Lysine 114 of antithrombin is of crucial importance for the affinity and kinetics of heparin pentasaccharide binding.

Lys(114) of the plasma coagulation proteinase inhibitor, antithrombin, has been implicated in binding of the glycosaminoglycan activator, heparin, by previous mutagenesis studies and by the crystal structure of antithrombin in complex with the active pentasaccharide unit of heparin. In the present work, substitution of Lys(114) by Ala or Met was shown to decrease the affinity of antithrombin for heparin and the pentasaccharide by approximately 10(5)-fold at I 0.15, corresponding to a reduction in binding energy of approximately 50%. The decrease in affinity was due to the loss of two to three ionic interactions, consistent with Lys(114) and at least one other basic residue of the inhibitor binding cooperatively to heparin, as well as to substantial nonionic interactions. The mutation minimally affected the initial, weak binding of the two-step mechanism of pentasaccharide binding to antithrombin but appreciably (>40-fold) decreased the forward rate constant of the conformational change in the second step and greatly (>1000-fold) increased the reverse rate constant of this step. Lys(114) is thus of greater importance for the affinity of heparin binding than any of the other antithrombin residues investigated so far, viz. Arg(47), Lys(125), and Arg(129). It contributes more than Arg(47) and Arg(129) to increasing the rate of induction of the activating conformational change, a role presumably exerted by interactions with the nonreducing end trisaccharide unit of the heparin pentasaccharide. However, its major effect, also larger than that of these two residues, is in maintaining antithrombin in the activated state by interactions that most likely involve the reducing end disaccharide unit.

The plasma proteinase inhibitor of the serpin family, antithrombin, is a major regulator of blood clotting. Its crucial role is evident from many observations that individuals with heterozygous antithrombin deficiency, resulting in lower amounts of active inhibitor, have an increased tendency to develop thrombosis (1). Moreover, deletion of the gene in mice leads to embryonic lethality, due to fibrin deposition in myocardium and liver and consumptive coagulopathy (2). Antithrombin inhibits most coagulation proteinases, although its main physiological targets are thrombin and factor Xa (1,3,4). Like all inhibitorily active serpins, antithrombin inactivates target proteinases by a unique mechanism. The proteinase initially recognizes a reactive bond, located in a surface-exposed loop of the serpin, and proceeds to cleave this bond. At the acyl-intermediate stage of this cleavage, the reactive bond loop is opened, which releases the strain on the loop. As a consequence, the liberated N-terminal part of the loop is rapidly inserted into a major ␤-sheet of the inhibitor, the A sheet. The proteinase is still attached to this segment by an acyl bond and is therefore transported to the opposite pole of the protein. In this location, the proteinase is squeezed against the main body of the inhibitor and is inactivated by the resulting distortion of a large part of its structure, including the active site (3)(4)(5)(6)(7).
In contrast to most serpins, which rapidly inhibit their target proteinases, antithrombin inactivates thrombin and factor Xa at only moderate rates. However, these reactions are greatly accelerated by the sulfated glycosaminoglycan, heparin, which thereby acts as an efficient anticoagulant (3,4). The rate acceleration is due to a specific pentasaccharide region of heparin binding to antithrombin (8,9). This binding occurs by a twostep mechanism, in which an initial, weak complex is formed in a rapid equilibrium in the first step. A conformational change that tightens the binding of the pentasaccharide region and activates the inhibitor is then induced in the second step (10,11). This conformational change is sufficient to accelerate the inhibition of factor Xa, whereas rapid thrombin inhibition is mainly dependent on approximation of enzyme and inhibitor by both binding to the same pentasaccharide-containing heparin chain of at least 18 saccharide units (3,4). The x-ray structure of a complex between antithrombin and a synthetic heparin pentasaccharide (12) indicates that the increased heparin affinity induced in the second binding step involves the D helix of the inhibitor being elongated by 1.5 turns and a new, short ␣-helix, the P helix, being formed ( Fig.  1). Moreover, the enhanced reactivity of antithrombin with factor Xa caused by heparin is most likely due to an increased exposure of the reactive bond loop. The x-ray structure has also identified several basic residues of antithrombin, predominantly Arg 47 , Lys 114 , Lys 125 , and Arg 129 , that participate in the binding by interacting with negative groups of the pentasaccharide (Fig. 1). The importance of Arg 47 and Arg 129 for the binding is consistent with the reduced heparin affinities of natural antithrombin variants in which these residues are altered (13,14). The roles of the four residues have been further investigated by mutations in recombinant variants of the in-hibitor (15)(16)(17)(18)(19)(20). Substitution of Arg 47 , Lys 125 , and Arg 129 led to 20 -30-, 30 -150-, and 400 -2500-fold, respectively, losses of affinity for pentasaccharide and full-length heparin (15,19,20), reflecting a larger contribution of Arg 129 than of the other two residues to the binding. Kinetic studies showed that both Arg 47 and Arg 129 are involved in the second step of heparin binding (19,20), predominantly by decreasing the reverse rate constant of this step and thus aiding in keeping antithrombin in its activated form. In contrast, they only moderately contribute to increasing the rate of induction of the conformational change. Analogous studies of the role of Lys 125 in the kinetics of heparin binding are lacking. Similarly, although mutation of Lys 114 has demonstrated the importance of this residue for heparin binding (16,17), its quantitative contribution to the binding and role in the binding kinetics are unknown.
In this work, we have investigated the role of Lys 114 in binding of heparin by mutation of this residue to Ala or Met. We find that these mutations decrease the affinity for pentasaccharide and full-length heparin by ϳ10 5 -fold, reflecting a considerably larger contribution to heparin binding of Lys 114 than of the other residues investigated so far. Like Arg 47 and Arg 129 , Lys 114 does not participate to any appreciable extent in the first step of heparin binding but acts predominantly in the second step. However, it functions by both substantially increasing (Ͼ40-fold) the forward rate constant and greatly decreasing (Ͼ1000-fold) the reverse rate constant of this step. It is therefore of crucial importance both for induction of the heparin-induced conformational change that activates antithrombin and for locking the inhibitor in the conformationally activated state.

EXPERIMENTAL PROCEDURES
Antithrombin Variants-Antithrombin variants with substitutions of Lys 114 by Ala or Met were produced by site-directed mutagenesis with the previously characterized N135A variant as base molecule and were expressed in a baculovirus system (18 -22). (It should be noted that a previously reported "K114A/N135A" mutant (18) that had the same heparin affinity as the control was actually K107A/N135A, i.e. had the wild type Lys at position 114. The K114A/N135A mutant used in the present work has the correct sequence and is not the "K114A/N135A" of Ref. 18). The N135A, K114A/N135A, and K114M/N135A variants were purified by affinity chromatography on a 5-ml HiTrap Heparin (Amersham Pharmacia Biotech) column at pH 7.4, as detailed in earlier work (18 -20, 22). However, to bind the K114A/N135A and K114M/N135A variants to the immobilized heparin, it was necessary to reduce the ionic strength of the filtered lysates by dilution with one-half volume of 20 mM sodium phosphate, 100 M EDTA, pH 7.4, prior to loading of the columns. These variants were further purified by anion exchange chromatography on a Mono Q HR 5/5 column (Amersham Pharmacia Biotech), eluted with a 30-ml gradient from 0.02 to 0.6 M NaCl in 20 mM sodium phosphate, 0.1% (w/v) polyethylene glycol 8000, pH 7.4.
A Lys 114 to Met antithrombin variant was also produced by sitedirected mutagenesis with an N135Q variant as base molecule and was expressed in a baby hamster kidney cell system (23,24). The N135Q and K114M/N135Q variants were purified by affinity chromatography on heparin-agarose, followed by successive chromatographies on DEAE-Sepharose and Sephacryl S-200 (Amersham Pharmacia Biotech), as described previously (25).
The purity of the antithrombin preparations was analyzed by SDSpolyacrylamide gel electrophoresis with the Tricine 1 or Laemmli buffer systems (26,27) and by nondenaturing polyacrylamide gel electrophoresis with the Laemmli buffer system. Concentrations of the three variants were determined from the absorbance at 280 nm with the use of the molar absorption coefficient of plasma antithrombin, 37,700 M Ϫ1 cm Ϫ1 (28).
Proteinases and Saccharides-Human ␣-thrombin was a gift from Dr. J. Fenton (New York State Department of Health, Albany, NY). Human factor Xa was purified as described elsewhere (29). The synthetic antithrombin-binding normal (9) and high affinity (compound 83 in Ref. 30) pentasaccharides and the nonreducing end trisaccharide unit of the pentasaccharide (DEF in Ref. 31) were generous gifts from Dr. M. Petitou (Sanofi Recherche, Toulouse, France). Full-length heparin with high affinity for antithrombin was isolated as described previously (11,22,32) and had a molecular mass of ϳ8000 Da (ϳ26 saccharides) and a reduced polydispersity.
Experimental Conditions-All experiments were carried out at 25.0 Ϯ 0.2°C. The buffer in most experiments was 20 mM sodium phosphate, 100 M EDTA, 0.1% (w/v) polyethylene glycol 8000, adjusted to pH 6.0 or 7.4. The ionic strength of this buffer is 0.025 and 0.05 at the two pH values, respectively, and NaCl was added if higher ionic strengths were desired. However, 10 mM sodium phosphate, 100 M EDTA, 0.1% (w/v) polyethylene glycol 8000 was used for measurements at I 0.025, pH 7.4.
Stoichiometries and Affinities of Heparin Binding-Stoichiometries and dissociation equilibrium constants for the binding of the different heparin forms to the antithrombin variants were measured by titrations monitored by the enhancement of intrinsic protein fluorescence accompanying the interaction, as described previously (19,20,22,25). Stoichiometries of full-length heparin binding to the N135A and N135Q variants were measured at I 0.15, pH 7.4, with antithrombin concentrations of 0.1-0.3 M. Stoichiometries of pentasaccharide or full-length heparin binding to the K114A/N135A, K114M/N135A, and K114M/ N135Q variants were determined at I 0.025, pH 6.0, and 1-2 M antithrombin. Affinities of trisaccharide, pentasaccharide, or fulllength heparin binding to the variants were determined at pH 6.0 or 7.4 and different ionic strengths with antithrombin concentrations of 50 -500 nM. The data were fitted to the equilibrium binding equation by nonlinear least-squares analysis (25).
Kinetics of Heparin Binding-The kinetics of binding of pentasaccharide or full-length heparin to the N135A and K114A/N135A antithrombin variants were analyzed under pseudo-first order conditions by monitoring the increase in protein fluorescence in an SX-17MV stopped-flow instrument (Applied Biophysics, Leatherhead, United Kingdom) as in earlier work (11,19,20,22). Experiments with both saccharides were done at I 0.075, although at pH 6.0 for the pentasaccharide and at pH 7.4 for full-length heparin. Saccharide concentrations varied between 0.2 and 13 M and were at least 10-fold higher than antithrombin concentrations. Progress curves were fitted to a single exponential function to give the observed pseudo-first order rate constant, k obs . Four traces were averaged for each rate constant determination, and reported k obs values are averages of at least four such determinations.
Stoichiometries and Kinetics of Proteinase Inactivation-Stoichiometries of inhibition of active site-titrated human ␣-thrombin by the antithrombin variants in the absence of heparin were measured essentially as detailed previously (19,20,25). Briefly, a series of samples of thrombin at a constant concentration of 0.1 or 0.5 M was incubated with increasing amounts of antithrombin variant in I 0.15, pH 7.4 buffer. The residual activity of the enzyme was then determined after 16 h (for 0.1 M thrombin) or 1-2 h (for 0.5 M thrombin) from the initial rate of hydrolysis of the substrate, S-2238 (D-phenylalanyl-L-pipecolyl-L-arginyl-p-nitroanilide; Chromogenix, Mölndal, Sweden). Stoichiometries of thrombin inhibition by the N135A, K114A/N135A, N135Q, and K114M/N135Q variants in the presence of heparin at I 0.05, pH 7.4, were measured by incubating 20 nM thrombin with increasing concentrations of antithrombin variant and 25 nM full-length heparin for 1 h and assaying the residual enzyme activity in the same manner. The inhibition stoichiometries were obtained from linear least-squares fits of plots of residual enzyme activity versus the molar ratio of inhibitor to enzyme (25).
Second order rate constants for inhibition of human ␣-thrombin or factor Xa by the N135A, K114A/N135A, N135Q, and K114M/N135Q variants in the absence and presence of pentasaccharide or full-length heparin were measured under pseudo-first order conditions, essentially as in earlier work (11,19,20,22,25,31,33). Uncatalyzed reactions with thrombin and factor Xa were analyzed at I 0.15 and 0.05, pH 7.4, and such reactions with factor Xa were also studied at I 0.025, pH 6.0. The rates of the pentasaccharide-catalyzed and full-length heparin-catalyzed reactions of the variants with the two proteinases were only measured at I 0.025, pH 6.0, and I 0.05, pH 7.4, respectively. Reaction mixtures contained 100 -2000 nM antithrombin and 5-10 nM proteinase, with or without 0.1-700 nM pentasaccharide or full-length heparin. The concentration of the saccharides was in most cases Յ10% of the antithrombin concentration. However, in the analyses of the pentasaccharide-catalyzed thrombin inhibition by the N135A and K114A/N135A variants, the pentasaccharide concentration approached and exceeded the antithrombin concentration, due to the small accelerating effect.
After different reaction times, aliquots were diluted 10-fold in I 0.15, pH 7.4 buffer, containing 100 M S-2238 for thrombin or Spectrozyme FXa (American Diagnostica, Greenwich, CT) for factor Xa, and the residual proteinase activity was determined. Observed pseudo-first order rate constants, k obs , were obtained by fitting the decrease of this activity with time to a single exponential decay function with an end point of zero activity (25). Second order rate constants for uncatalyzed reactions were obtained by dividing k obs with the antithrombin concentration. Most such rate constants for pentasaccharide-or full-length heparincatalyzed reactions were derived from the least-squares slope of the linear dependence of k obs on the concentration of the antithrombinsaccharide complex, calculated from measured dissociation constants (11,22,31,33). Alternatively, the rate constants for some catalyzed reactions were calculated from k obs measured at a single saccharide concentration by first subtracting k obs for the uncatalyzed reaction and then dividing by the calculated concentration of the antithrombinsaccharide complex (22,31,33), with several such values averaged. Uncatalyzed rate constants for thrombin inhibition by the N135A and K114A/N135A variants at I 0.025, pH 6.0, and I 0.05, pH 7.4, were obtained from the intercepts on the ordinate of the plots of k obs versus the concentration of the antithrombin-heparin complex from which also the heparin-catalyzed rate constants were derived.

Purification, Homogeneity, and Activity of Antithrombin
Variants-K114A and K114M antithrombin variants were expressed on an N135A background in a baculovirus system, as in our previous studies of the roles of Arg 47 and Arg 129 of the inhibitor in heparin binding (19,20). In addition, a K114M variant was expressed on an N135Q background in a baby hamster kidney cell system. The N135A or N135Q substitutions produce antithrombin forms, corresponding to ␤-antithrombin in plasma, which have high heparin affinity due to the absence of an oligosaccharide side chain on Asn 135 (21,22,34,35). The substitutions thus prevent the heterogeneity in heparin binding affinity associated with partial glycosylation of this residue (21,36). Moreover, the increased affinity facilitates purification of mutants with large defects in heparin binding by affinity chromatography. Both the K114A and K114M substitutions lead to loss of the positive charge of the original Lys, allowing an evaluation of the role of this charge in the binding. The K114A substitution additionally eliminates most of the Lys side chain, whereas the K114M substitution is more conservative, the side chain of Met being approximately isosteric with that of Lys. The two mutations might thus reveal a potential importance of a noncharged side chain in position 114 for the binding.
All variants were purified by heparin affinity chromatography. The N135A and N135Q control variants eluted free from contaminants at about 2.5 M and 2.0 M NaCl, respectively (21,22,35). The N135A variant was used without further purification, whereas the N135Q mutant was subjected to additional anion exchange and gel chromatographies. The K114A/N135A, K114M/N135A, and K114M/N135Q variants eluted at 0.2-0.3 M NaCl together with other proteins and therefore required further purification by anion exchange chromatography and, in the case of the K114M/N135Q variant, also gel chromatography. Both the N135Q and K114M/N135Q variants isolated were the nonfucosylated forms with highest heparin affinity (34,35). All variant preparations were more than 95% homogenous in SDS-polyacrylamide gel electrophoresis and nondenaturing polyacrylamide gel electrophoresis (not shown). The K114M mutants had slightly higher mobilities than the N135A or N135Q controls under nondenaturing conditions at alkaline pH, consistent with the loss of positive charge.
Stoichiometries of pentasaccharide or full-length heparin binding to the variants were measured at antithrombin concentrations substantially higher than K d by titrations monitored by the protein fluorescence enhancement reporting the binding. Because of the high heparin affinities of the N135A and N135Q control variants, the stoichiometries of these variants could be accurately measured at I 0.15, pH 7.4 (22,35). However, as suggested by the elution behavior in heparinaffinity chromatography, the Lys 114 mutations greatly affected heparin binding, and the affinities of the K114A/N135A, K114M/N135A, and K114M/N135Q variants therefore had to be increased by lowering the ionic strength to 0.025 and the pH to 6.0 (31) for the analyses to give sufficiently reliable results. The full-length heparin to antithrombin binding stoichiometry ranged from 0.6 to 0.7 for the two preparations of the N135A variant used in the study and was 0.8 for the N135Q variant preparation. Corresponding full-length heparin to antithrombin stoichiometries ranged from 0.35 to 0.55 for the four preparations of the K114A/N135A variant used, and it was 0.5 for the single preparation of the K114M/N135A variant. In the case of the K114M/N135Q variant, the stoichiometry of pentasaccharide binding was instead determined and was 0.75 for the preparation used. All three Lys 114 mutants showed fluorescence increases of ϳ30% on saturation with pentasaccharide or full-length heparin in the stoichiometry titrations.
Stoichiometries of thrombin binding to the mutants were measured by titrations monitored by the loss of thrombin activity in the absence of heparin at I 0.15, pH 7.4. The thrombin to antithrombin stoichiometry ranged from 0.5 to 0.6 for the N135A control variant and was 0.85 for the corresponding N135Q variant. Similarly, this stoichiometry ranged from 0.4 to 0.7 for the K114A/N135A variant, was 0.55 for the K114M/ N135A variant, and was 0.65 for the K114M/N135Q variant. The heparin and thrombin binding stoichiometries agreed within ϳ20% for each preparation. The subequimolar and comparable heparin and thrombin binding stoichiometries indicate that the preparations contained some inactive, most likely latent, inhibitor, as in previous studies of other recombinant antithrombin variants (19,20,22). In subsequent studies of heparin binding or proteinase inhibition, concentrations of active antithrombin in each preparation of the N135A, K114A/ N135A, and K114M/N135A variants were derived from the heparin and thrombin binding stoichiometries, respectively. However, in both types of analyses of the N135Q and K114M/ N135Q variants, concentrations of active protein were those obtained from the thrombin binding stoichiometry.
Affinity of Pentasaccharide and Heparin Binding-Dissociation equilibrium constants for pentasaccharide and full-length heparin binding to the K114A/N135A, K114M/N135A, and K114M/N135Q variants were measured by fluorescence titrations at protein concentrations comparable with K d . Because of the weak interaction, reliable K d values for pentasaccharide binding to the K114A/N135A and K114M/N135A variants could not be measured at pH 7.4, since prohibitively large amounts of material would have been required. However, a value for pentasaccharide binding to the K114M/N135Q variant was obtained at this pH and I 0.025 (Table I). Not even the pentasaccharide used in determining the crystal structure of the pentasaccharide-antithrombin complex (12), which has an increased affinity for plasma antithrombin due to an extra sulfate group (30,37), bound sufficiently tightly to the K114A/ N135A and K114M/N135A variants to allow accurate measurements at pH 7.4. Reliable values of K d for pentasaccharide binding to all three variants were instead measured at pH 6.0 and low ionic strengths (Table I), as in previous studies of the binding of truncated forms of the pentasaccharide to plasma antithrombin (31). These studies showed that the affinity of pentasaccharide binding is increased at pH 6.0, presumably because of the increased overall positive charge of the pentasaccharide binding site of antithrombin, whereas the mechanism of activation of the inhibitor appears unaltered. As shown for plasma and other recombinant antithrombins (11,19,20,22), full-length heparin bound more tightly than the pentasaccharide to the three Lys 114 variants, so that K d values could be determined with reasonable accuracy at pH 7.4, although only at low ionic strengths (Table I). In addition to these measured values, K d values at I 0.15 for pentasaccharide binding at pH 6.0 and full-length heparin binding at pH 7.4 to the K114A/N135A variant (Table I) were estimated by extrapolation of data obtained at lower ionic strengths (see Fig. 2). Analogously, the binding of both saccharides to the N135A and N135Q variants at all ionic strengths of Յ0.15 was too tight to allow direct measurements of K d , and control values at these ionic strengths (Table I) therefore instead had to be estimated by extrapolation of data obtained at higher ionic strengths (see Fig. 2; similar data, not shown in Fig. 2, were obtained for pentasaccharide binding to the N135Q variant). Such control values were not derived, however, for full-length heparin binding to the N135Q variant, since the ionic-strength dependence of this interaction was not analyzed.
Both the K114A and K114M mutations resulted in decreased affinities for pentasaccharide and full-length heparin, of comparable magnitude for the three variants. The losses in affinity for the pentasaccharide at pH 6.0 from those of the control variants were 1⅐10 6 -to 8⅐10 6 -fold, ϳ3⅐10 5 -fold, and ϳ1⅐10 5 -fold at I 0.025, 0.075, and 0.15, respectively, whereas the corresponding losses for full-length heparin at pH 7.4 were ϳ2⅐10 7 -, ϳ1⅐10 6 -, and ϳ1⅐10 5 -fold, respectively. Comparable reductions in affinity for the pentasaccharide, relative to that of the N135Q control, were measured for the K114M/N135Q variant at pH 7.4 and 6.0, I 0.025, consistent with the effects of the mutation being independent of pH. The decreases given are somewhat uncertain due to the extrapolations involved, the uncertainty being greater at the lower ionic strength due to the longer extrapolation. Nevertheless, the data reflect a dramatic decrease in the affinity of antithrombin for heparin as a result of the two different Lys 114 substitutions. Because of the comparable affinity losses for the three variants, regardless of the expression system used and the length of the noncharged side chain replacing the original Lys, most further experiments were only done with the K114A/N135A variant.
Affinity of Binding of the Nonreducing End Trisaccharide Unit of the Pentasaccharide-A trisaccharide comprising the nonreducing end of the pentasaccharide and denoted DEF (Fig.  1) has been shown to compete with the pentasaccharide for binding to plasma antithrombin and to induce a similar activating conformational change in the inhibitor, although binding more weakly than the pentasaccharide (31). The trisaccharide thus presumably interacts in the same manner with the pentasaccharide binding site of antithrombin when free as when forming part of the pentasaccharide. K d values of (140 Ϯ 20) ϫ 10 Ϫ9 M and (7 Ϯ 0.7) ϫ 10 Ϫ6 M (S.E., n ϭ 3) at I 0.025, pH 6.0, were measured for the binding of the trisaccharide to the N135Q control and the K114M/N135Q variant, respectively, by fluorescence titrations. The loss in affinity for the trisaccharide caused by the K114M mutation, ϳ50-fold, thus was only moderate in comparison with the Ͼ10 6 -fold loss of affinity for the pentasaccharide under the same conditions (Table I). This behavior indicates that most of the energy of binding of the pentasaccharide contributed by Lys 114 can be accounted for by interactions of this residue with the reducing end disaccharide unit of the pentasaccharide and that the interactions with the nonreducing end trisaccharide are substantially weaker. Ionic and Nonionic Contributions to Pentasaccharide and Heparin Binding-The sodium ion concentration dependence of K d for pentasaccharide binding at pH 6.0 and heparin binding at pH 7.4 to the N135A and K114A/N135A variants was evaluated from the measured K d values in Table I and additional data. Log(K d ) varied linearly with log[Na ϩ ] for the interaction of both saccharides with both the N135A control variant and the K114A/N135A variant (Fig. 2), in agreement with the equation, where Z is the number of ionic interactions involved in the binding, ⌿ is the fraction of Na ϩ bound per heparin charge and released on binding to antithrombin (estimated to be 0.8 (32)), and K d Ј is the dissociation constant at 1 M Na ϩ , reflecting the strength of the nonionic interactions (11,22,32). About five ionic interactions were found to contribute to pentasaccharide binding to the N135A control variant at pH 6.0 (Table II), as has been shown previously also at pH 7.4 (19,20,22). This finding contrasts somewhat with the results of a previous study (20), in which about six ionic interactions were found to participate in the binding of the pentasaccharide to the N135A variant at pH 6.0. In both this and the previous work (20), however, the nonionic contribution to the binding was appreciably larger at pH 6.0 than at pH 7.4 (19), corresponding to a decrease in K d Ј of 5-10-fold. The K114A mutation resulted in a loss of two to three (the value obtained being 2.5) charge interactions with the pentasaccharide and in a large decrease of the strength of the nonionic interactions, corresponding to an increase in K d Ј of more than 3 orders of magnitude (Table II). Analogous results were obtained for the binding of full-length heparin at pH 7.4. As shown previously at this pH (19,22), one more ionic interaction is involved in full-length heparin binding than in pentasaccharide binding to the control N135A variant.
A loss of about three charge interactions and an increase in K d Ј of about 3 orders of magnitude were observed for full-length heparin binding as a consequence of the K114A substitution (Table II).

Rapid Kinetics Studies of Pentasaccharide and Heparin
Binding-The kinetics of binding of the pentasaccharide to the N135A and K114A/N135A antithrombin variants were analyzed under pseudo-first order conditions by stopped-flow fluorometry at I 0.075, pH 6.0. This ionic strength and pH were chosen to allow analysis of pentasaccharide binding to both the N135A control and K114A/N135A variants under the same conditions, despite the very large affinity difference between the two variants. However, the rapid binding to the N135A control variant under these circumstances limited the analysis of this variant to low pentasaccharide concentrations (Յ1 M), since the dead time of the stopped-flow instrument precluded measurements of rate constants higher than ϳ500 s Ϫ1 . In the accessible concentration range, k obs for binding to the N135A variant increased linearly with pentasaccharide concentration (Fig. 3A), the slope of this plot giving the overall association rate constant, k on (10, 11) (Table III). However, the intercept on the ordinate, potentially yielding the overall dissociation rate constant, k off (10, 11), was experimentally indistinguishable from zero. k off for N135A binding was instead calculated from k on and the estimated K d and is therefore somewhat uncertain (Table III). In contrast to the N135A control, pentasaccharide binding to the K114A/N135A variant was slower and could be analyzed also at higher saccharide concentrations. The k obs values in-  The number of ionic interactions (Z) involved in the binding of pentasaccharide or full-length heparin to the antithrombin variants and the nonionic contribution (log(K d Ј)) to the binding were determined from the slopes and intercepts, respectively, of plots of log(K d ) versus log[Na ϩ ] (Fig. 2), as described (11,22,32). Errors represent Ϯ S.E. obtained by linear regression. H5, antithrombin-binding heparin pentasaccharide; H26, full-length heparin with high affinity for antithrombin and containing ϳ26 saccharide units. creased hyperbolically with increasing pentasaccharide concentration (Fig. 3B), consistent with the previously defined two-step, induced fit mechanism for heparin binding, involving a weak initial binding of heparin in a rapid equilibrium, followed by a conformational change of antithrombin (10,11,22).
AT*-H SCHEME 1 In this scheme, AT is antithrombin, H is heparin, AT-H is the initial weak complex, AT*-H is the final tight complex, K 1 is the dissociation equilibrium constant of the first binding step, and k ϩ2 and k Ϫ2 are the forward and reverse rate constants, respectively, of the conformational change step. This two-step binding mechanism leads to a hyperbolic dependence of k obs on the total heparin concentration, [H] o , according to the following equation (10,11).
For this mechanism, k on and k off (which in this mechanism is equal to k Ϫ2 ) can be obtained from the initial slope and intercept on the ordinate, respectively, of the hyperbolic plot (10,11). These values were therefore derived by linear regression of the approximately linear increase in k obs with pentasaccharide concentration in the 0.5-1.8 M range (Fig. 3B; Table III). The K114A mutation resulted in a ϳ120-fold decrease in k on and a very large (Ͼ1000-fold) increase in k off , an accurate quantification being impossible in the latter case due to the uncertain control value (Table III). The dissociation equilibrium constant for the pentasaccharide-K114A/N135A interaction was calculated as k off /k on and agreed well with the value measured by fluorescence titrations (Table III). A nonlinear least-squares fit to Equation 2 of the full hyperbolic concentration dependence of k obs for the binding of the pentasaccharide to the K114A/N135A variant gave a K 1 of 7 Ϯ 0.4 M, and a k ϩ2 of 36 Ϯ 1 s Ϫ1 . An assessment of the effect of the Lys 114 mutation on these parameters requires that the values for the N135A control be estimated, since the kinetic analyses of this variant could not be extended to sufficiently high pentasaccharide concentrations to allow a direct evaluation (see above). Previous studies of the interaction of the pentasaccharide with plasma antithrombin have shown that decreasing the pH and the ionic strength only affects K 1 and not k ϩ2 (31). Assumption of the same behavior for the N135A variant suggests that k ϩ2 would be 2100 Ϯ 300 s Ϫ1 (i.e. the value measured at I 0.15, pH 7.4 (19)) and that K 1 , calculated as k ϩ2 /k on (11), would be 4 Ϯ 0.7 M at I 0.075, pH 6.0. That these values are reasonable estimates was verified by simulations, based on Equation 2, showing that the plot of k obs versus pentasaccharide concentration for the N135A variant would have deviated detectably from linearity for K 1 Յ 3 M and k ϩ2 Յ 1600 s Ϫ1 . The data therefore suggest that the K114A mutation resulted in a major decrease in k ϩ2 (Ͼ40-fold) and at most a minor increase in K 1 (Ͻ3-fold) for pentasaccharide binding.
Unlike pentasaccharide binding, the kinetics of full-length heparin binding to both the N135A and K114A/N135A variants could be analyzed at I 0.075, pH 7.4. As for the pentasaccharide, the dependence of k obs for the interaction of full-length heparin with the N135A variant on heparin concentration could only be measured at low heparin concentrations and varied linearly with these concentrations (Fig. 3A). A value for k on was obtained from the slope of this dependence, whereas k off again had to be calculated from k on and the estimated value of K d (Table III). In contrast to what was observed with the pentasaccharide, k obs for full-length heparin binding to the K114A/N135A variant increased linearly with heparin concentration up to 10 M (Fig. 3B). Because of the low slope of this dependence, it was considered futile to extend the analyses to higher heparin concentrations to possibly obtain evidence for a hyperbolic behavior, in particular in view of the prohibitively large amounts of full-length heparin that would have been required. Only k on and k off could therefore be determined for full-length heparin binding to the K114A/N135A variant (Table  III). The K114A mutation resulted in a ϳ1200-fold decrease in k on and a Ͼ5000-fold increase in k off , the magnitude of the effect on the latter parameter again being uncertain due to the approximate k off for the control.
Contrary to what was observed for pentasaccharide binding, the calculated K d for full-length heparin binding to the K114A/ N135A variant was ϳ15-fold higher than the measured K d (Table III). A similar discrepancy was also apparent from measurements of k on , k off , and K d at I 0.075, pH 6.0 (Table III), suggesting that the agreement for the pentasacharide and disagreement for full-length heparin is not due to the different pH values used in the studies of the two saccharides. The effect of the Lys 114 mutation on k on and k off for full-length heparin binding under these latter conditions could not be assessed, because the binding to the N135A control was too rapid, even at the lowest experimentally feasible heparin concentrations. The disagreement between the calculated and measured values of K d for full-length heparin binding to the K114A/N135A variant is most likely due to a small contribution of a preequilibrium pathway for heparin activation of this mutant, as found previously for mutants of both Arg 47 and Arg 129 (19,20). In this pathway, heparin binds preferentially to a small amount of already conformationally activated antithrombin in equilibrium with unactivated inhibitor (31). Also a minor flux along the preequilibrium pathway, in addition to the dominating, induced fit pathway, may lead to an anomalously high k off being determined (19). The measured k off for full-length heparin binding to the K114A mutant (Table III) is therefore most likely an overestimation of the true value. Nevertheless, it is apparent that the mutation caused a substantial increase in k off .
Proteinase Inhibition-Second order rate constants for inhibition of thrombin and factor Xa by the N135A and K114A/ N135A variants in the absence and presence of pentasaccharide or full-length heparin were determined by discontinuous assays of residual activity. Essential proteinase inhibition rate constants were also checked for the N135Q and K114M/N135Q variants. Uncatalyzed rate constants were determined at I 0.15 and 0.05, pH 7.4, and I 0.025, pH 6.0, whereas pentasaccharide-and full-length heparin-catalyzed rate constants were only determined at I 0.025, pH 6.0, and I 0.05, pH 7.4, respectively. Under the latter conditions, the K d values for the binding of the two saccharides to the K114A/N135A and K114M/ N135Q variants were known and were close to the antithrombin concentration used, so that concentrations of antithrombin-saccharide complexes could be calculated with reasonable accuracy (22,25,31,33). It should be noted that the catalyzed rate constants for antithrombin inhibition of proteinases measured by the procedure used in this work reflect the rate constants for proteinase inhibition by the antithrombinsaccharide complexes, i.e. the rate constants at saturation of the inhibitor with the saccharides (25,33).
All uncatalyzed rate constants were essentially unaffected by the K114A mutation (Table IV). Similarly, in most cases no differences between the K114A/N135A or K114M/N135Q antithrombin variants and their N135A or N135Q controls in the rate constants for either pentasaccharide-or full-length heparin-catalyzed inhibition of the two proteinases were evident (Table IV). However, the pentasaccharide-catalyzed rate constant for factor Xa inhibition by the K114M/N135Q variant was ϳ2.5-fold lower than that of the N135Q variant. Moreover, the rate constants for full-length heparin-catalyzed inhibition of thrombin by the K114A/N135A and K114M/N135Q variants were ϳ14and ϳ2.5-fold higher, respectively, than those of the controls. Stoichiometries of thrombin inhibition by the four variants in the presence of full-length heparin at I 0.05, pH 7.4, were measured to clarify whether these latter differences could reflect a larger contribution of the substrate pathway to the measured rate constants for the control variants (3,38). A rate constant characterizing only the inhibitory pathway of the serpin mechanism can thus be calculated as k app SI, where k app is the measured rate constant and SI is the antithrombin to thrombin stoichiometry of inhibition (3,38). The N135A and K114A/N135A variants inhibited thrombin under the conditions of the kinetics measurements in the presence of fulllength heparin with stoichiometries of ϳ6 and ϳ3, respectively, reflecting substantial contributions of the substrate pathway that differed between mutant and control. Correction for these different extents of the substrate reaction still resulted in an appreciable ϳ7-fold difference in the rate constants of the inhibitory pathway for the two variants (ϳ15 ϫ 10 6 and ϳ120 ϫ 10 6 M Ϫ1 ⅐s Ϫ1 , respectively). In contrast, the stoichiometries of inhibition of the N135Q and K114M/N135Q variants under the same conditions, ϳ6 and ϳ2, respectively, imply that the true inhibition rate constants for these two variants (ϳ140 ϫ 10 6 and ϳ110 ϫ 10 6 M Ϫ1 ⅐s Ϫ1 , respectively) are essentially indistinguishable.

DISCUSSION
Early chemical modification studies gave indications for Lys 114 of antithrombin being involved in binding of heparin (39). Site-directed mutagenesis of this residue to Gln was later shown to decrease the ability of heparin or the heparin pentasaccharide to accelerate antithrombin inhibition of thrombin or factor Xa (16,17). The substitution also reduced the affinity of the inhibitor for heparin, although the extent of the decrease could not be quantified (16). A role of Lys 114 in heparin binding was further indicated by the x-ray structure of an antithrombin-pentasaccharide complex (12), which revealed that this residue makes contacts with several groups of the pentasaccharide (Fig. 1). In the present work, substitution of Lys 114 by Ala or Met was found to decrease the affinity of antithrombin for pentasaccharide at pH 6.0 and full-length heparin at pH 7.4 by a factor that varied between ϳ10 6 -and ϳ10 7 -fold at I 0.025 and ϳ10 5 -fold at I 0.15. The effects of mutations of Arg 47 , Lys 125 , or Arg 129 , the other residues of antithrombin implicated to be of major importance for heparin binding (12) (Fig. 1), on heparin affinity have previously been assessed with a ␤-antithrombin background the same as or comparable with that used in this work (15,19,20). By contrast to the Lys 114 mutations, the largest decrease in heparin or pentasaccharide affinity in these studies was only ϳ2500-fold at I 0.15, pH 7.4, observed for heparin binding to an Arg 129 mutant (20). The  Table I). c From Table I, for facile comparison. three Lys 114 mutants inactivated thrombin with stoichiometries comparable with those of the controls and were normally activated by heparin, as judged from the typical ϳ30% fluorescence increase seen on saturation with the polysaccharide. Moreover, the Lys 114 mutations did not change the uncatalyzed rate constants for antithrombin inhibition of thrombin or factor Xa or, as will be discussed further below, the catalyzed rate constants (i.e. the rate constants at saturation of antithrombin with pentasaccharide or full-length heparin) for factor Xa inhibition. These findings indicate that the Lys 114 mutations only eliminated a side chain of importance for heparin affinity and did not affect the native or activated conformations of the inhibitor. The reductions in affinity at I 0.15 correspond to a binding energy of ϳϪ29 kJ⅐mol Ϫ1 (i.e. ϳ50% of the total energy of binding of pentasaccharide or full-length heparin to the antithrombin control). These results thus show that Lys 114 of antithrombin is of appreciably greater quantitative importance for heparin binding than any other residue investigated so far. The large decrease in antithrombin affinity for pentasaccharide and full-length heparin caused by the K114A mutation was due to loss of both multiple (two to three) ionic interactions and nonionic interactions. The observed loss of more than one ionic interaction presumably means that the interaction of the positive charge of Lys 114 with a negative charge in the pentasaccharide region is a prerequisite for at least one other antithrombin residue to establish a comparable ionic interaction. Similar to what has been observed for the role of Arg 129 in heparin binding (20), the binding of Lys 114 and at least one more residue of antithrombin to heparin thus appears to be highly cooperative. This cooperative binding also comprises appreciable nonionic interactions (viz. hydrogen bonds, hydrophobic interactions, or van der Waals interactions) between the inhibitor and the saccharide, most likely involving both the Lys 114 side chain and other antithrombin residues of the cooperative network. The nonionic interactions are more important for the binding affinity than the ionic interactions, as evident from the former contributing 800 -2500-fold to the affinity decrease due to the K114A mutation at I 0.15, compared with the ionic interactions contributing only 40 -120-fold.
Only the kinetics of binding of the pentasaccharide to the K114A mutant could be fully characterized and indicated that the decrease in pentasaccharide affinity as a result of the mutation was due to both a large (ϳ120-fold) decrease of k on and an even larger (Ͼ1000-fold) increase in k off . The data obtained for full-length heparin are in agreement with comparable major effects on both k on and k off . The pentasaccharide bound to the K114A mutant by a two-step mechanism analo-gous to that established previously for binding of pentasaccharide and full-length heparin to plasma and recombinant antithrombins (Scheme 1) (10,11,19,20,22). The large decrease in k on caused by the K114A mutation was due primarily to a large (Ͼ40-fold) decrease of the forward rate constant of the conformational change in the second step, the effect on the first step being at most minor. Moreover, the reverse rate constant of the second step was greatly increased (Ͼ1000-fold) by the mutation, since this rate constant in the two-step binding mechanism is equal to k off (10,11). Lys 114 thus does not participate to any appreciable extent in the initial, weak binding of heparin. However, it is of major importance both for the rate of induction of the conformational change that activates antithrombin in the second step and for keeping the pentasaccharide anchored to the inhibitor and thereby maintaining the latter in the activated state. Its effects in these respects are larger than those of either Arg 47 or Arg 129 , which also predominantly participate in the second binding step (19,20). Arg 129 is of greatest importance of the two but only contributes ϳ5-fold to increasing the forward rate constant of the conformational change induced by the pentasaccharide and ϳ100-fold to decreasing the reverse rate constant of this change (20).
The second step of pentasaccharide and full-length heparin binding to antithrombin is normally strongly shifted toward the conformationally activated state of antithrombin, so that essentially all of the inhibitor exists in this state at saturation with the saccharides (11,20). However, due to the large effects of the K114A mutation on both rate constants of the second step of pentasaccharide binding, the equilibrium constant of this step, calculated as k ϩ2 /k Ϫ2 , is greatly reduced and is only ϳ1.4 at I 0.075, pH 6.0. Because the conformationally activated state therefore is only marginally favored under these conditions, not all of the inhibitor is converted to this state even at saturation with the pentasaccharide, which would be expected to lead to a decreased catalyzed rate of inhibition of factor Xa. However, the anticipated decrease of k Ϫ2 but minimal change of k ϩ2 at the lower ionic strength (0.025) in the measurements of the proteinase inactivation kinetics (11,31) would be expected to result in an equilibrium more in favor of the activated state. This favorable equilibrium presumably accounts for the fact that no clear reduction in the rate of pentasaccharidecatalyzed factor Xa inactivation by the K114A/N135A mutant, compared with that of the N135A control, was evident. By contrast, the 2.5-fold reduction observed for the K114M/N135Q mutant suggests that the mutation in this case shifted the equilibrium of the second step to only marginally favor, or even disfavor, the activated state. antithrombin variants with proteinases at 25°C Second-order association rate constants for uncatalyzed (k uncat ), pentasaccharide-catalyzed (k H5 ), and full-length heparin-catalyzed (k H26 ) reactions of the antithrombin variants with proteinases were determined as described in Refs. 11,19,22,and 33. Errors are given as Ϯ S.E. in the case of three or more measurements and as Ϯ range in the case of two measurements. The number of measurements are given in parentheses. ND, not determined. a Obtained from the intercepts on the ordinate of the plots of k obs versus the concentration of the antithrombin-heparin complex from which the heparin-catalyzed rate constants were derived.