Thrombin-Variable Region 1 (VR1) EVIDENCE FOR THE DOMINANT CONTRIBUTION OF VR1 OF SERINE PROTEASES TO THEIR INTERACTION WITH PLASMINOGEN ACTIVATOR INHIBITOR 1*

The importance of a specific variable region in dif- ferent serine proteases for the interaction with plasminogen activator inhibitor 1 (PAI-1) is studied. To that end, we have constructed a thrombin substitution variant, thrombin-VR1, in which the entire variable region 1 (VR1) of the protease domain (Phe-34 to Leu- 40) has been replaced by the corresponding sequence (Phe-294 to Phe-305) of tissue-type plasminogen ac- tivator. The substitution resulted in a 2000-fold increase of the second-order rate constant of inhibition by PAI-1 (kz= 2.2 % lo6 M” s-’) as compared to a-thrombin (k2= 1.1 x lo3 M” s-’). Inhibition of throm- bin-VR1 by PAI-1 is mediated by the formation of SDS-stable, enzyme-inhibitor complexes. The substi- tution did not affect the rate constant of inhibition by antithrombin 111, whereas clotting efficiency and the rate of inhibition by heparin cofactor I1 were decreased %fold. These results demonstrate the importance and specificity of the protease domain VR1 region for the interaction of PAI-1 with its target proteases.

The variable regions of the proteases are mainly arranged at the protein surface and include a number of surface-exposed loops (2). These loops have been suggested to contribute to the high degree of specificity, e.g the VR2 of a-thrombin contains the anion-binding exosite (residues 70-80, chymotrypsin numbering) (3) and has been implicated in the specific interaction with fibrinogen (4), hirudin (5), and the thrombin receptor on platelets (6). Furthermore, the interaction of PAI-1 with the plasminogen activators t-PA and urokinase (u-PA) is strongly dependent on the presence of the variable region VR1 on the protease domain of these enzymes (7,8). Both t-PA and u-PA contain a VR1 that is rich in positively charged amino acids, including the sequences KHRR (t-PA) and RRHR (u-PA). Deletion or charge reversal of these basic residues results in a more than 1000-fold reduction of the second-order rate constant of inhibition. Recently, we reported that a-thrombin is yet another target enzyme for PAI-1. Interestingly, efficient thrombin-inhibition by PAI-1 is only observed in the presence of either vitronectin (9) or heparin (10). Accordingly, we speculated on the requirement for these cofactors and the differences in the corresponding VR1 regions of a-thrombin and the plasminogen activators (11). This paper provides positive evidence for the dominant role of the VR1 region of serine proteases in the specific interaction with PAI-1. Finally, an explanation is presented for the function of the indicated cofactors in the inhibition of thrombin by PAI-1.
Proteins-Echis carinatus crude venom and human ATIII were from Sigma. Purified human HCII was a kind gift from Dr. R. Bertina (State University, Leiden, The Netherlands). Bowes melanoma t-PA (two-chain) was obtained from Biopool. Purified, active site-titrated human a-thrombin was from Dr. K. Mertens (this institute). Vitronectin was donated by Dr. K. T. Preissner (Bad Nauheim, Germany).
Purification of recombinant PAI-1 from Escherichia coli lysates, activation by guanidine-HCI, and titration of active inhibitor on active site-titrated t-PA have been described (9,11).
Construction of Prothrombin Variant VR1-The isolation of fulllength prothrombin cDNA from a human liver cDNA bank, employing standard polymerase chain reaction amplification with simultaneous introduction of an unique EcoRI site upstream and an XbaI site downstream of the prothrombin coding sequence, will be described elsewhere. The resulting cDNA was inserted as an EcoRI-XbaI fragment into the polylinker of the plasmid pcDNAl and entirely sequenced. The sequence was identical to the one previously reported (12). Mutagenesis was performed by the polymerase chain reaction overlap extension technique (13), using two, partially overlapping oligonucleotides with a 33-base pair mutated sequence (underlined) and a 30-base uair annealinn seauence at the 3'-end (5'-GCC

7
-G A A GCG C T C C C C G G G A C T G C G C C G G T G C T T G G C G A A A A G C A T CAC CTG CCA AGG TGA CAT GCC-3'). This Drocedure results in I the replacement of the amino~acid sequence RKSfiNEL (prothrombin residues 340-346) by AKHRRSPGERF (&PA residues 295-305). The mutated fragment was used to substitute the "wild-type" SstI-EglII Expression and Partial Purification of Recombinant Prothrombin Using the Recombinant Vaccinia Virus Expression System-Prothrombin and prothrombin-VR1 were expressed in a recombinant vaccinia virus expression system, essentially as described before (14). Prothrombin cDNA was inserted as an EcoRI-XbaI fragment into the vaccinia expression vector pATA-18, and thymidine kinase-negative recombinant virus was produced and selected as described (15). Purified recombinant virus stocks were assayed for prothrombin production by incubation of conditioned medium with E. carinatus venom. Subsequently, the amidolytic activity was determined using the chromogenic substrate S2238. CV-1 cells were grown to confluence and growth medium (Iscove's modified Dulbecco medium, 10% fetal bovine serum) was supplemented with 10 pg/ml vitamin K, one day prior to infection. Cells were washed with serum-free medium and infected for 2 h with recombinant vaccinia virus a t a multiplicity of infection of 5, after which the cells were washed twice with serumfree medium. Production was continued for 36 h in serum-free medium, supplemented with 10 pgfml vitamin K,. Conditioned medium was harvested and diluted with an equal volume of 10 mM Tris-HCI (pH 8.3), 30 mM EDTA, 5 mM benzamidine, and 0.1% (v/v) Triton X-100. Subsequently, diluted medium was applied to a Q-Sepharose Fast Flow column, equilibrated with 25 mM Tris-HCI (pH 8.3). The column was washed with 10 volumes of 25 mM Tris-HC1 (pH 7.5), 100 mM NaCl and, finally, recombinant (r-)prothrombin was eluted in one volume of 25 mM Tris-HC1 (pH 7.5), 500 mM NaCl. Analysis of the resulting, partially purified protein by SDS-PAGE, followed by silver staining, showed r-prothrombin to be the major band displaying a mobility equal to plasma prothrombin. Activation of r-Prothrombin to a-Thrombin-r-Prothrombin was activated for 30 min with E. carinatus venom (16), in 25 mM Tris-HCI (pH 7.81, 0.15 M NaCI, 10 mM CaCI2, 0.1% (v/v) Tween-80 a t 37 "C, and the reaction was terminated by adding EDTA to a final concentration of 15 mM. The mixture was applied to a heparin-Sepharose column, washed with 25 mM Tris-HC1 (pH 7.5), 0.25 M NaC1, 0.1% (v/v) Tween-80, and a-thrombin was eluted in one step with the same buffer, containing 0.6 M NaC1. Both under nonreducing and reducing conditions, we detected the same mobility by SDS-PAGE for r-thrombin, thrombin-VR1, and plasma a-thrombin. T h e concentration of active r-thrombin was determined by titration with hirudin (17), the concentration of which was first calibrated with active site-titrated a-thrombin.
Kinetic Data Analysis-Second-order rate constants of inhibition of thrombin by various serpins were determined at 37 "C in HST buffer (20 mM HEPES (pH 7.5), 150 mM NaCI, 0.1% (v/v) Tween-80 in polystyrene tubes that had been pretreated with 1% (w/v) polyethylene glycol 20,000. The inhibitors were preincubated a t 37 "C in 300 p1 of HST, and the reaction was started by the addition of 100 p1 of prewarmed 4 nM thrombin in HST. Final concentrations were: 1 nM thrombin and varying inhibitor concentrations and, when appropriate, 1 unit/ml heparin or 20 nM vitronectin. At various time intervals, 25-pl aliquots were withdrawn and assayed for residual thrombin amidolytic activity, after a 10-fold dilution in HST containing 1 mM S2238, a t 405 nm in a Titertek Twinreader. Progress curves were straight up to 30 min, indicating that the inhibition reaction was sufficiently quenched by this procedure. Control experiments without inhibitor indicated that no loss of thrombin activity occurred throughout the experiments. The second-order rate constants of inhibition (k2) were determined as follows: rate constants less than 10' M-' s-l were determined under pseudo first-order conditions, i.e. at initial inhibitor concentrations (I,,) of 50-250 times higher than enzyme concentration (Eo), from a linear plot of pseudo first-order rate constant of inhibition uersus I,. To determine rate constants larger than lo6 M" s-', enzyme and inhibitor were incubated a t equimolar concentrations ( E , = Io), in which case kinetics are second order according to the equation: l/Etl/Eo = K2.t (18). Thus, k, is obtained from a linear plot of 1/E,l/Eo uersus time. At intermediate rate constants, standard second-order kinetics was used with a 5-fold excess of inhibitor over enzyme, according to the equation: k, 3 t = 1/ (I, -E,).(ln(l + ((Io-Eo)/Et))ln(Io/Eo)) as previously described (9).
SDS-stable Complex Formation of Thrombin and Thrombin-VRI with Various Serpins-2.5 nM thrombin or thrombin-VRl (lz5I-labeled by the IODOGEN method) were incubated for 30 ,min a t 37 "C with a 2.2-fold excess of active serpins in HST, supplemented with 0.05% (w/v) bovine serum albumin. In addition, particular reactions contained 1 unit/ml heparin. Subsequently, an equal volume of sam-ple buffer (5% (w/v) SDS, 45% (v/v) glycerol, 0.05 M Tris-HC1 (pH 6.8), and 0.05% (w/v) bromphenol blue) was added and the mixtures were briefly heated a t 95 "C to quench the reaction. SDS-PAGE was performed on 7.5% (w/v) polyacrylamide gels (19), and the radiolabeled material was visualized by autoradiography.  Table I). Recombinant prothrombin and prothrombin-VRl were expressed upon infection of monkey kidney CV-1 cells with recombinant vaccinia virus. Both proteins were produced a t a level of approximately 4 gg/106 cellslday. Purification from the conditioned medium yielded proteins with a molecular weight identical to that of plasma prothrombin as judged by SDS-PAGE followed by immunoblotting (data not shown). Upon activation with E. carinatus venom these proteins were converted to meizo-thrombin and, similar to plasma prothrombin, subsequently to a-thrombin (16). Titration of the recombinant thrombins with hirudin showed no apparent differences in the dissociation constant as compared to plasma a-thrombin. Hence, hirudin was used as an active site titrant to calibrate the concentrations of the recombinant proteins to a-thrombin. Based on this calibration, the recombinant thrombins had similar amidolytic activities toward the chromogenic substrate S2238. However, the clotting activity showed a &fold decrease for r-thrombin-VRl (50 NIH units/ nmol) as compared to r-thrombin and plasma thrombin (both 150 NIH units/nmol).

Expression, Activation, and Partial Characterization
Interaction of r-Thrombin and r-Thrombin-VR1 with PAI-I-The serine proteases r-thrombin, thrombin-VRl, and t-PA were incubated at 37 "C with increasing concentrations of PAI-1 (Fig. 1). Under these conditions (no cofactors), rthrombin was not inhibited by PAI-1. In contrast, thrombin-VR1 was inhibited almost to the same extent as t-PA. Prolonged incubation did not alter the thrombin-VR1 profile, indicating that equilibrium had been reached. The difference with the t-PA profile could indicate a higher dissociation constant of the thrombin-VRl/PAI-1 inhibitory complex, since 1:l complexes are formed as shown below. Alternatively, some cleavage of PAI-1 might occur, as previously reported for a-thrombin (22), resulting in an increased apparent stoichiometry.
Second-order Rate Constants of Inhibition by Various Serpins-The second-order rate constants of inhibition of rthrombin and thrombin-VRl by the serpins PAI-1, ATIII, and HCII were determined both in the presence and in the absence of heparin or vitronectin (Table 11). The rate constants for r-thrombin are in good agreement with the pub-

CR1
VR1 CR2 , and t-PA ( X ) a t 1 nM were incubated with increasing concentrations of PAI-1 for 30 min a t 37 "C in HST buffer as described before (9). Residual amidolytic activity was determined using the chromogenic substrates S2238 (thrombins) and S2288 (&PA), respectively.

Second-order inhibition rate constants of r-thrombin and thrombin-VRl for different serpins
Second-order rate constants of inhibition, expressed in M" s-', were determined at 37 "C and represent the mean value of a t least four independent determinations. Standard deviations are 10% or less. The concentration of the cofactors used are: 1 unit/ml heparin, 20 nM vitronectin. lished data for plasma thrombin (10,11,23). In the absence of cofactors, a 2000-fold increase of the inhibition rate was observed for thrombin-VR1 as compared to r-thrombin. This rate constant of 2.2 X lofi M" s" is of the same magnitude as observed for the inhibition of a-thrombin by the other inhibitors in the presence of heparin. Clearly, replacement of the VR1 region of thrombin by that of t-PA does not affect the inhibition rate of thrombin by ATIII both in the absence and in the presence of heparin. For HCII, a 3-fold decrease in inhibition rate is observed in the presence of heparin. Furthermore, the effect of cofactors is markedly reduced for thrombin-VR1 as compared to a-thrombin; vitronectin enhancement is reduced from 200-to 2-fold, whereas the effect of heparin is reduced from 100-to 6-fold. Serpin-specific Complex Formation-The interaction of serine proteases with serpins typically results in the formation of equimolar, SDS-stable complexes. In the absence of heparin and a 2.2-fold excess of active PAI-1, virtually no complex formation was observed for r-thrombin, whereas thrombin-VR1 was fully complexed (Fig. 2, lanes 4 and 6). In the presence of heparin, about half of r-thrombin was encountered in SDS-stable complexes (Fig. 2, lane 5 ) . Complex formation of PAI-1 with thrombin-VR1 was not influenced by the presence of cofactors (Fig. 2, lanes 7 and 8). Furthermore, thrombin-VR1 is still fully capable of forming SDS-stable complexes with ATIII and HCII.

DISCUSSION
The pseudo-irreversible inhibition of serine proteases by their cognate serine protease inhibitors depends on the reac- tive site ("bait") region of the inhibitor (1). The interaction between this region and the active site of the protease results in a serpin-specific complex that is resistant to denaturants like SDS (1). At present, the exact nature of this complex under native conditions is still a matter of debate, being either of the reversible Michaelis-type or a covalent complex in which the Pl-Pl' bond is cleaved. Previously, we have shown that variations in the bait region (P3-PB') hardly affect the protease specificity of PAI-1 (9,11). Therefore, other molecular interaction sites are likely to play an essential role in the specific serpin-serine protease interactions and might control the rate of the reaction. In the case of thrombin, the anionbinding site, located as a surface-exposed loop in VR2 (21, has been implicated in the highly specific interactions with fibrinogen (4), the receptor on platelets (6), and the serpin HCII (24). It has been shown for the plasminogen activators t-PA and u-PA that the first variable region of these proteins (VR1) is essential for the high rate of inhibition by PAI-1 (7, 8). An inventory of the target proteases for PAI-1 indeed showed a remarkable coincidence between basic VR1-regions and susceptibility toward inhibition by PAI-1 (11). In this report, we present positive evidence that the interaction of PAI-1 with its target proteases is largely governed by the interaction of the serpin with the VR1 region of these proteases as exemplified by a 2000-fold increase of the second-order inhibition rate due to the replacement of the VR1 region of thrombin by the corresponding region of t-PA.
The function of the VRl/PAI-1 interaction might be explained by comparison with the results obtained on thrombin/ hirudin interactions (25). The association rate for thrombin/ hirudin is determined by a fast, ionic interaction between the acidic carboxyl-terminal region of hirudin and the basic anion-binding exosite of thrombin. Subsequently. a substantially slower, but very stable, interaction occurs between the aminoterminal region of hirudin and the active site region of thrombin, resulting in the formation of complexes with an extremely low dissociation constant (25). A similar ionic interaction may control the initial interaction between VR1 and PAI-1, yielding an optimal spatial orientation of the two proteins to facilitate the actual inhibitory interaction, mediated by the catalytic center of the protease and the bait region of the serpin. Spatially, an ionic alignment interaction that involves VR1 is feasible, since the two insertion-loops VR1 and VR2 (anion-binding exosite) are located near the active site of athrombin and constitute the two opposite walls of the fibrinogen binding cleft (3,5). Steric hindrance at this site by the longer VR1 loop of thrombin-VR1 compared to thrombin may explain the observed 3-fold decrease in rate of the interactions with fibrinogen and HCII, which both interact with the anionbinding exosite.
The function of the cofactors in this scheme is fairly obvious. They provide a template at which both protease and serpin align to facilitate the catalytic sitebait region interaction as exemplified by the well described template mechanism for ATIII/heparin (26). In accordance with this view, the stimulating effect of the cofactors vitronectin and heparin on thrombin-VRl inhibition by PAI-1 is considerably reduced as compared to the 2 orders of magnitude enhancement observed for a-thrombin. This observation is reminiscent of the effect of replacing the P1 leucine of HCII by arginine, resulting in a 100-fold increase in the inhibition rate of thrombin and a simultaneous decrease of the cofactor function of heparin (27). We propose that non-optimal molecular contacts between proteases and serpins enable cofactors to efficiently regulate this interaction. In conclusion, we provide additional, positive evidence that surface-exposed VR1 of target proteases determines the rate of interaction with PAI-1.