Distinct binding sites of Ala48-hirudin1-47 and Ala48-hirudin48-65 on alpha-thrombin.

The interaction of alpha-thrombin with Ala48-hirudin, Ala48-hirudin1-47, and Ala48-hirudin48-65 was analyzed. Mutations at Pro48 were found to cause only slight changes in the kon (human: 3.1 +/- 0.3 x 10(8) M-1 s-1; bovine: 1.03 +/- 0.3 x 10(8) M-1 s-1) and koff (human: 0.4 +/- 0.2 x 10(-3) s-1; bovine: 2.9 +/- 0.4 x 10(-3) s-1) rate constants for the formation of the thrombin-hirudin complex. The amino-terminal fragment Ala48-hirudin1-47 containing the three disulfide bridges and the carboxyl-terminal fragment Ala48-hirudin48-65 were derived from the Ala48 mutant by proteolysis with endoproteinase Lys-C. These fragments inhibit bovine alpha-thrombin clotting activity with IC50 values of 0.6 and 4.9 microM, respectively (2.4 nM for r-hirudin). By mapping the interaction of Ala48-hirudin-derived fragments with bovine alpha-thrombin by limited proteolysis with trypsin and pancreatic elastase distinct binding sites for each fragment were determined. The carboxyl-terminal fragment was found to bind to the proposed anion-binding exosite in the region B62-74, whereas the amino-terminal fragment binds to a region around the elastase cleavage site at residues 150-151 of the alpha-thrombin B-chain.

The amino-terminal fragment Ala4*hirudin '-47 containing the three disulfide bridges and the carboxyl-terminal fragment Ala48-hirudin48-6S were derived from the Ala4' mutant by proteolysis with endoproteinase Lys-C. These fragments inhibit bovine cy-thrombin clotting activity with I&, values of 0.6 and 4.9 FM, respectively (2.4 nM for r-hirudin). By mapping the interaction of Ala4'-hirudin-derived fragments with bovine cY-thrombin by limited proteolysis with trypsin and pancreatic elastase distinct binding sites for each fragment were determined. The carboxylterminal fragment was found to bind to the proposed anion-binding exosite in the region B62-74, whereas the amino-terminal fragment binds to a region around the elastase cleavage site at residues 150-151 of the cY-thrombin B-chain.
The thrombin-specific inhibitor hirudin is a polypeptide of 65 or 66 amino acid residues isolated from the leech Hirudo medicinalis (Markwardt, 1970;Bagdy et al., 1976;Dodt et al., 1986a;Tripier, 1988). Sequence analysis of different naturally occurring forms of hirudin revealed homologies of about 80%. The 6 cysteine residues were assigned to three disulfide bonds connecting Cys' with Cys14, Cys" with Cys", and Cys" with Cys3' (Dodt et al., 1985). The three-dimensional structure of hirudin in solution has been determined using 'H two-dimensional NMR (Clore et al., 1987;Haruyama and Wiithrich, 1989) and has been refined by analyzing r-hirudin and the Lys47 --, Glu mutant (Folkers et al., 1989). These studies indicate that hirudin consists of an amino-terminal compact domain (residues l-49) held together by the three disulfide linkages and a distorted carboxyl-terminal tail (residues 50-65) which does not fold back on the rest of the protein.
Hirudin reacts rapidly with a-thrombin (k,, = 1.3-7 x 10' M-' s-') forming tight, noncovalent complexes (& = lo-l4 M with human a-thrombin, Stone and Hofsteenge, 1986; Kd = lo-'* M with bovine a-thrombin, Dodt et al., 1988). The specificity of the interaction of ol-thrombin with macromolec-* This work was supported by the Deutsche Forschungsgemeinschaft (to J. D. and T. S.) and a personal grant of the Deutsche Gesellschaft fiir Chemisches Apparakwesen, Chemische Technik, und Biotechnologie (to S. K.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$. To whom correspondence should be addressed. ular substrates is assumed to reside in interactions at three distinct regions: the primary binding pocket for the P1 residue (Magnusson, 1971), an apolar binding site adjacent to the catalytic site (Berliner and Shen, 1977;Sonder and Fenton, 1984), and an anion-binding region (Fenton et al., 1989) which may be responsible for the specific interaction of thrombin with fibrinogen (Fenton et al., 1988, Henriksen andMann, 1988). As thrombin usually cleaves adjacent to arginine it was believed that a basic amino acid residue in hirudin binds in the primary binding pocket. The region around the Lys-Pro peptide bond at positions 47-48 displays strong homology with the a-thrombin cleavage site in prothrombin (Petersen et al., 1976). As other proteinase inhibitors of serine proteinases become bound in a substrate-like manner to the active site of their target proteinase, the prothrombin-like region of residues 40-48 could perform substrate-like interactions with thrombin.
Attempts have been made by site-directed mutagenesis of a hirudin gene to determine the basic amino acid residue which interacts with the primary binding pocket (Braun et al., 1988;Dodt et al., 1988). The results indicate, however, that hirudin does not require interactions in the primary specificity pocket to form tight complexes with thrombin. Nevertheless, the observed minor differences in Kd of complexes of Ly@-hirudin mutants with a-thrombin correlate with the decreased in uivo antithrombotic efficiency of these mutants (Degryse et al., 1989). As the carboxyl-terminal region of hirudin is especially rich in acidic amino acid residues it may therefore be involved in the interaction with the anion-binding exosite of the enzyme. The importance of an intact carboxyl-terminal region for inhibitory activity has been demonstrated by successive enzymatic removal of the acidic tail (Chang, 1983;Dodt et al., 1987). Dependence of the association rate constant for thrombin-hirudin complex formation on the acidic nature of the carboxyl-terminal region was observed with site-specific hirudin mutants (Braun et al., 1988). Peptide synthesis of hirudin45-65 confirmed the binding of the carboxyl-terminal region to a noncatalytic site of athrombin and inhibition of thrombin-mediated clotting, albeit with a specific anticoagulant activity 3-4 orders of magnitude lower than intact hirudin. Amino acid residues 56-65 and 53-64 were found to be minimally required for this action (Krstenansky and Mao, 1987;Mao et al., 1988;Maraganore et al., 1989). These studies as well as kinetic analysis of hirudin-thrombin interaction have revealed two or more binding sites in complex formation including at least one that involves substrate competition (Stone and Hofsteenge, 1986;Degryse et al., 1989).
In this paper we describe the generation of Ala4'-hirudin'm47 containing three disulfide bonds and Ala48-hirudin48-65 by changing Pro4' to alanine by site-directed mutagenesis. The experiments were performed (a) to examine the importance of the amino acid residue adjacent to the putative reactive site, and (b) to dissect hirudin inhibitory activity and In experiments without inhibitor, but with varying enzyme concentrations (l-50 nM), the clotting time for infinite enzyme concentration was calculated from linear regression analysis of plots of clotting time uersus reciprocal thrombin activity .

Oligonucleotide-directed
Mutagenesis-The Pro4' + Ala mutation was constructed by the gapped duplex technique according to Kramer and Fritz (1987). The oligonucleotide GGGGCTTTCGCGTCAGAG was used as mutagenic primer to change the CCG codon for Pro4' to the GCG codon for alanine.
This oligonucleotide was hybridized to the gapped duplex DNA, and the gap was filled in with T4 DNA polymerase and Escherichia coli ligase. After transfection of E. coli BMH71-18mutS with this construct and propagation of recombinant phages, wild-type phages to which the mutation is coupled were selected in E. coli MK30-3. Mutant phages were screened by dot-blot hybridization (Zoller and Smith, 1983) using the mutagenic primer. The mutation was confirmed by sequencing single-stranded DNA according to Sanger et al. (1977). The Hind111 hirudin gene fragment from the replicative form of MlSmpSwt containing the mutation was inserted into the Hind111 site of the expression vector (Dodt et al., 198613) and the resultant plasmid was used to transform E. coli BMH71-18.
Protein Purifications-Prothrombin (human and bovine) was isolated as previously described (Mann, 1976) and the prothrombin forms were converted to a-thrombin using the venom of Oxyuranus scutellatus (Owen and Jackson, 1973). The purity was ascertained by 10 steps of Edman degradation, and a-thrombin concentrations were determined by active-site titrations (Jameson et al., 1973). Recombinant Ala@-hirudin was directed to the periplasmic space of E. coli by the signal peptide of alkaline phosphatase and isolated as previously described for r-hirudin (Dodt et al.. 198613 (Morrison, 1982;Morrison and Stone, 1985).
The symbols uO, u,, and k., represent the initial velocity, steady-state velocity, and an apparent first-order rate constant; d is a displacement term to account for the fact that at t = 0 the fluorescence may not have been known accurately.
The values of k,, were plotted against the inhibitor concentration, weighted to the inverse squares of their standard errors, and fitted to Equation 2 to obtain k,, and k,ff (Morrison, 1982;Morrison and Stone, 1985 The inhibitor concentration which inhibits half of the enzyme activity represents the ICsO value. These values were obtained from plots of clotting times (ct) uersus inhibitor concentration  using Equation 3.
Chc, = 2 x (cto -CL) + ct, The abbreviations cto and ct, represent clotting times without inhibitor and for infinite enzyme concentrations, respectively.

Effect of PTO~~ + Ala Mutation in the Hirudin-Thrombin Interaction-Recombinant
Ala4'-hirudin was isolated by ionexchange chromatography and reverse-phase HPLC. Correct cleavage of the signal peptide was confirmed by sequencing (10 steps of Edman degradation) and determination of amino acid composition.
The data were found to be in agreement with those expected for Ala4'-hirudin.
The results of slow binding inhibition studies comparing the interaction of natural hirudin, r-hirudin, and Ala@-hirudin with human and bovine Lu-thrombin are summarized in Table  I. Previously, we were able to determine the association rate constant for complex formation by native hirudin, r-hirudin, and r-hirudin mutants with bovine ol-thrombin in slow binding inhibition experiments (Dodt et al., 1988). Enzyme concentrations from 10 to 20 pM were applied. In the present study, we also present data of the reaction of human (Ythrombin with native and r-hirudins (Table I)   (1986) and Braun et al. (1988) in slow, tight-binding inhibition experiments.
In general, the&values for human a-thrombinhirudin complexes were 3.9-21.5fold lower than those obtained for the complexes of hirudin with the bovine enzyme and, therefore, 2-fold lower human a-thrombin concentrations were applied to get progress curves for precise ka,, evaluations ( Fig. 1). Hirudin from leeches contains a sulfated tyrosine residue in position 63 (Petersen et al., 1976). Recombinant hirudin isolated from E. coli is lacking this structural element (Dodt et al., 1986b;Fortkamp et al., 1986;Loison et al., 1988, Braun et al., 1988 resulting in a lo-fold lower affinity of human and bovine a-thrombin for r-hirudin. Mutation of Pro4' to alanine positively influenced the affinity of human a-thrombin (& = 1.3 & 0.6 x 10-l' M) and slightly decreased the affinity of bovine ol-thrombin (& = 28 f 4 X lo-l2 M). Comparison of the kinetic data for the reaction of r-hirudin and Ala4'-hirudin with human a-thrombin indicates that the 3.6-fold increased affinity was due to a 1.6-fold increase in the rate of association and a 2.2-fold decrease in the rate of dissociation.
The 1.9-fold decreased affinity of bovine a-thrombin for Ala"'-hirudin compared to r-hirudin was caused predominantly by the 1.9-fold increase of the dissociation rate constant. These data indicate that the Pro4* ---, Ala mutation cause only minor differences in Kd of thrombin-hirudin complexes. However, based on the present knowledge, the reason for quite the reverse effect on Kd of the reaction of Ala4'-hirudin with the two thrombin species is not obvious.

Isolation
of Ala4X-Hirudin-derived Fragments--It has been shown that incubation of r-hirudin with trypsin or endoproteinase Lys-C does not lead to the cleavage of any of the three lysyl peptide bonds. However, it was possible to separate two peptides by reverse-phase HPLC from incubation mixtures of Ala4'-hirudin with endoproteinase Lys-C (Fig. 2). These peptides were rechromatographed twice and analyzed for amino acid composition and were sequenced 10 steps by Edman degradation.
The peptide eluting at 19.4 min corresponds to residues l-47 containing the three disulfide linkages and the amino-terminal valine, whereas the peptide eluting at 30.9 min contains positions 48-65 starting with an alanine.

Anticoagulant Activities of Ala'H-Hirudin-derived
Fragments-Unsulfated N"-acetylhirudin45-65 has been shown to inhibit fibrinogen cleavage by binding to a noncatalytic site of thrombin (Krstenansky and Mao, 1987;Mao et al., 1988;Maraganore et al., 1989). Therefore we analyzed the inhibitory activity of the Ala4"-hirudin fragments in a clotting assay using purified bovine a-thrombin and fibrinogen.  Ala'*-Hirudin'." 0.6 /.LM Ala'X-Hirudin'R-G' 4.6 PM Ala""-hirudin-derived fragments bind on oc-thrombin and to localize the binding regions of cu-thrombin in these fragments we mapped cu-thrombin and the complexes of cy-thrombin with r-hirudin, cu-thrombin with Ala4R-hirudin'-47 and LYthrombin with Ala48-hirudin48-6' with either trypsin or pancreatic elastase. For these experiments we incubated 1.6 pM a-thrombin with 35 nM trypsin and 400 nM elastase, respectively. These concentrations of trypsin and elastase completely hydrolyzed Lu-thrombin at one predominant peptide bond in 1 h at 37 "C as was predetermined by SDS gel electrophoresis from samples of a dilution series of the appropriate enzyme with 1.6 pM oc-thrombin (Fig. 3, A and B, lane 2). To identify the cleavage sites, digests of 1 nmol of (Ythrombin with either trypsin or elastase were separated by SDS gel electrophoresis under nonreducing conditions. The proteins, indicated by arrows in Fig. 3, were blotted onto Immobilon membranes for 3 h at a constant current of 1 mA/ cm'. The Coomassie-stained proteins bands were subjected to Edman degradation in a gas-phase sequenator. The sequencing results are provided in Table III The SDS-7 standard proteins (1.5 pg each) were also run under nonreducing conditions. The fragments obtained by hydrolysis of thrombin with elastase or trypsin are indicated by Al and A2 or Bl and B2, respectively.
These were blotted and subjected to sequence analysis.
already published by Sonder and Fenton (1986). By sequencing peptides Al and A2 elastase was found to hydrolyze CYthrombin at position 150-151 of the cu-thrombin B-chain as previously has been reported for the human enzyme (Kawabata et al., 1985;Fenton and Bing, 1986). In summary, the anion-binding exosite of cu-thrombin is defined by the trypsin cleavage site; and, based on a cu-thrombin B-chain model , the elastase cleavage site defines a region close to the fibrin groove.
In forming the a-thrombin-hirudin complex, the observed elastase as well as trypsin cleavage sites were inaccessible to hydrolysis by the corresponding proteinase ( Fig. 3, lanes   3). Mapping the complexes of n-thrombin and Ala4*-hirudinderived fragments for elastase and trypsin-sensitive sites we were able to determine distinct binding sites of these fragments on a-thrombin.
In the complex with the amino-terminal fragment Ala4s-hirudin'-47, cu-thrombin is shielded completely for cleavage by elastase (Fig. 3A, lane 4), whereas in the complex with the carboxyl-terminal fragment Ala4"-hirudin4"-'" the rate of elastase hydrolysis appeared to be diminished (Fig. 3A, lane 5). On the other hand, u-thrombin is not accessible to trypsin in the complex with Ala4R-hirudin4R-G" and the rate of trypsin cleavage appeared to be diminished for the complex with the amino-terminal fragment Ala4R-hirudin'-4'.
In summary, considering the I& values, the thrombin-hirudin interaction comprises at least two binding IVEGQDAEVG---B-chain starting with I-B1 a All protein bands were sequenced 13-15 steps.
sites with distinct affinities. The results indicate direct interactions of the COOH-terminal peptide Ala4s-hirudin4a65 with the proposed anion-binding exosite and interactions between the amino-terminal fragment Ala48-hirudin'-47 and a-thrombin in the region of the elastase cleavage site. The latter seems to represent the major interaction site which accounts for tight complexing.

DISCUSSION
The experiments presented here have separated two hirudin binding sites on cy-thrombin. Two fragments, Ala48-hirudin'-47 and Ala4s-hirudin48-65, were derived from the Ala4'-hirudin mutant. These were shown to combine with cy-thrombin at distinct sites, inhibiting clot formation. However, the thrombin inhibition activity of either fragment is 3OO-2OOO-fold smaller than that of the intact mutant and r-hirudin.
Evidence for the existence of an anion-binding exosite which determines thrombin specificity has been summarized (Fenton, 1981Fenton et al., 1989). Alignment of LYthrombin and a-chymotrypsin sequences shows that the cleavage site producing @-thrombin is located within a sequence in which 6 additional basic amino acid residues have been introduced into bovine a-thrombin. This sequence corresponds to the surface loop 63-85 in a-chymotrypsin (Birktoft and Blow, 1972) and represents the anion-binding exosite. The 100-200-fold increase of the Kd value for the complex of hirudin with P-thrombin suggests that the region around the P-cleavage site is important for binding hirudin (Stone et al., 1987). Additional evidence for the importance of this region was obtained with antibodies to the 62-73 region of the human cY-thrombin B-chain which acts as a competitive inhibitor of the thrombin-hirudin interaction (Noe et al., 1988). On the inhibitor side, the carboxyl-terminal portion of rhirudin containing 6 negatively charged residues has been implicated in binding to the anion-binding exosite of 01thrombin (Fenton, 1981). Evidence derives from mutagenesis of these acidic residues of hirudin which lowers the association rate constant for complex formation concomitant with lowering the charge of this region (Braun et al., 1988). In addition, chemically synthesized peptides of the carboxyl-terminal region were shown to recognize thrombin and to inhibit thrombin-catalyzed conversion of fibrinogen to fibrin. As the synthetic peptides only exhibit anticoagulant activity and do not inhibit hydrolysis of small peptide substrates, it was concluded that the hirudin-related peptides do not block the active site of a-thrombin (Krstenansky and Mao, 1987;Mao et al., 1988;Maraganore et al., 1989). Clotting inhibition by these peptides then must result from binding to a noncatalytic site on o(thrombin.
Indeed, the present study has demonstrated that the carboxyl-terminal part Ala48-hirudin48-65 and residues 62-74 of the thrombin B-chain combine in thrombin-hirudin complex formation.
The recognition between these sites appears therefore to be caused mainly by a complementary surface of charged amino acid residues.
However, considering the high affinity of hirudin for thrombin and the inhibition of the enzyme's catalytic as well as noncatalytic functions by intact hirudin it is obvious that the thrombin-hirudin interaction comprises an extended binding region not restricted to the anion-binding exosite. The work of Degryse et al. (1989) has identified a mixed kinetic mechanism for the thrombin-hirudin interaction indicating at least two binding sites. One of these appears to be involved in substrate competition.
The present study provides evidence that the amino-terminal fragment Ala48-hirudin'-47 binds to a distinct site on a-thrombin.
This site is located around the elastase cleavage site (Thr-B150) of the bovine enzyme corresponding to surface peptides in a-chymotrypsin (Fenton and Bing, 1986). The inhibitory activity of Ala48-hirudin'-47 may then be caused either directly by blocking a fibrinogen recognition site or indirectly by induction of a conformational change in thrombin leading to a masked binding site. Since we observed an additional inhibitory effect of Ala4'-hirudin'-47 on the hydrolysis of Tos-Gly-Pro-Arg-p-nitroanilide' it seems possible that the fragment interacts with the active site groove of a-thrombin.
An intriguing observation is the decreased rate of a-thrombin cleavage in complexes with Ala4s-hirudin'-47 by trypsin and in complexes with Ala48-hirudin48-65 by elastase. The latter may reflect a conformational change of the enzyme, as has been observed in CD spectra of complexes of cr-thrombin with hirudin and hirudin-related carboxyl-terminal peptides (Konno et al., 1988;Mao et al., 1988). Binding of the aminoterminal fragment Ala48-hirudin'-47 may also be associated with a significant conformational change of the enzyme displayed by the reduced trypsin cleavage rate.
The kinetic data for complex formation of hirudin with human and bovine cu-thrombin (Table I) show a reduced affinity of hirudin for the bovine enzyme. Recent analysis of anticoagulant activities of synthetic hirudin-related carboxylterminal fragments toward human and bovine cY-thrombin (Maraganore et al., 1989) has identified a lo-fold increased potency toward human a-thrombin.
Thus it may be speculated that the observed species differences in complexes of hirudin with human and bovine cY-thrombin are attributed exclusively to differences in the interaction with the carboxyl-terminal hirudin region.
The results presented here are consistent with the idea that hirudin binds to thrombin via an extended binding region by hydrophobic interactions in conjunction with general electrostatic complementarity.
There is striking evidence that hirudin's carboxyl-terminal tail shares the anion-binding exosite with other macromolecular substrates of thrombin and that the compact amino-terminal domain interacts with the active site. All these interactions together block a wide range of thrombin functions.
With the Ala4&-hirudin-derived fragments at hand tools for probing the interaction of thrombin with physiological substrates are available.