The Signature 3-O-Sulfo Group of the Anticoagulant Heparin Sequence Is Critical for Heparin Binding to Antithrombin but Is Not Required for Allosteric Activation*

Heparin and heparan sulfate glycosaminoglycans allosterically activate the serpin, antithrombin, by binding through a specific pentasaccharide sequence containing a critical 3-O-sulfo group. To elucidate the role of the 3-O-sulfo group in the activation mechanism, we compared the effects of deleting the 3-O-sulfo group or mutating the Lys114 binding partner of this group on antithrombin-pentasaccharide interactions by equilibrium binding and rapid kinetic analyses. Binding studies over a wide range of ionic strength and pH showed that loss of the 3-O-sulfo group caused a massive ∼60% loss in binding energy for the antithrombin-pentasaccharide interaction due to the disruption of a cooperative network of ionic and nonionic interactions. Despite this affinity loss, the 3-O-desulfonated pentasaccharide retained the ability to induce tryptophan fluorescence changes and to enhance factor Xa reactivity in antithrombin, indicative of normal conformational activation. Rapid kinetic studies showed that loss of the 3-O-sulfo group affected both the ability of the pentasaccharide to recognize native antithrombin and its ability to preferentially bind and stabilize activated antithrombin. By contrast, mutation of Lys114 solely affected the preferential interaction of the pentasaccharide with activated antithrombin. These findings demonstrate that the 3-O-sulfo group functions as a key determinant of heparin pentasaccharide activation of antithrombin both by contributing to the Lys114-independent recognition of native antithrombin and by triggering a Lys114-dependent induced fit interaction with activated antithrombin that locks the serpin in the activated state.

Antithrombin, a member of the serpin superfamily of protein protease inhibitors, is the principal physiologic regulator of blood coagulation proteases in vertebrates (1,2). Deficiencies of this blood plasma protein thus increase the risk of thrombotic disease (3), and complete deficiency appears to be incompatible with life (4). Antithrombin regulates the activity of its major target proteases, thrombin, factor Xa, and factor IXa, by inactivation of the enzymes through a branched pathway suicide substrate mechanism of inhibition that is characteristic of serpin-protease reactions (5). In this mechanism, the protease initially recognizes an exposed reactive loop of the serpin as a normal substrate and proceeds to cleave the loop and form the usual acyl-intermediate. However, once this cleavage has occurred, the serpin is induced to undergo a massive conformational change in which the N-terminal part of the reactive loop inserts into the major ␤-sheet of the serpin, causing the acyl-linked protease to be dragged to the opposite end of the protein inhibitor and inactivated through conformational distortion (6 -8).
In contrast to the rapid rates at which many serpins inactivate their target proteases, antithrombin inhibits coagulation proteases at slow nonphysiologic rates. However, these inhibition rates increase up to several thousand-fold in the presence of the sulfated glycosaminoglycans, heparin and heparan sulfate, which thereby act as efficient anticoagulants (9). The accelerating effects of heparin and heparan sulfate on antithrombin-protease reactions are dependent on the binding of a sequence-specific pentasaccharide to the serpin (10,11), present in about one-third of naturally occurring heparin chains and in a much smaller fraction of heparan sulfate chains (12). A characteristic structural marker of the specific antithrombinbinding pentasaccharide region is the 3-O-sulfonated central glucosamine residue (Fig. 1), which is absent or rare in other parts of the heparin molecule (10). The x-ray structure of the antithrombin-pentasaccharide complex and mutagenesis studies have shown that antithrombin basic residues in the N-terminal region, helix A, helix D, and the loop preceding helix D form a positively charged site to which the negatively charged pentasaccharide binds (13). Lys 114 , Lys 125 , and Arg 129 are the most important of these residues, with Lys 114 , the binding partner of the 3-O-sulfo group, contributing the greatest binding energy (9,14).
In this work, we have studied the interaction of a variant pentasaccharide lacking the 3-O-sulfo group with antithrombin. This variant has been previously reported to poorly bind antithrombin and to have greatly reduced anticoagulant activity, suggesting an essential role of the 3-O-sulfo group in binding and activating antithrombin (15)(16)(17)(18). However, quantitative studies were done at physiologic ionic strength and pH, where binding to antithrombin was extremely weak, and the extent of inhibitor activation was difficult to quantify. We show, by analyzing pentasaccharide binding to and activation of ␣and ␤-glycoforms of antithrombin over a wide range of ionic strength and pH, that deletion of the 3-O-sulfo group results in tremendous ϳ10 4 to 10 5 -fold losses in affinity for antithrombin under physiologic conditions due to the disruption of a coop-erative network of ionic and nonionic interactions that stabilize the complex. Despite these large losses in affinity and in contrast to past reports (16), loss of the 3-O-sulfo group does not affect the ability of the pentasaccharide to induce conformational activation of antithrombin. Rapid kinetic studies of the variant pentasaccharide interaction with antithrombin reveal marked effects of 3-O-desulfonation on both initial binding and subsequent conformational activation steps. These effects contrast with the effects of mutating the Lys 114 binding partner of the 3-O-sulfo group, which exclusively involve the second conformational activation step (14). Together, our findings show that the 3-O-sulfo group is of greater quantitative importance for the antithrombin-pentasaccharide interaction than any other pentasaccharide sulfo group or mutated antithrombin investigated so far due to its dual role in recognizing the native antithrombin conformation as well as in preferentially binding and stabilizing the activated antithrombin conformation. In the latter role, our results suggest that the 3-O-sulfo group must engage Lys 114 to position pentasaccharide and antithrombin residues for an induced fit interaction that enhances affinity and locks antithrombin in the activated state.

EXPERIMENTAL PROCEDURES
Antithrombin-N135Q and K114M/N135Q recombinant antithrombin variants that lacked glycosylation at Asn 135 , and thereby mimicked the natural ␤-form of plasma antithrombin, were expressed in BHK cells and purified as previously reported (14). Fully glycosylated ␣-antithrombin was isolated from human plasma as described (19). Protein concentrations were determined from the 280-nm absorbance using a molar absorption coefficient of 37,700 M Ϫ1 cm Ϫ1 (20).
Proteases and Saccharides-The synthetic pentasaccharide, DEFGH, corresponding to the antithrombin binding sequence in high affinity heparin chains, and a variant pentasaccharide lacking the 3-O-sulfo group of the central saccharide F were generously provided by Sanofi-Aventis (Toulouse, France) ( Fig.  1). The concentration of the natural pentasaccharide was based on stoichiometric titrations of antithrombin with the saccharide at inhibitor concentrations exceeding K D , as described previously (11,19). These concentrations agreed well with concentrations determined by weight and the molecular mass of the saccharides. The concentration of the 3-O-desulfonated pentasaccharide was determined based on the weight and molecular mass of the saccharide. Human ␣-thrombin was a gift from Dr. John Fenton (formerly of the New York State Department of Health). Human factor Xa (mostly ␣ form) was purchased from Enzyme Research (South Bend, IN). Concentrations of proteases were assessed from their activities in standard assays with peptidyl-p-nitroanilide chromogenic substrates and were based on calibration of these assays with active site-titrated enzymes (21).
Experimental Conditions-All experiments were carried out at 25°C. Studies of the 3-O-desulfonated and natural pentasaccharide interactions with N135Q antithrombin and plasma antithrombin or of the natural pentasaccharide interaction with K114M/N135Q antithrombin were done in buffers of 10 -20 mM sodium phosphate, 0.1 mM EDTA, 0.1% (w/v) polyethylene glycol 8000, adjusted to pH 6.0 or pH 7.4. The ionic strength of the 20 mM sodium phosphate buffer is 0.025 at pH 6, whereas 10 -20 mM sodium phosphate buffers at pH 7.4 correspond to ionic strengths of 0.025-0.05. NaCl was added to achieve higher ionic strengths. For experiments in pH 6 buffer, antithrombin was extensively diluted from a pH 7.4 buffer just prior to the experiment to minimize losses in activity due to reduced stability at the lower pH.
Affinities of Antithrombin-Pentasaccharide Interactions-Dissociation equilibrium constants for the binding of the different pentasaccharides to N135Q, K114M/N135Q, and plasma antithrombins were measured by titrations monitored by the enhancement of the intrinsic protein fluorescence accompanying the binding, as in past studies (11,14,19). Affinities of the natural and variant pentasaccharides for recombinant and plasma antithrombins were determined at pH 6.0 or pH 7.4 with antithrombin concentrations ranging from 50 to 2000 nM. Titrations were computer-fit by nonlinear leastsquares analysis to the quadratic binding equation, where ⌬F obs /F o and ⌬F max /F o are the observed and maximal changes in fluorescence relative to the initial fluorescence, respectively, [AT] o and [H] o are the total antithrombin and heparin concentrations, respectively, K D,obs is the observed dissociation constant, and n is the binding stoichiometry. The binding stoichiometry was assumed to be 1, and K D,obs and ⌬F max /F o were fitted as parameters. Corrections were made for nonbinding protein in the recombinant antithrombin preparations based on measured stoichiometries of thrombin inhibition as in past studies (14).
Kinetics of Pentasaccharide Binding-The kinetics of pentasaccharide binding to antithrombin were analyzed under pseudo-first order conditions as in past studies (11,14) by monitoring the increase in protein fluorescence that accompanies binding in an Applied Biophysics SX-17MV stopped-flow instrument. Experiments analyzing the binding of the natural and desulfonated pentasaccharides to N135Q antithrombin or of the natural pentasaccharide to K114M/N135Q antithrombin were done at I 0.04, pH 6.0. Pentasaccharide concentrations were at least 5-fold higher than antithrombin concentrations. Progress curves were fit by a single exponential function to give the observed pseudo-first order rate constant, k obs . Multiple curves were averaged to improve signal/noise at subsaturating saccharide concentrations, and reported k obs values reflect global average values from at least 20 reaction traces. For binding of the desulfonated pentasaccharide to antithrombin, fluorescence amplitudes were confirmed to increase nearly proportionally with saccharide concentration over the range examined, in accordance with the measured K D for the interaction. Simulations of the kinetics of heparin binding to antithrombin based on kinetic parameters derived from fitting progress curve data to the general two-step binding mechanism were done using Global Kinetic Explorer software (Kintek Corp., Austin, TX) (22).
Protease Inhibition-Stoichiometries of uncatalyzed and pentasaccharide-catalyzed reactions of N135Q, K114M/ N135Q, and plasma antithrombins with thrombin or factor Xa were determined as described previously (19). A fixed concentration of protease (100 -1000 nM) was incubated with increasing molar ratios of inhibitor to enzyme of 1-2 in 50 -100 l of I 0.025, pH 6 or I 0.15, pH 7.4 sodium phosphate buffers for times sufficient to achieve complete reaction based on measured reaction rate constants. When present, normal or variant pentasaccharides were added at a fixed level to yield comparable extents of saturation of antithrombin based on measured K D values, such that the pentasaccharide-accelerated reaction was dominant. Residual enzyme activity was determined by adding 900 -950 l of 100 M S-2238 for thrombin reactions or 100 M Spectrozyme FXa (American Diagnostica, Greenwich, CT) for factor Xa reactions and monitoring the initial rate of substrate hydrolysis from the linear increase in absorbance at 405 nm. Plots of residual enzyme activity versus molar ratio of inhibitor to enzyme were fit by linear regression and the stoichiometry of inhibition (SI) 2 obtained from the abscissa intercept.
Second order rate constants for inhibition of thrombin or factor Xa by N135Q, K114M/N135Q, and plasma antithrombins in the absence and presence of natural or 3-O-desulfonated pentasaccharides were measured under pseudo-first order conditions at I 0.025, pH 6.0 or I 0.15, pH 7.4 as in past studies (11,14). Reaction mixtures of 100 l contained 100 -1000 nM antithrombin and 5-50 nM protease and different concentrations of pentasaccharide. Reaction mixtures were incubated for increasing times at fixed pentasaccharide concentrations or for fixed times at increasing pentasaccharide concentrations at 25°C in polyethylene glycol-coated polystyrene cuvettes and then quenched with 900 l of 100 M S-2238 for thrombin or of 100 M Spectrozyme FXa for factor Xa in I 0.15 sodium phosphate buffer, pH 7.4, containing 50 -100 g/ml Polybrene. The residual protease activity was determined from the initial rate of substrate hydrolysis, as indicated above. Observed pseudofirst order rate constants (k obs ) were obtained from computer fits of the time-dependent or pentasaccharide concentrationdependent decrease in enzymatic activity by an exponential function with a zero activity end point (23). Uncatalyzed reactions of protease with antithrombin were shown to be free of traces of heparin by analyses in the absence or presence of 0.1 mg/ml Polybrene. Second order rate constants for uncatalyzed reactions were calculated by dividing k obs by the antithrombin concentration. Pentasaccharide-catalyzed reactions of N135Q and plasma antithrombins with thrombin were determined at saturating pentasaccharide concentrations for the natural pentasaccharide and at subsaturating concentrations for the 3-O-desulfonated pentasaccharide reactions. For the latter, the fitted k obs was corrected for the free antithrombin reaction and then divided by the concentration of antithrombin-pentasaccharide complex, as calculated from the quadratic binding equation using the measured K D for the interaction (19). Second order rate constants for pentasaccharide-catalyzed reactions of antithrombin with factor Xa were determined using two methods. For reactions at I 0.15, pH 7.4, the decrease in protease activity with increasing subsaturating saccharide concentration at a fixed reaction time was fit by an exponential decay function, and the second order rate constant was obtained from the equation, where t represents the fixed reaction time, [AT] o is the antithrombin concentration, and K D is the measured dissociation constant for the antithrombin-pentasaccharide interaction (23). In the second method used for reactions at I 0.025, pH 6, apparent second order rate constants were measured as a function of increasing saccharide concentrations over a range that approached saturation of antithrombin and were then fit by the quadratic binding equation (24). Apparent association rate constants were determined in this case for fixed reaction times from the equation, where v and v o represent the initial rates of substrate hydrolysis of enzyme after reaction for the fixed time, t, or for an unreacted enzyme control, respectively.

Binding of the 3-O-Desulfonated Pentasaccharide to
Antithrombin-To assess the effect of 3-O-desulfonation on the ability of the sequence-specific heparin pentasaccharide, DEFGH ( Fig. 1), to allosterically activate antithrombin, we initially compared the binding of natural and 3-O-desulfonated pentasaccharides to antithrombin. Binding was measured by titrating antithrombin tryptophan fluorescence changes that report binding as in past studies. Binding was examined with a ␤-type N135Q variant of antithrombin missing the N-linked carbohydrate chain at Asn 135 that interferes with heparin binding (25)(26)(27) and under conditions of low ionic strength (I 0.025), pH 6.0, and 25°C. These conditions were chosen to strengthen heparin binding affinity and thereby allow accurate measurement of the anticipated weak affinity of the 3-O-desulfonated pentasaccharide for antithrombin (15,28). The 3-Odesulfonated pentasaccharide produced a saturable increase in antithrombin tryptophan fluorescence like the natural pentasaccharide under these low ionic strength and pH conditions (Fig. 2). Nonlinear least squares fitting of variant pentasaccharide binding curves by the equilibrium binding equation yielded an average K D of 2.8 Ϯ 0.2 M. Binding of the natural pentasaccharide under the same conditions was stoichiometric and too tight to measure accurately. An estimated K D of 40 fM could be obtained by extrapolation of values measured at higher ionic strengths at the same pH ( Fig. 3) (see below), a value that agrees with the one estimated in previous studies for the N135Q anti-thrombin-pentasaccharide interaction at I 0.025, pH 6, 25°C (14). These results indicate a dramatic loss in affinity of the pentasaccharide for antithrombin of ϳ10 8 -fold at I 0.025, pH 6, when the 3-O-sulfo group is deleted (Table 1). Maximal fluorescence changes induced in antithrombin by the binding of natural and variant pentasaccharides in these experiments were similar (36 -37%), indicating that loss of the 3-O-sulfo group did not affect the ability of the bound saccharide to conformationally activate antithrombin (Table 1).
Ionic and Nonionic Contributions to Pentasaccharide Binding-To determine whether the large loss in pentasaccharide affinity for antithrombin caused by deletion of the 3-O-sulfo group was due to changes in the ionic or nonionic components of the binding interaction, we evaluated the dependence of the observed dissociation constant, K D,obs , for the interaction on ionic strength. Fluorescence titrations of the binding of the 3-O-desulfonated pentasaccharide to N135Q antithrombin at pH 6.0 and ionic strengths ranging from 0.025 to ϳ0.09 or the binding of the natural pentasaccharide at pH 6.0 and higher ionic strengths of 0.3-0.6 showed saturable binding curves with indistinguishable maximal fluorescence enhancements but with progressive decreases in saccharide affinity for antithrombin with increasing ionic strength. The ionic and nonionic components of the binding were resolved from the linear dependence of log K D,obs on log [Na ϩ ] for the two saccharide interactions (Fig. 2), in accordance with the equation (11,29), where Z represents the number of ionic interactions involved in the binding, is the fraction of Na ϩ ions bound per heparin charge that are released on binding to antithrombin (estimated to be 0.8 (11)), and KЈ D is the dissociation constant at 1 M Na ϩ , reflecting the strength of the nonionic interactions. The slopes and intercepts of these plots showed that about six (5.8 Ϯ 0.6) ionic interactions contributed to the binding of the natural pentasaccharide to N135Q antithrombin at pH 6.0 and that nonionic interactions contributed significantly to the binding (log KЈ D ϭ Ϫ5.8 Ϯ 0.2) ( Fig. 3 and Table 2). By contrast, binding of the 3-O-desulfonated pentasaccharide involved 2.0 Ϯ 0.2 ionic interactions and a considerably smaller contribution of nonionic interactions (log KЈ D ϭ Ϫ2.9 Ϯ 0.2). Deletion of the 3-Osulfo group thus causes a loss of ϳ4 ionic interactions and an ϳ800-fold decrease in the strength of nonionic interactions of the pentasaccharide with antithrombin at pH 6.
To extrapolate these findings to physiologic ionic strength and pH, we additionally analyzed binding of the 3-O-desulfonated pentasaccharide to N135Q antithrombin at pH 7.4 and   OCTOBER 2, 2009 • VOLUME 284 • NUMBER 40

JOURNAL OF BIOLOGICAL CHEMISTRY 27057
ionic strengths in the range 0.025-0.05. Dissociation constants increased from 35 Ϯ 5 M at I 0.025 to 79 Ϯ 18 M at I 0.05 at pH 7.4, values that were ϳ10-fold weaker than those measured at pH 6 at the same ionic strength. From the linear dependence of log K D,obs on log [Na ϩ ], we estimated that 1.5 Ϯ 0.3 ionic interactions and a log KЈ D of Ϫ2.4 Ϯ 0.3 for nonionic interactions contributed to the binding at this pH (Fig. 3), providing an estimated K D,obs at physiological ionic strength of 370 Ϯ 220 M. Corresponding values for natural pentasaccharide binding to N135Q antithrombin were 5.5 Ϯ 0.5 ionic interactions and log KЈ D of Ϫ5.1 Ϯ 0.1 for nonionic interactions, yielding an extrapolated K D,obs of 0.0013 Ϯ 0.0006 M at physiological ionic strength. Deletion of the 3-O-sulfo group thus results in an ϳ10 5 -fold decreased binding affinity at I 0.15, pH 7.4, 25°C due to the loss of ϳ4 ionic interactions and a 500-fold affinity loss due to nonionic interactions. Limited studies of the binding of natural and 3-O-desulfonated pentasaccharides to fully glycosylated plasma ␣-antithrombin at I 0.025, pH 6, indicated an extrapolated K D of ϳ1 pM and measured K D of 17 Ϯ 4 M for natural and variant pentasaccharide interactions, respectively, indicating comparable losses in pentasaccharide binding energy due to 3-O-desulfonation of ϳ60% for both ␣and ␤-antithrombin glycoforms.
Rapid Kinetics of Pentasaccharide Binding to Antithrombin-The kinetics of binding of the 3-O-desulfonated and natural pentasaccharides to N135Q antithrombin were analyzed at I ϳ0.04, pH 6.0, under pseudo-first order conditions by monitoring tryptophan fluorescence changes in a stopped-flow instrument as in past studies (11,14). The rapid binding of the natural pentasaccharide to N135Q antithrombin under these conditions limited the analysis to low pentasaccharide concentrations (Յ2 M), since rate constants higher than ϳ500 s Ϫ1 could not be reliably measured. A progressive saturation of observed pseudo-first order rate constants (k obs ) with increasing saccharide concentration was found for the natural pentasaccharide interaction (Fig. 4), which was indicative of a two-step binding

TABLE 2 Ionic and nonionic contributions to the interactions of natural and 3-O-desulfonated pentasaccharides with N135Q and K114M/N135Q antithrombins at pH 6 and pH 7.4, 25°C
The number of ionic interactions (Z) and the nonionic contribution (log KЈ D ) to the binding of pentasaccharides to antithrombins were determined from computerfitted values of the slope and intercept, respectively, of linear plots of log K D,obs versus log͓Na ϩ ͔ shown in Fig. 3. process in which an initial binding of the saccharide to antithrombin induces a subsequent conformational change in the inhibitor according to Scheme 1 (11,30),

Antithrombin
where K 1 represents the dissociation constant for an initial rapid equilibrium binding interaction, and k ϩ2 and k Ϫ2 are the forward and reverse rate constants for the subsequent conformational change step. The dependence of k obs on saccharide concentration ([H] o ) in Fig. 4 was fit by the rectangular hyperbolic equation that characterizes this induced conformational change binding mechanism, At low heparin concentrations much less than K 1 , this equation reduces to the following.
indicates that k Ϫ2 corresponds to the off-rate constant (k off ) and k 2 /K 1 represents the on-rate constant (k on ) for the overall binding interaction (11). The data were satisfactorily fit by Equation 5, which yielded values for K 1 of 0.52 Ϯ 0.05 M, for k ϩ2 of 530 Ϯ 20 s Ϫ1 , and, from the ratio k 2 /K 1 , a k on of 1000 Ϯ 100 M Ϫ1 s Ϫ1 . The fitted value of k Ϫ2 was indistinguishable from zero, indicating that k off was too low to determine accurately. Calculation of the expected k Ϫ2 from the relation, k Ϫ2 ϭ K D,obs ϫ k 2 /K 1 , confirmed an extremely low k off of 4 ϫ 10 Ϫ4 s Ϫ1 . In contrast to the results for binding of the natural pentasaccharide, k obs for the binding of the 3-O-desulfonated pentasaccharide to N135Q antithrombin showed an initial decrease followed by an increase as the concentration of the pentasaccharide was increased. The initial decline in k obs is diagnostic of an alternative preequilibrium pathway for pentasaccharide activation of antithrombin in which activation occurs by the saccharide selectively binding to a minor fraction of activated antithrombin that already exists in a preequilibrium with native antithrombin (28, 31) (Scheme 2). The preference for the preequilibrium activation pathway at low saccharide concentrations implies that the affinity of the variant saccharide for native antithrombin is greatly reduced from that of the natural pentasaccharide. The observation of a shift from a decrease to an increase in k obs at higher saccharide concentrations indicates that the binding mechanism reverts to the induced conformational activation pathway when saccharide concentrations are high enough to favor binding to native antithrombin. The two pathways are components of the general scheme for heparin activation of antithrombin (Scheme 2). In this scheme, K 1 , k ϩ2 , and k Ϫ2 are the parameters for the induced conformational activation pathway given above, k 3 is the rate constant for preequilibrium 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 saccharide binding to preequilibrium activated antithrombin (31). Assum-ing that the binding steps for both pathways are in rapid equilibrium, the dependence of k obs on [H] o for this mechanism is given by the equation, Of the six parameters in this equation, only five are independent, since they are related by the relation, where K 3 and K 2 are the equilibrium constants for the preequilibrium conformational activation and the heparin-induced conformational activation steps of the two binding pathways, respectively. Because of the limited range of saccharide concentrations over which kinetic data were obtainable, fitting of the data by Equations 7 and 8 did not allow all parameters to be uniquely determined. However, the observation that k obs showed no evidence of saturation over the range of heparin concentrations examined suggested that the value of K 1 for the induced conformational activation pathway was much greater than the highest heparin concentration employed. Under such conditions (K 1 Ͼ Ͼ [H] o ), Equations 7 and 8 reduce to the following.
Assuming our previously determined value for the preequilibrium conformational activation rate constant, k 3 , of ϳ10 s Ϫ1 under these conditions (28), and fixing k Ϫ3 at a value of 1000 s Ϫ1 , consistent with the estimated 1% of preequilibrium activated antithrombin (32), a good fit of the data by this equation could be obtained (Fig. 4). Fitted values for k 2 /K 1 (k on ) of 44 Ϯ 2 M Ϫ1 s Ϫ1 and for k Ϫ2 (k off ) of 76 Ϯ 3 s Ϫ1 were 25-fold decreased and 200,000-fold increased from corresponding values for natural pentasaccharide binding. The inability to detect saturation of k obs at variant saccharide concentrations of 4 M further implied values for K 1 of Ͼ4 M and for k 2 of Ͼϳ200 s Ϫ1 . The 25-fold decrease in k on could therefore be attributed to at least a 10-fold increase in K 1 and at most a ϳ2-fold decrease in k 2 .
The fitted values for k on and k off implied an overall dissociation constant (K D,obs ϭ K 1 K 2 /(1 ϩ K 2 )) of ϳ2 M, which was lower than the measured value for K D,obs of 6 M under these conditions (Table 1). However, fixing K D,obs in the range of 2-5 M produced satisfactory fits of the data with only marginal reduc-SCHEME 2 OCTOBER 2, 2009 • VOLUME 284 • NUMBER 40

JOURNAL OF BIOLOGICAL CHEMISTRY 27059
tions in k 2 /K 1 to at most 30 M Ϫ1 s Ϫ1 , a ϳ2-fold maximal increase in k Ϫ2 to 150 s Ϫ1 and plausible preequilibrium activation ratios (k Ϫ3 /k 3 ) between 0.1 and 1%. Kinetic simulations verified that reasonable fits of the data were obtained with the determined parameters even without making the rapid equilibrium assumption. Overall, these findings indicated that loss of the 3-O-sulfo group reduces pentasaccharide affinity for antithrombin through both a decrease in k on and an increase in k off , with the latter being overriding.
Effects of Lys 114 Mutation on Pentasacharide Binding-Since the x-ray structure of the antithrombin-pentasaccharide complex shows that the 3-O-sulfo group interacts with Lys 114 of antithrombin (13) and mutation of Lys 114 produces a substantial binding defect (14), we were interested in comparing the affinity and kinetics of binding of the natural pentasaccharide to a K114M/N135Q antithrombin variant under the same conditions used to analyze the binding of the 3-O-desulfonated pentasaccharide to N135Q antithrombin. K D,obs for the binding of the natural pentsaccharide to the mutant antithrombin was measured to be 190 Ϯ 3 nM at I 0.025 and 520 Ϯ 80 nM at I ϳ0.04, pH 6 and 25°C, in good agreement with values measured previously at I 0.025 (14). Values for Z of 2.5 and for log KЈ D of Ϫ3.5 were estimated from these limited data. Analysis of binding at pH 7.4 and ionic strengths between 0.025 and 0.05 gave K D values ranging from 5.3 Ϯ 0.3 to 12 Ϯ 2 M, allowing an extrapolated K D of 60 Ϯ 20 M to be determined at physiologic ionic strength and pH. Values for Z of 1.5 Ϯ 0.2 and for log KЈ D of Ϫ3.2 Ϯ 0.3 were obtained for the binding at pH 7.4 from this dependence (Fig. 3). Thus, although the antithrombin-pentasaccharide interaction is considerably weakened by the isosteric replacement of Lys 114 with an uncharged Met side chain, the reduction in affinity is ϳ10 7 -fold at I 0.025, pH 6 and ϳ50,000-fold at I 0.15, pH 7.4 (i.e. ϳ6 -10-fold less than that caused by removal of the pentasaccharide 3-O-sulfo group). The kinetics of binding of the natural pentasaccharide to K114M antithrombin at pH 6, I ϳ0.04 showed a hyperbolic dependence of k obs on saccharide concentration (Fig. 4), indicating that binding exclusively followed the induced conformational change pathway. Values of K 1 of 1.7 Ϯ 1.6 M, k 2 of 25 Ϯ 6 s Ϫ1 , and k Ϫ2 of 14 Ϯ 4 s Ϫ1 were obtained by fitting the data to Equation 5, indicating a k on of 15 Ϯ 15 M Ϫ1 s Ϫ1 and k off of 14 Ϯ 4 s Ϫ1 for the binding interaction. The calculated K D of 600 nM in this case, obtained from the formula, K 1 K 2 /(1 ϩ K 2 ), agreed well with the measured value of 520 Ϯ 80 nM. Loss of the positive charge of Lys 114 thus has no significant effect on the affinity for native antithrombin in the initial binding step (K 1 ) but results in major effects on both forward (k ϩ2 ) and reverse (k Ϫ2 ) rate constants in the subsequent conformational activation step, in agreement with previous studies of pentasaccharide binding to a K114A/N135A recombinant antithrombin at a higher ionic strength (14). Notably, the effects of Lys 114 mutation on k ϩ2 were much greater than those due to 3-O-desulfonation, whereas the effects of Lys 114 mutation on k Ϫ2 were of the same order of magnitude as those due to 3-O-desulfonation.

Effects of 3-O-Desulfonation and Lys 114 Mutation on Pentasaccharide-catalyzed Antithrombin-Protease Reactions-
The activating effects of natural and 3-O-desulfonated pentasaccharides on the ␤-type N135Q and K114M/N135Q antithrombins as well as plasma ␣-antithrombin were compared by measuring the second order rate constants (k a ) for reactions of unactivated and pentasaccharide-activated antithrombin with thrombin and factor Xa. Rate constants were measured by discontinuous assays of protease inhibition at I 0.025, pH 6.0, 25°C under pseudo-first order conditions as a function of increasing pentasaccharide concentration. The stoichiometries of antithrombin inhibition of thrombin and factor Xa (SI) in the absence and presence of the pentasaccharides were also measured under the same conditions in order to correct rate constants for a competing substrate reaction of antithrombin with proteases (33). This correction, made by multiplying the apparent association rate constant by the SI, is essential, since the reaction flux through the substrate pathway increases for pentasaccharide-catalyzed reactions as the ionic strength is decreased (11). Unactivated rate constants were similar to those reported previously under these conditions and were unaffected by the Lys 114 mutation (Table 3) (14). Rate constants for natural and variant pentasaccharide-activated reactions of N135Q antithrombin with thrombin and factor Xa measured at subsaturating saccharide concentrations, calculated to reflect the rate constants for protease inhibition by the antithrombin-saccharide complexes using measured K D values for complex formation, were indistinguishable after correcting for SI (Table 3). By contrast, the rate constants for natural pentasaccharide-activated reactions of N135Q and K114M/ N135Q antithrombins with the two proteases differed for factor Xa but were similar for thrombin (Table 3). To validate these findings at saturating levels of the saccharides, titrations of the increase in k a produced by natural and variant pentasaccharides were performed for N135Q and plasma antithrombin reactions with factor Xa. k a increased with increasing saccharide concentration in a saturable manner to reach indistinguishable end point values in all cases except for the variant pentasaccharide-activated reaction of plasma antithrombin, for which the end point was not well determined ( Fig. 5 and Table 3). Although K D values could not be determined for natural pentasaccharide activation of N135Q and plasma antithrombins because activation was stoichiometric, a K D of 2.7 Ϯ 0.1 M was determined for 3-O-desulfonated pentasaccharide activation of N135Q antithrombin, in excellent agreement with the value measured by equilibrium binding. Assuming the measured K D of 17 Ϯ 4 M for the variant pentasaccharide interaction with plasma antithrombin, a fitted end point corresponding to full activation was determined. k a for the natural pentasaccharideactivated reaction of K114M/N135Q antithrombin with factor Xa showed saturable increases to an end point k a that was ϳ75% of that measured for pentasaccharide activation of N135Q antithrombin and with a K D of 190 Ϯ 30 nM, which agreed with the value measured by equilibrium binding. Together, these results indicated that deletion of the 3-O-sulfo group does not affect the ability of the bound pentasaccharide to conformationally activate antithrombin, whereas mutation of Lys 114 modestly reduces the activating effect.
To determine whether these findings could be extrapolated to physiologic pH and ionic strength, we measured the association rate constants and SIs for natural and 3-O-desulfonated pentasaccharide-activated reactions of all forms of antithrombin with fac-tor Xa at I 0.15, pH 7.4. Similar to the findings at low pH and ionic strength, the 3-O-desulfonated pentasaccharide produced an activated rate constant for the N135Q antithrombin reaction with factor Xa that was experimentally indistinguishable from the value produced by the natural pentasaccharide when inhibition rates were measured from 1-10 M variant saccharide, and the antithrombin-saccharide complex concentration was calculated using the extrapolated K D for the variant pentasaccharide interaction under these conditions. Notably, plasma ␣-antithrombin resulted in a normal activated rate constant if a K D value for the ␣-antithrombin-variant saccharide interaction of ϳ600 M was assumed. Such a value is comparable with that previously reported under these conditions (15). Also, like the findings at low pH and ionic strength, the natural pentasaccharide produced an activated rate constant for the K114M/N135Q antithrombin reaction with factor Xa that was only modestly reduced from that for the corresponding reaction with N135Q antithrombin at I 0.15, pH 7.4.

DISCUSSION
Our studies provide new insights into the role of a signature 3-O-sulfo group of the anticoagulant heparin pentasaccharide sequence in the mechanism by which heparin and heparan sul-

TABLE 3 Association rate constants (k a ) and SI values for reactions of N135Q, K114M/N135Q, and plasma antithrombins with thrombin and factor Xa in the absence and presence of normal (H5) or 3-O-desulfonated pentasaccharides (⌬3OSO 3 -H5)
fate glycosaminoglycans allosterically activate antithrombin. By comparing the interactions of natural and 3-O-desulfonated pentasaccharides with both ␣and ␤-type glycoforms of antithrombin over a wide range of ionic strength and pH, we have shown that deletion of the 3-O-sulfo group produces a massive loss in binding affinity that ranges from ϳ10 7 -to 10 8 -fold at pH 6.0, I 0.025 to ϳ10 4 -to 10 5 -fold at pH 7.4, I 0.15 and that corresponds to ϳ60% of the binding energy. Significantly, our results show that the large affinity loss arises from a pH-independent loss of ϳ4 of 6 ionic interactions as well as a ϳ500 -800-fold decrease in the affinity due to nonionic interactions, corresponding to similar ϳ30% losses in ionic and nonionic binding energy at physiologic ionic strength. These findings demonstrate that the 3-O-sulfo group is a key mediator of the cooperative network of interactions that contribute to the high affinity antithrombin-pentasaccharide interaction (9). Such cooperativity has been evident in previous studies from the finding that multiple ionic and nonionic interactions are lost when any one of the three hotspot basic residues of antithrombin in the pentasaccharide binding site are mutated. Remarkably, the antithrombin binding defect resulting from loss of the pentasaccharide 3-O-sulfo group is comparable with or greater than that produced by the loss of other critical sulfo groups in the pentasaccharide, including the 6-O-sulfo group of residue D and the N-sulfo group of residue H (15,28,34), as well as that caused by mutating any of the three hotspot residues, Lys 114 , Lys 125 , and Arg 129 , in the pentasaccharide binding site of antithrombin (14,35,36). Not surprisingly, mutation of the binding partner of the 3-O-sulfo group in the x-ray structure of the antithrombin-pentasaccharide complex, Lys 114 , produces a binding defect that closely approaches (i.e. is ϳ6 -10-fold smaller than) that due to 3-O-desulfonation.
Despite the greatly reduced affinity of the 3-O-desulfonated pentasaccharide for antithrombin, the variant pentasaccharide was nevertheless able to induce full conformational activation of either ␣or ␤-antithrombin glycoforms, as judged from the normal tryptophan fluorescence enhancement and normal enhancement of antithrombin reactivity with factor Xa caused by variant pentasaccharide binding. Previous studies measured the affinity of the 3-O-desulfonated pentasaccharide for the ␣-glycoform of antithrombin at physiologic pH and ionic strength by equilibrium dialysis and found a K D value of ϳ500 M, representing a loss of binding energy comparable with what we have found (15). Surprisingly, later studies reported that the variant saccharide showed a marginal ability either to induce tryptophan fluorescence changes in antithrombin or to enhance rates of antithrombin inhibition of factor Xa (16), results in marked contrast to our findings. Although the reason for such discrepant findings is unclear, it should be noted that factor Xa inhibition rates were measured indirectly from fluorescence changes of an active site-bound probe rather than directly as in our studies, and both tryptophan and probe fluorescence measurements were made at high protein concentrations expected to cause substantial inner filter fluorescence quenching. We further note that the allosteric activating effects of the heparin pentasaccharide on antithrombin that we have measured in the present study are independent of previously reported calcium effects on heparin activation of antithrombin.
Such effects reflect the ability of calcium to enhance factor Xa binding to full-length heparins and thereby to promote fulllength heparin bridging of antithrombin and factor Xa in a ternary complex (21,24).
Notably, mutation of Lys 114 results in somewhat smaller losses in both ionic and nonionic binding energy than 3-Odesulfonation, underscoring a role for the 3-O-sulfo group in the antithrombin-pentasaccharide interaction that is independent of Lys 114 . This role cannot involve the established ability of the 3-O-sulfo group to induce the adjacent iduronate residue G to favor the high energy skew boat conformation that binds antithrombin with ϳ1000-fold higher affinity than lower energy chair conformations (37,38). This is because deletion of the 3-O-sulfo group reduces the fraction of skew boat conformer of the iduronate residue only modestly, from 64 to 40% (37), a reduction expected to decrease pentasaccharide affinity less than 2-fold under physiologic conditions (38). Moreover, recent studies suggest that 3-O-sulfonated pentasaccharide sequences in longer chain heparins that have the iduronate replaced with a glucuronate retain high affinity for antithrombin (34).
Rapid kinetic studies provided insights into the source of both Lys 114 -independent and -dependent effects of pentasaccharide 3-O-desulfonation. Loss of the 3-O-sulfo group was found to significantly weaken the pentasaccharide interaction with native antithrombin in the initial binding step (K 1 increased Ͼ10-fold), whereas Lys 114 mutation had minimal effects on this step. A role for the pentasaccharide 3-O-sulfo group in recognizing the native antithrombin conformation is supported by previous studies that have shown that the initial binding step is mediated largely by an electrostatically driven interaction of the nonreducing end trisaccharide DEF with native antithrombin (28). Notably, removal of saccharide D from pentasaccharide DEFGH was found to be sufficient to block binding to native antithrombin and favor direct binding to activated antithrombin through the preequilibrium activation pathway. Our present findings similarly show that loss of the 3-O-sulfo group reduces binding to native antithrombin sufficiently to favor activation of antithrombin through the preequilibrium pathway at low saccharide concentrations. Our results therefore suggest that both saccharide D and the 3-O-sulfo group in saccharide F are critically important for the initial electrostatic binding of trisaccharide DEF to native antithrombin.
Loss of the 3-O-sulfo group has a more pronounced effect on the preferential interaction of the pentasaccharide with activated antithrombin in the second conformational activation step, causing a 10 5 -to 10 6 -fold weakening of this interaction that presumably reflects the loss of interaction with Lys 114 . The second step defects minimally involve the ability to induce conformational activation (Ͻ2-fold decrease in k ϩ2 ) and mostly reflect a major loss in the ability to stabilize the activated conformation (k Ϫ2 increased ϳ10 5 -fold). However, the activated conformation remains favored over the native conformation (k ϩ2 /k Ϫ2 Ͼ Ͼ 2), thus accounting for the observed normal enhancements in antithrombin tryptophan fluorescence and reactivity with factor Xa produced by the 3-O-desulfonated pentasaccharide. Mutation of Lys 114 similarly greatly weakens the preferential interaction with activated antithrombin in the conformational activation step by ϳ10 6 -fold. However, this defect involves both a reduced ability to induce conformational activation of antithrombin (k ϩ2 decreased ϳ20-fold) and to stabilize the activated conformation (k Ϫ2 increased ϳ40,000-fold). Consequently, the activated conformation is only marginally favored over the native conformation (k ϩ2 /k Ϫ2 ϭ 1-2), explaining the modestly reduced ability of pentasaccharide to enhance the reactivity of the mutant antithrombin with factor Xa. The 3-O-sulfo group thus contributes similarly to Lys 114 to stabilizing antithrombin in the activated state but differs from Lys 114 in its contribution to inducing the activated state. Implicit in these findings is that the defects in antithrombin activation produced by pentasaccharide 3-O-desulfonation or Lys 114 mutation are fully accounted for by alterations in the equilibrium between native and activated states and not to any changes in the native and activated states themselves.
The x-ray structures of free and pentasaccharide-complexed antithrombin reveal that the unstructured loop preceding the N-terminal end of helix D is induced to form a new P helix in the activated serpin that positions Lys 114 for ionic and hydrogenbonding interactions with the 3-O-sulfo group of the pentasaccharide (13,39). However, Lys 114 forms interactions not only with the F saccharide 3-O-sulfo group but also ionic and hydrogen bonds with the carboxylate and ring oxygen of saccharide G and the N-sulfo group of saccharide H. The importance of these latter interactions is evident from the observation that deletion of the reducing end GH disaccharide causes a loss in pentasaccharide binding energy comparable with that resulting from mutation of Lys 114 (28). Formation of the P helix thus represents part of the induced fit structural changes in antithrombin that allow Lys 114 and other basic residues to make complementary interactions with the F, G, and H saccharides in the conformational activation step. Lys 114 binding to saccharides F, G, and H is a key factor both in driving conformational activation and in stabilizing the activated conformation, since mutation of Lys 114 greatly affects both forward and reverse conformational activation rate constants (14). Loss of the 3-O-sulfo group probably produces a similar inability to stabilize the activated conformation due to the disruption of Lys 114 interactions with saccharides F, G, and H, as suggested by the comparable major effects on k Ϫ2 resulting from loss of the 3-O-sulfo group or mutation of Lys 114 . The interactions lost in this step may include the basic residues, Arg 46 and Arg 47 , that cooperate with Lys 114 to bind the GH disaccharide in the conformational activation step (40) as well as Phe 122 , which accounts for much of the nonionic binding energy produced in this step (41). The relatively small effect of the 3-O-sulfo group deletion on k ϩ2 suggests that the 3-O-sulfo group is not important for inducing the activating conformational change. The initial binding of the DEF unit of the pentasaccharide to antithrombin may therefore induce helix P to form and position Lys 114 for interactions with saccharides G and H independent of the 3-O-sulfo group, but only when the 3-O-sulfo group engages Lys 114 can stabilizing interactions with these saccharides be made. The 3-O-sulfo group-Lys 114 interaction thus appears pivotal for establishing much of the cooperative network of antithrombin-pentasaccharide interactions that are responsible for the induced fit interaction of the pentasaccharide with activated antithrombin and the consequent stabilization of the activated conformational state.
Our findings have implications for understanding the greatly reduced antithrombin affinity and activating ability of so-called low affinity heparins lacking the pentasaccharide sequence (42,43). Although this has been thought to reflect the absence of the 3-O-sulfo group marker of this sequence, low affinity heparins show a smaller loss in affinity for antithrombin at physiologic ionic strength and pH (ϳ1000-fold) than that caused by deletion of the 3-O-sulfo group, and such heparins only partially induce conformational activation of the serpin that cannot be caused by loss of the 3-O-sulfo group. The difference in affinities is understandable, because low affinity heparin is a fulllength heparin that can bind antithrombin through multiple overlapping sequences extending beyond the pentasaccharide binding site (11,14,42,44,45), resulting in an apparent greater binding affinity. Clues to the reduced activating effect of low affinity heparin come from our previous finding that when the rare E glucuronic acid residue of the minimal DEF-activating sequence is replaced by the more common iduronate 2-O-sulfate, only partial conformational activation similar to that of low affinity heparin is observed (28). The preponderance of such trisaccharide sequences in low affinity heparin (46) may thus account for its reduced activating effect, presumably because these sequences preferentially bind and stabilize the native antithrombin conformation over the activated conformation. Given our finding that loss of the glucosamine 3-Osulfo group does not affect the ability of heparin and presumably also of heparan sulfate to activate antithrombin despite its major effect on affinity, it is noteworthy that mice lacking the enzyme responsible for 3-O-sulfonation of antithrombin binding sequences in heparan sulfate do not show anticoagulant defects (47). Our findings thus suggest the possibility that the abundance of heparan sulfate chains in vivo may allow sufficient activation of antithrombin through sequences lacking the 3-O-sulfo group so as to achieve normal hemostasis.