Dabigatran and Argatroban Diametrically Modulate Thrombin Exosite Function

Thrombin is a highly plastic molecule whose activity and specificity are regulated by exosites 1 and 2, positively-charged domains that flank the active site. Exosite binding by substrates and cofactors regulates thrombin activity by localizing thrombin, guiding substrates, and by inducing allosteric changes at the active site. Although inter-exosite and exosite-to-active-site allostery have been demonstrated, the impact of active site ligation on exosite function has not been examined. To address this gap, we used surface plasmon resonance to determine the effects of dabigatran and argatroban, active site-directed inhibitors, on thrombin binding to immobilized γA/γA-fibrin or glycoprotein Ibα peptide via exosite 1 and 2, respectively, and thrombin binding to γA/γ′-fibrin or factor Va, which is mediated by both exosites. Whereas dabigatran attenuated binding, argatroban increased thrombin binding to γA/γA- and γA/γ′-fibrin and to factor Va. The results with immobilized fibrin were confirmed by examining the binding of radiolabeled thrombin to fibrin clots. Thus, dabigatran modestly accelerated the dissociation of thrombin from γA/γA-fibrin clots, whereas argatroban attenuated dissociation. Dabigatran had no effect on thrombin binding to glycoprotein Ibα peptide, whereas argatroban promoted binding. These findings not only highlight functional effects of thrombin allostery, but also suggest that individual active site-directed thrombin inhibitors uniquely modulate exosite function, thereby identifying potential novel mechanisms of action.


Introduction
The substrate specificity of thrombin is dependent on exosites 1 and 2, positively-charged domains that flank the active site [1,2]. These exosites modulate the reactivity of thrombin by providing initial binding sites for substrates, inhibitors or cofactors, sterically hindering other interactions, and by allosterically modifying the active site. Exosite 1 is more versatile because it (a) serves as a docking site for substrates, (b) redirects thrombin activity by binding thrombomodulin, and (c) mediates thrombin inhibition by hirudin and heparin cofactor II. In contrast, exosite 2 has a more limited role because it mainly serves to tether or localize thrombin. For example, exosite 2 contributes to thrombin binding to platelet glycoprotein (Gp) Ibα [3,4] and, Radiolabeled thrombin. To 10 μl of 0.2 M sodium borate, pH 8.0, and 10 μl (1 mCi) of Na 125 I (McMaster University Nuclear Reactor, Hamilton, ON) was added 5 μl of 1.5 mM Bolton-Hunter reagent in DMSO. The reaction was initiated by adding 10 μl of 5 mg/ml chloramine T in 20 mM sodium phosphate, 150 mM NaCl, pH 7.4 (PBS). After 1 min at 23°C, the reaction was stopped by addition of 10 μl of 12 mg/ml sodium metabisulfite in PBS. To the reaction mixture was added 150 μg of thrombin and after incubation for 1 hr at 23°C, the reaction was terminated by addition of 100 μl of 0.2 M glycine in 0.2 M sodium borate, pH 8.0. Labeled thrombin was isolated on a PD-10 column (GE Healthcare, Baie d'Urfe, PQ), and eluted with 20 mM Tris-HCl, 150 mM NaCl, pH 7.4 (TBS) containing 0.01% Tween-80 (TBS-Tw). The specific activity of the labeled thrombin was 3.8 x 10 8 cpm/mg.
Active site-blocked thrombin. Thrombin (3-3.5 mg/ml) was incubated with 2-fold molar excess of FPR chloromethyl ketone at 23°C. After 30 min, absence of residual activity of 2 μM FPR-thrombin was determined at 405 nm with 200 μM Chz-Th in a Spectramax plate reader (Molecular Devices, Sunnyvale CA). The sample was washed with TBS and concentrated using an Amicon Ultra 4 ml-3000 MW centrifugal cartridge (Millipore) [14].

SPR.
Studies were performed using a Biacore T200 (GE Healthcare). γ A /γ A -fibrinogen, γ A /γ 0 -fibrinogen, and factor Va were immobilized to separate flow cells of a CM5 sensor chip using the amine coupling kit from GE Healthcare [32]. Briefly, after injecting the 1-ethyl-3-(3-dimethylaminopropyl)-carbodiimide hydrochloride/N-hydroxysuccinimide mixture into the flow cell at a rate of 10 μl/min for 420 s, 50 μg/ml of γ A /γ A -or γ A /γ 0 -fibrinogen or 28 μg/ml factor Va in 10 mM acetate buffer, pH 4.5, was injected at a rate of 5 μl/min until about 8000 response units (RU) were immobilized. High levels of protein were adsorbed because previous studies indicate that only 30% of bound fibrinogen is in an accessible orientation, and because of the large difference in molecular weight between the bound ligands and the analyte thrombin [28]. Flows cells were then washed with 1 M ethanolamine for 420 s, followed by the running buffer composed of 20 mM HEPES, 150 mM NaCl, pH 7.4 (HBS) containing 0.01% Tween-80 (HBS-Tw) and 5 mM CaCl 2 . To convert the immobilized fibrinogen to fibrin, flow cells were subjected to successive 60-min injections of up to 1 μM thrombin [24,28,32]. Quantitative conversion to fibrin was evidenced by saturable decline in RU following thrombin treatment [33].
To measure binding to immobilized γ A /γ A -fibrin, γ A /γ 0 -fibrin or factor Va, increasing concentrations of up to 5 μM thrombin or FPR-thrombin were injected into the flow cells. To measure the effects of dabigatran, argatroban, and DAPA on thrombin binding to fibrin or factor Va, 250-500 nM thrombin or FPR-thrombin was incubated with 0-5 μM inhibitor and injected into the flow cells. Because of the fast on-and off-rates, all SPR experiments were performed at a flow rate of 25 μl/min [28]. Association times were 120 s and dissociation was monitored by washing the flow cells with HBS-Tw for 600 s. Between runs, flow cells were regenerated with a 45 s wash with HBS-Tw containing 0.5 M CaCl 2 .
The K d values for one-site binding of thrombin to γ A /γ A -fibrin were determined from plots of RU values at equilibrium (Req) versus thrombin or FPR-thrombin concentration using Biacore T200 Evaluation software v 1.0. The rapid approach to equilibrium at the low flow rate nullified any potential mass transport concerns. Nonspecific binding was accounted for by subtracting the control RU values obtained from the unmodified flow cell. To calculate two-site binding of thrombin to γ A /γ 0 -fibrin or factor Va, plots were subjected to kinetic analysis using a two-site binding model. EC 50 values for the effect of dabigatran, argatroban, or DAPA on thrombin binding were determined by plotting Req values against the concentration of active site inhibitor and using non-linear regression to fit these to a rectangular hyperbola.
To examine the effect of the active site inhibitors on exosite 2-mediated binding, 1 mg of GpIbα269-286ppp in HBS was biotinylated at its NH 2 -terminus by incubation with a 10-fold molar excess of Hook-Sulfo-NHS-LC biotin (G-Biosciences, St. Louis, MO) for 1 h at 23°C. After separation from unincorporated reagent by chromatography on Sephadex G15, fractions containing biotinylated peptide were identified by monitoring absorbance at 204 nm. Streptavidin (Sigma) was attached to flow cells of a CM5 chip to about 12000 RU and biotin (b)-GpIbα269-286ppp was then adsorbed to about 150 RU above the streptavidin background. Aliquots of thrombin or RA-thrombin up to 4 μM in HBS-Tw were injected into the flow cell at 25 μl/min for 80 s, followed by HBS-Tw for 200 s. Between runs, flow cells were regenerated with 1 M NaCl for 30 s. To measure the effect of dabigatran, argatroban, or DAPA on this interaction, 250 nM thrombin was injected in the presence of 0-5000 nM dabigatran, argatroban, or DAPA. EC50 values were determined as above. To quantify the effects of inhibitors on binding affinity, 0-2 μM thrombin was injected in the presence of 2.5 μM active site inhibitor at 25 μl/min for 60 s. Binding affinities were calculated as described above.
Effect of dabigatran, argatroban and DAPA on 125 I-thrombin binding to fibrin clots. To a series of 1.5-ml micro-centrifuge tubes containing 0-1 μM dabigatran or argatroban, 0-4 μM DAPA, or 0-0.1 μM hirudin were added 10 nM 125 I-thrombin, 3 μM γ A /γ A -fibrinogen or 1 μM γ A /γ 0 -fibrinogen, 2 mM CaCl 2 and 1 unit/mL batroxobin in TBS-Tw in a total volume of 100 μl [24,28,32]. Batroxobin was used to induce clotting because its activity is unaffected by the inhibitors (not shown). After incubation for 1 h at 23°C, clots were pelleted by centrifugation at 14,000 × g for 4 min and 50-μl aliquots of supernatant were removed and counted for radioactivity using a gamma counter (Wizard 2 2470, Perkin-Elmer) to determine the concentration of unbound 125 I-thrombin. EC 50 values for the effect of the inhibitors were calculated by fitting plots of bound 125 I-thrombin versus the inhibitor concentration by non-linear regression analysis to an equation for a rectangular hyperbola. K i values were then determined using the Cheng-Prusoff equation and the dissociation constant of 125 I-thrombin for fibrinogen as determined in a separate experiment.
Dissociation of 125 I-thrombin from fibrin clots. Dissociation of 125 I-thrombin from preformed fibrin clots was performed as described [28]. Briefly, clots were formed around plastic inoculation loops by clotting a 130 μl solution containing 3 μM γ A /γ A -fibrinogen and 30 nM factor XIII in TBS-Tw containing 2 mM CaCl 2 with 10 nM 125 I-thrombin. After incubation for 1 h, clots were immersed in 50 ml plastic tubes containing 10 ml of 100 nM dabigatran, argatroban or DAPA, or 2 M NaCl in TBS-Tw at 23°C. At defined intervals, aliquots were removed, counted for radioactivity, and returned to the tubes. Plots of residual radioactivity versus time were analyzed using a 2-phase exponential decay equation [28].
Statistical analysis-All experiments were performed at least three times. Results are presented as the mean ± standard deviation (SD). Paired data were compared using Student ttests; whereas group means were compared using one-way ANOVA followed by Tukey's test for multiple comparisons. For all analyses, p values less than 0.05 were considered statistically significant.

SPR analysis of the interaction of thrombin with immobilized fibrin or factor Va
Increasing concentrations of thrombin or FPR-thrombin were injected into flow cells containing immobilized γ A /γ A -fibrin, γ A /γ 0 -fibrin, or factor Va [24,28]. Because sensorgram profiles rapidly reached equilibrium, Req values were determined and plotted versus the analyte concentration to determine K d (Fig 1). Non-specific binding is not a major concern because prothrombin exhibited only weak affinity for immobilized fibrin, and thrombin did not bind to immobilized ovalbumin (not shown). As expected, more thrombin and FPR-thrombin bound to γ A /γ 0 -fibrin than to γ A /γ A -fibrin. Binding to γ A /γ A -fibrin was via a single lower affinity site, whereas binding to γ A /γ 0 -fibrin occurred through higher and lower affinity interactions, demonstrating involvement of both exosites (Table 1) [28,32]. Affinities are similar to those reported previously using SPR and fibrin clots [7,24]. More thrombin and FPR-thrombin also bound to factor Va than to γ A /γ A -fibrin. The higher affinities for factor Va than for γ A /γ Afibrin are consistent with interaction mediated by both exosites [8][9][10]. The affinities of FPRthrombin for fibrin and factor Va were~50% higher than those of active thrombin (p < 0.001).
In contrast to the results with dabigatran, argatroban and DAPA increased thrombin binding to γ A /γ A -fibrin, γ A /γ 0 -fibrin, and factor Va in a concentration-dependent manner with EC 50 values for argatroban of 62.4 ± 4.8 nM, 59.4 ± 5.1 nM, and 23.4 ± 8.8 nM, respectively, and for DAPA of 514.1 ± 24.0 nM, 515.9 ± 31.0 nM, and 565.1 ± 200 nM, respectively (Fig 2). Although argatroban and DAPA are reported to bind thrombin with similar affinities, the EC 50 values for argatroban were more than 10-fold lower than those for DAPA. As observed with dabigatran, neither argatroban nor DAPA had any effect on FPR-thrombin binding to γ A /γ A -fibrin, γ A /γ 0 -fibrin, or factor Va (Fig 2). Therefore, reversible engagement of the active site by small molecules allosterically modulates the exosite-mediated interactions of thrombin with fibrin and factor Va.
bound at equilibrium after background correction (Req) was determined and plotted against the thrombin concentration. Data points represent the mean ± S.D. of 3 experiments, and the lines represent nonlinear regression analysis.
doi:10.1371/journal.pone.0157471.g001 Table 1. Dissociation constants (K d ) for the binding of thrombin to fibrin, factor Va and GpIbα269-286ppp. The binding of thrombin or FPR-thrombin to immobilized γ A /γ A -fibrin, γ A /γ 0 -fibrin, factor Va or GpIbα269-286ppp was quantified using SPR. K d values for thrombin binding to γ A /γ A -fibrin and GpIbα269-286ppp were determined by steady state analysis, whereas bivalent binding of thrombin to γ A /γ 0 -fibrin and factor Va was determined by kinetic analysis for a two-site model using BIAevaluation software. K d values are shown as mean ± S.D. for three separate experiments.  Thrombin bound to the immobilized GpIbα269-286ppp peptide with a K d of 306 ± 11 nM ( Fig  3A; Table 1); an affinity intermediate between that reported for a similar peptide and for glycocalicin [30,34]. In contrast, RA-thrombin, an exosite 2 variant, did not bind (not shown), consistent with the reported specificity of this interaction for exosite 2 [30]. Thrombin binding to the GpIbα269-286ppp peptide was then quantified in the presence of increasing concentrations of inhibitors. Whereas dabigatran had no effect on thrombin binding, argatroban and DAPA increased the amount of thrombin bound by 17% and 25%, respectively ( Fig 3B; Table 2). Binding to GpIbα269-286ppp was then quantified in the presence of 2.5 μM dabigatran, argatroban or DAPA. Whereas thrombin bound with similar affinity in the presence of dabigatran (K d 306 ± 11 nM; not shown) as in its absence, the affinity of thrombin for GpIbα269-286ppp significantly (p < 0.001 by ANOVA) increased by 25-35% in the presence of argatroban or DAPA (K d values of 229 ± 7 nM and 197 ± 5 nM, respectively). These findings demonstrate that occupation of the active site by some, but not all, small molecules increases the affinity of ligands for exosite 2, similar to the observation for exosite 1-dependent interactions.

Effects of dabigatran, argatroban and DAPA on 125 I-thrombin binding to fibrin clots
Although the affinities of thrombin and FPR-thrombin for γ A /γ A -and γ A /γ 0 -fibrin determined by SPR are comparable with those obtained with fibrin clots [28], it was important to confirm that the allosteric response observed using SPR also occurred in a functional assay with threedimensional fibrin clots. 125 I-thrombin bound γ A /γ A -and γ A /γ 0 -fibrin clots with K d values of 3.6 ± 0.3 μM and 1.2 ± 0.2 μM (not shown), respectively; values comparable with those obtained using SPR and with those published previously [32]. Up to 70% of the 125 I-thrombin bound to fibrin, demonstrating that labeling had little effect on the capacity of thrombin to bind to fibrin. Dabigatran reduced 125 I-thrombin binding to γ A /γ A -and γ A /γ 0 -fibrin clots in a concentration-dependent manner with K i values of 3.8 ± 1.5 nM and 26.0 ± 4.0 nM, respectively (Fig 4). The K i value for the thrombin interaction with γ A /γ A -fibrin is comparable with the value of 4.5 nM obtained by chromogenic assay [35]. The K i value is higher with γ A /γ 0 -fibrin than γ A /γ Afibrin, likely reflecting the higher affinity, bivalent interaction mediated by both exosites [7]. As Req values of thrombin bound in the absence (Bound o ) and presence (Bound) of dabigatran, argatroban or DAPA are plotted against the input inhibitor concentrations. Open symbols represent the effects of the inhibitors on FPR-thrombin binding to γ A /γ A -fibrin, γ A /γ 0 -fibrin, or factor Va. Data points represent the mean ± S.D. of 3 experiments, and the lines represent nonlinear regression analysis.
doi:10.1371/journal.pone.0157471.g002 Table 2. Effect of dabigatran, argatroban, or DAPA on thrombin binding to γ A /γ A -fibrin, γ A /γ 0 -fibrin, factor Va, or GpIbα269-286ppp. The effect of 0-5000 nM dabigatran, argatroban, or DAPA on the binding of thrombin to immobilized γ A /γ A -fibrin, γ A /γ 0 -fibrin, factor Va or GpIbα269-286ppp was quantified using SPR. EC 50 values were determined by plotting the Req values against the inhibitor concentration and fitting by nonlinear regression analysis using BIAevaluation software. Change is the calculated difference between thrombin binding in the absence of inhibitors and that determined at saturating inhibitor concentration. Values are shown as mean ± S.D. for three separate experiments.

Dabigatran
Argatroban DAPA   observed using SPR, dabigatran produced a 2-fold greater reduction in thrombin binding to γ A /γ A -fibrin than to γ A /γ 0 -fibrin (p < 0.05); a difference that likely reflects the higher affinity of thrombin for γ A /γ 0 -fibrin. To verify the role of exosite 1 in binding, the effect of hirudin also was examined. As expected, hirudin fully inhibited 125 I-thrombin binding to γ A /γ A -and γ A /γ 0fibrin. In agreement with the SPR observations, argatroban and DAPA enhanced the binding of 125 I-thrombin to γ A /γ A -and γ A /γ 0 -fibrin clots to a similar extent (~15-25%) (Fig 4, Table 3). Collectively, therefore, the data obtained with fibrin clots support the SPR results and suggest that active site engagement induces allosteric changes in the thrombin exosites, with the direction and magnitude determined by the ligand.

Effects of dabigatran, argatroban and DAPA on 125 I-thrombin dissociation from fibrin clots
To determine whether the alteration in the affinity of thrombin for fibrin in the presence of active site inhibitors influences the amount of thrombin associated with fibrin, dissociation from fibrin clots was examined [28]. Clots formed from γ A /γ A -fibrinogen in the presence of 125 I-thrombin were incubated in buffer in the absence or presence of active site inhibitors and the amount of 125 I-thrombin released from the clot over time was quantified. Biphasic analysis was used to evaluate the fast non-adsorbent and slow bound phases of dissociation of thrombin from fibrin. In the control, the bound fraction dissociated with a half-life of 20.5 ± 3.1 h, whereas in the presence of 2 M NaCl where ionic interactions are abrogated, the rate of dissociation was significantly faster (p = 0.02) at 14.2 ± 1.2 h (Fig 5). The half-life values in the presence of 2.5 μM dabigatran, argatroban, and DAPA were 17.7 ± 2.3 h, 25.5 ± 2.9 h, and 23.7 ± 2.9 h, respectively. Although these values were not significantly different from the control, there was a trend for more rapid dissociation in the presence of dabigatran. However, the half-lives in the presence of argatroban and DAPA were significantly longer than that in the presence of dabigatran (p = 0.01 and 0.02, respectively). These data are consistent with the effects of the inhibitors on the affinity of thrombin for fibrin, and confirm the differences in response between dabigatran and argatroban or DAPA.

Discussion
The purpose of this study was to determine whether occupation of the active site of thrombin by small molecules induces functional changes at its exosites. We chose fibrin, factor Va and GpIbα269-286ppp as targets for three reasons. First, their use enabled studies with unmodified, active thrombin. Second, the mode of interaction of thrombin with these ligands has been well the inhibitor concentrations. Data points represent the mean ± S.D. of 2-3 experiments, each performed in duplicate, and the lines represent nonlinear regression analysis.
doi:10.1371/journal.pone.0157471.g004 Table 3. Effect of dabigatran, argatroban, DAPA or hirudin on 125 I-thrombin binding to fibrin clots. Binding of 125 I-thrombin to γ A /γ A -or γ A /γ 0 -fibrin clots in the absence or presence of dabigatran, argatroban, DAPA, or hirudin was determined by measuring the amount of 125 I-thrombin in the supernatants of compacted clots prepared by incubating γ A /γ A -or γ A /γ 0 -fibrinogen with a catalytic amount of batroxobin. EC 50 values were determined by nonlinear regression analysis and converted to K i values by correction for K d of 125 I-thrombin for fibrin. Change denotes the maximal change in the percentage of bound 125 Ithrombin compared with that in the absence of inhibitor. characterized; thus, binding of thrombin to γ A /γ A -fibrin and GpIbα269-286ppp is solely dependent on exosite 1 and exosite 2, respectively, whereas thrombin binding to γ A /γ 0 -fibrin and factor Va requires both exosites. Third, the use of native ligands such as fibrin and factor Va would complement previous results obtained with peptide ligands, and the use of fibrin would enable confirmation in functional assays. Dabigatran and argatroban were chosen because they are highly specific for thrombin and their interactions with the active site of thrombin have been extensively investigated and because both agents are in clinical use. DAPA was chosen to complement the results with argatroban because both incorporate an Arg group that is essential for specificity [36,37]. Dabigatran saturably reduces thrombin binding to γ A /γ A -fibrin and factor Va by 28-48%, but has no effect on the interaction of thrombin with Gp1bα269-286ppp. Surprisingly, argatroban and DAPA increase thrombin binding to γA/γA-fibrin, γ A /γ 0 -fibrin, factor Va and GpIbα269-286ppp by 13-47%. The observations with fibrin were confirmed using threedimensional fibrin clots and were not dependent on the affinity of the inhibitors for thrombin because dabigatran, argatroban and DAPA inhibit thrombin with similar nanomolar K i values. DAPA and argatroban also slow the dissociation of thrombin from fibrin clots in a manner consistent with their enhancement of the affinity of thrombin for fibrin. Thus, these studies demonstrate allosteric effects of active site binding on thrombin, and provide evidence that responses are ligand specific.
Investigation into the allosteric regulation of thrombin has shown that binding of a variety of effectors to the exosites modulates the active site. This phenomenon has been attributed to a complex allosteric network within thrombin that is subject to modulation through surface residues [13,16]. Within the active site, residues in the S 1-4 subsites and the 60-and autolysis-loops contribute to this allosteric network [16,17,38,39]. These regions within the active site make unique contact with individual substrates and are perturbed by exosite ligands, thereby rendering them central to allosteric regulation. It is interesting to compare the interaction of dabigatran, argatroban and DAPA with the active site of thrombin because dabigatran induces effects that differ from those of argatroban and DAPA. Although dabigatran and argatroban engage the S1, S2, and S4 subsites in the active site, they do so in distinct ways [35][36][37]40]. Thus, the differences in exosite response elicited by the various inhibitors likely reflect unique points of contact within the allosteric network. In an analogous fashion, different exosite 1 ligands evoke unique changes in thrombin activity [14,24,39,41,42], suggesting that the exosites are composed of distinct subdomains [2,[43][44][45][46]. Consistent with the results observed here with fibrin and factor Va, argatroban has previously been shown to increase the affinity of hirudin peptide for exosite 1 on thrombin [47]. These observations were not limited to exosite 1 because argatroban and DAPA in the current study also increased the affinity of thrombin for GpIbα peptide; an interaction mediated solely by exosite 2. The structural explanations for how different small molecules can bind to active site of thrombin and evoke distinct responses at the exosites remain to be determined.
The observation that binding of dabigatran, argatroban or DAPA to the active site of thrombin induces changes at the exosites is consistent with the hypothesis that changes transmitted over the allosteric network are bidirectional. Most studies have focused on the effect of exosite ligands on active site function. However, calorimetry and fluorescence studies using peptide ligands demonstrate that structural changes that result from ligand binding at the active site or exosite 1 can run in both directions [21,47,48]. This property suggests that covalent occupation of the active site by FPR should alter exosite function; a concept supported by some studies [38,[49][50][51][52], but not by others [21,[53][54][55]. Results presented here suggest that noncovalent adducts also can modify exosite function, and represent another way by which thrombin can convert between different functional and conformational states [23]. Although extensive investigations of thrombin structure have not provided a unifying model of thrombin allostery, they demonstrate that thrombin is a highly dynamic molecule with numerous conformations and networks of internal communication.
The capacity of dabigatran to attenuate thrombin binding to fibrin and factor Va reveals a unique mechanism by which this inhibitor might function as an anticoagulant. Thus, in addition to inhibiting the catalytic activity of thrombin, dabigatran may displace thrombin from fibrin or platelets. This would reduce the protection of thrombin from inhibition afforded by such binding, and could further attenuate the thrombogenic potential of the thrombus [56,57]. Such displacement would be similar to that produced by exosite-directed ligands, which also displace thrombin from fibrin or platelets [3,58]. Because argatroban promotes thrombin binding to fibrin, it is likely that not all direct thrombin inhibitors benefit from this secondary effect. This raises the possibility that in addition to simple target potency, future efforts at structurebased drug design should consider the secondary effects of inhibitors; effects that may be altered by refining contacts within the active site.
In summary, our results suggest that active site inhibitors modulate thrombin function mediated by the exosites. This reverse direction of allosteric signaling provides greater insight into the dynamic nature of thrombin. Furthermore, different active site-directed agents evoke unique responses, demonstrating that their inhibitory function can be refined. Thus, exploitation of its allosteric network endows thrombin with additional mechanism of regulation without compromising its repertoire of substrates. This demonstrates that the versatility of thrombin is a consequence of the intricate connections between the active site and the exosites.

Author Contributions
Conceived and designed the experiments: CHY ARS BAL JCF JIW. Performed the experiments: CHY ARS BAL. Analyzed the data: CHY ARS BAL JCF JIW. Wrote the paper: CHY JCF JIW.