg/kg, the anticoagulative activity may be sufficient to substitute the initial co-medication with heparin

Bachmann,F (ed.) (1996) Handbook of Experimental Pharmacology: Fibrinolytics and Antifibrinolytics. Springer, Heidelberg, in press. Bode,W., Mayr,I., Baumann,U., Huber,R., Stone,S.R. and Hofsteenge,J. (1989) EMBO J., 8, 3467–3475. Diefenbach,Ch. et al. (1988) Am. J. Cardiol., 61, 966–970. DiMaio,J., Gibbs,B., Munn,D., Lefebvre,J., Ni,F. and Konishi,Y. (1990) J. Biol. Chem., 265, 21698–21703. Dodt,J., Seemu¨ller,A., Maschler,R. and Fritz,H. (1985) Biol. Chem. HoppeSeyler, 366, 379–384. Erlanger,B.F., Kokowsky,N. and Cohen,W. (1965) Arch. Biochem. Biophys., 95, 271–278. Hansen,A.P., Petros,A.M., Meadows,R.P., Netterheim,D.G., Mazar,A.P., Olejniczak,E.T., Xu,R.X., Pederson,T.M., Henkin,J. and Fesik,S.W. (1994) Biochemistry, 33, 4847–4864. Klement,P., Borm,A., Hirsh,J., Maraganore,J., Wilson,G. and Weitz,J. (1992) Thromb. Haemostasis, 68, 64–68. Lijnen,H.R., Wnendt,S., Schneider,J., Janocha,E., Van Hoef,B., Collen,D. and Steffens,G.J. (1995) Eur. J. Biochem., 234, 350–357. Maraganore,J.M., Bourdon,P., Jablonski,J., Ramachandran,K.L. and Fenton,J.W.,II (1990) Biochemistry, 29, 7095–7101. Rydel,T.J., Ravichandran,K.G., Tulinsky,A., Bode,W., Huber,R., Roitsch,C. and Fenton,J.,II (1990) Science, 249, 277–280. Schneider,J. (1991) Thromb. Res., 64, 677–689. Schneider,J., Wnendt,S., Heinzel-Wieland,R., Saunders,D., Wilffert,B. and Steffens,G.J. (1996) Eur. J. Pharmacol., 302, 69–77. Schneider,J., Hauser,R., Hennies,H.-H., Korioth,J., Steffens,G. and Wnendt,S. Thromb. Haemostasis, in press. Segel,I.H. (1975) Biochemical Calculations, 2nd edn. Wiley, New York. Skrzypczak-Jankun,E., Carperos,V.E., Ravichandran,K.G., Tulinsky,A., Westbrook,M. and Maraganore,J.M. (1991) J. Mol. Biol., 221, 1379–1393. Spraggon,G., Phillips,C., Nowak,U.K., Ponting,C.P., Saunders,D., Dobson,C.M., Stuart,D.I. and Jones,E.Y. (1995) Structure, 3, 281–691. Steffens,G.J., Wnendt,S., Schneider,J., Saunders,D. and Heinzel-Wieland,R. (1994) Deutsches Patenamt, Patentschrift DE4323754C1 Stubbs,M.T. and Bode,W. (1995) Trends Biochem. Sci., 20, 23–28. Svendsen,L., Blombaeck,B., Blombaeck,M. and Olsson,P. (1972) Thromb. Res., 1, 267–278. Vriend,G. (1990) J. Mol. Graphics, 8, 52–56. Vu,T.-K.H., Hung,D.T., Wheaton,V.I. and Coughlin,S.R. (1991a) Cell, 64, 1057–1068. Vu,T.-K.H., Wheaton,V.I., Hung,D.T., Charo,I. and Coughlin,S.R. (1991b) Nature, 353, 674–677. Wnendt,S., Janocha,E., Steffens,G. and Heinzel-Wieland,R. (1996a) Deutsches Patentamt, Offenlegungsschrift DE4442665. Wnendt,S., Schneider,J., Janocha,E. and Steffens,G.J. (1996b) Protein Engng, 9, 213–223. Yao,S.-K. et al. (1992) Am. J. Physiol., 262, H374–H379. Yue,S.Y., DiMaio,J., Szewczuk,Z., Purisima,E.O., Ni,F. and Konishi,Y. (1992) Protein Engng, 5, 77–82. Zeymer,U. and Neuhaus,K.-L. (1995) Drug Safety, 12, 234–239. Received May 13, 1996; revised September 18, 1996; accepted September 30, 1996

173