Importance of the P2 Glycine of Antithrombin in Target Proteinase Specificity, Heparin Activation, and the Efficiency of Proteinase Trapping as Revealed by a P2 Gly → Pro Mutation*

A sequence-specific heparin pentasaccharide activates the serpin, antithrombin, to inhibit factor Xa through an allosteric mechanism, whereas full-length heparin chains containing this sequence further activate the serpin to inhibit thrombin by an alternative bridging mechanism. To test whether the factor Xa specificity of allosterically activated antithrombin is encoded in the serpin reactive center loop, we mutated the factor Xa-preferred P2 Gly to the thrombin-preferred P2 Pro. Kinetic studies revealed that the mutation maximally enhanced the reactivity of antithrombin with thrombin 15-fold and decreased its reactivity toward factor Xa 2-fold when the serpin was activated by heparin pentasaccharide, thereby transforming antithrombin into an allosterically activated inhibitor of both factor Xa and thrombin. Surprisingly, the enhanced thrombin specificity of the mutant antithrombin was attenuated when a full-length bridging heparin was the activator, due both to a reduced rate of covalent reaction of the mutant serpin and thrombin and preferred reaction of the mutant serpin as a substrate. These results demonstrate that the reactive center loop sequence determines the specificity of allosterically activated antithrombin for factor Xa and that the conformational flexibility of the P2 Gly may be critical for optimal bridging of antithrombin and thrombin by physiologic heparin and for preventing antithrombin from reacting as a substrate in the bridging complex.

A sequence-specific heparin pentasaccharide activates the serpin, antithrombin, to inhibit factor Xa through an allosteric mechanism, whereas full-length heparin chains containing this sequence further activate the serpin to inhibit thrombin by an alternative bridging mechanism. To test whether the factor Xa specificity of allosterically activated antithrombin is encoded in the serpin reactive center loop, we mutated the factor Xa-preferred P2 Gly to the thrombin-preferred P2 Pro. Kinetic studies revealed that the mutation maximally enhanced the reactivity of antithrombin with thrombin 15-fold and decreased its reactivity toward factor Xa 2-fold when the serpin was activated by heparin pentasaccharide, thereby transforming antithrombin into an allosterically activated inhibitor of both factor Xa and thrombin. Surprisingly, the enhanced thrombin specificity of the mutant antithrombin was attenuated when a full-length bridging heparin was the activator, due both to a reduced rate of covalent reaction of the mutant serpin and thrombin and preferred reaction of the mutant serpin as a substrate. These results demonstrate that the reactive center loop sequence determines the specificity of allosterically activated antithrombin for factor Xa and that the conformational flexibility of the P2 Gly may be critical for optimal bridging of antithrombin and thrombin by physiologic heparin and for preventing antithrombin from reacting as a substrate in the bridging complex.
Antithrombin is the principal serpin family protein inhibitor of coagulation proteinases in blood (1). Like other serpins, antithrombin inhibits its target proteinases by undergoing a large scale conformational change during reaction of the serpin as a regular proteinase substrate, resulting in the trapping of proteinase in a stable complex with the serpin (2). However, antithrombin differs from most other serpins in that its inhibitory activity is regulated by the polysaccharide cofactor, heparin, which acts to enhance the rate of antithrombin-proteinase reactions several thousand-fold (1,2). Two distinct mechanisms appear to mediate the cofactor effect of heparin, the relative contribution of these mechanisms depending on the proteinase inhibited. The dominant mechanism with factor Xa as the target proteinase involves an allosteric activation of antithrombin by a sequence-specific heparin pentasaccharide (3)(4)(5)(6). The pentasaccharide activates antithrombin by inducing conformational changes in a proteinase binding loop of the serpin known as the reactive center loop, which presumably allows the loop to optimally interact with factor Xa (7)(8)(9)(10)(11)(12). With thrombin as the target proteinase, the allosteric activation mechanism appears to be a minor contributor to the heparin cofactor activity, since the pentasaccharide minimally enhances the rate of antithrombin inhibition of thrombin. Heparin chains at least 18 saccharides in length and containing the pentasaccharide are instead required to substantially accelerate thrombin inhibition by the serpin. Available evidence suggests that the longer chain heparin accelerates the inhibition reaction by an alternative bridging mechanism in which the binding of both antithrombin and thrombin to heparin promotes the interaction between serpin and proteinase in a ternary complex (6,(13)(14)(15)(16).
Based on these findings it has been implied, although without experimental proof, that the conformational changes induced in the antithrombin reactive center loop by pentasaccharide and full-length heparins allow the loop to optimally interact with factor Xa but not with thrombin because the reactive center loop sequence specifically recognizes factor Xa (2). In keeping with such a hypothesis, the P4-P1 IAGR sequence of the loop resembles the IEGR and IDGR P4 -P1 sequences in prothrombin, the natural substrate of factor Xa, whereas the loop sequence does not match well the specificity requirements of thrombin except for the P1 Arg residue (17). To test whether the IAGR reactive center loop sequence of antithrombin is responsible for the factor Xa specificity of allosterically activated antithrombin, we chose to mutate the antithrombin P2 Gly to Pro. This choice was made because (i) a P2 Pro is preferred in natural thrombin substrates (17) and (ii) replacement of a P2 Gly for Pro in synthetic tripeptide substrates produces several hundred-fold enhancements in thrombin specificity (18). While the effects of a P2 Gly 3 Pro mutation in antithrombin on thrombin inhibition were previously found to be modest, the specificity changes resulting from allosteric activation of the mutant antithrombin by heparin pentasaccharide were not examined (19,20). Consistent with the predictions of our hypothesis, our results show that the P2 Gly 3 Pro mutation transforms antithrombin into a serpin that can be activated by the heparin pentasaccharide to inhibit both thrombin and factor Xa with comparable enhanced rates approaching that of activated wild-type antithrombin inhibition of factor Xa. Surprisingly, the P2 Gly 3 Pro mutation converts antithrombin into a worse inhibitor of thrombin when a fulllength heparin is used to activate the serpin, due largely to the preferred reaction of the mutant antithrombin as a substrate of thrombin in the ternary bridging complex with the full-length heparin. Such results suggest that while the P2 Gly of anti-thrombin is not optimal for interaction with thrombin when the serpin is activated by heparin, the conformational flexibility of this residue may be important to allow antithrombin to efficiently trap thrombin in a stable complex when full-length heparin chains, such as would occur in vivo, bridge the serpin and proteinase in a ternary complex.

Expression and Purification of Recombinant Wild-type and G392P
Antithrombins-The cDNA for N135Q antithrombin was used as the template for mutating the P2 Gly residue to Pro in order to mimic the high heparin affinity ␤-form of plasma antithrombin, which lacks the Asn 135 carbohydrate chain (21). Site-directed mutagenesis was carried out in an M13mp19 vector containing the N135Q antithrombin cDNA template as described (9), using the antisense oligonucleotide 5Ј-GTT TAG CGA ACG GGG AGC AAT CAC AAC-3Ј to generate the mutation (the underlined codon corresponds to the mutation). BHK cells were cotransfected with the expression vector, pMAStop, carrying either the reference or mutant antithrombin cDNAs together with the selection plasmids, pRMH140 and pSV2dhfr (all provided by Dr. Gerd Zettlmeissl of Behringwerke, Marburg, Germany), and stable transfected cell lines were selected by resistance to neomycin (Life Technologies) and methotrexate (Sigma) as described (9,(22)(23)(24). The mutant and "wild-type" recombinant antithrombins were isolated from serum-free cycles of medium collected from the stably transfected BHK cells grown to confluence in roller bottles. G392P/N135Q antithrombin (referred to henceforth as G392P antithrombin) was expressed in high yields (average of 12 mg/liter) as determined by radial immunodiffusion assays (The Binding Site, San Diego, CA). The variant and wild-type antithrombins were purified by heparin-agarose chromatography (25) to resolve the high heparin affinity glycoform corresponding in affinity to plasma ␤-antithrombin (23,26). A lower heparin affinity glycoform of the expressed antithrombins resulting from core fucosylation of the Asn 155 glycosylation site (26,27) was separated by this procedure and not further characterized. Further purification of variant and wild-type recombinant antithrombins was done by DEAE-Sepharose chromatography followed by Sephacryl S-200 size exclusion chromatography (25). Concentrations of the recombinant antithrombins were determined from the absorbance at 280 nm using a molar absorption coefficient of 37,700 M Ϫ1 cm Ϫ1 (28).
Proteinases and Heparin-Human ␣-thrombin was a gift of Dr. John Fenton, New York State Department of Health (Albany, NY). Human factor Xa (predominantly ␣-form) was obtained by activation of purified factor X followed by purification on SBTI-agarose as described (29) or generously provided by Dr. Paul Bock (Vanderbilt University). Proteinase concentrations were based on active site titrations, which indicated Ͼ90% and Ͼ70% active enzyme for thrombin and factor Xa, respectively (30). The ␣-methyl glycoside of a synthetic heparin pentasaccharide corresponding to the antithrombin binding sequence in heparin was generously provided by Dr. Maurice Petitou (Sanofi Recherche). A fulllength heparin containing the pentasaccharide with an average molecular weight of ϳ8000 (ϳ26 saccharides) was isolated from commercial heparin by size and antithrombin affinity fractionation (25). Concentrations of pentasaccharide and full-length heparins were determined by stoichiometric titrations of antithrombin with the saccharides monitored by changes in protein fluorescence (6).
Experimental Conditions-All experiments were done at 25°C in sodium phosphate buffers containing 20 mM sodium phosphate, 0.1 mM EDTA, and 0.1% polyethylene glycol 8000 and 0.10 M NaCl (I 0.15), 0.25 M NaCl (I 0.3), or 0.40 M NaCl (I 0.45) adjusted to pH 7.4, unless otherwise noted. Errors associated with reported measurements represent Ϯ 2 S.E. except when otherwise noted.
Stoichiometry of Antithrombin-Proteinase Reactions-Thrombin (5-100 nM) or factor Xa (50 nM) were incubated with increasing molar ratios of wild-type or variant antithrombins in the absence or presence of heparin levels approximately equimolar with the maximum inhibitor concentration (100 nM to 2 M). Incubation times were sufficient to complete the reaction based on measured second order rate constants. Residual enzyme activity was determined by 100-fold dilution of an aliquot of the incubation mixture into 100 M D-Phe-Pip-Arg-p-nitroanilide (S-2238, Chromogenix) for thrombin or 200 M Spectrozyme FXa (American Diagnostica) for factor Xa, and the initial rate of substrate hydrolysis was measured from the linear absorbance increase at 405 nm. Alternatively, thrombin activity was measured by dilution of enzyme into 50 M tosyl-Gly-Pro-Arg-7-amido-4-methylcoumarin (Sigma) with monitoring of substrate hydrolysis from the fluorescence increase at excitation and emission wavelengths of 380 and 440 nm, respectively.
All substrates contained 50 -100 g/ml Polybrene to neutralize any added heparin. The inhibition stoichiometry was obtained by extrapolating linear least squares fits of the decrease in residual enzyme activity with increasing molar ratios of inhibitor to enzyme up to the ratio yielding complete enzyme inhibition (25).
SDS Gel Electrophoresis-The products of antithrombin-proteinase reactions were analyzed by SDS-PAGE using the Laemmli discontinuous buffer system (31) and a 10% polyacrylamide gel under nonreducing conditions as described (25). 5 M thrombin was reacted with 10 M antithrombin for 5 min, and 2 M factor Xa was reacted with 6 M antithrombin for 30 min in the absence or presence of full-length heparin equimolar with antithrombin in I 0.15 buffer. Thrombin reactions were quenched with 250 M FPR-chloromethylketone (Calbiochem), and factor Xa reactions were quenched with 500 M EGR-chloromethylketone (Bachem) prior to preparing samples for electrophoresis.
Characterization of the Antithrombin-Heparin Interaction-Binding of heparin to antithrombin was quantified by titrations of 20 -100 nM inhibitor with a 4 -5-fold molar excess of polysaccharide in I 0.3 buffer, monitored from the intrinsic protein fluorescence enhancement that accompanies polysaccharide binding, as in previous studies (6,15). Titrations were individually computer fitted by the quadratic equilibrium binding equation to obtain the maximal fluorescence change. Normalized titrations were then globally fit to obtain the K D and binding stoichiometry that best fit all of the binding data (25). The kinetics of heparin binding to antithrombin were analyzed under pseudo-first order conditions by monitoring the increase in protein fluorescence, which signals binding, in an Applied Photophysics SX-17MV stoppedflow instrument as described (6,8). Progress curves were fit by a single exponential function to obtain the observed pseudo-first order binding rate constant, k obs , with values from 6 to 14, such curves averaged for each heparin concentration. A heparin concentration range sufficient to partially saturate an initial low heparin affinity interaction was employed at I 0.15, whereas only subsaturating heparin levels were examined at I 0.3 (6).
Kinetics of Antithrombin-Proteinase Association-Second-order rate constants for the association of antithrombin with proteinases were measured under pseudo-first order conditions by using a molar excess of inhibitor over enzyme of at least 10 times the inhibition stoichiometry, similar to previous studies (6,32). Reactions contained 50 -300 nM antithrombin and 0.1-10 nM proteinase with or without catalytic levels of pentasaccharide or full-length heparins, ranging from 0.25 to 9 nM. For the pentasaccharide-accelerated antithrombin-thrombin reaction, pentasaccharide levels were stoichiometric with antithrombin (100 nM) and ranged from 25 to 200 nM. The time-dependent decrease in enzyme activity, measured from the initial rate of chromogenic or fluorogenic substrate hydrolysis of quenched samples as in determinations of reaction stoichiometry, was computer-fitted by a single exponential function to obtain the observed pseudo-first order rate constant, k obs . Apparent second order association rate constants for uncatalyzed and catalyzed reactions were obtained from the least squares slope of the linear dependence of k obs on the antithrombin or the heparin concentration, respectively, in accordance with the equation that applies when where k uncat and k H are the second order rate constants for uncatalyzed and heparin-catalyzed reactions, respectively, [AT] o and [H] o represent the total antithrombin and heparin concentrations, and K AT, H is the dissociation constant for the antithrombin-heparin interaction. The term multiplying k H represents the antithrombin-heparin complex concentration, which under the conditions of these experiments ([AT] o Ͼ Ͼ K AT, H ) was closely approximated by [H] o . For the pentasaccharideaccelerated antithrombin-thrombin reaction, k obs was fit by the equations for saturable binding given previously to obtain the second order rate constant for the reaction of antithrombin fully complexed with pentasaccharide (6). Apparent second order rate constants were corrected for the different fractions of wild-type and mutant serpin reacting through the inhibitory pathway by multiplying by the measured stoichiometry of proteinase inhibition (2,33).
Rapid Kinetic Analysis of the Heparin-accelerated Antithrombin-Thrombin Reaction-Resolution of the two-step reaction of antithrombin-heparin complex with thrombin was done by analyzing the kinetics of reactions in I 0.15 or I 0.3 buffers in the Applied Photophysics stopped-flow instrument under pseudo-first order conditions, as in pre-vious studies (13,34). Reactions contained 0.06 -75 nM thrombin and a 40 -400-fold molar excess of antithrombin-heparin complex, generated by saturating 0.025-7.5 M heparin with a 1.1-2-fold molar excess of antithrombin. The exact concentrations of complex were calculated using measured K D values. Reactions were continuously monitored from the exponential decrease in the rate of hydrolysis of the fluorogenic substrate, tosyl-Gly-Pro-Arg-7-amido-4-methylcoumarin, present at 5 M. k obs values from 4 -12 progress curves were averaged for each inhibitor-heparin complex concentration. The dependence of k obs on the antithrombin-heparin complex concentration was fit by the hyperbolic equation (13), where k represents the limiting rate constant for conversion of a noncovalent heparin-antithrombin-thrombin ternary complex to a covalent serpin-proteinase complex with release of heparin, K T, ATH is the dissociation constant for formation of the ternary complex, Kinetics of Dissociation of Antithrombin-Proteinase Complexes-Rate constants for dissociation of serpin-proteinase complexes were measured as in previous studies (35,36). Briefly, complexes were formed by reacting 5-10 M antithrombin with either 0.5 M thrombin and 1 M heparin for 30 min or 1 M factor Xa and 1 M heparin for 10 min at 25°C. Thrombin reactions were done in I 0.45 buffer to minimize any substrate reaction, whereas factor Xa reactions were done in I 0.15 buffer. Complexes were diluted to 0.3 to ϳ6 nM in 1 ml of 400 M S2238 or 400 M Spectrozyme FXa in I 0.15 buffer containing 100 g/ml polybrene at 37°C. The initial rate of complex dissociation was continuously monitored from the acceleration of the rate of substrate hydrolysis at 405 nm for 80 min. (Ͻ1% complex dissociation). The time-dependent absorbance changes were fit by a second order polynomial equation to obtain the initial rate of complex dissociation, as described (35,36). Initial rates were plotted against the concentration of serpinproteinase complex, and the first order dissociation rate constant was calculated from the least squares slope of this linear plot.

Stoichiometry of Proteinase Inhibition and Formation of SDS-stable Complexes-G392P
antithrombin inhibited thrombin and factor Xa in the absence of heparin with ϳ1:1 inhibition stoichiometries and with the appearance of stable complexes by SDS-PAGE, similar to the properties of the wild-type inhibitor (Figs. 1 and 2). The inhibition stoichiometries exceeded 1:1 when G392P or wild-type antithrombin reactions were conducted in the presence of either a full-length heparin or a heparin pentasaccharide fragment, which constitutes the binding site for antithrombin within the larger polysaccharide ( Fig.  2 and Table I). Such increases in inhibition stoichiometry result from heparin promoting a competing substrate reaction of antithrombin with the proteinase (6,37,38), as was evident from the appearance on SDS-PAGE of cleaved antithrombin concomitant with antithrombin-proteinase complex in these reactions (Fig. 1). Notably, the heparin-induced increases in stoichiometry were more substantial for G392P antithrombin reactions than for wild-type inhibitor reactions (Table I). This was particularly pronounced for the full-length heparin-catalyzed inhibition of thrombin by antithrombin, in which case the mutant inhibitor reaction stoichiometry of 25 Ϯ 3 mol of inhibitor/mol of enzyme greatly exceeded the stoichiometry of 2.2 Ϯ 0.1 mol of inhibitor/mol of enzyme measured for the wild-type inhibitor reaction (Fig. 2). Increasing the ionic strength lowered the inhibition stoichiometry for the heparin-catalyzed mutant antithrombin-thrombin reaction to 5.1 Ϯ 0.5 mol of inhibitor/mol of enzyme at I 0.3 and to ϳ2 mol inhibitor/mol enzyme at I 0.45, whereas the wild-type inhibitor reaction stoichiometry was reduced to ϳ1 mol inhibitor/mol enzyme under these conditions (Fig. 2), in keeping with the salt dependence of the heparin enhancement of the reaction stoichiometry found in past studies (6,38).
Interaction with Heparin-To determine whether the G392P mutation affected the ability of heparin to bind and activate antithrombin, the affinity and rates of heparin binding to G392P and wild-type antithrombins were measured from the protein fluorescence enhancement, which reports this binding (6,7). Indistinguishable K D values of 9.6 Ϯ 2.9 and 7.9 Ϯ 1.2 nM and heparin binding stoichiometries of 1.14 Ϯ 0.14 and 0.97 Ϯ 0.09 were measured from global analysis of 11 titrations of the mutant antithrombin and eight titrations of the reference antithrombin interactions, respectively, in I 0.3 buffer at inhibitor concentrations ranging from ϳ2 to 10 times K D (not shown). Moreover, the 31 Ϯ 4% average fluorescence enhancement induced in G392P antithrombin by heparin binding was similar to the 37 Ϯ 3% enhancement induced in the wild-type inhibitor. Analysis of observed pseudo-first order rate constants for heparin binding to normal and mutant inhibitors by stopped-flow flurometry at I 0.15 as a function of the heparin concentration (6,8) indicated identical K D values of 7.1 Ϯ 2.9 and 7.1 Ϯ 3.0 M for an initial low heparin affinity interaction with mutant and normal inhibitors and indistinguishable limiting rate constants of 990 Ϯ 360 and 900 Ϯ 330 s Ϫ1 for the subsequent conformational activation of G392P and wild-type antithrombins. The initial linear dependence of k obs on heparin concentration at I 0.3 similarly showed the same overall second-order association rate constants of 18 Ϯ 1 and 16 Ϯ 1 M Ϫ1 s Ϫ1 for heparin binding to wild-type and mutant inhibitors. Together, these results indicated that heparin binding and conformational activation of the mutant antithrombin was normal.
Kinetics of Proteinase Inhibition-The changes in specificity of antithrombin resulting from the P2 Gly 3 Pro mutation were examined by measuring the second order rate constants for association of mutant and normal inhibitors with the target proteinases, thrombin and factor Xa with and without activa- tion by pentasaccharide and full-length heparins. These rate constants were determined from the slopes of linear plots of the pseudo-first order association rate constant (k obs ) versus the antithrombin or the heparin concentration (equal to the antithrombin-heparin complex concentration under the experimental conditions) or, in the case of the pentasaccharide-accelerated antithrombin-thrombin reaction, from the saturable dependence of k obs on the heparin concentration at a fixed antithrombin concentration (Fig. 3). To compare the rates of association of mutant and wild-type inhibitors with proteinase along the inhibitory pathway, it was necessary to correct apparent second order rate constants derived from the data of Fig. 3 for the different contributions of the substrate pathway to these reactions by multiplying by the inhibition stoichiometry (2,33). Apparent and corrected second order rate constants obtained from these kinetic analyses are summarized in Table I. G392P antithrombin inhibited thrombin with a 3.5-fold faster apparent second order rate constant than wild-type antithrombin in the absence of heparin, whereas the mutant serpin inhibited the enzyme with a 9.4-fold faster rate constant than the wild-type serpin when allosterically activated by heparin pentasaccharide. Surprisingly, the variant antithrombin inhibited thrombin with a 5-fold slower rate constant than the wild-type serpin when a full-length heparin was the activator. However, correction of these rate constants for the different extents to which heparin-activated mutant and wild-type serpins reacted as substrates of thrombin revealed that G392P antithrombin reacted faster with thrombin along the inhibitory pathway than the wild-type serpin in the presence of either heparin activator, although the reaction rate enhancement was much greater when the pentasaccharide was activator. Thus, G392P antithrombin complexes with pentasaccharide and fulllength heparins inhibited thrombin with 15-and 2.2-fold faster corrected association rate constants, respectively, than the corresponding wild-type inhibitor-heparin complexes. By contrast, G392P antithrombin was a poorer inhibitor of factor Xa than wild-type antithrombin with or without heparin activation, with 3-5-fold slower apparent rate constants for inhibition of factor Xa being measured. Correction for the different extents of substrate reaction of heparin-activated mutant and wildtype inhibitors with factor Xa indicated that activation of G392P antithrombin by either heparin still lowered the factor Xa inhibition rate constant relative to the activated wild-type serpin but to a lesser extent of 2-fold. Together, these results indicated that the G392P mutation increased the specificity of antithrombin for thrombin at the expense of decreasing its

TABLE I Apparent second order rate constants (k app ) and stoichiometries of inhibition for antithrombin-proteinase reactions
Stoichiometries of inhibition (SI) in units of mol of inhibitor/mol of proteinase were determined from linear plots of the decrease in enzyme activity with increasing molar ratio of inhibitor to enzyme as in Fig. 2. Reported values are averages Ϯ S.E. from at least three titrations except for antithrombin-factor Xa reactions in the presence of pentasaccharide, in which case the average from two titrations Ϯ range is reported. Apparent second order association rate constants (k app ) were determined from the slopes of plots of k obs against antithrombin or heparin concentration in Fig. 3, except for the pentasaccharide-catalyzed antithrombin-thrombin reaction, in which case the dependence of k obs on pentasaccharide concentration was fit by a saturation function (6). Reported errors represent Ϯ2 S.E. Other details are given under "Materials and Methods." H5 refers to heparin pentasaccharide and H26 to full-length 26-saccharide heparin. specificity for factor Xa, with the increased thrombin specificity being maximal when the serpin was allosterically activated by the heparin pentasaccharide.

Reaction
Step Affected by the G392P Mutation-The enhanced thrombin specificity and stimulated substrate reaction of heparin-activated G392P antithrombin with thrombin could result from alterations in the affinity of an initial noncovalent antithrombin-thrombin encounter complex or from changes in the rate of conversion of the noncovalent complex to a stable covalent complex (13,15). To distinguish between these possi-bilities, the kinetics of thrombin inhibition by mutant and wild-type inhibitors activated by the full-length heparin were compared by stopped-flow fluorometry using a fluorogenic reporter substrate to continuously monitor proteinase inhibition as in past studies (34). k obs increased in a saturable manner as the effective antithrombin-heparin complex concentration (i.e. the concentration corrected for the competitive effect of the synthetic substrate) was increased for both wild-type and mutant inhibitor reactions at two different ionic strengths (Fig. 4). Fitting of data for reactions at I 0.15 by a hyperbolic function yielded indistinguishable dissociation constants of 297 Ϯ 28 and 220 Ϯ 40 nM for the initial ternary encounter complex of thrombin with antithrombin-heparin complex but greatly differing limiting rate constants of 3.1 Ϯ 0.1 and 0.43 Ϯ 0.03 s Ϫ1 for conversion of the ternary complex to a stable antithrombinthrombin complex. Similarly, at I 0.3 wild-type and mutant inhibitor reactions were characterized by indistinguishable dissociation constants of 1800 Ϯ 400 and 1200 Ϯ 100 nM for ternary complex formation but significantly different limiting rate constants of 5.4 Ϯ 0.7 and 2.4 Ϯ 0.1 s Ϫ1 for conversion to stable complex. These results indicated that the G392P mutation did not affect the initial ternary complex formation step but instead affected the subsequent stable complex formation step.
The sole effect of the G392P mutation on the second-step rate constant indicated that the ϳ2-fold enhanced overall association rate constant and ϳ11-fold increased stoichiometry for the mutant relative to the wild-type inhibitor reaction (Table I) are reflected in the altered second-step rate constant. The influence of reaction stoichiometry on the second-step rate constant was evident from the parallel changes in the inhibition stoichiometry (4.9-fold decrease) and in the second-step rate constant (5.8-fold increase) for the mutant inhibitor reaction when the ionic strength was increased from I 0.15 to 0.3. Parallel, although much smaller, changes in inhibition stoichiometry (1.8fold decrease) and second-step rate constant (1.3-fold increase) also occurred in the wild-type inhibitor reaction over this range of ionic strength. Correction of the second-step rate constants for the different fractions of the mutant and wild-type antithrombins that reacted along the substrate pathway by multiplying by the inhibition stoichiometry (2, 33) yielded values of 6.8 Ϯ 0.5 and 11 Ϯ 2 s Ϫ1 for wild-type and mutant inhibitor reactions at I 0.15 and values of 6.5 Ϯ 1.4 and 12 Ϯ 2 s Ϫ1 for these reactions at I 0.3. These results demonstrated that the ϳ2-fold greater specificity of G392P than of wild-type antithrombin for thrombin in the presence of full-length heparin is due to a faster rate of conversion of a noncovalent antithrombin-thrombin-heparin bridging complex to a covalent antithrombin-thrombin complex, which is essentially independent of ionic strength.
Kinetics of Antithrombin-Proteinase Complex Dissociation-To determine whether the G392P mutation affected the stability of the covalent antithrombin-proteinase complex, the rate constants for dissociation of mutant and wild-type antithrombin complexes with thrombin and factor Xa were measured by continuously monitoring the appearance of enzyme activity as the complex dissociates after dilution into a reporter chromogenic substrate, as in previous studies (35,36). Slower rates of complex dissociation were observed for mutant antithrombin complexes than for control antithrombin complexes with either thrombin or factor Xa over a wide range of complex concentrations (Fig. 5). From the slopes of linear plots of the initial rate of complex dissociation versus complex concentration, dissociation rate constants of 7.5 Ϯ 0.2 ϫ 10 Ϫ7 and 4.5 Ϯ 0.1 ϫ 10 Ϫ7 s Ϫ1 were determined for wild-type and mutant antithrombin complexes with thrombin, whereas rate constants of 6.5 Ϯ 0.2 ϫ 10 Ϫ7 and 3.5 Ϯ 0.2 ϫ 10 Ϫ7 s Ϫ1 were obtained for complexes of wild-type and mutant antithrombins with factor Xa (Fig. 5). These results indicated that the G392P mutation increased complex stability for both thrombin and factor Xa to about the same extent (1.7-1.9-fold).

DISCUSSION
In the present study, we have provided the first evidence to support the idea that the allosteric activation of antithrombin by a sequence-specific heparin pentasaccharide results in accelerated inhibition of factor Xa and not thrombin due to the reactive center loop sequence of antithrombin being specific for factor Xa. Thus, by mutating the reactive center loop P2 residue from the factor Xa-preferred Gly to the thrombin-preferred Pro (17), we were able to transform pentasaccharide-activated antithrombin from a 50-fold better inhibitor of factor Xa than thrombin to an inhibitor with comparable specificity for thrombin and factor Xa approaching that of activated wild-type antithrombin for factor Xa. Our finding that the P2 Gly to Pro mutation greatly enhances the differential recognition of native and activated states of antithrombin by thrombin from 1.6to ϳ7-fold (Table I) is consistent with the wild-type serpin sequence poorly recognizing thrombin in either conformational state and with the mutant serpin sequence showing enhanced discrimination between the two conformational states due to an increase in recognition determinants for thrombin. The 15fold maximal enhancement in thrombin specificity of allosterically activated G392P antithrombin contrasts with the larger 45-300-fold enhancements in thrombin specificity achieved by replacing a P2 Gly with Pro in synthetic tripeptide substrates (18). Since the extent of these larger specificity enhancements was correlated with how well flanking residues matched the specificity requirements of thrombin, the smaller thrombin specificity enhancement observed with G392P antithrombin supports the conclusion that the reactive center loop sequence of wild-type antithrombin is poorly recognized by thrombin in both native and allosterically activated conformations. Moreover, it suggests that larger enhancements in thrombin specificity can be engineered through additional changes in the reactive center loop sequence. It follows that the ϳ200-fold enhanced specificity of wild-type antithrombin for factor Xa upon pentasaccharide activation arises from the favorable P4 to P1 IAGR recognition sequence in the loop and its greater accessibility to the enzyme when the reactive center loop is exposed by allosteric activation (10 -12, 39).
Our results also provide evidence that the reactive center loop of antithrombin is not involved in enhancing the recognition of thrombin when the serpin is activated by a full-length bridging heparin, in agreement with previous work showing that the enhanced thrombin reactivity of full-length heparinactivated antithrombin is almost fully accounted for by thrombin recognizing a bridging site on heparin adjacent to bound antithrombin (6,(13)(14)(15)(16). Rather, the large attenuation of the enhancement in antithrombin specificity for thrombin from 15-fold with the pentasaccharide activated serpin to only 2-fold with the full-length heparin activated serpin suggests that the requirements for thrombin recognition by the antithrombin reactive center loop are more strict in the ternary bridging complex. The reduced recognition of thrombin by the mutant antithrombin in the bridging complex may reflect an effect of the P2 Gly 3 Pro mutation on the conformational flexibility of the antithrombin reactive center loop. Such flexibility may be important for an optimal P2-S2 interaction when thrombin is constrained to interact with both antithrombin and heparin in the ternary bridging complex (15). This flexibility may also be important for the recognition of factor Xa by antithrombin when physiologic full-length heparin chains activate the serpin, since the bridging mechanism contributes to this recognition in the presence of physiologic levels of calcium ion (40).
Kinetic studies established that the P2 Pro mutation enhanced the specificity of antithrombin for thrombin by increasing the rate of conversion of a noncovalent Michaelis complex to a covalent inhibited complex, whereas the mutation had only small effects on the rate of dissociation of the covalent complex. Moreover, the mutation stimulated a substrate reaction of the serpin with both proteinases in the presence of heparin, which reduced the rate of the covalent reaction step. Such findings are in accord with the branched pathway suicide substrate mechanism by which serpins have been proposed to inhibit their target proteinases (Scheme 1) (38,(41)(42)(43). According to this mechanism, the proteinase (E) reacts with the serpin (I) as it would with a normal substrate to first form a noncovalent Michaelis complex (E⅐I) and then a covalent acyl-enzyme intermediate (E-I C ) in which the P1-P1Ј scissle bond is cleaved, and the serpin P1 residue is covalently linked to the proteinase serine. The cleavage triggers a large scale conformational change in the serpin, which traps the proteinase as an acylenzyme complex (E-I C *) (44 -46), but this trapping competes with the deacylation of the acyl-enzyme intermediate to complete the substrate reaction and generate free proteinase and cleaved inhibitor (I C *).
In the context of this mechanism, our findings would indicate that the antithrombin P2 residue contributes to serpin specificity by lowering the activation energy for the rate-limiting acylation of antithrombin by thrombin in the covalent reaction step (36,47), similar to the action of specificity-determining residues in normal proteinase substrates (48). The observation that the P2 residue of antithrombin influences serpin specificity through primary effects on the rate of formation of the covalent complex and minimal effects on the rate of complex dissociation are in keeping with similar effects of the antithrombin P1Ј residue on specificity observed in previous studies (34,36). Such findings strongly argue for a mechanism of kinetic stabilization of the serpin-proteinase complex wherein reactive center loop interactions with proteinase are disrupted in the final complex through insertion of the loop into sheet A, resulting in a greatly decreased rate constant for deacylation of the complex (36). The small decreases in the rate constants for dissociation of antithrombin complexes with both thrombin and factor Xa contrast with the opposing effects of the G392P mutation on the association rate constants for these enzymes. The decreases in dissociation rate constants are thus not likely to be due to direct interactions of the P2 residue with the corresponding S2 subsite of the proteinase in the complex and may instead reflect a common stabilizing effect of the P2 Pro on the serpinproteinase complex due to its reducing reactive center loop flexibility. Such an effect of decreased conformational flexibility of the loop on complex stability would support a proteinase trapping mechanism in which immobilization of the loop through insertion into ␤-sheet A stabilizes the proteinase in a distorted conformational state by decreasing the conformational freedom of the tethered proteinase (36,(51)(52)(53).
The greater stimulation of a substrate mode of reaction of G392P antithrombin with proteinases in the presence of heparin, most prominent with thrombin as proteinase and a fulllength heparin as activator, implies that the G392P mutation affects the partitioning of the acyl-intermediate between inhibitor and substrate modes of reaction. Heparin itself stimulates FIG. 5. Effect of the P2 Gly 3 Pro mutation on the kinetics of dissociation of antithrombin-proteinase complexes. Comparison of the initial rate of dissociation of complexes of wild-type (E) or G392P mutant (q) antithrombin with thrombin (left panel) or with factor Xa (right panel) as a function of the antithrombin-proteinase complex concentration is shown. Dissociation of complexes was initiated by 100 -2000-fold dilution into proteinase substrate and continuously monitored from the acceleration of substrate hydrolysis as detailed under "Materials and Methods." Solid lines are linear regression fits of the data from which the dissociation rate constant was determined from the slope. SCHEME 1 a substrate reaction of wild-type antithrombin with proteinases (6,37,38), which can be accounted for by heparin's ability to antagonize the serpin conformational change, which traps proteinase as an acyl-enzyme complex (49). This antagonism may arise both from the pentasaccharide stabilizing the reactive center loop in an exposed conformation, which precludes its burial into ␤-sheet A, as well as from heparin bridging interactions with proteinase hindering the movement of the enzyme during the trapping serpin conformational change (50,51). An important finding was that heparin stimulated the substrate reaction of G392P antithrombin to a greater extent than the wild-type serpin with both thrombin and factor Xa. This cannot arise from an enhanced rate of deacylation of the acyl-enzyme intermediate, since the P2 Pro mutation had opposite effects on antithrombin specificity for thrombin and factor Xa. This observation instead suggests that the G392P mutation must further slow the rate of the trapping conformational change. The slowing of the conformational change could result from the reduced conformational flexibility of the reactive center loop upon replacing the flexible Gly with a rigid Pro. Such an explanation is plausible given recent findings that the trapping conformational change involves full insertion of the loop and tethered proteinase into ␤-sheet A of the serpin (51), thus necessitating major changes in loop conformation. However, an effect of the G392P mutation on the deacylation rate is also likely to account for the more pronounced stimulation of a substrate reaction of G392P antithrombin with thrombin than with factor Xa. Thus, an enhanced rate of deacylation with thrombin but decreased rate of deacylation with factor Xa could explain the different extents to which the substrate pathway is stimulated by heparin in the reactions of the mutant serpin with these two proteinases.
In summary, the present study supports the conclusion that the factor Xa specificity of allosterically activated antithrombin is encoded in the reactive center loop sequence of the serpin, based on our ability to engineer an allosterically activated antithrombin with reduced specificity for factor Xa and enhanced specificity for thrombin by mutation of the P2 reactive center loop residue. Our findings further support a branched pathway suicide substrate mechanism for antithrombin-proteinase reactions and a role for the P2 Gly residue in this mechanism in determining the rate of formation of an acylintermediate but not in stabilizing the trapped acyl-intermediate. Finally, our results suggest that the conformational flexibility of the P2 Gly may be critical for optimal acylation and for efficient trapping of the acyl-intermediate in a ternary bridging complex with heparin. Indeed, the loss of such conformational flexibility in G392P antithrombin contributes to the mutant serpin reacting predominantly as a substrate of thrombin in the presence of physiologically relevant full-length heparin chains. The importance of the P2 Gly in antithrombin's anticoagulant function is supported by the conservation of this Gly in the seven vertebrate antithrombins, namely human, cow, sheep, pig, rabbit, mouse, and chicken, that have been sequenced. Based on the present and past studies, this function may include both the promotion of antithrombin inhibition of its target procoagulant proteinases when the serpin is bound to its glycosaminoglycan activator as well as the prevention of antithrombin inhibition of the anticoagulant proteinase, activated protein C (54).