With over 800 members in mammals, G protein-coupled receptors (GPCRs) are the largest family of cell-surface signaling proteins. They are receptors for an extraordinary range of structurally diverse agonists in the extracellular fluid, including endogenous hormones, neurotransmitters, and paracrine regulators, as well as multiple exogenous ligands (1, 2). Due to their critical importance in the control of most patho-physiological processes, GPCRs are the primary target for over 30% of the clinically used drugs (3, 4). The established mechanism of GPCR activation is that agonist binding results in conformational changes in the receptor that activate the Gα subunits of heterotrimeric G proteins, leading to the dissociation of Gβγ dimers from Gα. Activated Gα and Gβγ then initiate downstream signaling processes (5). To control the duration and magnitude of this signaling, activated receptors are phosphorylated by G protein-coupled receptor kinases (GRKs) or other kinases, and then interact with β-arrestins, which mediate receptor desensitization and endocytosis (6). Depending on the receptor and the agonist, internalized receptors are then sorted to lysosomes for degradation, or move to the plasma membrane for another cycle of activation (7, 8). However, a common feature of GPCRs is that a single receptor can interact with multiple endogenous and exogenous ligands, each of which may activate the receptor in different ways. For example, a large number of endogenous opioid neuropeptides as well as many different opiate drugs interact with opioid receptors, and different opioids and opiates result in divergent processes of receptor activation and regulation (9). Thus, the simplistic view of receptor activation and regulation has been revised by the appreciation that different agonists of the same receptor can result in distinct patterns of signaling and regulation.

The early two-state model of receptor function suggested that a receptor adopts active conformation upon ligand binding. This model considered only one active state, leading to a single functional readout. However, increased understanding of receptor signaling has revealed that different ligands can initiate distinct signaling events through the same GPCR. The heterogeneity of signaling events by a single GPCR can include different maximum responses from a single pathway (i.e., full or partial agonism) or activation of distinctly different signaling pathways by different agonists. The capacity of different agonists to initiate signaling of the same GPCR by distinct mechanisms is referred to as biased agonism or signaling (10, 11), and has been described for many GPCRs, including opioid receptors (12), angiotensin receptors (13), and glutamate receptors (14). This phenomenon of signaling bias is not surprising because GPCRs are flexible proteins that interact with multiple ligands and regulatory proteins, all of which may influence the capacity of the receptor to signal by particular mechanisms. Indeed, recent advances in our understanding of the structure of GPCRs in various activation states has revealed that a single GPCR can exist in multiple active conformations that may favor coupling to different signaling pathways (15, 16).

This review focuses on the capacity of different proteases and synthetic ligands to induce biased signaling of protease-activated receptors (PARs). The PARs are a family of four GPCRs (PAR_1–4) that belong to group A rhodopsin-like GPCR subfamily. The first family member, PAR₁, was identified as a receptor for thrombin, a serine protease coagulation factor (17). PAR₂ was subsequently identified as a receptor for the serine protease trypsin (18), followed by PAR₃, another thrombin receptor (19), and PAR₄, a receptor for both thrombin and trypsin (20). PARs are expressed in many tissues and cell types, where they regulate multiple patho-physiological processes, including hemostasis, inflammation, pain, cellular proliferation, and healing (21–23). However, in addition to thrombin and trypsin, a large number of proteases have been identified that can cleave PARs. In some cases, these proteases cleave at the same sites as thrombin or trypsin and thereby initiate common signaling events. However, in other cases, proteases cleave PARs at distinct sites, and either activate distinct signals (biased agonism), or disarm the receptor by removing or destroying tethered ligand domains (receptor antagonism). We will review mechanisms by which various proteases and synthetic agonists activate PARs, and will discuss the implications of protease-biased signaling of PARs for patho-physiological control and therapeutic targeting.

Unlike other GPCRs, the endogenous ligands for PARs reside within the extracellular N-terminus of the receptors. Receptor cleavage at the defined sites within the N-terminus by proteases such as thrombin and trypsin reveals these tethered ligands that, once exposed, can bind to regions in the second extracellular loops of the cleaved receptors, initiating conformational changes in the receptors that activate downstream signals (23). This is the canonical mechanism of PAR activation (Figure 1A). There are subtle differences in the mechanisms by which different proteases initiate the canonical pathways of receptor activation, which depend on the protease and PAR in question. For example, thrombin first binds to PAR₁ and PAR₃; this action facilitates receptor cleavage and exposure of the tethered ligand sequence. Mutation of the binding site reduces the efficacy with which thrombin activates these receptors, and mutation of the cleavage site prevents receptor activation (17, 19). On the other hand, trypsin activates PAR₂ directly, without first binding to the receptor (18, 24). Accessory proteins can also influence the capacity of proteases to activate PARs. In particular, proteins that anchor proteases to the plasma membrane can enhance proteolytic activation. For example, during tissue damage and inflammation, tissue factor (TF) binds coagulation factors (F) VIIa, which in turn activates FX to FXa. FXa and its co-factor FVa promote conversion of prothrombin to thrombin, and subsequent PAR₁ activation (25). Besides promoting thrombin activation, FVIIa and FXa both can signal directly through PAR₁ and PAR₂, although the efficiency and potency of receptor activation is substantially enhanced when they are coupled with TF (26). Similarly, the proteolytic activity of the anticoagulant activated protein C (APC) toward PARs is largely regulated by its association with the endothelial protein C receptor (EPCR) at the surface of endothelial cells (27, 28).

Support for the tethered ligand mechanism of PAR activation is provided by the observation that synthetic peptides, referred to as activating peptides (APs), that mimic the tethered ligand domain can also activate certain PARs directly, without the requirement for proteolysis (Figure 1C). Peptides mimicking the tethered ligands of PAR₁, PAR₂, and PAR₄ can directly activate these receptors, although with a considerably lower potency than the activating proteases, especially in the case of PAR₄ (17, 18, 20). The higher EC₅₀ values of APs compare to those of proteases possibly reflect the differences between a tethered ligand and an untethered ligand in solution. PAR₃ is not activated by tethered ligand-derived peptides, and appears to be unable to signal directly, but rather to serve as a co-factor for other PARs, such as PAR₁ and PAR₄ (29, 30).

Activating peptides have been considered to mimic the effects of proteases and have been widely used to probe the functions of PARs without the use of proteases, which can cleave multiple other proteins that may influence outcomes. However, this is not always the case because in some circumstances proteases and APs agonists can exert different effects. For example, in human brain microvascular endothelial cells, thrombin activation of PAR₁ triggers endothelial barrier permeability, whereas PAR1-AP (SFLLRN-NH₂) has no significant effect (31). In addition, the signaling properties of a PAR₂ mutant with substitutions within the trypsin-revealed tethered ligand domain differ from those of APs with the same substitutions, suggesting distinct activation modes by tethered versus soluble peptides (32). The divergent signaling effects of proteases and APs provide evidence for biased signaling of PARs.

In addition to the diversity of signals that can originate from the same receptor after activation by proteases or synthetic agonists (i.e., biased signaling), many other factors also affect patho-physiological outcome of PAR activation. These factors include the availability of activated proteases as well as the existence of regulatory and accessory proteins in different tissues and cell types.

The availability of active, functional proteases is a key requirement of PAR signaling, and the predominant endogenous proteases that activate PARs may vary in different patho-physiological states. For instance, the compliment of available active proteases varies markedly during the course of inflammation and healing, depending on the presence of immune cells, which are the source of many proteases, and on the existence of endogenous inhibitors. The compliment of proteases of mast cells, neutrophils, eosinophil, and macrophages, which participate in different phases of inflammation, varies considerably. For example, as the first responders to microbial infection, neutrophil produce elastase, cathepsin G, and proteinase-3 (33), whereas macrophages, which mediate chronic inflammation, release plasmin, matrix metalloproteinases (MMPs), and cathepsin S (34).

Although PARs can be activated by distinct proteases under different conditions, proteases that cleave PARs at the same sites would be expected to activate the same canonical signaling pathway and to induce common patho-physiological outcomes. However, the consequences of PAR cleavage could vary considerably if the activated proteases cleave PARs at distinct sites and are biased agonists or even antagonists, as discussed below. Indeed, at any one time multiple proteases would likely be activated and capable of cleaving PARs at distinct sites with unique outcomes. Thus, the active conformation of PARs may vary depending on the milieu of available proteases, which may differ between health and disease conditions. For example, during the initial stages of inflammatory processes such as inflammatory bowel diseases or chronic obstructive pulmonary disease, infiltration of neutrophils leads to increased level of elastase (35, 36), a biased agonist of both PAR₁, and PAR₂ (37, 38) (discussed below). Further complexity is provided by the presence of endogenous protease inhibitors that control the activity of proteases (39).

The outcome of PAR activation by the same protease or synthetic agonist can also vary between tissues and cell types. For example, thrombin and PAR₁-AP cause relaxation of the intact coronary artery but contraction when the endothelium is removed, indicating distinct outcomes of PAR₁ activation in endothelial versus vascular smooth muscle cells, possibility due to formation of different signaling complexes (40).

Compared to other GPCRs, the N-terminal domains of PARs are particularly susceptible to proteolysis. Although the reasons for this susceptibility are not fully understood, they probably relate to the presence of protease binding sites on the receptors, the existence of multiple scissile bonds, and the lack of groups that would sterically hinder proteolysis. However, the outcome of PAR activation depends on the site of proteolytic cleavage. Those proteases that cleave PARs at the conserved activating sites reveal tethered ligands that trigger the canonical signaling pathways (Figure 1A). Proteases that cleave PARs at distinct sites can act as biased agonists by triggering signals that are distinct from those activated by the canonical pathways (Figures 1B,D). In some cases, these alternative signaling mechanisms appear to involve exposure of distinct tethered ligands (Figure 1B). However, in other instances receptor cleavage per se may generate a conformational change that is sufficient to activate the receptor (Figure 1D). Alternatively, proteases can destroy or remove tethered ligand domains, forming N-terminally truncated receptors that are unresponsive to further activation by other proteases (Figure 1E).

Activated PARs can couple to multiple G protein-dependent (Gαq, Gα12/13, Gαi, Gαs, and Gβγ) and β-arrestin-dependent pathways. Although in many instances a particular protease or synthetic agonist can activate more than one of these pathways, in some cases proteases and synthetic agonists activate a single pathway. By comparing and categorizing the signaling pathways that are initiated by different proteases and synthetic agonists with the overall outcome of receptor activation, it is possible to identify the primary signaling pathways responsible for PAR-mediated patho-physiological responses (Tables 1–3). Moreover, a comprehensive understanding of the mechanisms and outcomes of PAR signaling by different proteases and synthetic agonists can guide the development of agonists and antagonists that may selectively activate or inhibit disease-relevant pathways. This approach has implications for development of pathway-specific therapies.

Although thrombin cleaves PAR₃ at K³⁸/T³⁹, there is little evidence that the cleaved receptor is capable of signaling. Thrombin and a synthetic peptide corresponding to the putative tethered ligand failed to generate PAR₃-dependent Ca²⁺ signals (19). However, this PAR₃-derived AP is able to activate PAR₁ and PAR₂ (72). Instead of signaling in its own right, PAR₃ appears to be a co-factor for the activation of other PARs, including PAR₄ and PAR₁ (27, 29, 30). Co-expression of PAR₃ with PAR₄ in COS-7 cells leads to over 10-fold increase in the efficiency of thrombin cleaving PAR₄ compare to PAR₄ expressed alone (EC₅₀ from 0.3 to 0.05 nM) (29). However, with the appreciation of PAR biased signaling, further studies are warranted to examine the full repertoire of potential PAR₃ signals.

In contrast to other PARs, fewer proteases have been identified that can cleave PAR₃. Besides thrombin, APC is the only protease to be identified that exhibits proteolytic activity toward PAR₃ (27, 74). In immortalized human and mouse podocytes, which have higher expression of PAR₂ and PAR₃ than PAR₁ and PAR₄, the maximum inhibitory effect of APC-dependent podocyte apoptosis requires cleavage of N-terminal domain of PAR₃ by APC at the same position as thrombin (27). Cleavage of PAR₃ by APC promotes dimerization between PAR₂ and PAR₃, and this is required for APC-dependent cytoprotective effect. A recent study reported a novel mechanism of APC activation of PAR₃ via a cleavage at a non-canonical site (R⁴¹/G⁴² X instead of K³⁸/T³⁹) (74). Peptide hydrolysis experiments revealed a slow kinetics of APC cleavage of PAR₃ (50% peptide cleavage reached after ~5 h), suggesting that other modulators may facilitate APC cleaving PAR₃ in cells; indeed, the efficiency of this cleavage increased proportional to the expression of EPCR (74). This novel cleavage by APC generates a new tethered ligand domain starting with G⁴²APPNS. Consistent with a tethered ligand activating mechanism, APC-AP (G⁴²APPNSFEEFPFS⁵⁴-NH₂) is able to prevent thrombin-induced endothelial barrier disruption. Interestingly, an extended peptide generated from thrombin cleavage site (⁴⁰TFGAPPNSFEEFPFS⁵⁴-NH₂) fails to do so. This suggests that APC and thrombin activation of PAR₃ mediates different signaling profile, and thus is involved in different cellular response, respectively. However, similar to thrombin activation of PAR₃, existence of another receptor, such as PAR₁ is necessary for APC/PAR₃-mediated signaling (74).

PAR₄ was identified by a homology search using amino acid query sequence derived from known sequences of PAR₁, PAR₂, and PAR₃ (20). The putative protease cleavage site was identified (R⁴⁷/G⁴⁸), and the EC₅₀ of thrombin toward PAR₄ was much higher compare to other thrombin sensitive receptors PAR₁ and PAR₃ (5 nM for PAR₄, and 0.2 nM for PAR₁ and PAR₃) (17, 19, 20). Further investigation suggests that this may be due to the lack of the thrombin binding site within the amino terminus of the receptor. Consistent with this observation, γ-thrombin, another isoform of thrombin that lacks a receptor binding site exhibits similar affinity on PAR₄ compare to α-thrombin (20). Different from other PARs that can be cleaved preferentially by trypsin or thrombin, PAR₄ exhibits similar sensitivity toward both enzymes. Both thrombin and trypsin activities toward PAR₄ can be abolished by R⁴⁷/A mutation of the receptor, suggesting that both enzymes cleave the receptor at the same site (20).

Compared to PAR₁ and PAR₂, little is known about biased signaling of PAR₄, and to date no additional activating cleavage sites have been identified. However, although proteases such as plasmin can activate PAR₄ by cleaving the receptor at the canonical site, thrombin and plasmin cleave PAR₄ with different kinetics, possibility due to different mechanisms of action or the affinity of the proteases for the receptor (20). In mouse platelets, PAR₃ binds to thrombin and thereby acts as a co-factor to facilitate thrombin cleavage and activation of PAR₄ (29). However, in both transfected cell lines as well as platelet, instead of acting as a co-factor, the presence of PAR₃ inhibits plasmin-mediated PAR₄ activation leading to a decrease in intracellular Ca²⁺ mobilization and platelet aggregation (79). The mechanism that underlies these findings is not clear. However, since thrombin cleaves both PAR₃ and PAR₄ whereas plasmin can only cleave PAR₄, the conformation of the PAR₃–PAR₄ receptor pair might be different after plasmin or thrombin cleavage. In addition, difference in binding affinities and kinetics of plasmin and thrombin for PAR₄ may also result in distinct receptor-protease complex formation and lead to variation in downstream responses.

Cathepsin G is a neutrophil serine protease that plays an important role in inflammation. Cathepsin G can evoke PAR₄-dependent Ca²⁺ signals in human platelets and in PAR₄-transfected fibroblasts (81). Cathepsin G and PAR₄ are upregulated in ulcerative colitis patients, and inhibition of cathepsin G and PAR₄, but not PAR₁ or PAR₂, is protective (124). Although no direct evidence suggests that PAR₄ is the target for cathepsin G in colitis, this study highlights the possibility of targeting cathepsin G or PAR₄ as novel therapeutic approach.

Recently, mannose-binding lectin-associated serine protease-1 has been shown to cleave PAR₄ but not PAR₁ or PAR₂ in endothelial cells, and to induce PAR₄-dependent Ca²⁺ responses and activation of NF-κB and p38 MAPK pathways. However, the exact cleavage site remains to be determined (82).

Although most studies have examined signaling by monomeric PARs, considerable evidence suggests that PARs may form homo- or hetero-dimers and function as a complex. Dimerized PARs could adopt unique conformations and activate different signaling pathways compare to monomer (125).

As a receptor that exhibits little or no activity when expressed alone, PAR₃ has been examined as a co-factor for other PARs. PAR₃ can modulate the activity of PAR₁ by potentiating its response to thrombin, thereby increasing endothelial barrier permeability without altering Ca²⁺ responses (30). This receptor dimer pair favors coupling to Gα13 over Gαq, whereas both pathways are similarly activated by PAR₁ monomer (30). Thus, PAR₁ may exhibit distinct signaling profiles in response to the same ligand when coupled to PAR₃.

In mouse platelets, dimerization between PAR₃ and PAR₄ leads to negative regulation of PAR₄-mediated Ca²⁺ mobilization and PKC activation without affecting Gα12/13 and Gαi activation, suggesting that PAR₄ signaling is biased away from Gαq activation when coupled to PAR₃ (126).

In human podocytes, APC cleavage of PAR₃ leads to the formation of PAR₂ and PAR₃ heterodimers, which is essential for the anti-apoptotic actions of APC (127). Although the signaling pathways that regulate this activity remain to be defined, the observation that both PAR₂ and PAR₃ activating peptides were able to produce similar effects suggest that the formation of this hetero-dimer may stimulate signaling pathways that are similar to those activated by PAR₂ monomers.

Another example of the contribution of receptor dimerization to biased PAR signaling is dimerization between PAR₁ and PAR₂. PAR₁–PAR₂ dimerization has been demonstrated in both overexpression system and endogenous expression system (128, 129). When PAR₁ forms dimer with PAR₂, the thrombin-revealed PAR₁ tethered ligand can trans-activate PAR₂ and trigger PAR₂-dependent Gαi/Rac signaling, while PAR₁-mediated Gαq and Gα12/13 signaling is switched off (130). In addition, recruitment of β-arrestin to the PAR₁–PAR₂ dimer exhibits distinct kinetic compared to each protomer, suggesting a potential alteration in β-arrestin-dependent ERK1/2 signaling (128). During the early development of sepsis, the effect of thrombin is vascular disruption whereas at the later phase of sepsis, with increasing expression of PAR₂, thrombin induces a vascular protective effect that is mediated by PAR₂/Rac1 activation (130).

Protease-activated receptors may act as co-factors or may dimerize either constitutively or in a ligand-dependent manner. The formation of dimers may play a role in organization of receptors at the cell surface, and may allosterically modulate the activation of either monomer or act as a complex that generates unique signaling outcomes.

Different proteases can lead to biased PARs activation, and PAR signaling can also be modulated by co-factors and receptor dimerization. In addition, the interaction of PARs with different G proteins also has marked influence on the outcome of protease signals. In response to a single protease, PARs are able to couple to multiple G proteins. Although how this occur is not clear, recent studies using bioluminescence resonance energy transfer (BRET) approach suggest dynamic regulation of PAR₁ and PAR₂ coupling with multiple G proteins. In Cos-7 cells, both receptors spontaneously form pre-assembled complexes with Gαi, whereas they only couple to Gα12 following ligand stimulation, and with slow kinetics (131, 132). Further investigation revealed the existence of two different PAR populations that are responsible for coupling to different Gα protein. The existence of distinct receptor populations may be interpreted as receptor clustering in different membrane microdomains such as membrane raft and caveolin-containing vesicles (131). GPCRs located in lipid raft-enriched domains may assure certain conformations of the receptor that preferentially couple to specific G proteins compared to receptor located in non-raft membrane compartment (133). It would be of great interest to examine whether divergent PARs signaling in various cell types depends on the lipid component of the membrane and whether altering membrane lipid composition leads to a shift in PAR signaling profile, thereby contributing to bias signaling. This mechanism may also account for the tissue specificity of PAR signaling.

A growing number of proteases have been identified that can cleave PARs at distinct sites, leading to diverse signals. One striking observation is the capacity of different proteases to stimulate receptor endocytosis. It is well established that cleavage of PARs at canonical activation site leads to rapid receptor internalization in a β-arrestin-dependent (PAR₂) or -independent (PAR₁) manner (96, 134). Receptor endocytosis not only contributes to signaling, but is also the first step in the degradation of activated PARs, which irrevocably terminates their ability to signal. Thus, if cleaved PARs remain at the cell surface, how is signaling regulated?

The contribution of PARs to important patho-physiological processes, including hemostasis, inflammation, pain, and proliferation, has been extensively investigated through studies of PAR-deficient mice and by use of proteases and synthetic agonists/antagonists of the canonical signaling pathways (21, 22). The capacity of certain proteases to acts as biased PAR agonists or even antagonists adds further complexity to this system, and the relevance of protease-biased signals to complex patho-physiological processes is far from clear. A major difficulty relates to the identities of the proteases that activate PARs under physiological conditions and during disease. Since proteases are regulated through post-translational control of activity (e.g., by zymogen processing and endogenous inhibitors), studies should include assessment of enzymatic activity rather than gene or protein expression. A major advance in this regard is the use of activity-based probes that covalently interact with activated proteases, allowing their localization by whole animal or cellular imaging and identification by proteomic approaches (99). This approach has been used to detect activated cathepsin S in macrophages of tumors and the inflamed colon, as well as in spinal microglial cells during colitis (67, 147). However, the use of probes is likely to reveal that multiple proteases become activated during physiological and pathological events, many of which could activate or disarm PARs. Additional information can be provided by studies of protease knockout mice or through use of selective inhibitors. However, a detailed understanding of the importance of biased signaling of PARs would probably require genetic or pharmacological strategies to selectively disrupt particular biased pathways, and such tools are currently lacking.

Although the patho-physiological importance of biased signaling of PARs is not fully understood, biased agonism could explain certain paradoxes about the patho-physiological contribution of PARs. For example, PARs can have both pro-inflammatory and anti-inflammatory roles, which may depend on the animal models, species, tissues, or the protease that drive the response. In an ovalbumin-induced model of allergic inflammation of the mouse airway, PAR₂ deletion is protective, suggesting that PAR₂ contributes to the development of immunity and to allergic inflammation of the airway (148). However, in a lipopolysaccharide-induced pulmonary neutrophilia model, PAR₂ shows a protective effect (149, 150). The underlying mechanism of these observed differences is unclear. However, differences in the repertoire of proteases that are activated in acute versus chronic inflammation, leading to distinct mechanisms of PAR signaling, could be one explanation. Lipopolysaccharide-induced pulmonary neutrophilia is an acute inflammation characterized by influx of neutrophils and activation of elastase and proteinases 3, biased agonists of PAR₁ and PAR₂. On the other hand, ovalbumin-induced inflammation is characterized by infiltration of eosinophil and macrophages, leading to activation of distinct proteases (151). Thus, the predominant active proteases for PAR₂ may be different in these two models, which could potentially activate different signaling pathways that lead to opposite responses. The contrasting pro-inflammatory and cytoprotective actions of thrombin and APC, respectively, may also be attributed to PAR₁ biased signaling. In this instance, the relative concentration of the proteases as well as the occupancy of EPCR by its ligand are critical in determining the PAR₁ signaling pathways (28).

Considerable progress has been made in defining the mechanisms by which proteases and synthetic agonists activate PARs. Proteases, peptides, and small molecules have been identified that can activate PARs by distinct mechanisms, leading to the stimulation of divergent pathways of receptor signaling and trafficking. The information derived from these studies has provided insights into the signaling pathways that are responsible for certain patho-physiological processes.

However, there are many unanswered questions about biased signaling of PARs. The ability of a PAR cleaved by different proteases or bound to various synthetic agonists to differentially signal probably arises from distinct receptor conformations. However, the structures of PARs in these different stabilized states remain to be determined. Although some proteases activate PARs by exposure of a tethered ligand, this is not always the case and the mechanism by which proteolysis per se can activate signaling is unknown. There is tantalizing evidence to suggest that biased signaling may underlie contrasting patho-physiological consequences of PAR activation, depending on the available proteases and the nature of PAR signaling. However, the proteases that are responsible for PAR activation in particular cell types in different conditions remain to be identified, and the signaling pathways that give rise to particular patho-physiological outcomes are not fully defined. Finally, very little is known about the mechanisms that regulate protease-biased signaling of PARs, particularly by those proteases that fail to promote the recruitment of β-arrestins and endocytosis of the activated receptors.

Whether protease-biased signaling of PARs can be exploited therapeutically remains an open question. The development of receptor antagonists or agonists that target disease-relevant PAR signaling pathways without affecting beneficial signaling events could provide a route for enhanced selectivity, with fewer on target side-effects. Future challenges will be to identify the primary pathways that mediate PAR-dependent physiological and patho-physiological events, and to develop receptor agonists and antagonists that selectively target these pathways. A deeper understanding of the mechanisms of initiation, regulation, and termination of protease-signaling will have profound implication in developing therapeutics for many critical conditions, including sepsis, thrombosis, inflammation, and pain processes.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.