The Heparin Binding Properties of Heparin Cofactor II Suggest an Antithrombin-like Activation Mechanism*

The serpin heparin cofactor II (HCII) is a glycosaminoglycan-activated inhibitor of thrombin that circulates at a high concentration in the blood. The antithrombotic effect of heparin, however, is due primarily to the specific interaction of a fraction of heparin chains with the related serpin antithrombin (AT). What currently prevents selective therapeutic activation of HCII is the lack of knowledge of the determinants of glycosaminoglycan binding specificity. In this report we investigate the heparin binding properties of HCII and conclude that binding is nonspecific with a minimal heparin length of 13 monosaccharide units required and affinity critically dependent on ionic strength. Rapid kinetics of heparin binding indicate an induced fit mechanism that involves a conformational change in HCII. Thus, HCII binds to heparin in a manner analogous to the interaction of AT with low affinity heparin. A fully allosteric 2000-fold heparin activation of thrombin inhibition by HCII is demonstrated for heparin chains up to 26 monosaccharide units in length. We conclude that the heparin-binding mechanism of HCII is closely analogous to that of AT and that the induced fit mechanism suggests the potential design or discovery of specific HCII agonists.

The serpin heparin cofactor II (HCII) is a glycosaminoglycan-activated inhibitor of thrombin that circulates at a high concentration in the blood. The antithrombotic effect of heparin, however, is due primarily to the specific interaction of a fraction of heparin chains with the related serpin antithrombin (AT). What currently prevents selective therapeutic activation of HCII is the lack of knowledge of the determinants of glycosaminoglycan binding specificity. In this report we investigate the heparin binding properties of HCII and conclude that binding is nonspecific with a minimal heparin length of 13 monosaccharide units required and affinity critically dependent on ionic strength. Rapid kinetics of heparin binding indicate an induced fit mechanism that involves a conformational change in HCII. Thus, HCII binds to heparin in a manner analogous to the interaction of AT with low affinity heparin. A fully allosteric 2000-fold heparin activation of thrombin inhibition by HCII is demonstrated for heparin chains up to 26 monosaccharide units in length. We conclude that the heparin-binding mechanism of HCII is closely analogous to that of AT and that the induced fit mechanism suggests the potential design or discovery of specific HCII agonists.
Thrombin is the ultimate protease in the blood coagulation cascade and has many pro-coagulant functions (for review see Ref. 1). Thrombin inhibition is thus critical for the prevention of thrombosis, and is achieved by the circulating serpins antithrombin (AT) 1 and heparin cofactor II (HCII) in a glycosaminoglycan (GAG)-dependent fashion (for review see Ref. 2). The antithrombotic effect of therapeutic heparin, however, is mediated primarily by its interaction with AT (3), because of the presence of a specific AT-binding pentasaccharide sequence in one-third of heparin chains (4). The failure of heparin to appreciably activate HCII in vivo derives from the absence of such a high affinity HCII-specific sequence (5,6), and thus high heparin levels would need to be achieved before HCII could contribute to the antithrombotic effect of heparin. HCII could thus be considered an untapped source of anti-thrombin activity.
Although HCII deficiency has not been established as a significant risk factor for thrombosis (7), recent studies suggest that HCII may play a role in preventing arterial thrombosis (7)(8)(9)(10) and may be uniquely capable of inhibiting clot-bound thrombin (11). Clot-bound thrombin is resistant to heparin anticoagulant therapy because of the obligate co-occupation of AT and thrombin on the same heparin chain (12,13). However, HCII inhibition of thrombin does not require the formation of a heparin bridge to thrombin (14); rather, HCII interacts directly with a part of thrombin that does not mediate its sequestration on fibrin (15). The discovery or design of a specific agonist of HCII may thus provide a significant improvement in anticoagulant therapy by targeting clot-bound thrombin, resistant to inhibition by heparin-activated AT.
The recently solved structures of native HCII and HCII in its Michaelis complex with thrombin support the hypothesis of an AT-like heparin activation mechanism for HCII (15). AT binds to a specific pentasaccharide sequence of heparin by an induced fit, conformational change mechanism (16) (Fig. 1A), which involves an initial weak interaction followed by a rapid conformational change and a 1000-fold improvement in binding affinity (K d from 37 M to 20 nM) (17). Such a mechanism could be in operation for HCII as its heparin-binding region (Fig. 1B), and native and activated conformations (Fig. 1C) are nearly identical to AT. If HCII shares the heparin-binding mechanism of AT it becomes possible to envision the discovery or design of GAGs that interact with HCII as the pentasaccharide does with AT. In this study we determined the heparin binding characteristics of HCII that support an induced fit, AT-like mechanism of heparin binding and resolve a 2000-fold allosteric activation toward thrombin inhibition by HCII. These findings confirm HCII as a target for the design of specific antithrombotic agents.

EXPERIMENTAL PROCEDURES
Materials-Unfractionated heparin with an average molecular mass of 15,000 Da (UFH) was purchased from Rovi Pharmaceutical Laboratories (Madrid, Spain). Depolymerized heparins of lengths varying from 6 to 26 monosaccharide units were prepared as before (18). The main disaccharide composition (75%) was the trisulfated IdoA,2S-GlcNS,6S. Human thrombin was purchased from Sigma, and human HCII was purified from freshly frozen plasma as described previously (19). 2-(p-Toluidinyl)naphthalene-6-sulfonic acid (TNS) was purchased from Molecular Probes and was reconstituted in dimethyl sulfoxide.
Equilibrium Binding Studies-Binding of heparin to HCII was followed by monitoring TNS fluorescence, essentially as previously done with AT variants (20). Briefly, fluorescence emission spectra (360 -500 nm) were collected on a Perkin Elmer LS 50B fluorometer exciting at 330 nm from solutions containing 250 nM HCII, 10 M TNS in 50 mM Tris, pH 7.4, 0.1% polyethylene glycol 8000, and ionic strength adjusted with NaCl (125 mM NaCl referred to as "physiological ionic strength buffer"). An average of two spectra was recorded for each titration point, and the emission at 448 nm was taken as the fluorescence maximum. The data were corrected for dilution and fitted to Equation 1, where ⌬F and ⌬F max are fractional and maximal fractional fluorescence change, [HCII] o and [H] o are total HCII and heparin concentrations, and K d is the apparent dissociation constant as described previously (21) to obtain values for the dissociation constants, using the program GraphPad (Sigma). Saturation of fluorescence signal was not achieved for heparin chains smaller than 8 monosaccharide units in length, even when no NaCl was added to the buffer. The dissociation constants obtained using the extrinsic fluorescence method were independent of the concentrations of either TNS or HCII, because identical dissociation constants were obtained for UFH when either concentration was doubled (data not shown). In addition, TNS fluorescence was found to be insensitive to the heparin additions in the absence of HCII (data not shown). Stoichiometric binding (i.e. one molecule of heparin binds only one molecule of HCII) was verified for the 26-mer heparin at 0 mM NaCl (data not shown), and thus the binding site density is inferred to be no greater than one for all studies. The dependence of binding affinity on heparin length was determined as previously (22) using Equation 2, where K d,app is the measured, apparent dissociation constant derived from Equation 1 for heparin chains of N monosaccharide units, and K d,int is the intrinsic dissociation constant for a heparin chain of minimal binding site size L. The slope of the linear fit of the dependence of equilibrium association constant versus heparin length thus gives the intrinsic dissociation constant, and the x intercept is the minimal binding chain length minus one monosaccharide unit. The effect of ionic strength on binding affinity was determined as previously (21) using Equation 3, where K d is the apparent dissociation constant at a particular ionic strength (I), K NI is the apparent nonionic contribution to the apparent dissociation constant, Z is the number of ionic interactions, and is a constant representing the fraction of counter ions on heparin released for each ionic interaction with a protein (estimated at 0.8 (13)). The intrinsic nonionic dissociation constant, K NI,int , was determined from the linear plot of the dependence of apparent K NI on heparin chain length, according to Equation 2. Rapid Kinetics Studies-Rapid binding experiments were conducted under pseudo first order conditions in heparin (minimal 5-fold excess of heparin over HCII) using a BioLogic MOS-450 stopped flow fluorometer (Grenoble, France). The fluorescence signal from the extrinsic fluorescence probe TNS (100 M) was followed by exciting at 330 nm using an emission cut-off filter of 360 nm. The method was validated by determining the rates of 16-mer binding in 50 mM Tris, pH 7.4 (0 NaCl), where an accurate equilibrium dissociation constant had already been obtained. Good agreement between the equilibrium value and that FIG. 1. The induced fit heparinbinding mechanism of AT. A, AT (ribbon diagram) binds heparin (ball-andstick) via a two-step mechanism. The circulating native state of AT (left panel) is in a conformation that is poorly reactive toward its target proteases and interacts weakly with heparin (middle panel, K 1 ϭ 37 M). When a specific pentasaccharide sequence is encountered, AT undergoes a rapid conformational change (k 2 ) to a high affinity state (right panel, overall K d ϭ 20 nM). The conformational change includes the N-and C-terminal elongation (magenta) of helix D (cyan) and the expulsion of the hinge region of the reactive center loop (yellow) from ␤-sheet A (red). Tight binding is significantly determined by the ability of the specific pentasaccharide to lock AT in the high affinity conformation by effectively reducing the rate of reverse conformational change (k Ϫ2 ). B, the heparin binding region of AT (left panel) is nearly identical to that of HCII (right panel) despite the apparent convergent evolution. The residues (ball-and-stick) on AT that interact with the specific pentasaccharide are Arg 46 , Arg 47 , Lys 114 , Lys 125 , and Arg 129 , with Arg 132 and Lys 133 interacting with longer heparin chains and contributing minimally to affinity. Basic residues are found on HCII at each of these positions. C, the native (left panel) and activated (right panel) conformations of HCII (colored as above) resemble the corresponding conformations of AT. The hinge region of the native conformation is partially inserted into ␤-sheet A and is expelled in the active conformation with the consequential closure of ␤-sheet A to the five-stranded state. calculated from the linear plot of observed rate constant versus 16-mer concentration (see Table II) suggested that the binding of TNS to HCII was rapid relative to the binding of heparin to HCII. The rapid binding of TNS to HCII was also established directly by rapidly mixing TNS with HCII and monitoring change in TNS fluorescence. Only 1% of the equilibrium signal was observed, indicating that the reaction was 99% completed in the dead time of the instrument (5.3 ms). As found for the equilibrium titrations, the fluorescence of TNS was not affected by heparin alone under stopped flow conditions (data not shown). Kinetics of UFH binding to HCII were measured at physiological ionic strength (50 mM Tris, pH 7.4, 0.1 polyethylene glycol 8000, 125 mM NaCl), with 100 M TNS present in all solutions. The values for k obs were obtained from the fit of the average of at least six traces to the equation for single exponential decay and were plotted against total heparin concentration ([H] 0 ). All of the data were then fit to Equation 4, where K 1 is the initial rapid dissociation constant, k 2 and k Ϫ2 are the rates of forward and reverse conformational change steps, respectively, according to the kinetic mechanism outlined in Fig. 1A. On and off rates were also measured by rapid dilution experiments, where one syringe was filled with 1 M HCII, 50 M UFH, and 100 M TNS, and another was filled with 100 M TNS alone, both in the 125 mM NaCl buffer. 2-, 3-, 4-, and 5-fold dilutions were conducted, and the observed rate constant, k obs , was fit to Equation 5, where k off and k on are the off and on rates, and [HCII] o and [UFH] o are the total concentrations of HCII and UFH, respectively.

Effect of Heparin on Rates of Thrombin Inhibition-
The effect of heparin concentration (from 1 to 100 nM) on the rate of thrombin inhibition by HCII was determined for heparin fractions ranging in size from 8 to 26 monosaccharide units in length under pseudo first order conditions in HCII. Solutions contained 50 mM Tris, pH 7.4, 125 mM NaCl, and 0.1% polyethylene glycol 8000, and HCII and thrombin concentrations were 1 M and 25 nM, respectively, in all experiments. Heparin chains of this size range are incapable of bridging HCII and thrombin, and thus any effect on the observed rates of thrombin inhibition is due to allostery alone. To determine whether the rate enhancement was due to the interaction of heparin with HCII or thrombin, the observed rate constants of thrombin inhibition were plotted against 10-mer heparin concentrations up to 100 M. The resulting fit to Equation 1 (substituting the fluorescence signal with k obs ) suggested that the interaction of heparin with HCII alone affected the observed rate of thrombin inhibition (more details in "Results"). Thus, it was possible to obtain the maximal catalyzed rate of thrombin inhibition (k cat ) for all heparin fragments used by fitting the plot of k obs versus total heparin concentration, [H] 0 (up to 100 nM) according to Equation 6, where [HCII] 0 is the initial HCII concentration, K d is the apparent dissociation constant for the HCII-heparin interaction (extrapolated for heparin chains smaller than 16 units in length), and k uncat is the rate constant for the uncatalyzed reaction.

RESULTS
The Effect of Heparin Size and Ionic Strength on Affinity for HCII-Characterization of the binding affinity of HCII for heparin and other GAGs has been hampered by the lack of intrinsic fluorescence signal and by the poor affinity of the interaction (23,24). We adapted an extrinsic fluorescence method used previously to determine the binding constants of AT variants with heparin (20). For AT, where a known conformational change is associated with heparin binding, a 50% quench in TNS fluorescence and a small blue shift are observed. Similarly, TNS fluorescence is altered in a saturable manner by the addition of heparin to solutions containing TNS and HCII. Interestingly, the fluorescence change depended on the length of heparin chain used in the titration, with chains 12 monosaccharide units and smaller resulting in a 40 -70% fluorescence enhancement with a 1.5-nm blue shift, and chains 14 monosaccharide units and larger causing a 30 -50% fluorescence quench with a 7.5-nm red shift ( Fig. 2A). Both signals were saturable and allowed for the determination of apparent dissociation constants (K d ) (example titrations are given in Fig. 2B). The value obtained for UFH at physiological ionic strength using this method (26.6 M) is in good agreement with previously obtained values (24,25). Thus, K d values were determined for heparin chains of 8,10,12,14,16,18,20,26, and 50 (UFH) units in 50 mM Tris buffer (Table I), with ionic strength adjusted by adding NaCl. Even under low ionic strength conditions, saturation of fluorescence signal could not be achieved for any heparin chain shorter than 8 monosaccharide units in length. Analysis of the dependence of apparent equilibrium association constant (K a,app ϭ 1/K d,app ) on heparin chain length, according to Equation 2, resulted in experimentally indistinguishable heparin-binding site lengths (L) of 12.5, 14.5, and 11.2 monosaccharide units at 0, 20, and 125 mM NaCl, respectively (Fig. 2C). The minimal heparin length that can fully occupy the heparin-binding site of HCII is thus ϳ13 monosaccharide units (the average value is 12.7). Consistent with a 13-mer being the minimal heparin size to fully occupy the heparin-binding site on HCII is the observation that a TNS fluorescence enhancement is observed for heparin chains smaller than 14 monosaccharide units in length, with a quench observed for all heparin lengths 14 monosaccharide units and larger. The saturable fluorescence enhancement for 10-and 12-mer heparin fragments was found to be followed by a fluorescence quench upon further addition (data not shown). This behavior supports the contention that full occupancy of the heparin-binding site on HCII yields a fluorescence quench under these experimental conditions. The slopes of the linear fits in Fig. 2C also provide values for the intrinsic dissociation constant, K d,int , which is defined as the dissociation constant for the minimal 13-mer heparin chain at the three ionic strengths studied. The values obtained were 1.1 Ϯ 0.2, 2.2 Ϯ 0.1, and 695 Ϯ 167 M for 0, 20, and 125 mM NaCl, respectively, and correspond well to those directly determined for the 14-mer at 0 and 20 mM NaCl (0.4 and 2.4 M) and the extrapolated value at 125 mM NaCl (275 M). The apparent increase in binding affinity for heparin chain lengths greater than 13 monosaccharide units can be understood as the statistical effect of increasing the number of overlapping HCII sites available, with each monosaccharide addition resulting in an additional possible binding site (N Ϫ L ϩ 1), according to Equation 2.
Although it was not possible to directly measure the K d values for heparins smaller than 16 monosaccharide units in length at physiological ionic strength, it was, however, possible to calculate dissociation constants from the double log plot of the dependence of K d on ionic strength ( Fig. 2D and Table I).
The linear fit also provides information about the number of ionic interactions and the strength of the nonionic contribution, according to the polyelectrolyte theory of Record et al. (26). The number of ionic interactions (Z) is calculated from the slope, and the nonionic contribution to the apparent dissociation constant (K NI ) is derived from the y intercept (Equation 3 and Table I). The number of ionic interactions increases from 2 for the 10-mer, to 4 for the 12-mer, to 4 -5 for all other heparin lengths up to the 26-mer, where one or two additional ionic interactions are observed. Thus, our observation that 13 monosaccharide units is the minimal heparin size for full occupation of the heparin-binding site of HCII corresponds to between four and five ionic interactions and is identical to the value obtained for the interaction of AT with its specific pentasaccharide (Table I). The binding affinity appears to derive largely from the ionic contribution, with the apparent nonionic dissociation constants (K NI ) in the millimolar range. The decrease in the apparent K NI with increasing heparin chain length is a product of the statistical effect of increasing the number of overlapping HCII-binding sites, as discussed above. When the apparent nonionic dissociation constants, K NI,app , are fit to Equation 2 (Fig. 2D, inset), an intrinsic K NI of 460 mM is obtained. This value is close to the theoretical maximum of 1 M and is indicative of a nonspecific binding model where nonionic interactions do not contribute appreciably to the overall binding affinity. Thus, heparin binding to HCII utilizes the same number of ionic interactions as does heparin binding to AT, and the low affinity of the heparin-HCII interaction is explained by the weak nonionic contribution.
Rapid Kinetics Studies Support an Induced Fit Heparinbinding Mechanism-Long before crystal structures of AT revealed the conformational changes associated with heparin binding (27-29), a conformational change step was inferred  from stopped flow studies (16). The hyperbolic dependence of the rate of binding on the concentration of heparin (under pseudo first order conditions in heparin) indicated that a conformational change step was the limiting rate at high heparin concentrations. The shared native and activated conformations of AT and HCII suggest a similar induced fit mechanism for the interaction of HCII with heparin. We established that the extrinsic fluorescence method employed in equilibrium fluorescence studies could also be applied under stopped flow conditions by determining the on and off rates for the binding of 16-mer heparin to HCII in 50 mM Tris buffer (Table II). From the linear plot of k obs versus concentration of the 16-mer, we calculated a K d of 120 nM (data not shown), which was similar to that obtained under equilibrium conditions (220 nM). Thus, to test the hypothesis that HCII binds heparin via an AT-like induced fit mechanism, we determined the effect of heparin concentration on the rate of UFH binding to HCII at physiological ionic strength (Table II and Fig. 3). The plot of the observed pseudo first order rate constants versus total heparin concentration (Fig. 3A) is unmistakably hyperbolic, and when fit to Equation 2 yields values for the initial dissociation constant, K 1 , of 104 M, and of the forward and reverse conformational change steps, k 2 and k Ϫ2 , of 111 and 12.4 s Ϫ1 , respectively. The on rate, k on , calculated from the hyperbolic plot was 1.07 M Ϫ1 s Ϫ1 , and is thus 20 times slower than k on for heparin binding to AT (20 M Ϫ1 s Ϫ1 ) (17). For AT, k Ϫ2 is indistinguishable from the off rate, k off , because the first binding step, K 1 , represents a rapid equilibrium (16). Rapid dilution experiments yielded similar results for both k on and k off (0.69 M Ϫ1 s Ϫ1 and 13.6 s Ϫ1 , respectively), and a calculated dissociation constant of 19.7 M (Fig. 3B and Table II (17) and is therefore 34 times slower than the k off value obtained for the HCII-heparin interaction. Thus, the ϳ1000-fold weaker affinity of HCII for heparin than of AT for heparin (containing the pentasaccharide sequence) is due to a 20 -30-fold reduction in the on rate and a ϳ34-fold increase in the off rate. Interestingly, the forward conformational change (k 2 ) for HCII is only 6-fold slower than for AT, indicating that, even in the absence of an HCII-specific heparin sequence, heparin binding is effective at inducing the activating conformational change in HCII.
The Effect of Heparin Size on the Acceleration of Thrombin Inhibition-The dependence of observed rate of thrombin inhibition by HCII on heparin chain length was determined for heparin concentrations ranging from 1 to 100 nM ( Fig. 4A and Table III). The experiments were conducted under pseudo first order conditions in HCII (1 M for HCII and 25 nM for thrombin) at physiological ionic strength. Because of the concentrations of HCII and thrombin used and the low affinities at physiological ionic strength, most of the heparin added to the inhibition reactions is free, with only a small fraction bound to HCII or thrombin. Fig. 4A is a typical plot of the linear dependence of k obs on heparin concentration. Heparins smaller than 30 monosaccharide units in length are incapable of bridging thrombin to HCII (15,23); thus any effect on the rate of thrombin inhibition will therefore be due either to an allosteric effect on HCII or on thrombin or a combined allosteric effect. The crystal structure of the Michaelis complex between HCII and thrombin (15) suggests that occupancy of exosite II on thrombin by small heparin chains would not affect the rate of throm-

FIG. 3. Rapid kinetics studies indicate that HCII binds heparin via an induced fit mechanism.
A, the hyperbolic dependence of pseudo first order rates of heparin binding on concentration is indicative of an induced fit mechanism. The values were obtained using UFH at physiological ionic strength for UFH concentrations ranging from 3.3 to 120 M, and a representative trace at 3.3 M is shown in the inset. B, on and off rates were also determined by following the fluorescence enhancement obtained from the rapid dilution of the preformed HCIIheparin complex (example trace shown as inset). a Conducted in the absence of NaCl. b From rapid dilution experiments. H5 is the AT-specific pentasaccharide, and H26 is high affinity heparin, each with plasma ␣-AT at physiological ionic strength from Turk et al. (17). -, not determined or not applicable. K d,calc is the dissociation constant calculated from kinetic constants, and K d,meas is the measured equilibrium dissociation constant. bin inhibition by HCII. This conclusion is supported by studies conducted on thrombin exosite mutants (14), and wild-type thrombin in the presence of a DNA aptamer specific for the heparin-binding site of thrombin (23). We would thus predict that only the heparin interaction with HCII would affect the observed rate of thrombin inhibition in the presence of small heparin chains. To test this, we continued to determine k obs in the presence of higher 10-mer heparin concentrations (up to 100 M) and plotted the data as before (Fig. 4B). The data fit a nonlinear, hyperbolic curve, which gave an apparent K d of 41 M, which is close to our extrapolated K d for the 10-mer-HCII interaction and is inconsistent with the previously determined K d for thrombin with 10-mer heparin (22). Therefore, because the thrombin-heparin interaction does not appreciably affect the observed rate of thrombin inhibition by HCII, we can now calculate maximal catalyzed second order rate constants (k cat ) of thrombin inhibition by HCII in the presence of the heparins from 10 to 26 units in length, taking into account the calculated and observed dissociation constants. The calculated k cat values are given in Table III and are remarkably similar in magnitude, ranging from 3.06 to 19 ϫ 10 6 M Ϫ1 s Ϫ1 with an average value of (8.1 Ϯ 5.1) ϫ 10 6 M Ϫ1 s Ϫ1 . Because the K d values for the interaction of the 10-, 12-, and 14-mer were extrapolated from values obtained at low ionic strength, the k cat values obtained for heparins of 16 -26 monosaccharide units in length are less subject to error. These values are even more similar to each other, with low and high values of 3.06 and 6.6 ϫ 10 6 M Ϫ1 s Ϫ1 and an average of (5.3 Ϯ 1.6) ϫ 10 6 M Ϫ1 s Ϫ1 and are in good agreement with the maximal rate obtained from Fig. 4B with the 10-mer (2.3 ϫ 10 6 M Ϫ1 s Ϫ1 ). Thus, when K d is taken into account, there is no effect of heparin length on the catalyzed rate of HCII inhibition of thrombin, and the allosteric effect alone accounts for a 2000-fold increase in rate.

DISCUSSION
Many conflicting reports have been published on the minimal heparin length required to activate HCII. Scully et al. (30) reported in 1987 that the minimal length of 26 monosaccharide units was required to accelerate thrombin inhibition by HCII but found some activity for the smallest fraction used corresponding to 20 monosaccharide units. This was echoed a year later by Sie et al. (5) who concluded that a 26-unit heparin chain was minimal but noted that low levels of activation were achieved with heparins as small as 10 units long. In 1989, Bray et al. (31) published a more definitive report on the effect of depolymerized heparin chains ranging from 8 -20 units (in 2-unit increments) and heparin fractions of 7.2, 7.8, 12.4,and 18.8 kDa in average molecular mass, corresponding to heparins of roughly 24, 26, 40, and 60 units in length. These heparins were further fractionated for antithrombin affinity, so that they all contained at least one copy of the AT-specific pentasaccharide sequence. The catalytic efficiencies of these heparin fractions at activating HCII inhibition of thrombin were plotted relative to UFH, revealing a "gradual and progressive rise in catalytic activity" with increasing heparin size, with measurable activity for the 8-mer, the smallest fragment used. This contrasted strikingly to the dependence of heparin size on the rate of thrombin inhibition by antithrombin, where "absolutely no activity of oligosaccharides was observed until an M r of 5400 (corresponding to a heparin chain of 18 monosaccharide units) was reached." They concluded that heparin acceleration of thrombin inhibition by AT was dependent on a template mechanism but that this could not be the case for HCII because an  8-mer heparin could not possibly serve to bridge HCII and thrombin. Our current study is in agreement with that of Bray et al. (31) and further demonstrates that the observed dependence of catalytic activity on heparin length is caused entirely by the size dependence of heparin affinity for HCII. It is now clear that small heparin chains are capable of activating HCII inhibition of thrombin by a fully allosteric mechanism, and when the affinity of HCII for heparin is accounted for, all heparin sizes from 10-to 26-mer produce the same rate of thrombin inhibition (ϳ5 ϫ 10 6 M Ϫ1 s Ϫ1 ). This is consistent with earlier studies concluding that heparin bridging of HCII and thrombin contributes only a fraction of the catalytic effect of large heparin chains (Ն30 monosaccharide units in length) (14,23).
One of the surprising conclusions from this work is that HCII binds heparin by an induced fit mechanism analogous to that of AT. Although the native conformations and heparin-binding sites of HCII and AT are nearly identical (Fig. 1), they have evolved independently (32,33). This, coupled with the lack of an identified HCII-specific sequence in heparin, has led many to conclude that HCII binds heparin without the potential for the specificity conferred by conformational change (5). Our recent structures of native and active HCII suggested that an AT-like global conformational change necessarily takes place in response to heparin binding (15) and thus that induced fit was likely. In fact, the binding of HCII to heparin is very similar to the binding of AT to low affinity heparin (LAH), i.e. heparin devoid of chains containing the pentasaccharide sequence. LAH binds to AT with a dissociation constant (19 M) (34) similar to that of heparin to HCII determined here (26 M). Activation of AT by heparin toward factor Xa inhibition is also primarily allosteric (35), presumably because of the release of the reactive center loop from ␤-sheet A, and is thus analogous to the allosteric activation of HCII toward thrombin inhibition, which is also dependent on reactive center loop release. When the effect of affinity is accounted for, LAH stimulates AT inhibition of factor Xa to a similar degree as the specific pentasaccharide itself (0.29 versus 0.46 M Ϫ1 s Ϫ1 ) but at concentrations 1000-fold higher (34). This parallels the effect of heparin size (therefore affinity) on the rate of thrombin inhibition by HCII reported here. Thus, the activation of HCII by heparin is analogous to the activation of AT by LAH, suggesting the possible existence of an HCII-specific GAG sequence or its potential design.
However, despite several attempts to identify such a sequence by fractionating GAGs on HCII-conjugated Sepharose, no GAG fraction with significantly (ϳ1000-fold) higher affinity has ever been identified (6, 36 -38). From these studies it is clear that a high degree of sulfation and flexibility is required, but no dependence on the position or type of anionic groups on binding affinity (as for the 3-O-sulfate for AT) has been found. This fact does not exclude the possibility of the existence of an HCII-specific GAG sequence; rather, it highlights the difficulty in identifying specific sequences that are not highly represented in the GAG chains. The AT-specific sequence was easy to identify by affinity fractionation because it occurs in onethird of heparin chains. Heparin is also highly homogeneous compared with other GAGs such as heparan sulfate, and, because it is secreted by mast cells, does not suffer from tissuespecific sequence differences. AT is the only heparin-binding protein for which such exquisite specificity has been demonstrated, but the specific pentasaccharide sequence might not have been discovered if affinity fractionation were only conducted on endothelial cell heparan sulfate.
In summary, we have determined the heparin binding properties of HCII and conclude that, despite weak affinity and lack of an identified high affinity sequence, HCII binds heparin in an AT-like fashion and is capable of inhibiting thrombin by an allosteric mechanism. The design of specific HCII agonists is thus conceivable but will ultimately depend on improving the nonionic contribution to the overall binding affinity.