Regulation of the Protein C Anticoagulant and Antiinflammatory Pathways

Abstract

Protein C is a vitamin K-dependent anticoagulant serine protease zymogen in plasma which upon activation by the thrombin-thrombomodulin complex down-regulates the coagulation cascade by degrading cofactors Va and VIIIa by limited proteolysis. In addition to its anticoagulant function, activated protein C (APC) also binds to endothelial protein C receptor (EPCR) in lipid-rafts/caveolar compartments to activate protease- activated receptor 1 (PAR-1) thereby eliciting antiinflammatory and cytoprotective signaling responses in endothelial cells. These properties have led to FDA approval of recombinant APC as a therapeutic drug for severe sepsis. The mechanism by which APC selects its substrates in the anticoagulant and antiinflammatory pathways is not well understood. Recent structural and mutagenesis data have indicated that basic residues of three exposed surface loops known as 39-loop (Lys-37, Lys-38, and Lys-39), 60-loop (Lys-62, Lys-63, and Arg-67), and 70-80-loop (Arg-74, Arg-75, and Lys-78) (chymotrypsin numbering) constitute an anion binding exosite in APC that interacts with the procoagulant cofactors Va and VIIIa in the anticoagulant pathway. Furthermore, two negatively charged residues on the opposite side of the active-site of APC on a helical structure have been demonstrated to determine the specificity of the PAR-1 recognition in the cytoprotective pathway. This article will review the mechanism by which APC exerts its proteolytic function in two physiologically inter-related pathways and how the structure-function insights into determinants of the specificity of APC interaction with its substrates in two pathways can be utilized to tinker with the structure of the molecule to obtain APC derivatives with potentially improved therapeutic profiles.

1. Introduction

Protein C is a single chain vitamin K-dependent plasma serine protease zymogen that upon activation by the thrombin-thrombomodulin (TM) complex down-regulates the clotting cascade by a feedback loop inhibition mechanism [1–3]. Activated protein C (APC) circulates in plasma as a light and heavy chain molecule held together by a single disulfide bond [2]. The N-terminal light chain of APC contains the non-catalytic γ-carboxyglutamic (Gla) domain followed by two epidermal growth factor (EGF)-like domains [2]. The catalytic domain of APC, with a trypsin-like primary specificity pocket, is located on the C-terminal heavy chain of the molecule [2,4]. The Gla domain, with nine vitamin K-dependent γ-carboxylated Glu residues, mediates the Ca²⁺-dependent interaction of APC with protein S on negatively charged membrane surfaces [2,5]. Protein S is a vitamin K-dependent plasma cofactor which promotes the anticoagulant function of APC in the proteolytic degradation of the procoagulant cofactors factors Va (FVa) and VIIIa (FVIIIa) [3,5]. The APC cleavage of these procoagulant cofactors shuts down thrombin generation through both intrinsic and extrinsic pathways [1–3]. Insight into the importance of the APC anticoagulant pathway in the regulation of blood coagulation can be gleaned from the observation that a heterozygous protein C deficiency is associated with a high risk of venous thrombosis and its homozygous deficiency causes purpura fulminans, which is fatal unless treated by protein C replacement therapy [6]. A complete protein C deficiency in mice results in lethal perinatal consumptive coagulopathy, as demonstrated by the targeted gene disruption [7].

In addition to its anticoagulant function, APC also exhibits potent cytoprotective, antiinflammatory and profibrinolytic properties [8–11]. The protective cellular activities of APC require the Gla domain-dependent interaction of the protease with endothelial protein C receptor (EPCR) on the surface of vascular endothelial cells [10,12,13]. The importance of the EPCR-dependent APC regulation of the inflammatory pathways has been demonstrated in animal models of septic shock where blocking either the thrombin-TM activation of protein C or blocking the interaction of APC with EPCR by specific monoclonal antibodies converts a sub-lethal dose of E. coli to a lethal phenotype with the characteristic multiple organ failure observed in severe sepsis [8,14]. The protective anticoagulant and antiinflammatory activities of APC have led to FDA approval of recombinant APC as a therapeutic drug for treating severe sepsis [15]. The mechanism by which APC functions in the antiinflammatory pathway is not fully understood. It has been hypothesized that the interaction of APC with EPCR renders the protease capable of cleaving the exodomain of protease activated receptor 1 (PAR-1), thereby eliciting cytoprotective and antiinflammatory signaling responses in vascular endothelial cells [9,10,16]. Nevertheless, since thrombin is the only known physiological activator of protein C which can activate PAR-1 with 3–4 orders of magnitude higher catalytic efficiency to elicit proinflammatory responses [17,18], it is not yet clear how APC initiates protective responses in endothelial cells through the cleavage of the same receptor. This article will provide an overview of the mechanism through which thrombin activates protein C and how APC recognizes its different substrates and cofactors in the anticoagulant and protective cellular signaling pathways. Insights into the molecular determinants of the APC specificity in these two pathways can lead to rational design of second generation APC therapeutics with improved safety profiles.

2. Protein C Activation

Protein C circulates in plasma as a single chain zymogen [2]. Thrombin removes a 12-residue peptide from the activation peptide of protein C to convert the zymogen to a two-chain active protease held together by a disulphide bond [1,2]. The physiological activation of protein C by thrombin occurs on vascular endothelial cell surfaces and requires the cofactor function of the integral membrane glycoprotein, TM, and the divalent cation, Ca²⁺ [1,19]. TM accelerates the rate of protein C activation by thrombin in the presence of Ca²⁺ ~1000-fold [1,19,20]. The activation of protein C by the thrombin-TM complex is further improved ~10-fold if the zymogen is bound via its Gla domain to EPCR on the surface of vascular endothelial cells [21]. The mechanisms by which the three cofactors (TM, EPCR, and Ca²⁺) improve the rate of protein C activation by thrombin by approximately 4 orders of magnitude have been extensively studied. The extracellular domain of TM contains six epidermal growth factor-like domains [1]. It has been shown that the isolated 4–6 domains of TM (TM4-6) can bind to a basic exosite on thrombin (exosite-1), thereby enabling the protease to recognize and rapidly activate protein C [1,20,22]. In addition to promotion of the activation of protein C, TM4-6 also inhibits the activity of thrombin toward the procoagulant substrates, fibrinogen, PAR-1, and cofactors V and VIII [1,23,24]. The interaction of these procoagulant substrates with exosite-1 is a prerequisite for their recognition and subsequent activation by thrombin [1,23,24]. The high-affinity interaction of TM4-6 with exosite-1 competitively inhibits the binding of procoagulant substrates to this site, thus the cofactor essentially converts thrombin from a procoagulant to a potent anticoagulant protease upon binding [1].

The exact molecular mechanism by which TM4-6 improves the catalytic efficiency of thrombin toward protein C is not fully understood and has been the subject of investigation by several laboratories for many years [1,25–28]. An attractive hypothesis that has emerged more than two decades ago is that TM may function by inducing allosteric changes in the active-site pocket of thrombin to alleviate its potential inhibitory interactions with protein C in the presence of Ca²⁺ [1,29]. Interestingly, unlike its stimulatory role in the presence of TM, physiological levels of Ca²⁺ potently inhibits the activation of protein C by thrombin when TM is absent [1,25]. Since thrombin has no known Ca²⁺-binding site, it has been hypothesized that the TM-altered conformer of thrombin optimally recognizes the Ca²⁺-altered conformer of protein C [1,29,30].

We have localized the Ca²⁺-binding site critical for protein C activation by the thrombin-TM complex to the 70-80-loop (chymotrypsinogen numbering system [31]) of the zymogen [32], the same loop that also binds Ca²⁺ in trypsin [33]. The presence of two conserved acidic residues (Glu-70 and Glu-80) is the signature for Ca²⁺ binding to this loop, and the carboxylic oxygen atoms of both residues contribute to the ligation of Ca²⁺ in this loop [33]. In thrombin, residue 70 is a Lys, thus the loop is stabilized by a salt-bridge between Lys-70 and Glu-80, providing a structural basis for thrombin functioning independent of Ca²⁺ [31]. Although the x-ray crystal structure of protein C has not been resolved, the same residues are also thought to be involved in ligating Ca²⁺ in the 70-80-loop of the zymogen. In support of this hypothesis, we have shown that the substitution of Glu-80 of protein C with a Lys results in a functionally active mutant which is activated normally by the thrombin-TM complex independent of Ca²⁺, presumably by the Glu-70 and Lys-80 establishing a salt-bridge in the mutant zymogen [32]. The mutant APC also exhibits normal amidolytic activity independent of Ca²⁺ [32].

The stabilization of the 70-80-loop of protein C by the Ca²⁺ ion is essential for the TM-dependent recognition and activation of the zymogen by thrombin [28,34]. Structural and mutagenesis data have indicated that basic residues of this loop (Arg-74, Arg-75, and Lys-78) together with basic residues of two nearby surface loops known as 39-loop (Lys-37-39) and 60-loop (Lys-62, Lys-63, and Arg-67) constitute an anion binding exosite which interacts with the acidic residues of the fourth EGF-like domain of TM in the protein C activation complex in the presence of Ca²⁺ [22,34]. The three dimensional positions of these residues in the x-ray crystal structure of the catalytic domain of APC are presented in Fig. 1. There are several lines of evidence supporting the hypothesis that the conformation of the 70-80-loop is allosterically linked to the thrombin recognition site on the activation peptide of protein C [32,35,36]. Thus it has been hypothesized that the binding of Ca²⁺ to 70-80-loop induces a conformational change in the activation peptide of protein C, thereby improving its complimentarity with the TM-altered conformation of the active-site pocket of thrombin [1,29,30]. The inhibitory effect of Ca²⁺ in the absence of TM further supports the hypothesis that the metal ion-induced conformer of the zymogen is not optimal for docking into the active-site pocket of free thrombin [1,25,29].

Fig. 1.

Crystal structure of the catalytic domain of APC. The side chains of basic residues of three loops (39, 60, and 70–80) and acidic residues of 162-helix (Glu-167 and Glu-170) are shown. The coordinates (Protein Data Bank code 1AUT) were used to prepare the figure [4].

Recently, based on structural data, we used a mutagenesis approach to understand the mechanism by which Ca²⁺ regulates the activation of protein C by thrombin in the absence and presence of TM. Initially, we conducted an Ala-scanning of the basic residues of all three loops of protein C (39, 60 and 70–80) constituting the anion-binding exosite of the zymogen [28,34]. The kinetic characterization of these mutants in activation studies by thrombin suggested that the basic residues of the protein C exosite are required for the TM-dependent recognition of the zymogen by thrombin, however, these residues played a negative role in the recognition of protein C by thrombin when TM was absent [28,34]. Interestingly, the activation of the Arg-67 to Ala mutant of protein C by thrombin was improved 20-fold independent of both TM and Ca²⁺ [28]. The examination of the three dimensional position of the 70-80-loop in the x-ray crystal structure of APC suggested that this loop is located between two anti-parallel β strands comprised of residues 64–69 and 81–91 (Fig. 2). It was noticed that the NH1 guanidyl group of Arg-67 on the former strand is located within 2.67 Å from the OD2 carboxylic oxygen of Asp-82 on the latter strand (Fig. 2). We hypothesized that the formation of a salt-bridge between these two residues may modulate the unique Ca²⁺-dependence of protein C activation by thrombin in the absence and presence of TM [28]. To test this hypothesis, we substituted both Arg-67 and Asp-82 of the substrate with Cys residues, thereby engineering a new disulfide bond between the two residues [28]. Interestingly, we discovered that the activation of the Cys-67/Cys-82 mutant of protein C by thrombin was improved 60–80-fold independent of both TM and Ca²⁺ [28]. In an earlier study, we had demonstrated that the substitution of Arg-35 of thrombin with a Glu (R35E) leads to a 20–25-fold improvement in the rate of wild-type protein C activation by the mutant thrombin also independent of TM and Ca²⁺ [30]. We discovered that the activation of the Cys-67/Cys-82 mutant of protein C by R35E thrombin is enhanced by three orders of magnitude independent of both TM and Ca²⁺ [28]. Thus, the accelerating effect of the two mutations, one in the zymogen and the other in the enzyme, was additive [28]. This effect is essentially identical to the accelerating effect of TM and Ca²⁺ in the wild-type system [1,28]. Thus, it appears that the primary function of TM in protein C activation by thrombin involves the alleviation of the inhibitory interactions of Arg-67 of protein C and Arg-35 of thrombin in the presence of Ca²⁺. Structural data suggest that the guanidyl group of Arg-35 is pointing toward the active-site pocket of thrombin [31], possibly impeding access of the active-site pocket by the Ca²⁺-stabilized conformer of protein C in the absence of TM.

Fig. 2.

Crystal structure of the catalytic domain of APC. (A) The three dimensional positions of Arg-67 and Asp-82 on two anti-parallel β structures of APC are shown. (B) Intra-atomic distances between NH1 guanidyl of Arg-67 and carboxylic oxygens of Asp-82.

Based on these results, we proposed a model (Fig. 3) for protein C activation by thrombin and TM which supports and extends the previous model presented by Esmon et al [1,28]. This model predicts that in the absence of Ca²⁺, the guanidyl group of Arg-67 of protein C is in a salt-bridge/hydrogen bond contact with Asp-82, a residue immediately outside of the 70-80-loop of protein C [28]. The binding of Ca²⁺ to the 70-80-loop of protein C is associated with a conformational change, leading to disruption of the electrostatic interaction between Arg-67 and Asp-82 of the protein. The side chain of Arg-67 in protein C, which remains exposed in the presence of Ca²⁺, makes an inhibitory for interaction with thrombin, nevertheless, this residue is also required for the ability of the zymogen to interact with an acidic site on the fourth EGF-like domain of TM in the activation complex [22,28,34]. In the Ca²⁺-stabilized conformer of protein C, the side chain of Arg-67 is free and not engaged in a salt-bridge interaction with Asp-82, thereby impeding the docking of protein C into the active-site groove of thrombin possibly due to its repulsive interaction with Arg-35 of thrombin [28].

Fig. 3.

Hypothetical model of protein C activation by the thrombin-TM complex. (A) Arg-67 and Asp-82 are involved in electrostatic interactions in the zymogen protein C in the absence of Ca²⁺. The activation peptide (AP) of protein C is not complimentary to fit optimally into the active-site pocket of thrombin in the absence of C^a2+. (B) The binding of Ca²⁺ to the 70-80-loop of protein C disrupts the salt-bridge/hydrogen bond between Arg-67 and Asp-82, thereby relocating Arg-67 to an inhibitory position for interaction with thrombin in the absence of TM. The metal ion also induces conformational change in the activation peptide of protein C, which is still not complimentary for the active-site pocket of free thrombin. (C) The acidic EGF5 domain of TM binds to exosite-1 of thrombin and the acidic EGF4 domain binds to the basic exosite of the Ca²⁺-stabilized protein C (including Arg-67), thereby altering the conformation of the active-site and/or the extended binding pocket residues of thrombin (including Arg-35), thus facilitating the docking of the activation peptide of protein C in the catalytic groove of thrombin (see the text for further details).

The activation peptide of protein C, at its primed site, contains several basic residues which are also shown to play an inhibitory role in zymogen recognition by thrombin [37]. It is possible that the repulsive interaction of these residues with Arg-35 also contributes to the inability of thrombin to accommodate the activation peptide of protein C in its active-site pocket in the presence of Ca²⁺. It appears that the allosteric modulation of the extended active-site pocket residues of thrombin (including the side chain of Arg-35) by TM overcomes these inhibitory interactions [28,38,39,40]. Thus, Arg-35 of thrombin and Arg-67 of protein C cooperatively drive the unique TM and Ca²⁺ dependence of protein C activation by thrombin [28]. It is worth noting that two Asp residues, present at the P3 and P3′ sites of the protein C activation peptide, have been shown to play an inhibitory role in interaction of protein C with thrombin, presumably due to their repulsive interactions with Glu-192 of thrombin [41–43]. It is likely that TM also plays a role in alleviating these inhibitory interactions [1,43].

3. Anticoagulant Function of APC

APC down-regulates the clotting cascade by proteolytically degrading the procoagulant cofactors FVa and FVIIIa which are essential cofactors for factors Xa and IXa in the extrinsic and intrinsic pathways of thrombin generation, respectively [1–3,44]. Both cofactors are homologous glycoproteins, synthesized as single chain precursors with domain structures A1-A2-B-A3-C1-C2 [45–47]. During activation by the physiological activator thrombin, the B domain from both cofactors is released and the resulting A1-A2 heavy chain subunits remain non-covalently associated with the A3-C1-C2 light chain in a divalent cation-dependent mechanism [48–50]. While A1-A2 subunits in FVa remain contiguous, the cleavage at Arg-372 of FVIII by thrombin converts the heavy chain of the cofactor into two separate subunits [51]. The A1 subunit of FVIIIa retains a stable metal-ion dependent linkage with the A3-C1-C2 subunit, whereas the A2 subunit remains weakly associated with the dimer through electrostatic interactions [52]. The C2 subunit in both cofactors binds to negatively charged membranes [53–56].

The mechanism by which APC recognizes the two cofactors in the anticoagulant pathway has been extensively investigated by several laboratories in recent years [57–61]. It has been demonstrated that the same basic residues of three surface loops (39-, 60- and 70-80-loops) (Fig. 1) that are critical for the recognition of the zymogen protein C by the thrombin-TM complex are also involved in determining the recognition specificity of APC with FVa [61–63] and FVIIIa [64] in the anticoagulant pathway. In addition to basic residues of these loops, the autolysis loop (148-loop) of APC is also highly basic, with 5 Lys/Arg between residues 143–154. Mutagenesis studies have indicated that these residues may also be critical for the APC recognition of FVa in anticoagulant pathway [60,65–67].

The inactivation of FVa by APC requires two cleavages at Arg-306 and Arg-506 sites, located on the A1 and A2 domains of the cofactor, respectively [58,59]. It has been demonstrated that APC cleaves these two sites in a sequential kinetic order, with an initial cleavage occurring rapidly at the Arg-506 site followed by a slower cleavage occurring at the Arg-306 site [58]. While the cleavage of the Arg-506 site is relatively fast and independent of a cofactor, the cleavage of the Arg-306 site is accelerated 20-fold by protein S, a vitamin K-dependent cofactor for APC in plasma [57]. Mutagenesis data have indicated that the interaction of the basic exosite of APC with a complimentary site, flanking the Arg-506 site of FVa, is responsible for the ability of APC to preferentially cleave this bond with a higher catalytic specificity [62,63]. The site on APC that determines the cleavage specificity of the Arg-306 peptide bond has not been identified. Since the cleavage of the Arg-306 site by APC is membrane-dependent, it has been hypothesized that protein S binding to APC leads to the relocation of the active-site topography of the membrane-bound APC, thereby allowing the protease to recognize and cleave the Arg-306 scissile bond on the membrane surface [68,69]. The affinity of protein S for membrane is very high, thus protein S also improves the affinity of APC for the negatively charged membrane surfaces [70].

In the case of FVIIIa, APC cleaves the cofactor at the homologous Arg-336 and Arg-562 sites, located on the A1 and A2 subunits, respectively [51,52]. Unlike FVa, however, in which a single cleavage at Arg-506 site only partially inactivates the cofactor [57,59], the APC cleavage of either site in FVIIIa leads to a near-complete inactivation of the cofactor [71]. In vitro data has indicated that protein S and FV function synergetically as cofactors to accelerate the rate of FVIIIa degradation by APC [72]. Both the A1 and A2 subunits of FVIIIa contain acidic C-terminal sequences that might potentially provide contact sites for the basic exosite of APC [48,64]. It is however worth noting that FVIIIa has a low concentration and a short half-life in plasma due to the spontaneous A2 subunit dissociation which leads to a complete loss of cofactor activity [48,73]. Thus a physiological role for APC in the regulation of the procoagulant activity of FVIIIa has not been fully established. By contrast, the critical role of APC in the regulation of the clotting cascade through the proteolytic degradation of FVa has been well-established [26,58]. A natural common variant of FVa in which the protease recognition site at residue Arg-506 is mutated to Gln is resistant to efficient inactivation by APC [74,75]. This mutation, which is named FVa Leiden [75], occurs with a high frequency of ~5–10% in the general population and is associated with a high incidence of venous thrombosis [58,74].

In addition to the heavy chain, an interactive site for APC on the light chain of both FVa and FVIIIa has been reported [76,77]. This site has been mapped to the C-terminal end of the A3 domain in both cofactors [76,77]. Peptide sequences derived from the C-terminal end of the A3 domain of both FV and FVIII inhibit the APC inactivation of the cofactors [76]. The association of FVa and FVIIIa with their target proteases factor Xa and factor IXa in the prothrombinase and Tenase complexes, respectively, prevents the APC recognition and inactivation of cofactors [78–80]. This is possible by both factors Xa and IXa having overlapping binding sites with APC on the cofactors, thereby protecting the susceptible cleavage sites from recognition by APC upon binding to cofactors on negatively charged membrane surfaces [78–80].

4. Intracellular Signaling Function of APC

In addition to its anticoagulant activity, APC also exhibits potent cytoprotective and antiinflammatory properties. This has been demonstrated both in in vitro endothelial cell culture systems and in vivo animal models of inflammation [8–10,81,82]. The mechanism of the protective cellular signaling activity of APC is poorly understood. It has been hypothesized that when the Gla domain of APC binds to EPCR it acquires a different specificity, thus activating PAR-1 and thereby initiating protective signaling events in endothelial cells [9,10]. The specific Gla domain residues that differentially mediate the interaction of APC with either protein S in the anticoagulant pathway or EPCR in the cytoprotective pathway has been mapped by a mutagenesis approach [83]. It appears that several residues at the N-terminal Gla domain play critical roles in determining the specificity of APC binding to EPCR and several others in the C-terminal end of the Gla domain are involved in specific interactions with protein S at the membrane surface [83].

In a recent study, we demonstrated that Glu-167 and Glu-170, two residues unique for the 162-helix of APC on the catalytic domain (Fig. 1), which are not conserved on the homologous helix of other vitamin K-dependent coagulation proteases, constitute a specific PAR-1-binding exosite on APC that facilitates the recognition and cleavage of the receptor by the protease on the endothelial cell surface [84]. We showed that the specific interaction of APC with an unknown site of the PAR-1 extracellular domain through this exosite is essential for the EPCR- and PAR-1-dependent cytoprotective signaling function of APC [84]. Both APC mutants exhibited normal anticoagulant activity in both FVa-inactivation and plasma-based clotting assays, suggesting that neither one of the acidic residues of 162-helix are involved in interaction of APC with FVa and FVIIIa [84]. Previous results have demonstrated that the substitutions of the basic residues of either 39-loop (Lys-37, Lys-38 and Lys-39) or the Ca²⁺-binding 70-80-loop (Arg-74 and Arg-75) (Fig. 1) specifically abrogate the anticoagulant activity, but not the protective cell signaling activity of APC [85]. Thus, the specificity of the APC for interaction with its natural substrates in the alternative anticoagulant and cytoprotective pathways are determined by two distinct exosites with different polarity. The acidic residues of the 162-helix, on the left side of the active-site toward the back of the molecule, determine the specificity of the protease interaction with PAR-1, and the basic residues of 39- and 70-80-loops on the right side of the active-site, determine the specificity of the protease interaction with the procoagulant cofactors (Fig. 1).

As indicated above, recombinant APC has been approved by the FDA in the US for treating severely septic patients, reducing the overall mortality rate by nearly 20% [15]. However, there are at least two relatively major problems with the APC-therapy that need to be addressed. The first problem is that the use of APC is associated with an increased risk (~ 2%) of bleeding in certain patients [10,15]. This problem is greater than it appears to be since it also limits the threshold of APC dosage in treatments, thus potentially masking the beneficial effect of APC in a higher number of patients. The second major problem is the high cost of the APC-therapy, with a 96 hour continuous infusion of recombinant APC [drotrecogin alfa (activated)^™ Eli Lilly] costing ~$10,000, thus limiting its wide utilization due to socioeconomic reasons [86]. Recent results, using a mouse non-anticoagulant APC derivative in which all three basic residues of the 39-loop and two basic residues of the 70-80-loop have been replaced with Ala residues (APC-5A), have demonstrated that the cellular signaling activity of APC is primarily responsible for its beneficial protective property in animal models of severe sepsis [87,88]. If similar results hold true for human APC, such mutants could eliminate the increased bleeding risk of the APC-therapy in severe sepsis.

Recently, we showed that the 70-80-loop stabilized Cys-67/Cys-82 mutant of human APC, described above (Fig. 2), has a dramatically reduced anticoagulant, but near normal protective intracellular signaling activity in endothelial cells [89]. Moreover, our preliminary in vivo data suggests that the non-anticoagulant APC Cys-67/Cys-82 improves the neurological outcome of experimental stroke in mice [90] and exhibits cardio-protective activities in a mouse model of ischemia/reperfusion injury [91]. In the latter model, the 162-helix mutant of APC which possesses normal anticoagulant activity but lacks a PAR-1-dependent cytoprotective activity, did not exhibit a cardio-protective activity [91]. In a recent study, another APC mutant with no cytoprotective activity but improved anticoagulant and antithrombotic activity exhibited diminished protective activity in an endotoxin-induced murine sepsis model [92]. These mutants provide excellent tools to initiate further in vivo studies to understand the extent to which the anticoagulant activity versus direct cellular activity of APC contributes to the beneficial effect of APC in severe sepsis. If it turns out that the direct cellular effect of APC is responsible for its protective effect in severe sepsis, it may be possible to use the structure-function knowledge described above to develop new APC variants with no anticoagulant but improved specific activity toward PAR-1, thereby reducing the bleeding side effect and the high cost of the APC-therapy in the septic setting.

4. Mechanism of the Protective Intracellular Signaling Function of APC

The mechanism by which the APC activation of PAR-1 elicits protective intracellular responses is not fully understood. The complexity of this question is underscored by the observation that thrombin activates PAR-1 with at least three orders of magnitude higher catalytic efficiency to initiates disruptive, hyperpermeability and proinflammatory responses in endothelial cells [17,18]. The PAR-1-dependent proinflammatory properties of thrombin have been reported to be mediated through the protease activating the nuclear factor (NF)-κ B pathway in endothelial cells [93–95]. It has been demonstrated that the APC-EPCR complex can inhibit the thrombin- or proinflammatory cytokines-mediated activation of NF-κB and down-regulate the expression of proinflammatory cytokines by activation of PAR-1 in endothelial cells [93–95]. The mechanism by which the activation of PAR-1 by the two proteases initiates two opposite responses in in vitro cellular models is not known. Recently, we investigated whether the level of receptor activation by thrombin and APC determines the type of response in endothelial cells. Thus, we engineered a chimeric meizothrombin (thrombin intermediate which has both Gla and Kringle-1 and -2 domains of prothrombin) in which the Gla domain of the thrombin intermediate was substituted with the corresponding domain of APC (Fig. 4) [96]. This meizothrombin derivative retained its thrombin-like high specific activity toward PAR-1 and interacted with EPCR with normal affinity [96]. Interestingly, we discovered that PAR-1 cleavage by this meizothrombin derivative elicits a protective signaling response in endothelial cells, suggesting that the binding of Gla domain of APC to EPCR rather than the protease type cleaving the receptor determines the type of PAR-1 response in endothelial cells [96]. To investigate the mechanism of this effect, we studied the effect of PAR-1 cleavage by thrombin in endothelial cells which were pretreated with the catalytically inactive Ser-195 to Ala substitution mutant of protein C. Both APC and the zymogen protein C interact with EPCR with identical affinities. The results revealed that when EPCR is occupied by its ligand protein C, the cleavage of PAR-1 by thrombin elicits only protective signaling responses in endothelial cells [96].

Fig. 4.

Cartoons of wild-type and mutant prothrombin activation by factor Xa. (A) The proteolytic cleavage of prothrombin at Arg-320 by factor Xa yields an activation intermediate that is an active product called meizothrombin. A second cleavage at Arg-271 by factor Xa is required to separate the catalytic domain of prothrombin from the non-catalytic domain (Gla, Kringle-1 and Kringle-2 domains) to yield thrombin. Further cleavages can occur at Arg-155 and Arg-286 in a feed-back reaction by both thrombin and meizothrombin. (B) The substitution of the Gla domain of prothrombin with the corresponding domain of protein C and its Arg-155, Arg-271 and Arg-286 residues with 3 Ala’s yields a mutant of prothrombin (3A-prothrombin/PC-Gla) which can be activated by factor Xa through the cleavage of Arg-320 to yield meizothrombin/PC-Gla which cannot be further processed by either factor Xa or the resulting mutant meizothrombin.

To provide further support for our hypothesis, we pre-incubated human umbilical vein endothelial cells or human pulmonary artery endothelial cells with a physiologically relevant concentration of the zymogen protein C (wild-type or the S195A mutant) and then used the thrombin receptor agonist peptides (TRAP) SFLLRN or TFLLRN to activate PAR-1. Both PAR-1 agonist peptides elicited a barrier disruptive response in endothelial cells that could be effectively reversed to a protective response if endothelial cells were pretreated with the zymogen protein C prior to their stimulation with TRAP [96,97]. Further studies demonstrated that when EPCR was occupied by protein C, similar to APC, the thrombin activation of PAR-1 inhibits the activation of RhoA and enhances the activation of Rac1 pathways in the TNF-α-stimulated endothelial cells [95]. Furthermore, thrombin inhibited NF-κB pathway by a PAR-1-dependent mechanism if cells were pretreated with the zymogen protein C [95]. Based on these results, we concluded that the cleavage of PAR-1 by thrombin on intact endothelium expressing EPCR would initiate potent protective intracellular responses in the presence of physiological concentrations of the zymogen protein C [95–97]. Thus, the in vitro data proposing a PAR-1-dependent proinflammatory effect for thrombin in cellular models may have no physiological relevance.

How the occupancy of EPCR by protein C changes the PAR-1-dependent signaling specificity of thrombin is not known. PAR-1 can signal through interaction with different members of the G protein subfamilies including G_i, G_q and G_12/13 [98]. It has been hypothesized that thrombin disrupts endothelial barrier function through the activation of PAR-1 that is coupled to G_q and/or G_12/13, but APC signals through the cleavage of PAR-1 coupled to G_i to reverse the hyperpermeability induced by proinflammatory cytokines [96,98]. However, in a series of studies, we discovered that the ligand occupancy of EPCR recruits PAR-1 to a protective pathway by coupling PAR-1 to G_i in endothelial cells independent of the protease cleaving the receptor [95–97]. Our studies revealed that both PAR-1 and EPCR associate with caveolin-1 in the cholesterol rich lipid-rafts and that the occupancy of EPCR by either APC or protein C leads to its dissociation from caveolin-1, a process which appears to change the specificity of PAR-1 signaling from a disruptive to a protective response in cultured endothelial cells (Fig. 5). Our further studies revealed that when EPCR is occupied, the disruptive hyperpermeability activity of thrombin is mediated through the activation of PAR-4 [96]. The activation of PAR-4 requires a higher concentration of thrombin since, unlike PAR-1, PAR-4 lacks a hirudin-like sequence to bind exosite-1 of thrombin [18].

Fig. 5.

Cartoons of PAR-1 activation by APC and thrombin when EPCR is either free or occupied by its ligand protein C. (A) EPCR is associated with caveolin-1 (Cav-1) in the lipid-rafts of endothelial cells when the receptor is not occupied by the Gla domain of protein C/APC. Thrombin cleavage of PAR-1 elicits disruptive signaling responses through coupling the receptor to G_12/13 and G_q proteins. (B) The occupancy of EPCR by protein C (PC) results in the dissociation of EPCR from caveolin-1, thereby switching the specificity of PAR-1 signaling by coupling it to the G_i protein. Thus, thrombin cleavage of PAR-1 initiates protective response when EPCR is occupied. (C) The same as (B) except that the EPCR and PAR-1 dependent protective signaling response is elicited by APC.

Given our in vitro cellular data that both thrombin and APC exert their PAR-1-dependent protective effects by an EPCR-dependent mechanism (Fig. 5), the appropriate question that needs to be addressed is: since in animal models of inflammation thrombin is expected to be generated and thrombin is the only known physiological activator of protein C, how then does APC exert its protective effect in vivo through the activation of PAR-1 in the presence of thrombin, which has three orders of magnitude higher catalytic efficiency toward PAR-1? Clearly, numerous animal models of inflammation have established an EPCR and PAR-1-dependent protective effect for APC and neither the active-site inhibited protease nor the S195A mutant of protein C exhibit protective activity in these models [10]. One possibility is that thrombin partitioned to the vasculature may not be capable of activating the microvascular endothelial cell surface PAR-1 due to its high-affinity interaction with TM which occupies exosite-1 of thrombin. In support of this hypothesis, we recently demonstrated that all three receptors, TM, EPCR and PAR-1 are colocalized within lipid-rafts of endothelial cells [99]. We also discovered that the interaction of exosite-1 of thrombin with the acidic hirudin-like sequence present on the exodomain of PAR-1 is required for recognition and cleavage of the receptor by thrombin [100]. Noting that the affinity of thrombin for endothelial cell TM (K_D < 1 nM) [101] is much higher than that of PAR-1 (in μM range for the exodomain of the receptor) [102] and the effective concentration of TM in the microcirculation can be as high as 500 nM [103,104], it is possible that thrombin partitioned onto the microvascular endothelial cell surface would all bind TM, with the complex being capable of only activating the EPCR-bound protein C but not PAR-1. Thus, in the microcirculation, the activation of the endothelial cell PAR-1 may proceed solely via the APC pathway. Noting the membrane lipid-raft localization of all three receptors, this pathway would be sufficiently robust since the activation of protein C by the thrombin-TM complex would be mechanistically linked to the activation of PAR-1 by APC, as has been previously proposed [105].

Another possibility is that the PAR-1-dependent endothelial cell signaling does not contribute significantly to the antiinflammatory role of APC, but rather that leukocytes are major targets for the protective effect of APC in severe sepsis. In this venue, it has recently been demonstrated that APC has an RGD sequence that can bind to leukocyte integrins, thereby inhibiting the migration of neutrophils into tissues [106]. Results of several other recent studies have indicated that the EPCR-dependent barrier protective activity of APC is mediated through a crosstalk with other G- and non-G-protein coupled receptors [94,107,108]. Thus, the activation of PAR-1 by APC and thrombin can differentially couple PAR-1 to different trans-activators in various cells. In support of this hypothesis, it has been shown that PAR-1 crosstalk with sphingosine 1-phosphate receptor 1 (S1P₁) elicits a protective response [107], however, a crosstalk between PAR-1 and S1P₃ evokes proinflammatory responses [109]. Another recent study showed that the protective effect of PAR-1 requires transactivation of the PAR-2 signaling pathway [110]. Two other studies demonstrated that the APC-mediated signaling via both apolipoprotein E receptor 2 (ApoER2) and angiopoietin (Ang)/Tie2 pathways also contribute to cytoprotective and antiinflammatory properties of APC [111,112]. Finally, to put the complexity of this question into perspective, a very interesting recent study showed that extracellular histones are major mediators of endothelial dysfunction, organ failure and death during severe sepsis and that the protective effect of APC in an LPS-induced murine sepsis model is primarily mediated through the direct APC degradation of histones independent of both EPCR and PAR-1 [113]. Thus, much further research work is required to understand the exact mechanism by which APC modulates the inflammatory pathways in the severe sepsis syndrome.

5. Regulation of the Proteolytic Activity of APC in Plasma

The proteolytic activity of APC in plasma is primarily regulated by the serpins, protein C inhibitor (PCI), plasminogen activator inhibitor 1 (PAI-1) and α₁-proteinase inhibitor (α₁-PI) [114–116]. The general non-serpin inhibitor α₂-macroglobulin also contributes to regulation of APC activity in plasma [117]. However, APC reacts very slowly with all of these inhibitors, thus it has a long circulating half-life of approximately 20–25 min in plasma [115,117]. Another mechanism, which may be involved in the regulation of APC in plasma, is the EPCR-mediated endocytosis that can facilitate the transcytosis of the protease and its clearance from circulation [118,119]. Among the three serpins, the reactivity of APC with α₁-PI is the slowest (~10 M⁻¹s⁻¹). However, owing to its high concentration (~40 μM), α₁-PI is believed to contribute to the regulation of APC activity in plasma [115]. The reactivity of APC with the other two serpins is also relatively slow (10²–10³ M⁻¹s⁻¹). However, the two cofactors heparin and vitronectin markedly accelerate the reactivity of APC with PCI and PAI-1, respectively [114,120].

Therapeutic high molecular weight heparins can accelerate the inhibition of APC by PCI by three orders of magnitude in the presence of physiological levels of Ca²⁺ [121], possibly suggesting a negative outcome for heparin-therapy in severe sepsis. The heparin-mediated acceleration of the PCI inhibition of APC requires the interaction of heparin with the same basic exosite that is also critical for protein C activation by the thrombin-TM complex and the anticoagulant activity of the protease [121]. PCI may also play an important role in the regulation of protein C activation by thrombin since it rapidly inhibits thrombin when the protease forms a complex with TM [122]. It has been demonstrated that TM provides a binding site for PCI in the thrombin-TM complex, thereby facilitating the optimal recognition of the serpin by the protein C activation complex [123,124]. How vitronectin accelerates the PAI-1 inhibition of APC has not been investigated. Vitronectin is abundant plasma and platelet glycoprotein and its ability to promote the interaction of APC with PAI-1 may lead to the depletion of the plasma pool of PAI-1, thereby increasing the effective concentration of plasminogen activators in plasma and possibly accounting for the reported profibrinolytic property of APC [11,116,120]. Insight into mechanisms by which serpins neutralize the proteolytic activity of APC in plasma can lead to the design of novel protease variants with reduced serpin reactivity, thus providing another approach for potentially improving the therapeutic profile of APC by protein engineering. Indeed, such knowledge has been utilized to engineer APC derivatives which exhibit resistance to inactivation by PCI and α₁-PI without alteration in their anticoagulant activity [125].