Partial activation of antithrombin without heparin through deletion of a unique sequence on the reactive site loop of the serpin.

Native antithrombin (AT) has an inactive reactive site loop conformation unless it is activated by a unique pentasaccharide fragment of heparin (H(5)). Structural data suggests that this may be due to preinsertion of two N-terminal residues of the reactive site loop of the serpin into the A-beta-sheet of the molecule. Relative to alpha(1)-antitrypsin, the reactive site loop of AT has three additional residues, Arg(399), Val(400), and Thr(401), at the C-terminal P' end of the loop. To determine whether a longer reactive site loop of AT is responsible for loop preinsertion in the native conformation, mutants of the serpin were expressed in which these residues were individually or in combination deleted. Kinetic analysis suggested that deletion of two residues, Val(400) and Thr(401), changed the solution equilibrium of the serpin in favor of the active conformation, thereby enhancing the inhibition of factor Xa by an order of magnitude independent of H(5). Interestingly, the reactivity of this mutant with thrombin was impaired by the same order of magnitude in the absence, but not in the presence of H(5). These results suggest that a longer reactive site loop in AT is responsible for its inactive native conformation toward factor Xa, while at same time AT requires this feature to regulate the activity of thrombin.

Antithrombin (AT) 1 is the primary serpin inhibitor in plasma that regulates the activities of the serine proteinases of both the intrinsic and extrinsic pathways of the blood coagulation cascade (1)(2)(3). Similar to other inhibitory serpins, AT inhibits its target proteinases by binding to their active sites through an exposed reactive center loop and undergoing a conformational change which traps the enzymes in inactive, stable complexes (4,5). Unlike most other inhibitory serpins, however, AT has a reactive site loop that has an inactive conformation (6 -9).
A unique high affinity pentasaccharide (H 5 ) fragment of heparin can bind and allosterically activate AT to promote its reactivity with factor Xa (fXa) by several hundred-fold (6,10). Surprisingly, however, the allosteric activation of AT by H 5 has no effect on the reactivity of the serpin with thrombin. In this case, longer chain heparins containing H 5 plus at least 13 additional saccharides are required to efficiently accelerate the inhibition reaction by an alternative ternary complex bridging or template mechanism (11,12). The molecular basis for differential reactivity of fXa and thrombin with the activated conformation of AT is not known.
Structural data suggest that the inactive native conformation of the reactive site loop of AT is caused by preinsertion of two N-terminal P14 and P15 (nomenclature of Schechter and Berger (13)) residues of the loop into the A-␤-sheet of the serpin and that the binding of the cofactor to AT causes the expulsion of this inserted region and, thereby, activation of the serpin (6,8,14,15). The structural feature(s) in the reactive site loop of the serpin that may be responsible for preinsertion of the two N-terminal residues into the A-␤-sheet of the molecule has not been identified. Moreover, it is not known why activation of AT by H 5 specifically promotes AT inhibition of fXa, but not that of thrombin.
To address these questions, the amino acid sequence of the reactive site loop of AT was compared with those of other inhibitory serpins to determine whether a unique feature exists in the reactive site loop of AT that may be responsible for its inactive native conformation. It was noted that relative to ␣ 1 -antitrypsin, the archetypical inhibitor of the serpin superfamily, and other inhibitory serpins that have a basic heparinbinding D-helix ( Fig. 1), the reactive site loop of AT has three insertion residues, Arg 399 , Val 400 , and Thr 401 , at the most Cterminal end of the loop between the P5Ј site and s1C-sheet (16). To determine the contribution of these residues to the inactive, partially buried reactive site loop conformation of AT, several mutants of the serpin were generated in which these residues were individually or in combination of two and three deleted from human AT cDNA and expressed in mammalian H293 cells. Following purification to homogeneity, the properties of these mutants were characterized with respect to their ability to bind heparin and react with fXa and thrombin in both the absence and presence of high affinity heparin and the H 5 fragment of high affinity heparin. The results suggest that mutagenesis of all three residues leads to loss of affinity of the mutant for heparin, as well as its inhibitory property toward both fXa and thrombin. However, deletion of one or two residues shifts the conformational equilibrium of the AT mutants toward the activated state, thereby promoting the inhibition of fXa independent of H 5 . Interestingly, however, the abilities of mutants to inhibit thrombin are impaired in the absence, but not in the presence of H 5 , suggesting that these residues in the native conformation of AT enable the serpin to interact with the catalytic pocket of thrombin. Consistent with the conformational equilibrium model for AT (6,8,17), improvement in the reactivity of a double mutant lacking Val 400 and Thr 401 (des-VT) with fXa (ϳ12-fold) is identical to the extent of impairment of the mutant's reactivity with thrombin. This leads to a similar rate-accelerating effect for H 5 in des-VT inhibition of both fXa and thrombin. These results suggest that a longer reactive site loop of AT at the PЈ end of the loop is responsible for the partially buried loop of the serpin in the native conformation, which renders it inactive toward fXa. AT, however, requires this feature to regulate the activity of thrombin in the native conformation, which explains the differential reactivity of two proteinases with the native and H 5 -activated conformers of the serpin. The significance of this unique structural feature with respect to the physiological function of AT is discussed.

Expression and Purification of Recombinant Proteins-Recombinant
human antithrombin (rAT) and its deletion derivatives in which the three insertion residues, Arg 399 , Val 400 , and Thr 401 , were individually (des-R, des-V, and des-T), or in combination of two (des-VT) or three (des-RVT) deleted by standard PCR mutagenesis methods, expressed in H293 cells using RSV-PL4 expression/purification vector system as described previously (18,19). Accuracy of the mutations was confirmed by sequencing prior to expression. Wild type and mutant serpins were purified from cell culture supernatants by immunoaffinity chromatography using the HPC4 antibody linked to Affi-Gel 10 (Bio-Rad) followed by a HiTrap-Heparin (Amersham Biosciences, Inc.) ion exchange chromatography with a gradient elution from 0.1 to 2.0 M NaCl in 20 mM Tris-HCl, pH 7.4 as described previously (19). Concentrations of the AT derivatives were determined from the absorbance at 280 nm using a molar absorption coefficient of 37,700 M Ϫ1 cm Ϫ1 and by stoichiometric titration of the serpins with calibrated heparin as monitored from changes in intrinsic protein fluorescence (20).
Human plasma AT (pAT), the active AT-binding H 5 fragment of heparin and full-length high affinity heparin containing the H 5 with an average molecular mass of ϳ21,000 Da (ϳ70 saccharides, H 70 ) were generous gifts from Dr. Steven Olson (University of Illinois, Chicago). Concentrations of heparins were based on the AT binding sites and were determined by stoichiometric titration of AT with the polysaccharides, with monitoring of the interaction by changes in protein fluorescence (10,21). Recombinant human thrombin was expressed and purified as described previously (22). Human fXa was purchased from (Hematologic Technologies, Essex Junction, VT). The chromogenic substrates, Spectrozyme FXa (SpFXa) and Spectrozyme PCa (SpPCa) were purchased from American Diagnostica (Greenwich, CT). Polybrene was purchased from Sigma.
Fluorescence Measurements-An Aminco-Bowman series 2 spectrophotometer (Spectronic Unicam, Rochester, NY) was used for protein fluorescence measurements at 25°C. The excitation and emission wavelengths were 280 and 340 nm, respectively. The bandwidths were set at 1 nm for excitation and 8 nm for emission. Heparin titration was performed by the addition of a 2-5 l of high concentration of stock solution of heparin into a 50 nM concentration of each AT sample in 0.1 M NaCl, 0.02 M Tris-HCl, pH ϭ 7.4, (TBS buffer, ionic strength ϭ 0.12) containing 0.1% polyethylene glycol (PEG) 8000. Addition of heparin diluted samples less than 5% of the original volume (500 l). Data from at least three experiments were analyzed as the ratio of change in the fluorescence intensity of the sample containing heparin to the initial intensity of the control protein lacking heparin. The affinity of each AT derivative for heparin was calculated by nonlinear least squares computer fitting of the data by the quadratic binding equation as described previously (21).
Kinetic Methods-The rate of inactivation of fXa and thrombin by the AT derivatives in both the absence and presence of heparin was measured under pseudo-first order conditions by a discontinuous assay method as described previously (22,23). Briefly, in the absence of heparin, a 1 nM concentration each proteinase was incubated with 200 -500 nM AT in TBS buffer containing 0.1 mg/ml bovine serum albumin (BSA), 0.1% PEG 8000, and 2.5 mM CaCl 2 . All reactions were carried out at room temperature in 50-l volumes in 96-well polystyrene plates. After a period of time (1-60 min depending on the rate of the reactions), 50 l of the chromogenic substrate SpFXa or SpPCa (500 M) in TBS was added to each well, and the remaining enzyme activities were measured by a V max Kinetics Microplate Reader (Molecular Devices, Menlo Park, CA). The reaction conditions with both proteinases in the presence of a saturating concentration of H 5 (500 nM) were the same except that concentrations of the AT derivatives ranged from 5 to 200 nM and the incubation time was reduced to 0.5-20 min. The observed pseudo-first order rate constants (k obs ) were determined by computer fitting of the time-dependent change of the proteinase activity to a single exponential function and the second order association rate constants (k 2 ) for uncatalyzed and catalyzed reactions were obtained from the slopes of linear plots of k obs versus the concentrations of AT as described previously (22,23).
The heparin concentration dependence of wild type and mutant AT inhibition of both fXa and thrombin were evaluated by incubating a 0.2 nM concentratino of each proteinase with AT (1.5-2.5 nM) in the presence of increasing concentrations of H 70 (0 -8.2 M) in TBS as described previously (22,23). Following incubation for 10 -120 s at room temperature, 50 l of SpFXa or SpPCa in TBS containing 1 mg/ml Polybrene was added, and the k obs and k 2 values for the heparin-catalyzed inactivation were determined as described above.

RESULTS
Expression and Purification of the Antithrombin Derivatives-Wild type and mutant antithrombin derivatives were expressed in H293 cells and purified to homogeneity by a combination of HPC4 immunoaffinity and HiTrap-Heparin column chromatography as described previously (18,19). Except for des-RVT, which eluted at ϳ0.3 M NaCl, recombinant wild type (rAT) and all other mutants of AT were eluted at ϳ0.7-0.8 M NaCl from the HiTrap-Heparin column (data not shown). Since des-RVT did not bind heparin with high affinity and exhibited insignificant inhibitory activity toward either fXa or thrombin in both the absence and presence of heparin, it was not included in the further studies described below. SDS-PAGE analysis of all other mutants under nonreducing conditions suggested that the AT derivatives have been purified to homogeneity and that all migrate with relative molecular masses identical to human pAT (Fig. 2).
Binding to Heparin-It is known that the binding of high affinity heparin to AT is associated with an ϳ30 -40% intrinsic protein fluorescence enhancement (21). The dissociation constants (K D ) of the heparin binding to each mutant were determined by monitoring the fluorescence emission spectra as de- suggested that the affinity of mutants for binding to heparin was either normal or improved. The most improvement in affinity (2-fold) was observed for the des-VT double deletion mutant. These results suggested that the mutants have been properly folded. Furthermore, consistent with previous observation by others (7,17), a high affinity interaction with heparin suggested that mutants might have adopted activated conformations.
Inactivation of FXa and Thrombin-k 2 values for the association of the wild type and mutant serpins with fXa and thrombin in both the absence and presence of H 5 or H 70 are presented in Tables I and II. As expected from the purity on SDS-PAGE, rAT and pAT exhibited identical reactivities with both proteinases in either the absence or presence of cofactors. Except for des-R, which exhibited a near wild type reactivity with fXa, all other mutants inhibited fXa better than the wild type serpin in the absence of the cofactors (Table I). The improvement in the reactivities of the mutants with fXa ranged from ϳ2-3-fold for des-V and des-T to ϳ12-fold for the double mutant des-VT. Interestingly, the 200 -300-fold rate-accelerating effect of H 5 in AT inhibition of fXa was reduced by a similar extent in inhibition by the mutant serpins. Thus, in contrast to ϳ300-fold catalytic effect of H 5 in rAT inhibition of fXa (Table  I), this value was reduced to only 35-fold in inhibition by the des-VT mutant. Similarly, the extent of the cofactor effect of the full-length heparin, H 70 , in inhibition of fXa by the mutant serpins was decreased (Table I). We previously demonstrated that the cofactor effect of H 70 in AT inhibition of fXa is mediated by ϳ300-fold enhancement through activation and ϳ200 -300-fold enhancement through template mechanism in the presence of a physiological concentration of Ca 2ϩ (22,23). Thus, analysis of data in Table I suggests that the observed decrease in the rate-accelerating effect of H 70 in inhibition of fXa by the mutant serpins is not due to a decrease in the extent of the template effect of heparin, but rather due to mutants being in constitutively activated conformations. This is derived from the observation that the ratio of the cofactor effect of H 70 to that of H 5 is not significantly affected with the mutant serpins (Table I).
Unlike the reactions with fXa, the ability of mutant serpins to inhibit thrombin was impaired 4 -15-fold (Table II). However, the cofactor function of H 5 and H 70 partially or completely restored the impairment in the reactivities of the mutants with thrombin. The extent of impairment in the reactivity of des-VT with thrombin was nearly identical to the extent of improvement in the reactivity of this mutant with fXa. Such a property resulted in the interesting observation that the rate accelerating effect of H 5 in fXa (35-fold) and thrombin (44-fold) inhibition by this mutant was essentially similar (Tables I and II). Consistent with the results observed for fXa, the extent of the template effect of H 70 (ratio of H 70 to H 5 ) in thrombin inhibition by the AT derivatives was not significantly altered. Taken together, these results suggested that reducing the length of the reactive site loop of AT by one and two residues has a reciprocal effect on the ability of the serpin to inhibit the two proteinases in the absence of cofactors, thus improving the inhibition rates with fXa and impairing them with thrombin.
Unlike all other mutants, the reactivity of des-R with fXa was not improved and was impaired most with thrombin. Based on previous results of our own (24) as well as those of others (25), it is believed that Arg 399 interacts productively with the acidic Glu 39 (chymotrypsin numbering) of both fXa and thrombin, thus its deletion impairs the rate of reaction with both proteinases. In support of this, the reactivity of wild type AT with the Glu 39 3 Lys mutant of thrombin (24) was impaired ϳ5-10-fold independent of H 5 (data not shown). Thus, the lack of improvement in the reactivity of this mutant with fXa is irrelevant for the conformational state of the reactive site loop of the mutant serpin.
Consistent with previous results (10,22), a bell-shaped dependence on the H 70 concentrations (optimal 20 -200 nM) was observed for all AT derivatives in reaction with both fXa and thrombin (data not shown). Similar to previous results with the wild type AT (10), inhibition stoichiometries of ϳ1 in the absence of heparin, and ϳ1-1.5 in the presence of heparin, were observed for all mutant serpins in reaction with both fXa and thrombin. This was consistent with the observation that both proteinases formed SDS-stable complexes with all AT derivatives to a similar extent (data not shown), suggesting that the mutagenesis has not affected the substrate pathway of the reactions.

DISCUSSION
It has been hypothesized that AT exists in equilibrium between two conformationally active and inactive states (6,8,17). In the inactive state, which predominates the solution equilibrium of the serpin, the two N-terminal P14 and P15 residues of the reactive site loop of AT are inserted into the A-␤-sheet of the molecule in the native conformation of the serpin (6,8). This unique structural feature, which is only observed in AT renders the serpin inactive toward reaction with fXa (8), but it is not influential in reaction with thrombin (10). Recent crystal structure determination of AT in complex with H 5 suggests that the binding of H 5 on the D-helix of the serpin results in expulsion of the two N-terminal P14 and P15 residues (6), which is accompanied by a 200 -300-fold enhancement in the reactivity of AT with fXa with no significant effect on its reaction with thrombin (10). The molecular basis for differential reactivity of fXa and thrombin with the native and activated conformations of AT is not known. Relative to ␣ 1 -antitrypsin, the reactive site loop of AT has three additional residues, Arg 399 , Val 400 , and Thr 401 , at the C-terminal PЈ end of the loop (Fig. 1). Since the conformation of the reactive site loop in AT is mobile and linked to the heparin-binding D-helix of the serpin (3,8), it was hypothesized in this study that there might be a size limitation for the length of the loop between the A and C sheets, and any excess in its length causes a constraint on the loop, leading to insertion of two N-terminal residues into ␤-sheet A and thus preventing the reactive site loop from accepting an optimal conformation to fit into the active site pocket of fXa. To test this hypothesis, several single, as well as a double, deletion mutants of AT were prepared and characterized.
Consistent with the hypothesis, kinetic analysis suggested that deletion of two residues from the reactive site loop of AT shifted the equilibrium by at least an order of magnitude in favor of the activated state in the mutant serpin. This is derived from the observation that the des-VT mutant inhibited fXa 12-fold better than the wild type serpin in the absence of H 5 , but at a comparable rate in the presence of the cofactor. Interestingly, the reactivity of the double mutant with thrombin was also impaired by the same order of magnitude in the absence, but not in the presence of H 5 . These results suggest that the interaction of thrombin with AT is sensitive to the length of the reactive site loop of wild type serpin, but not to its conformational state (either preinserted or exposed). It follows, therefore, that the native equilibrium fraction of the des-VT mutant, which has a shorter reactive site loop, exhibited impaired reactivity with thrombin only in the absence of H 5 . This implies that thrombin, similar to fXa, can react only with the activated conformation of the mutant serpin, since the length of the exposed loop in the mutant des-VT in the presence of H 5 is similar to that in wild type AT in the native conformation. Therefore, relative to fXa, the basis for the higher reactivity of thrombin with the native serpin lies in the ability of thrombin, but not fXa to react with the loop preinserted AT, independent of H 5 . This appears to be due to the PЈ insertion residues giving the reactive site loop of AT an optimal length and mobility to fit into the deep canyon-like active site pocket of thrombin in the native conformation. These results strongly support the conformational equilibrium hypothesis of AT and for the first time also provides an explanation for why fXa and thrombin differentially react with the native and heparin-activated conformations of the serpin.
The observation that a longer reactive site loop of AT is primarily responsible for the slow reactivity of the serpin with fXa, but is also required for AT to regulate the proteolytic activity of thrombin in the native conformation, has important implications for the physiological function of AT in circulation. AT, by virtue of its unique reactive site loop, can inhibit thrombin in the native conformation, but requires activation before it can effectively inhibit fXa. It follows, therefore, that fXa in vivo can be inhibited primarily by the small fraction of AT that is bound to the heparin-like molecules on the vasculature (6,26). Otherwise, in the presence of a high concentration of circulating active AT (2.3 M), maintenance of normal hemostasis would have been compromised due to rapid inactivation of fXa, which is generated at a minute concentration (at picomolar range) during the propagation phase of the coagulation cascade (27). On the other hand thrombin may be generated at very high concentrations (up to several hundred nM) under various (patho)physiological conditions (28). If AT were inactive toward thrombin in its native conformation, then thrombosis would have been a common problem. The unique reactive site loop of AT therefore enables the serpin to effectively regulate the activity of several clotting enzymes without compromising hemostasis.
Finally, the results of this study suggest that, unlike the mutagenesis of the P site reactive site loop of AT, which interferes with the mechanics of loop insertion, thereby converting all highly reactive mutants to efficient substrates (9), the mutagenesis of the PЈ residues of the C-terminal end of the reactive site loop of AT does not influence the substrate pathway of the reaction. Rather, it constitutively activates the mutant serpin. Such a mutagenesis approach holds a great promise for development of recombinant AT derivatives capable of specifically inhibiting fXa independent of polysaccharide cofactors. The second-order rate constants (k 2 ) in both the absence and presence of H 5 were determined by incubation of 1 nM fXa with 5-500 nM AT derivatives in TBS containing 0.1 mg/ml BSA, 0.1% PEG 8000, and 2.5 mM Ca 2ϩ for 0.5-30 min as described under "Experimental Procedures." The values in the presence of optimal concentration of H 70 (ϳ50 nM) were determined by the same procedure except that 0.2 nM fXa was incubated with AT derivatives for 10 -20 s. All values are averages of at least three independent measurements Ϯ S.D. The second-order rate constants (k 2 ) in both the absence and presence of H 5 were determined by incubation of 1 nM thrombin with 50 -500 nM AT derivatives in TBS containing 0.1 mg/ml BSA, 0.1% PEG 8000, and 2.5 mM Ca 2ϩ for 2-60 min as described under "Experimental Procedures." The values in the presence of optimal concentration of H 70 (ϳ50 nM) were determined by the same procedure except that 0.2 nM enzyme was incubated with AT derivatives for 10 -20 s. All values are averages of at least three independent measurements Ϯ S.D. The Reactive Site Loop of Antithrombin