Mechanism of Heparin Activation of Antithrombin

To determine the role of individual saccharide residues of a specific heparin pentasaccharide, denoted DEFGH, in the allosteric activation of the serpin, antithrombin, we studied the effect of deleting pentasaccharide residues on this activation. Binding, spectroscopic, and kinetic analyses demonstrated that deletion of reducing-end residues G and H or nonreducing-end residue D produced variable losses in pentasaccharide binding energy of approximately 15-75% but did not affect the oligosaccharide's ability to conformationally activate the serpin or to enhance the rate at which the serpin inhibited factor Xa. Rapid kinetic studies revealed that elimination of the reducing-end disaccharide marginally affected binding to the native low-heparin-affinity conformational state of antithrombin but greatly affected the conversion of the serpin to the activated high-heparin- affinity state, although the activated conformation was still favored. In contrast, removal of the nonreducing- end residue D drastically affected the initial low-heparin-affinity interaction so as to favor an alternative activation pathway wherein the oligosaccharide shifted a preexisiting equilibrium between native and activated serpin conformations in favor of the activated state. These results demonstrate that the nonreducing-end residues of the pentasaccharide function both to recognize the native low-heparin-affinity conformation of antithrombin and to induce and stabilize the activated high-heparin-affinity conformation. Residues at the reducing-end, however, poorly recognize the native conformation and instead function primarily to bind and stabilize the activated antithrombin conformation. Together, these findings establish an important role of the heparin pentasaccharide sequence in preferential binding and stabilization of the activated conformational state of the serpin.

To determine the role of individual saccharide residues of a specific heparin pentasaccharide, denoted DEFGH, in the allosteric activation of the serpin, antithrombin, we studied the effect of deleting pentasaccharide residues on this activation. Binding, spectroscopic, and kinetic analyses demonstrated that deletion of reducing-end residues G and H or nonreducing-end residue D produced variable losses in pentasaccharide binding energy of ϳ15-75% but did not affect the oligosaccharide's ability to conformationally activate the serpin or to enhance the rate at which the serpin inhibited factor Xa. Rapid kinetic studies revealed that elimination of the reducing-end disaccharide marginally affected binding to the native low-heparin-affinity conformational state of antithrombin but greatly affected the conversion of the serpin to the activated high-heparinaffinity state, although the activated conformation was still favored. In contrast, removal of the nonreducingend residue D drastically affected the initial low-heparin-affinity interaction so as to favor an alternative activation pathway wherein the oligosaccharide shifted a preexisiting equilibrium between native and activated serpin conformations in favor of the activated state. These results demonstrate that the nonreducing-end residues of the pentasaccharide function both to recognize the native low-heparin-affinity conformation of antithrombin and to induce and stabilize the activated high-heparin-affinity conformation. Residues at the reducing-end, however, poorly recognize the native conformation and instead function primarily to bind and stabilize the activated antithrombin conformation. Together, these findings establish an important role of the heparin pentasaccharide sequence in preferential binding and stabilization of the activated conformational state of the serpin.
Antithrombin, a member of the serpin (serine proteinase inhibitor) superfamily of proteins, is the primary physiological inhibitor of blood coagulation proteinases, especially thrombin and factor Xa (for reviews, see Refs. 1 and 2). Antithrombin inhibits these proteinases by forming tight, possibly covalent, complexes in which an exposed reactive center loop of the inhibitor is bound at the enzyme active site in the manner of a substrate. The rate of this inhibition is slow in comparison with most other serpin-proteinase reactions. However, the inhibition rate greatly increases up to several thousand-fold in the presence of the glycosaminoglycan, heparin (1,2). This property of heparin is responsible for the widespread use of the polysaccharide as an anticoagulant drug.
The extent to which heparin accelerates antithrombin-proteinase reactions is dependent on a unique pentasaccharide sequence (DEFGH 1 ; see Fig. 1), present in about one-third of naturally occurring heparin chains (3)(4)(5)(6). Heparin molecules containing this sequence bind antithrombin with high affinity and induce an activating conformational change in the inhibitor (7)(8)(9)(10). Those polysaccharide chains that lack this sequence bind antithrombin with ϳ1000-fold lower affinity and only weakly activate the inhibitor (11). Whereas the antithrombin conformational change is necessary and sufficient for the inhibitor to accelerate the inactivation of factor Xa, it is not sufficient for accelerated thrombin inhibition (12). The latter acceleration additionally requires a longer heparin chain to bridge the proteinase and the inhibitor in a ternary complex (12)(13)(14)(15)(16).
The heparin binding site of antithrombin has been tentatively mapped to residues of helix A, helix D, and the Nterminal region of the inhibitor, which are contiguous in the x-ray structure (1,2,(17)(18)(19)(20)(21). Structure-activity studies of variants of the pentasaccharide have additionally revealed key functional groups in the oligosaccharide responsible for tight binding to the inhibitor. Four anionic groups, namely two Nsulfates and two O-sulfates in the glucosamine residues D, F, and H of the pentasaccharide (Fig. 1) are critical for complex formation and are thought to make ionic interactions with basic residues in the heparin binding site of antithrombin (6,14,22,23,31,41). Additionally, nonionic interactions make a considerable contribution (about 60% at physiological pH and ionic strength) to the total binding energy (12). The pentasaccharide binds to antithrombin in a two-step process, in which a low-affinity recognition complex is first formed, which then induces an activating conformational change in the inhibitor, leading to a high-affinity complex (10,12). Structural and functional studies suggest that the activating conformational change in antithrombin involves structural changes in the heparin binding site, which transform the reactive center loop from a partially buried to a fully exposed conformation (17, 21, 24 -26).
The mechanism of heparin activation of antithrombin is as yet ill-defined at the molecular level. Although residues in the pentasaccharide and in antithrombin that are critical for their high-affinity interaction are known, it has not been elucidated which of these residues are involved in the initial recognition and in the subsequent allosteric activation of the serpin. To assess the roles of individual pentasaccharide residues in the two-step activation mechanism, we have studied the effect of truncating pentasaccharide residues at either the reducing or the nonreducing-end (see Fig. 1) on the ability of the oligosaccharide to bind and induce the conformational change in antithrombin. These effects have been correlated with the ability of the truncated pentasaccharides to activate the inhibitor for rapid inhibition of factor Xa. In agreement with preliminary studies of antithrombin interactions with these oligosaccharides (38), the present study establishes that the nonreducingend trisaccharide, DEF, of the pentasaccharide is capable of fully activating antithrombin and, consequently, that the reducing-end residues are not essential for this activation, although they stabilize the activated conformation. The study further demonstrates a critical role of the pentasaccharide sequence in preferential binding of the activated conformational state of the serpin, in keeping with the pentasaccharide functioning as a classical allosteric activator.

MATERIALS AND METHODS
Proteins-Human antithrombin was purified from outdated plasma as described previously (15,27). Molar concentrations of the inhibitor were calculated from absorbance measurements at 280 nm using a molar absorption coefficient of 37,700 M Ϫ1 cm Ϫ1 (28). Human factor Xa was prepared by activation of purified factor X, followed by purification on SBTI-agarose, as described previously (29). Factor Xa preparations were predominantly the ␣-form as judged by SDS-polyacrylamide gel electrophoresis and were Ͼ90% active by comparisons of active site and protein concentrations (29).
Oligosaccharides-The methyl glycosides of pentasaccharide and pentasaccharide variants were synthesized, and their structures were confirmed as described previously (30,31,42). A 2-10 mM solution of each variant was prepared based on weight in distilled, deionized water. Stoichiometric titrations of high concentrations of antithrombin (at least 10 times the K D ) with DEFGH, DEFGHЈ, and DEFG* (Fig. 1), monitored by the endogenous fluorescence enhancement, showed fair agreement with the weight concentrations. The former concentrations were used for these tighter binding oligosaccharides, whereas concentrations based on weight were used for all other oligosaccharides.
Experimental Conditions-All experiments were conducted at 25°C and in 20 mM sodium phosphate buffer, containing 0.1 mM EDTA and 0.1% (w/v) PEG 8000, adjusted to the required pH. In the absence of added salt, the ionic strengths of buffers at pH 6.0 and 7.4 were 0.025 and 0.05, respectively. Sodium chloride was added to concentrations of 100, 250, 350, 500, 600, and 700 mM to achieve higher ionic strengths. Because of the reported decrease in stability of the inhibitor activity at lower pH (32), antithrombin solutions in pH 6.0 buffer were prepared by freshly diluting the protein (Ͼ100-fold) from concentrated stocks in pH 7.4, I 0.15 buffer. No significant losses in inhibitor activity were noted over the time frame of the experiments performed in this study.
Spectroscopic Studies-Fluorescence emission spectra of antithrombin and its complexes with oligosaccharides were obtained at 25°C in pH 6.0, I 0.025 buffer with 1 M antithrombin and at least 5 M oligosaccharide. The spectra were recorded with an SLM 8000 spectrofluorometer in the ratio mode at 1-nm wavelength intervals with excitation at 280 nm (4-nm bandpass), an emission bandpass of 2 nm, and 10-s integrations of the fluorescence signal at each wavelength. Corrections for Raman bands and any background signal from the buffer were made by subtracting buffer spectra. Control experiments showed no significant fluorescence of the variant oligosaccharides alone at concentrations used in the formation of complexes. The corrected spectra consisted of signal from both the free (Ͻ20%) and bound antithrombin. The normalized fluorescence spectrum of each complex was calculated by subtracting the expected spectrum of free antithrombin, based on measured dissociation constants for complex formation, and scaling up the resultant spectrum to 1 M complex.
Equilibrium Binding Studies-Equilibrium dissociation constants for antithrombin-oligosaccharide complexes were determined by titrating the oligosaccharide into a solution of antithrombin and monitoring the increase in intrinsic protein fluorescence accompanying the interaction, as described previously (9,27). Antithrombin concentrations were in the range of 1-2 times the K D , except for K D values Ͼ1 M, where the concentration was ϳ1 M. The increase in fluorescence signal with increasing oligosaccharide concentration was fit to the quadratic equilibrium binding equation, assuming a 1:1 binding stoichiometry (9,27). The nonionic and ionic contributions to the total pentasaccharide binding energy were resolved by analyzing the NaCl concentration dependence of the observed dissociation constant (K D, obs ) according to the equation (12,16,33), log K D, obs ϭ log K D, nonionic ϩ Z⌿ log [Na ϩ ], where K D, nonionic is the salt-independent dissociation constant, Z is the total number of charge-charge interactions involved in the association of the protein with heparin, and ⌿ is the fraction of monovalent counterions bound per heparin ionic charge that are released upon protein binding. Least-squares analysis of the linear dependence of log K D, obs on log [Na ϩ ] yielded the nonionic component of the binding energy from the intercept, log K D, nonionic , and the value of Z from the slope after dividing by the value of 0.8 for ⌿ (16).
Competitive binding experiments with oligosaccharides DEFG* and FGHٞ ( Fig. 1) were performed at a fixed FGHٞ concentration by monitoring the residual increase in fluorescence as a function of increasing DEFG* concentration. The increase in fluorescence was analyzed by the quadratic binding equation in the same way as for the other oligosaccharide interactions. Competitive binding was assessed by comparing apparent dissociation constants measured with those calculated based on the equation, Rapid Kinetic Studies-The rate of oligosaccharide binding to plasma antithrombin was measured in pH 6.0, I 0.025 buffer at 25°C in an Applied Photophysics stopped-flow fluorometer, as described previously (10,12). Pseudo-first-order conditions in which the oligosaccharide and antithrombin molar concentrations were maintained at a ratio of at least 5:1 and in most cases Ն10:1 were employed. The interaction was monitored from the increase in intrinsic protein fluorescence with an excitation wavelength of 280 nm and an emission filter that transmitted light only at wavelengths above 310 nm. Excitation slits corresponded to an 8-nm bandpass. The fluorescence traces were acquired for about 10 half-lives and could be satisfactorily fit by a single exponential function for DEFGH, DEFG*, DEF, and DEFGHЈ interactions, which provided the amplitude of the fluorescence change and the observed pseudo-first-order rate constant, k obs . The traces for variant EFGHЉ interacting with antithrombin could only be fit satisfactorily by a double exponential function that yielded two pseudo-first-order rate constants and two fluorescence amplitudes. Typically, 12-18 fluorescence traces were acquired for each set of concentrations and averaged.
Factor Xa Inhibition Studies-The accelerating effects of oligosaccharides on the kinetics of antithrombin inhibition of factor Xa were measured under pseudo-first-order conditions. A fixed 10 nM concentration of factor Xa was incubated with 2-12 M antithrombin and oligosaccharide in pH 6.0 buffer in a 50-l total reaction volume. The oligosaccharide concentration was typically less than 20 nM, except for EFGHЉ and FGHٞ, which were present in the 20 -240 nM range, and at least three different oligosaccharide concentrations were examined. After incubation for various times, reactions were quenched with 950 l of 100 M Spectrozyme FXa in 20 mM sodium phosphate buffer, containing 100 mM sodium chloride, 0.1 mM EDTA, 0.1% (w/v) PEG 8000, at pH 7.4. The residual factor Xa activity was then measured from the initial rate of substrate hydrolysis at 405 nm. The observed pseudofirst-order rate constant (k obs ) at each oligosaccharide concentration was determined from the exponential decay of factor Xa activity (27). The second-order rate constant for factor Xa inhibition by antithrombin alone (k uncat ) and those for the inhibition by antithrombin-oligosaccharide complexes (k H ) were obtained by least-squares analysis of the linear dependence of k obs on antithrombin-oligosaccharide complex concentration according to the equation,  Table I.

Characterization of the Binding of Truncated Pentasaccharides to Antithrombin by Intrinsic Protein Fluorescence Changes-
The series of truncated pentasaccharides shown in Fig. 1 were used to assess the role of individual saccharide residues in the allosteric activation of antithrombin. This series consisted of tetrasaccharide DEFG* and trisaccharide DEF, representing reducing-end truncations, and tetrasaccharide EFGHЉ and trisaccharide FGHٞ, representing nonreducing-end truncations. These two groups of oligosaccharides contained individual saccharides whose structure was slightly modified from those in pentasaccharide DEFGH, as denoted by the primes and asterisks, due to the difficult synthesis of sufficient quantities of the systematic series of truncated pentasaccharides. Extensive structure-activity studies of pentasaccharide DEFGH have shown that such structural modifications mini-mally affect pentasaccharide binding and activation of antithrombin in most cases (41,42). 2 For the one case in which a structural modification did significantly enhance pentasaccharide activity, namely the addition of a 3-O-sulfate on saccharide H in EFGHЉ, a reference pentasaccharide containing this structural modification (DEFGHЈ) served as a control (42), thereby allowing the effect of the D residue to be determined.
Under the conditions previously used to study the antithrombin-pentasaccharide interaction (pH 7.4, I 0.15, 25°C), binding of the tetrasaccharides but not the trisaccharides to antithrombin was measurable from the saturable enhancement in antithrombin fluorescence produced when the inhibitor was titrated with up to 100 M oligosaccharide. Lowering the ionic strength to Յ0.05 or the pH of the buffer to 6.0 sufficiently enhanced the affinity of the oligosaccharides for antithrombin so that binding constants were measurable for all oligosaccharides under the same conditions. Table I summarizes the dissociation constants and maximal fluorescence changes measured for the binding of the truncated pentasaccharides to antithrombin under these various conditions. At pH 7.4, I 0.15, 25°C, tetrasaccharide DEFG* bound antithrombin with a 300-fold lower affinity than pentasaccharide DEFGH (⌬⌬G 0 3.5 kcal/mol), in reasonable agreement with previous reports of a ϳ100-fold lower affinity of tetrasaccharide DEFG than DEFGH (23) and of the small contribution of the 2-O-sulfate of saccharide G missing in DEFG* (41). The decreased affinity of DEFG* for antithrombin is therefore mostly due to deletion of the reducing-end saccharide H. Tetrasaccharide EFGHЉ bound antithrombin with a 60,000-fold lower affinity than reference pentasaccharide DEFGHЈ (⌬⌬G 0 ϳ6 kcal/ mol). That this loss in binding energy was due to deletion of the nonreducing-end D saccharide and not to the replacement of the N-sulfate of residue H with an O-sulfate in EFGHЉ was suggested by the finding that DEFGHЈ with an O-sulfate instead of an N-sulfate in residue H bound antithrombin with an affinity indistinguishable from DEFGHЈ (not shown). Previous studies similarly found a much greater contribution of residue D than residue H to pentasaccharide binding (6,23).
The ϳ2.5 kcal/mol greater contribution of residue D than residue H to pentasaccharide binding energy was increased to ϳ4 kcal/mol when the ionic strength was lowered from I 0.15 to 0.05 at pH 7.4 and was further augmented to ϳ6.0 kcal/mol when both the pH and ionic strength were lowered to pH 6.0, I 0.025. The D residue interaction thus appears to be largely responsible for the increased pentasaccharide binding energy at lower pH. Contrasting the large effects of deleting residues D and H on the pentasaccharide binding energy, the further deletion of residues E and G produced relatively smaller losses in binding energy (Table I), again in agreement with previous studies (6,23).
Binding to antithrombin of the oligosaccharides DEFG* and DEF lacking the reducing-end residues of the pentasaccharide enhanced the intrinsic protein fluorescence of the serpin to the same extent (ϳ30%) as that of the reference pentasaccharides DEFGH and DEFGHЈ under all conditions examined (Fig. 2). By contrast, binding of pentasaccharides missing the nonreducing-end residues induced different extents of fluorescence change than the reference pentasaccharides. Tetrasaccharide EFGHЉ thus consistently produced a larger fluorescence enhancement of ϳ45%, whereas trisaccharide FGHٞ induced a much reduced fluorescence enhancement of 7-14%. Similar results were obtained from the maximal fluorescence enhance- Sulfate groups marked with asterisks in the heparin pentasaccharide, DEFGH, are critical for tight binding to antithrombin (22). Truncated pentasaccharides designated with primes or asterisks have individual saccharides whose structure differs from pentasaccharide DEFGH. These differences either minimally affect pentasaccharide activity or enhance this activity as discussed under "Results" (41,42). The modification which enhances activity, an additional 3-O-sulfate on residue H in EFGHЉ, was controlled for by studying pentasaccharide DEFGHЈ possessing this same modification. ments attained in equilibrium binding titrations (Table I). The wavelength of maximum emission intensity for antithrombin complexed with all oligosaccharides was very similar to that for antithrombin alone, in agreement with previous reports (9).
Origin of the pH-dependent Change in Pentasaccharide Binding Affinity-To determine whether the higher pentasaccharide affinity for antithrombin at pH 6.0 than at pH 7.4 was due to additional charge-charge interactions, the dissociation constant for the antithrombin-pentasaccharide interaction was measured as a function of the NaCl concentration both at pH 7.4 and 6.0. The log K D, obs increased linearly with increasing log [Na ϩ ] both at pH 7.4 and 6.0 (not shown), in accordance with the expected behavior of the protein-heparin interaction (12,16,33). The slopes of these lines indicated that 5.2 Ϯ 0.1 and 4.0 Ϯ 0.2 charge-charge interactions contributed to the binding at pH 6.0 and 7.4, respectively, whereas the intercepts gave values of Ϫ4.92 Ϯ 0.05 and Ϫ4.58 Ϯ 0.02 for log K D,nonionic at these respective pH values (see "Materials and Methods"). The values obtained at pH 7.4 are similar to those reported previously (12). These results indicate that the increase in pentasaccharide binding affinity when the pH is lowered from 7.4 to 6.0 results primarily from ϳone additional ionic interaction at pH 6.0, which is presumably made by residue D.
Rapid Kinetic Studies of DEFGH, DEFG*, and DEF Binding to Antithrombin-The kinetics of binding of oligosaccharides DEFGH, DEFG*, and DEF to antithrombin was studied by continuously monitoring the protein fluorescence changes accompanying their binding by stopped-flow fluorimetry under pseudo-first-order conditions in pH 6.0 buffer. Binding was observable in the stopped-flow time frame as a monophasic exponential process in all cases. The dependence on oligosaccharide concentration of the observed pseudo-first-order rate constant (k obs ) for binding of these oligosaccharides to antithrombin is shown in Fig. 3. In all cases, a progressive saturation of k obs with increasing oligosaccharide concentration was observed, indicative of a two-step binding process, in which an initial binding of the oligosaccharide to antithrombin induces a subsequent conformational change in the inhibitor (induced-fit pathway in Scheme 1) (10,12). The data of Fig. 3 were satisfactorily fit by the rectangular hyperbolic Equation 1, which characterizes the induced conformational change pathway (10).
In this equation, K 1 is the dissociation constant for the initial binding step, k 2 and k Ϫ2 are the forward and reverse rate constants for the conformational change step, and [H] o is the initial heparin oligosaccharide concentration. Table II shows the results of nonlinear regression fitting of Fig. 3 data by Equation 1. The k Ϫ2 values obtained from the fitted intercepts of these saturation curves were not distinguishable from 0 in the case of the intact pentasaccharide but were clearly nonzero for the truncated pentasaccharides, DEFG* and DEF. k Ϫ2 for the DEFGH interaction was calculated from the kinetic constants, K 1 and k 2 , and the dissociation constant, K D, obs , from the relation, k Ϫ2 ϭ K D, obs ϫ k 2 /K 1 . 3 For the truncated pentasaccharides where k Ϫ2 was measurable, a comparison of calculated and measured values of K D, obs showed excellent agreement (Table II).
Comparison of the kinetic constants obtained for the pentasaccharide at pH 6.0, I 0.025 with those previously measured at pH 7.4, I 0.15 (12) indicated that the increase in pentasaccharide binding affinity at lower pH and ionic strength resulted primarily from an effect on the conformational change step and specifically to a ϳ1000-fold reduction in the rate constant for  Spectra (average of two acquisitions) were recorded in pH 6.0, I 0.025 buffer at 25°C, corrected for buffer background and for the emission due to free antithrombin, and then normalized to the same concentration based on measured dissociation constants in Table I. See "Materials and Methods" for further details. reversal of the conformational activation step (k Ϫ2 ). The affinity for the initial antithrombin-pentasaccharide recognition complex (K 1 ) was also enhanced about 20-fold by the change in conditions, whereas the forward rate constant for conformational activation, k 2 , was not affected.
Tetrasaccharide DEFG* lacking the pentasaccharide reducing-end residue showed a modestly decreased affinity of the recognition complex of 5-fold and practically no effect on the rate of conformational activation. In contrast, the rate at which conformational activation was reversed substantially increased by more than 1000-fold. Further elimination of residue G from the reducing-end similarly only modestly affected the affinity of the initial recognition complex (ϳ2-fold) or the rate of conformational activation but greatly increased the rate of reversal of the conformational activation step a further ϳ20-fold. The primary effects of deleting reducing-end residues, G and H, on k Ϫ2 resulted in a progressive destabilization of the conformational equilibrium for inhibitor activation, governed by the ratio k 2 / k Ϫ2 . This ratio thus decreased from ϳ7 ϫ 10 5 for the intact pentasaccharide DEFGH to 190 for tetrasaccharide DEFG* and to 8 for trisaccharide DEF (Table II).
Rapid Kinetic Studies of DEFGHЈ, EFGH", and FGHٞ Binding to Antithrombin-The observed pseudo-first-order rate constant, k obs , for binding of pentasaccharide DEFGHЈ to antithrombin at pH 6.0, I 0.025, 25°C progressively increased to a limiting value with increasing concentration of oligosaccharide (Fig. 4A), similar to the pattern observed for binding of oligosaccharides DEFGH, DEFG*, and DEF to antithrombin. Fitting of this data by Equation 1 for the induced conformational change binding mechanism resulted in a K 1 value of 1.2 Ϯ 0.4 M, a k 2 value of 1210 Ϯ 230 s Ϫ1 , and a k Ϫ2 value indistinguishable from 0. k Ϫ2 was estimated to be ϳ10 Ϫ5 s Ϫ1 based on measured values of K 1 and k 2 and an approximated value for K D, obs (Table I). The K 1 and k 2 parameters for DEFGHЈ are very similar to those for heparin pentasaccharide, DEFGH, indicating that the greater affinity of DEFGHЈ than of DEFGH for antithrombin is due to a reduced k Ϫ2 value. Both 3-Osulfation of reducing end residue H and deletion of this residue therefore affect the same kinetic parameter. Replacement of the N-sulfate of residue H in DEFGHЈ with an O-sulfate had no effect on the binding kinetics (not shown).
Stopped-flow traces of the binding to antithrombin of tetrasaccharide EFGHЉ truncated at the nonreducing-end at pH 6.0 showed two striking differences from the kinetics of binding of pentasaccharides DEFGH and DEFGHЈ and of oligosaccharides DEFG* and DEF truncated at the reducing-end (Fig. 4B). First, individual traces were not fit satisfactorily by a single exponential function but instead required a double exponential function to achieve a good fit. Second, k obs for the fast phase I fluorescence changes accompanying the binding interaction by stoppedflow fluorimetry at 25°C in pH 6.0, I 0.025 buffer, as detailed in "Materials and Methods." Solid lines represent nonlinear least-squares fits of the data by Equation 1, which describes the induced conformational change binding mechanism (Scheme 1).

FIG. 3.
Effect of truncating reducing-end residues of the pentasaccharide on the kinetics of pentasaccharide binding to antithrombin. Shown is the oligosaccharide concentration dependence of pseudo-first-order rate constants (k obs ) for the binding to antithrombin of pentasaccharide DEFGH (A) and of truncated variants DEFG* (B) and DEF (C). k obs values were measured by following intrinsic protein SCHEME 1

Mechanism of Heparin Activation of Antithrombin
progressively decreased rather than increased with increasing EFGHЉ concentration, whereas k obs for the slow phase II remained practically independent of the oligosaccharide concentration (Fig. 4C). That the biphasic changes in fluorescence indeed reflected the binding of tetrasaccharide EFGHЉ to antithrombin was confirmed by the observation that the fluorescence amplitudes for each phase showed parallel saturable increases with EFGHЉ concentration (Fig. 4D), which could be fit by similar K D values of 6.0 Ϯ 0.8 M for phase I, 9.7 Ϯ 1.1 M for phase II, and 7.2 Ϯ 0.7 M for the sum of the two phases, values that were indistinguishable from the K D measured by equilibrium binding titrations (Table I).
The hyperbolic decrease in the phase I k obs with increasing EFGHЉ concentration indicated that EFGHЉ was binding and activating antithrombin in this phase by an alternative twostep pathway, in which oligosaccharide binding occurred only subsequent to the activation of the inhibitor in a pre-equilibrium conformational change step (Scheme 1, pre-equilibrium pathway) (34,35). In keeping with such a mechanism, the ⌬F max of ϳ30% associated with this phase was similar to that produced by pentasaccharide activation of antithrombin. The appearance of the additional kinetic phase II indicated that the two-step conformational activation of antithrombin in phase I was followed by a further conformational change in the inhibitor in phase II, occurring with a rate constant of ϳ1.5 s Ϫ1 . Phase I data were analyzed by Equation 2 for a ligand-induced protein conformational change proceeding through the preequilibrium pathway (34,35).
In this equation, k 3 is the rate constant for pre-equilibrium conformational activation of antithrombin in the absence of heparin, k Ϫ3 is the reverse rate constant for this activation, and K 4 is the dissociation constant for oligosaccharide binding to the activated antithrombin formed in the pre-equilibrium step (Scheme 1). Fitting of the data in Fig. 4C by this equation yielded a limiting k obs at high oligosaccharide concentration, corresponding to k 3 , of 6.4 Ϯ 0.5 s Ϫ1 . Large errors associated with measurements of k obs at lower oligosaccharide concentrations, due to smaller fluorescence changes, resulted in an indeterminate value of the ordinate intercept and a consequent inability to accurately determine k Ϫ3 and K 4 parameters.
No time-dependent changes in fluorescence were observable for the binding of trisaccharide FGHٞ to antithrombin in pH 6.0 buffer despite the use of oligosaccharide concentrations, resulting in significant complex formation. However, an increase in fluorescence was evident immediately after mixing antithrombin with the trisaccharide, as shown by comparison with the base-line fluorescence established by mixing antithrombin with buffer. These results indicated that binding of FGHٞ to antithrombin was complete within the 1.5-ms dead time of the stopped-flow instrument.
Competitive Binding of FGHٞ and DEFG* to Antithrombin-To determine whether FGHٞ bound to the same site on antithrombin as the pentasaccharide, competitive binding experiments were performed. Tetrasaccharide DEFG* was used as the competitor because of the much greater affinity of DEFGH for antithrombin than of FGHٞ at pH 6.0. Equilibrium binding of DEFG* to antithrombin in the absence and presence of fixed concentrations of FGHٞ at ϳ1 and ϳ4 times the K D for the FGHٞ-antithrombin interaction was monitored from the ϳ20% greater increase in inhibitor fluorescence produced by DEFG* than by FGHٞ binding (Table I) Accelerating Effects of Truncated Pentasaccharides on Factor Xa Inhibition by Antithrombin-The second-order rate constant for the uncatalyzed inhibition of factor Xa by antithrombin (k uncat ) was reduced by ϳ15-fold at pH 6.0 from that at pH 7.4 (Table III), consistent with the increased levels of catalytically inactive serine proteinase resulting from protonation of histidine 57 of the catalytic triad (36). Second-order rate constants for the accelerated inhibition of factor Xa by antithrombin-oligosaccharide complexes (k H ) at pH 6.0 were evaluated from the slopes of the linear dependence of the pseudo-firstorder inhibition rate constant on the concentration of antithrombin-oligosaccharide complex (see "Materials and Methods") and are presented in Table III. Antithrombin complexes with DEFGH, DEFG*, and DEF showed indistinguishable k H values of 4 -5 ϫ 10 4 M Ϫ1 s Ϫ1 , representing about 300-fold rate enhancements, indicating that deletion of reducing-end residues G and H did not affect the ability of the oligosaccharides to activate antithrombin. Similar k H values for antithrombin complexes with DEF and DEFGH were also found at pH 7.4 (Table III). Complexes of DEFGHЈ and EFGHЉ with antithrombin resulted in k H values similar to that of the antithrombin-DEFGH complex, indicating no effect of H residue 3-O-sulfation or D residue deletion on antithrombin activation. In contrast, binding to antithrombin of trisaccharide FGHٞ significantly reduced k H to ϳ10% of the pentasaccharide accelerated rate constant, indicating that removal of both residues D and E from the nonreducing-end greatly compromised the ability to activate antithrombin.

TABLE II
Kinetic parameters for the interaction of antithrombin with heparin oligosaccharides Kinetic parameters for the binding of oligosaccharides to antithrombin by the induced conformational change pathway in Scheme 1 were determined from fits of the dependence of k obs on oligosacharide concentration by the hyperbolic Equation 1 as shown in Fig. 3. See "Materials and Methods" for details.

Oligosaccharide
Conditions  Table I. b Calculated from the relation, K D,obs ϭ K 1 ϫ k Ϫ2 /(k 2 ϩ k Ϫ2 ). This reduces to K 1 ϫ k Ϫ2 /k 2 when k Ϫ2 Ӷ k 2 . c Data taken from previous work (12).

DISCUSSION
We have elucidated the role of individual saccharide residues of a specific heparin pentasaccharide in the allosteric activation of the serpin, antithrombin, by determining the effects of deleting pentasaccharide residues on this activation. These effects had to be analyzed at a lower ionic strength and pH than was used in past studies to reliably measure the affinities and rates of interaction of the truncated pentasaccharides with antithrombin. Such conditions did not appear to alter the activation mechanism, since the interaction of the pentasaccharide with antithrombin under these conditions was characterized by the same two-step binding process, fluorescence enhancement, and acceleration of factor Xa inhibition previously found at physiological ionic strength and pH (10,12). Notably, the bind-ing energy contribution of residue D was more dominant under these experimental conditions due to an additional electrostatic interaction of residue D with antithrombin. The origin of this additional electrostatic binding energy is presently unclear, since histidine residues close to the heparin binding site do not appear to participate in ionic interactions with heparin at pH 6 (37).
While both reducing and nonreducing-end residues contribute to pentasaccharide binding energy, our results show that these residues are not absolutely essential to activate antithrombin for rapid inhibition of factor Xa. Thus, all truncated pentasaccharides, with the exception of oligosaccharide FGHٞ, which appears to bind antithrombin nonproductively (see below), were capable of inducing the tryptophan fluorescence The solid lines are fits of the data by a double-exponential function. Panel C shows the variation of the two pseudo-first-order rate constants obtained from the double-exponential fits of kinetic data for the antithrombin-EFGHЉ interaction as a function of the EFGHЉ concentration. The solid line represents the fit of the data by Equation 2 for the pre-equilibrium conformational change binding mechanism in Scheme 1. Panel D shows the increase in amplitudes for the two kinetic phases with increasing EFGHЉ concentration at 0.3 M antithrombin. Error bars represent Ϯ S.E. The solid lines depict the fit of the data by the quadratic equilibrium binding equation (27). enhancement associated with antithrombin activation and of maximally accelerating antithrombin inhibition of factor Xa at saturation. Previous assertions that residues D and H are essential for full conformational activation of the serpin (39) were based on indirect measurements of the rate of factor Xa inactivation at a single concentration of inhibitor-oligosaccharide complex. In the present study, direct measurements of factor Xa inactivation over a wide range of oligosaccharide concentrations approaching inhibitor saturation clearly show that deletion of residues D or H alone or G and H together has no significant effect on pentasaccharide activation of antithrombin.
Rapid kinetic studies allowed us to distinguish whether defects in pentasaccharide binding to antithrombin were due to a reduced ability of the pentasaccharide to bind the native lowheparin-affinity state of antithrombin or to induce the inhibitor into the activated high-heparin-affinity state (Scheme 1, induced-fit pathway) (10,12). Deletion of reducing-end residue H or both G and H from the pentasaccharide only modestly affected binding to the low-heparin-affinity state of the serpin but greatly impaired conformational activation to the highheparin-affinity-state. This impairment was due to increases in the rate constant for reversal of the conformational activation step (k Ϫ2 in Scheme 1), which destabilized the activated conformation relative to the native conformation by as much as ϳ100,000-fold (Table II). Such findings suggest that the binding energy of residues G and H is utilized to form interactions mostly or exclusively with the activated antithrombin conformation. The loss of such interactions would thus be expected to increase the rate constant for reversal of the conformational change, since this rate constant should be inversely related to the number of interactions that must be disrupted in going from the activated conformation back to the native conformation. Although loss of residues G and H greatly destabilized the activated antithrombin conformation, the conformational equilibrium constant, k 2 /k Ϫ2 , remained large enough (Ն8) to still favor the activated conformation (Ն89%). This explains why oligosaccharides lacking residue H or both G and H appeared to fully activate antithrombin for accelerated factor Xa inhibition when present at saturating levels. While residues G and H of the pentasaccharide are thus not required for conformational activation, they are nevertheless critical for stabilizing the activated conformation induced by residues D, E, and F.
Contrasting with these effects of deleting pentasaccharide reducing-end residues, deletion of the nonreducing-end residue D resulted in an altered mechanism of oligosaccharide binding and activation of antithrombin. This altered binding mecha-nism was characterized by a preferential binding of the oligosaccharide to the small amount of conformationally activated antithrombin in pre-equilibrium with native unactivated antithrombin, which thereby resulted in the unactivated inhibitor being pulled into the activated state (Scheme 1, pre-equilibrium pathway) (34,35). Binding by this alternative pathway was complete in a rapid kinetic phase associated with a normal fluorescence enhancement and resulted in a normal acceleration of factor Xa inhibition. The observation that a further slow increase in antithrombin fluorescence followed the initial rapid fluorescence change with a rate that was independent of oligosaccharide concentration suggests that a further conformational change is induced in antithrombin following its activation by the variant pentasaccharide lacking residue D (EFGHЉ).
The different pathway and altered mode of EFGHЉ binding to antithrombin suggests a key role for the D residue in pentasaccharide recognition of the native antithrombin conformation and in anchoring of the pentasaccharide in the heparin binding site of both native and activated inhibitor conformations. Thus, the failure of EFGHЉ to bind by the induced conformational change mechanism implies a major defect in binding to the native inhibitor conformation. If much of the binding energy of residue D under these conditions is utilized in forming the initial recognition complex, then loss of this residue would be expected to drastically weaken the affinity of this complex. The presence of residues G and H, which are only able to bind the activated antithrombin conformation with significant affinity, would additionally favor selective binding of EFGHЉ to activated antithrombin in pre-equilibrium with unactivated antithrombin. Since at most 8 kcal/mol of binding energy is utilized in forming the initial recognition complex at pH 6 and since the binding energy contribution of residue D greatly exceeds 8 kcal/mol at this pH (Tables I and II), a significant amount of the binding energy of residue D also appears to be available to enhance the binding of this residue to the activated antithrombin conformation and thereby to assist in stabilizing the activated conformation.
Further truncation of the pentasaccharide from the nonreducing-end produced yet different effects on oligosaccharide binding and activation of antithrombin. A key observation was that the kinetics of FGHٞ binding were complete within the ϳ1.5-ms dead time of the stopped-flow instrument. Such behavior suggests that FGHٞ cannot be binding by the pre-equilibrium mechanism observed for EFGHЉ binding, since the rate of binding by this pathway should have been limited by the 6 s Ϫ1 rate constant for conformational activation of antithrombin in the absence of bound oligosaccharide. It follows that FGHٞ must be binding to antithrombin by the induced conformational change pathway. This could happen if FGHٞ bound nonproductively to the DEF sites of native antithrombin involved in recognizing the pentasaccharide. Binding of FGHٞ in the pentasaccharide site was confirmed from the observation that the trisaccharide competes with tetrasaccharide DEFG* for binding to this site. Nonproductive binding of FGHٞ to the DEF interaction sites of native antithrombin would be compatible with the glucosamine/hexuronic acid specificity of the DEF sites. Such nonproductive binding to unactivated antithrombin would be favored because of the sizable binding energy available from the D and possibly also the F interaction sites (6,23) in the unactivated conformation together with the unfavorable energy requirement for activating antithrombin to allow productive binding. Moreover, such nonproductive binding would explain the reduced ability of FGHٞ to activate antithrombin, as judged both from a decreased protein fluorescence enhancement (ϳ10%) and lower enhancement of the rate of factor Xa inhibition (about 10% that of the pentasaccharide). The extent of this reduced activation is reminiscent of the reduced activating effect of low-affinity heparin that lacks the pentasaccharide activating sequence (11). The more favorable, nonproductive mode of binding of FGHٞ to antithrombin may also explain why so little binding energy is lost when residue E is deleted as compared with the deletion of residue D. Together, the results of the present and past studies favor the model for pentasaccharide binding and activation of antithrombin depicted in Fig. 5. In this model, binding of the rigid nonreducing end residues D, E, and F (40) occurs first to the native antithrombin conformation, with residue D making a primary contribution to the binding energy, although residues E and F may also contribute. Residues G and H make very weak or no interactions with this conformation. Both modeling studies (43) and a preliminary x-ray structure of the antithrombin-pentasaccharide complex (44) suggest that this initial binding of residues D, E, and F is at the C-terminal end of helix D with the reducing-end oriented toward the N terminus of this helix. Residues D, E, and F then induce antithrombin to undergo an activating conformational change in which the inhibitor reactive center loop is exposed to allow rapid inhibition of factor Xa (21,25). The activated conformation produces a complementary fit of saccharides D, E, and F in the heparin binding site, which enhances their binding to the activated conformation and thereby stabilizes this conformation. In particular, more favorable interactions of the unique 3-O-sulfate of residue F in the activated conformation may provide the driving force for this conformational change, since loss of the 3-O-sulfate mostly abolishes conformational activation (39). However, increased interactions of residues D and E with the activated antithrombin conformation are also likely, given the substantial binding energy resulting from the interaction of residue D and the decreased activation resulting from nonproductive binding of oligosaccharide FGHٞ in the DEF interaction sites. Further stabilization of the activated conformation results from the generation of additional complementary sites of interaction for the more flexible reducing-end residues G and H in this conformation (40). Conformational changes in the reducing-end saccharides, in particular the conformationally flexible iduronate residue G, may be required to align the charges in these residues for an optimal fit with the activated conformation (40). According to our model, the pentasaccharide functions as a classical allosteric modifier by preferentially binding and stabilizing the activated antithrombin conformation. Unique to this model are the different roles of nonreducing and reducing-end pentasaccharide residues in recognizing the native low-heparin-affinity state and in preferentially binding and stabilizing the activated high-heparin-affinity state.  EFGH؆ (B). A, the nonreducing-end residues, D, E, and F, first form a low-affinity recognition complex with partially complementary sites in unactivated antithrombin. The inhibitor then undergoes a conformational change in this complex, which induces a full complementary fit of the trisaccharide with the activated conformation and generates complementary sites for binding reducing-end saccharides G and H. The reducing-end residues are then able to bind and lock the inhibitor in the activated state. B, the deletion of saccharide D reduces the affinity of the nonreducing-end saccharides, E and F, for native antithrombin sufficiently so that binding to the small fraction of activated antithrombin in pre-equilibrium with the native inhibitor becomes the preferred mode of binding. Because EFGHЉ best complements the activated conformation, binding of EFGHЉ pulls the native inhibitor into the activated state. A subsequent further conformational change induced in antithrombin by EFGHЉ (see "Discussion") is not depicted in this model.