Role of Thrombin Anion-binding Exosite-I in the Formation of Thrombin-Serpin Complexes*

Site-directed mutagenesis was used to investigate the role of basic residues in the thrombin anion-binding exosite-I during formation of thrombin-antithrombin III (ATIII), thrombin-protease nexin 1 (PN1), and thrombin-heparin cofactor II (HCII) inhibitor complexes, in the absence and presence of glycosaminoglycans. In the absence of glycosaminoglycan, association rate constant (k on) values for the inhibition of the mutant thrombins (R35Q, K36Q, R67Q, R73Q, R75Q, R77 a Q, K81Q, K109Q, K110Q, and K149 e Q) by ATIII and PN1 were similar to wild-type recombinant thrombin (rIIa), whereask on values were decreased 2–3-fold for HCII against the majority of the exosite-I mutants. The exosite-I mutants did not have a significant effect on heparin-accelerated inhibition by ATIII with maximal k on values similar to rIIa. A small effect was seen for PN1/heparin inhibition of the exosite-I mutants R35Q, R67Q, R73Q, R75Q, and R77 a Q, where k on values were decreased 2–4-fold, compared with rIIa. For HCII/heparin, k onvalues for inhibition of the exosite-I mutants (except R67Q, R73Q, and K149 e Q) were 2–3-fold lower than rIIa. Larger decreases in k on values for HCII/heparin were found for R67Q and R73Q thrombins with 441- and 14-fold decreases, respectively, whereas K149 e Q was unchanged. For HCII/dermatan sulfate, R67Q and R73Q had k onvalues reduced 720- and 48-fold, respectively, whereas the remaining mutants were decreased 3–7-fold relative to rIIa. The results suggest that ATIII has no major interaction with exosite-I of thrombin with or without heparin. PN1 bound to heparin uses exosite-I to some extent, possibly by utilizing the positive electrostatic field of exosite-I to enhance orientation and thrombin complex formation. The larger effects of the thrombin exosite-I mutants for HCII inhibition with heparin and dermatan sulfate indicate its need for exosite-I, presumably through contact of the “hirudin-like” domain of HCII with exosite-I of thrombin.

Hemostasis is dependent on the intricate balance between thrombin's procoagulant activity, which promotes clot formation, and its anticoagulant activity in complex with thrombomodulin that activates protein C and restricts excessive clotting (1,2). Serine protease inhibitors (serpins) further regulate the procoagulant activity of thrombin. The most important inhibitors of thrombin are antithrombin III (3), protease nexin 1 (4), and heparin cofactor II (5). These serpins inhibit thrombin by forming tight 1:1 complexes between the serpin P 1 -P 1 Ј reactive center and the active site of thrombin 1 (5)(6)(7). Complex formation for ATIII 2 and PN1 occurs by a two-step mechanism where there is an initial formation of a loose complex followed by a conformational change that yields an essentially irreversible complex (8). The final complex exists either as a tetrahedral intermediate or as an acyl enzyme in which the P 1 residue is covalently bound to Ser 195 of thrombin 3 after hydrolysis of the P 1 -P 1 Ј scissile bond (9).
The inhibitory activities of the three serpins are accelerated by the presence of glycosaminoglycans such as heparin (5)(6)(7)10), whereas the inhibitory activity of HCII is also accelerated by dermatan sulfate (11,12). Site-directed mutagenesis studies of thrombin (13,14), ATIII (15), and PN1 (16) have shown that heparin-accelerated inhibition by ATIII and PN1 occurs by a template mechanism. As heparin binds to ATIII and PN1 with a higher affinity than for thrombin (17), the template mechanism involves the binding of heparin to a conserved site on the serpin (18,19). The serpin-heparin complex then quickly associates with the heparin binding site on thrombin (anion-binding exosite-II) through electrostatic interactions (20). However, the mechanism for glycosaminoglycan-accelerated inhibition of thrombin by HCII seems to be different to that of ATIII and PN1. Inhibition studies of ␣-, ␤ T -(cleaved at Arg 77a ), and ␥ Tthrombin (cleaved at Arg 67 , Arg 77a , and Lys 149e ) by HCII and ATIII in the absence and presence of glycosaminoglycan (21) have shown that anion-binding exosite-I of thrombin is important for rapid inhibition by HCII. HCII contains an N-terminal acidic region (residues 56 -75) similar to the C-terminal tail of hirudin (22). The HCII acidic domain is required for inhibition of thrombin in the presence of glycosaminoglycan by interacting with thrombin's exosite-I, and it is also postulated to bind to the glycosaminoglycan binding site in the absence of glyco-saminoglycans (22)(23)(24). Thus, HCII inhibition of thrombin with glycosaminoglycan is consistent with a "double-bridge" mechanism where (i) glycosaminoglycan binds both to the glycosaminoglycan binding site of HCII and exosite-II of ␣-thrombin, forming a ternary complex, and (ii) the displaced acidic domain of HCII interacts with thrombin's exosite-I, thereby facilitating rapid complex formation (21,24). Sheehan et al. (20) examined the HCII inhibition of thrombin exosite-I and exosite-II mutants in the presence of glycosaminoglycans. These data suggest that dermatan sulfate acceleration of HCII inhibition is primarily due to only an "allosteric" mechanism in which dermatan sulfate binding to the serpin promotes an interaction with exosite-I (essentially the same as in (ii) above). It was also suggested that the ternary complex (described in (i) above) played a minor role during the heparin-accelerated HCII inhibition of thrombin (20).
This paper investigates the interaction of three heparinbinding serpins with the thrombin anion-binding exosite-I. Past work has not fully examined the contribution of all the basic residues of thrombin's exosite-I with ATIII and HCII (20,25) or PN1 (26) in the absence and presence of glycosaminoglycans. An extensive study will define those residues involved in interactions between thrombin's exosite-I and various domains on ATIII, HCII, and PN1. These studies further emphasize mechanistic differences in how different glycosaminoglycan-dependentserpinsrecognizethrombin,especiallythroughexosite-Idependent mechanisms. This work clearly establishes a very prominent, a less prominent, and a nonessential role for thrombin's exosite-I in HCII-, PN1-, and ATIII-thrombin inhibition reactions, respectively.
Thrombin Inhibition by Serpins in the Absence of Glycosaminoglycan-The value for the association rate constant (k on ) for thrombin(s) and the serpins HCII, ATIII, and PN1 in the absence of glycosaminoglycan were determined using pseudo-first order kinetics. For HCII, the value of k on was determined by preincubating a 400-l volume containing 1 M HCII and 0.1 g/ml polybrene in 1 ϫ assay buffer at 37°C. Adding thrombin to a final concentration of 50 nM started the reaction. At various time points, a 20-l aliquot was removed and added to 230 l of 200 M S2266 in assay and incubated at 37°C for 4 min in a Thermomax plate reader (Molecular Devices) to measure the initial velocity of residual thrombin. A control reaction was also performed with the same components except that HCII was omitted. The amount of uninhibited thrombin present is proportional to its initial velocity and obeys the equation A t ϭ A 0 e -k't , where t is the time the sample was taken, A t and A 0 are the velocities at times t and zero, respectively, and kЈ is the apparent first order rate constant. The value for the second order rate constant k on was determined by dividing kЈ by the inhibitor concentration.
For the serpins ATIII and PN1, the value for k on was determined using progress curve kinetics. Cuvettes containing 100 -200 M S2238 and serpin (15 M ATIII or 150 nM PN1) were preheated at 37°C for 10 min and then transferred to a Hewlett-Packard spectrophotometer. Reactions were started by the addition of thrombin to 100 pM, and the release of p-nitroaniline from the hydrolysis of S2238 was followed by measuring the absorbance at 400 -410 nm. Progress curves for the formation of product are described by the following equation (31)(32)(33): where P is the amount of p-nitroanilide at time t, kЈ is the apparent first order rate constant, and v 0 and v s are the initial and steady-state velocities, respectively. Progress curves fitted to the equation gave estimates for kЈ, v 0 , and v s . The association rate constant k on was calculated using the following equation (33): where [S] is the concentration of S2238 and K m is the Michaelis constant. At least two progress curves were performed for each enzymeinhibitor combination, with the calculated values for k on reported as a weighted mean.
Glycosaminoglycan-accelerated Inhibition of Thrombin by Serpins-Progress curve kinetics were used to estimate the value of k on for the interaction of native and variant thrombins with the serpins HCII, ATIII, and PN1 in the presence of the glycosaminoglycans heparin (heterogeneous) and dermatan sulfate under pseudo-first order conditions. The dependence of the k on value for the inhibition of recombinant native and plasma thrombin on the concentration of glycosaminoglycans was determined by incubating ATIII (200 nM) or PN1 (2 nM) with heparin concentrations ranging from 1 to 1000 nM, in the presence of S-2238 (50 -400 M) for 10 min at 37°C. Reactions were followed by the release of p-nitroaniline on a Hewlett Packard spectrophotometer after the addition of thrombin to a final concentration of 200 pM. The effects of both heparin and dermatan sulfate on the k on value were examined for HCII. HCII (25 nM) was incubated with varying concentrations of heparin (25 nM-25 M) and dermatan sulfate (2-50 M), and the reactions were started as described above. Values for k on were calculated using Equations 1 and 2. The optimum concentrations of glycosaminoglycan and S2238 were used to determine k on for all the recombinant thrombin mutants.
Molecular Modeling-The heparin cofactor II-thrombin complex was constructed essentially as described previously (34) using the Homology module of the Insight II molecular modeling package (Version 2.3.0, BIOSYM Technologies, Inc., San Diego). The x-ray crystal structure of ␣-thrombin complexed with residues 55-65 of hirudin, and D-Phe-Pro-Arg-chloromethyl ketone was obtained from the Brookhaven Protein Data Bank (Ref. 35; Brookhaven entry 1DWE). The coordinates of the backbone atoms in the two thrombin molecules were aligned in Insight II with a root mean square deviation of 0.595. Connolly solvent-accessible surfaces for heparin cofactor II, thrombin, D-Phe-Pro-Arg-chloromethyl ketone, and hirudin were created in Insight II using a probe radius of 1.4 Å.

Inhibition of Thrombin Exosite-I Mutants by ATIII and PN1 in
the Absence of Heparin-Values for the association rate constant (k on ) for the inhibition of native plasma thrombin (pIIa), recombinant native thrombin (rIIa), and mutant thrombins by ATIII and PN1 were determined under pseudo-first order conditions using progress curve kinetics. Plasma thrombin and rIIa were inhibited by ATIII (Table I) and PN1 (Table II) with k on values similar to previously reported values (16,36).
Inhibition of Thrombin Exosite-I Mutants by ATIII and PN1 in the Presence of Heparin-The optimum heparin concentration required to achieve the maximum k on value for the inhibition of pIIa and rIIa by the serpins ATIII and PN1 was determined by comparing k on values over a range of heparin 4 Standard nomenclature is used to describe mutant thrombins. The first letter shows the amino acid in the wild type to be replaced, and the second letter shows the amino acid used for the substitution (e.g. the mutant thrombin R67Q represents the substitution of Arg 67 with Gln. 5 T. Myles and S. R. Stone, manuscript in preparation. concentrations (1-1000 nM). Both plasma and recombinant thrombins showed typical bell-shaped inhibition rate curves with dependence of the k on values for PN1 and ATIII on heparin concentration (Fig. 1). The inhibition of pIIa and rIIa by ATIII gave maximal k on values (0.75 ϫ 10 8 M Ϫ1 s Ϫ1 and 1.19 ϫ 10 8 M Ϫ1 s Ϫ1 , respectively) at 50 nM heparin ( Fig. 1, Table I).
Maximal k on rates for PN1 were seen with pIIa (2.56 ϫ 10 9 M Ϫ1 s Ϫ1 ) and rIIa (2.52 ϫ 10 9 M Ϫ1 s Ϫ1 ) at 20 nM heparin ( Fig. 1, Table II). PN1 and ATIII were then used at their optimal concentration of heparin for inhibition studies on the thrombin exosite-I mutants.
There were no major differences in the k on values for the exosite-I mutants when compared with rIIa for ATIII, with k on varying no more than ϳ1.4-fold (Table I). The increase in k on for each thrombin exosite-I mutant in the presence of heparin and ATIII was similar to rIIa (9,675-fold), varying from 8,626fold (K149 e Q) to 15,172-fold (R67Q). Likewise, many of the thrombin exosite-I mutants did not have any significant effects on inhibition by PN1 in the presence of 20 nM heparin (Table  II). However, the exosite-I mutants R35Q and R77 a Q showed decreases in k on (3.7-and 3.2-fold, respectively) when compared with rIIa, whereas there were 2-fold decreases in k on for the inhibition of R67Q, R73Q, and R75Q by PN1 with heparin. Rate enhancement by heparin for the inhibition of R35Q and R77 a Q by PN1 was only 553-and 743-fold compared with rIIa (2,049-fold), whereas the rest of the mutants ranged from 1,018-fold (R67Q) to 2,194-fold (K81Q).  (Table III), which were approximately 2-fold higher to previously published values (21). This variance is most likely   (Table III).

Inhibition of Thrombin Exosite-I Mutants by HCII in the Presence of Glycosaminoglycans-Maximal k on values for the inhibition of rIIa and pIIa by HCII were determined by com-
paring k on values with various concentrations of these glycosaminoglycans (Fig. 1). Heparin titration of HCII gave optimum k on values of 7.6 ϫ 10 7 M Ϫ1 s Ϫ1 and 4.5 ϫ 10 7 M Ϫ1 s Ϫ1 at 500 nM heparin for rIIa and pIIa, respectively (Fig. 1, Table III). Dermatan sulfate at 30 M gave optimal k on values for rIIa and pIIa of 2.2 ϫ 10 7 M Ϫ1 s Ϫ1 and 1.8 ϫ 10 7 M Ϫ1 s Ϫ1 , respectively (Fig. 1, Table III). The k on values were slightly different than previously reported (20,21), likely due both to differences in glycosaminoglycans and in buffer/temperature conditions. HCII was then used at its optimal concentration of heparin and dermatan sulfate for inhibition studies on the thrombin exosite-I mutants.
The degree to which heparin-accelerated HCII inhibition of the exosite-I mutants R35Q, K36Q, R75Q, R77 a Q, K81Q, K109Q, K110Q, and K149 e Q varied from 45,714-fold (K81Q) to 67,692fold (R75Q) compared with rIIa (67,568-fold) (Table III). Interestingly, heparin accelerated the inhibition of the two mutants, R67Q and R73Q, by HCII by only 347-and 15,882-fold, respectively (Table III). The k on values for all other exosite-I mutants were approximately 2-3-fold lower than that for rIIa. These decreases in k on values in the presence of heparin paralleled the differences observed in the absence of heparin. In contrast, HCII inhibition of R73Q displayed a larger decrease (14-fold) in the k on value in the presence of heparin, whereas a very large decrease (441-fold) was observed with R67Q.
Dermatan sulfate-catalyzed HCII inhibition of the exosite-I mutant K149 e Q was similar to recombinant wild-type thrombin (Table III). The effect of the other exosite-I mutations was generally greater on the degree of dermatan sulfate acceleration compared with heparin (Table III). In particular, R67Q and R73Q were only accelerated 61-and 1323-fold, respectively, in the presence of dermatan sulfate. Additionally, the reduction of the k on values of HCII inhibition of the other exosite-I mutations was about 2-fold greater in the presence of dermatan sulfate than in the presence of heparin (Table III). DISCUSSION The k on values for the inhibition of recombinant native thrombin by the serpins were not very different from those observed with plasma thrombin. The recombinant thrombins were expressed in a baculovirus system and were devoid of Asn 60g -linked carbohydrate (37,38). Thus, the absence of car-bohydrate on recombinant thrombin does not influence the inhibitory activity of the serpins in the absence of glycosaminoglycan.
The k on values for inhibition of the exosite-I mutants by ATIII and PN1 in the absence of heparin did not differ substantially from rIIa, suggesting that there are no significant interactions between these serpins and exosite-I. However, the inhibitory activity of HCII was affected by most of the exosite-I mutants, where k on values were decreased ϳ2-3-fold. This is comparable with decreases in inhibition of ␥ T -thrombin and thrombin Quick I (Arg 67 3 Cys) by HCII (21,39,40). Although the slightly reduced k on values for the exosite-I mutants with HCII without glycosaminoglycan suggest that the acidic Nterminal domain is unlikely to make direct contact with exosite-I, these decreases indicate the possibility that a significant electrostatic field generated by the bound acidic domain of HCII can interact with the positive electrostatic potential of exosite-I. Therefore, exosite-I mutations that decrease this electrostatic field strength could slightly reduce the association rate constant between HCII and thrombin, which is consistent with our data.
The inhibitory activities of all three serpins are greatly accelerated by the presence of glycosaminoglycans. Interestingly, k on values for the interaction of recombinant native thrombin with ATIII and HCII in the presence of heparin were consistently higher (1.8-fold) than those with plasma thrombin. This suggests that the Asn 60g carbohydrate may slightly retard the association of HCII and ATIII with thrombin in the presence of heparin. This could be due to either charge repulsion or steric hindrance.
ATIII-Thrombin Exosite-I Interactions in the Presence of Heparin-The inhibition of thrombin by ATIII occurs by a two-step mechanism that is accelerated by heparin (7,8). This acceleration proceeds primarily by the formation of ATIII-heparin complexes, which then bind to exosite-II, using complementary electrostatic fields in a manner similar to the interaction between thrombin and the C-terminal tail of hirudin (13,14). Extensive site-directed mutagenesis of the thrombin exosite-I (R35Q, K36Q, R67Q, R73Q, R75Q, R77 a Q, K81Q, K109Q,K110Q, and K149 e Q; Fig. 2, top panel) has shown that inhibition by ATIII in the absence and presence of heparin does not require any of the basic residues that form exosite-I. Our results complement past studies that suggested the interaction between ATIII/heparin and thrombin uses exosite-II for electrostatic interactions and that exosite-I has no valuable contribution to complex formation (Fig. 2, middle panel), either in "electrostatic steering" or in direct interactions by salt-bridging (13,14).

PNI-Thrombin Exosite-I Interactions in the Presence of
Heparin-PN1 also inhibits thrombin in a two-step mechanism (8), and inhibition is heparin accelerated by a template mechanism similar to ATIII (26). Thrombin inhibition by PN1 in the presence of heparin shows a small effect for the exosite-I mutants R35Q, R67Q, R73Q, R75Q, and R77 a Q (Fig. 2), where there were decreases in the k on values of 3.7-, 2.2-, 2.0-, 1.8-, and 3.2-fold, respectively. The association rate constant for throm-bin-PN1 interaction in the presence of heparin was 2.5 ϫ 10 9 M Ϫ1 s Ϫ1 , which is more rapid than that usually observed for neutral molecules. The magnitude of the k on value suggested that electrostatic forces accelerated the interaction. The heparin-PN1 complex could use the negative electrostatic field generated by the heparin molecule to interact with the positive electrostatic field of exosite-II to accelerate the interaction and to promote a correct orientation for productive complex formation (41). Complex formation could be further enhanced if heparin-bound PN1 could also utilize the electrostatic field generated by thrombin's exosite-I through electrostatic steering. Our results indicate that several of the exosite-I residues may participate to some extent to optimally align the PN1-heparin complex during thrombin inhibition.
HCII-Thrombin Exosite-I Interactions in the Presence of Glycosaminoglycan-Thrombin inhibition by HCII is unique compared with the serpins ATIII and PN1 for three reasons: (a) the P 1 residue is a leucine as opposed to the P 1 arginine of ATIII and PN1; (b) the glycosaminoglycans heparin and dermatan sulfate increase the rate of inhibition; and (c) the unique Nterminal acidic domain (21,22,24). Like other thrombin macromolecular substrates, the N-terminal domain (homologous to the C terminus of hirudin) is rich in acidic and polar residues that are postulated to bind exosite-I (Fig. 3). There are two current models that explain the mechanism for glycosaminoglycan-accelerated inhibition of thrombin by HCII: the doublebridge mechanism (21,24) and the allosteric only model (20). In both models, it is assumed that the N-terminal acidic domain is displaced from the glycosaminoglycan binding site of HCII, thereby allowing the N-terminal domain to bind exosite-I (20,21,24,42).
The role of thrombin's exosite-I during inhibition by HCII has been partially characterized by studies using dysthrombin Quick I (39, 40) and exosite-I mutants R73E and R75E (20,25). In the current study, the effects of the majority of exosite-I mutants (R35Q, K36Q, R75Q, R77 a Q, K81Q, K109Q,K110Q, and K149 e Q) were minimal (Fig. 2, top and middle panels), which suggests that most of the basic residues in exosite-I make no major contacts with the acidic domain of HCII with or without glycosaminoglycans. In contrast, Arg 67 and Arg 73 play a major role in complex formation with HCII in the presence of glycosaminoglycans (Fig. 2, bottom panel). It is interesting to note that the hirudin peptide (similar to the acidic domain of HCII) strongly interacts with both Arg 67 and Arg 73 of thrombin (Fig. 2, bottom panel). A large decrease in k on was found with R67Q thrombin in the presence of both glycosaminoglycans and HCII (441-fold for heparin, 721-fold for dermatan sulfate), which confirms that the structural integrity of the thrombin 70 -80 loop (43) is important for interactions between the HCII acidic domain and exosite-I. More moderate decreases of k on were seen for the mutant R73Q when inhibited in the presence of heparin (14-fold) and dermatan sulfate (48-fold), also suggesting that Arg 73 may interact with the N-terminal domain of HCII (Fig. 2, bottom panel). The reasons for an overall greater effect by the thrombin exosite-I mutations in the presence of dermatan sulfate could indicate how different glycosaminoglycans might influence the binding mode of the N-terminal domain to exosite-I. Finally, dermatan sulfate is more dependent on the allosteric mechanism for HCII-thrombin inhibition than is heparin and might be more affected by alterations in exosite-I of thrombin (see below).
The Mechanism for HCII-Thrombin Inhibition in the Presence of Glycosaminoglycans-The mechanism for glycosaminoglycanaccelerated inhibition of thrombin by HCII is dependent on the displacement of the N-terminal domain from the HCII glycosaminoglycan binding site so that it is free to interact with exosite-I of thrombin. The difference between the proposed double-bridge model (21,24) and the allosteric model (13) is the role of glycosaminoglycan acting as a secondary bridge between HCII and thrombin's exosite-II. When our thrombin exosite-I site-directed mutagenesis data are compared with those obtained from Sheehan et al. (20,25), some interesting observations can be made. First, the effect of the K149 e Q mutation in this study is similar to that for the K149 e A mutant (25), which suggests that this residue plays no significant part in the interaction of HCII with thrombin. However, the effect of the two exosite-I mutants R73E and R75E have more pronounced effects when compared with residues substituted with glutamine replacements used in this study (20,25). In particular, the association rate constant for R73E in the presence of heparin was decreased 376-fold compared with 14-fold for R73Q. Likewise for the R73E substitution in the presence of dermatan sulfate, the decrease is 620-fold compared with a 48-fold decrease for R73Q. Thus, charge reversal has a major effect on the association rate constant. It seems likely that the charge reversal has affected a number of contacts in the exosite-I important for the formation of a stable complex.
On the basis of a large effect of exosite-I mutations and the small effects by exosite-II mutations, Sheehan et al. (20) concluded that the role of exosite-II for glycosaminoglycan-stimulated inhibition of thrombin by HCII was unimportant. They proposed that the major determinant for inhibition by HCII was allosteric, i.e. glycosaminoglycans bind to HCII and displace the N terminus that can then interact with exosite-I. However, many of the decreases in k on for the exosite-II mutants in the presence of heparin (14,20) are of similar magnitude to the decreases in k on seen for the exosite-I mutants in this study. Taking into account the charge reversal of the exosite-II mutants, it seems that both exosites play an important part in heparin-mediated inhibition of thrombin by HCII. In contrast, exosite-II mutants had no effect on dermatan sul-fate-accelerated inhibition of thrombin by HCII (20). This suggests that the two glycosaminoglycans employ different mechanisms for the inhibition of thrombin by HCII. Heparin employs the double-bridge model, whereby it binds to HCII displacing the acidic N-terminal region from the glycosaminoglycan binding site, allowing it to interact with the thrombin exosite-I through complementary electrostatic fields and also by acting as a template, bridging the HCII to exosite-II. Dermatan sulfate does not use exosite-II at all; therefore, the mode of thrombin inhibition is strictly allosteric.