The serine protease thrombin regulates platelet aggregation, endothelial cell activation and other important responses in vascular biology. Thrombin's actions on cells raise an intriguing question. How does thrombin, a protease, act like a traditional hormone and elicit cellular responses? Understanding thrombin signalling will provide insight into haemostasis and inflammation, and, probably, embryonic development. Because thrombin and platelets have a central role in myocardial infarction and other pathological processes, understanding how thrombin activates platelets and other cells may suggest new strategies for therapy.
Protease-activated receptors (PARs) provide one answer to the question of how thrombin produces signals. PARs are G-protein-coupled receptors that use a fascinating mechanism to convert an extracellular proteolytic cleavage event into a transmembrane signal: these receptors carry their own ligands, which remain cryptic until unmasked by receptor cleavage. Recent advances in our understanding of PARs provide a working model for thrombin signalling in human platelets, reveal a surprising variation in the paradigm for PAR activation and evoke testable hypotheses regarding the roles of PARs in thrombosis and inflammation. It is therefore timely to review progress in our understanding of thrombin signalling and PARs in the context of vascular biology.
When and where is thrombin generated?
Thrombin is the main effector protease of the coagulation cascade, a series of zymogen conversions that is triggered when circulating coagulation factors contact tissue factor. Tissue factor is a type-I integral membrane protein that functions as an obligate cofactor for activation of zymogen factor X by factor VIIa. Factor Xa (with the assistance of cofactor factor Va) then converts prothrombin to active thrombin. Other zymogen conversions provide both amplification and negative feedback loops that regulate thrombin production. Thrombin is short lived in the circulation and, in the context of a normal endothelium, its actions tend to terminate its production. Thus thrombin is thought to act near the site at which it is produced1, 2.
Tissue factor is expressed by epithelial cells, macrophages and other cell types that are normally separated from blood and circulating coagulation factors. Classically, thrombin generation is triggered when disruption of vascular integrity allows plasma coagulation factors to contact extravascular tissue factor. Thus the coagulation cascade provides a mechanism for converting mechanical information in the form of tissue damage and/or vascular leak into biochemical information in the form of the active protease thrombin.
Tissue factor is expressed at low levels on circulating monocytes and leukocyte-derived microparticles. These sources of intravascular tissue factor can be tethered to activated platelets and endothelial cells and concentrated in this way at sites of injury or inflammation3, 4. This alters the local balance between activation and inhibition of the coagulation cascade and triggers thrombin production. Tissue factor is also expressed at low levels by cytokine-stimulated endothelial cells, perhaps to promote thrombin generation at sites of inflammation5.
What are thrombin's actions on cells?
Thrombin converts circulating fibrinogen to fibrin monomer, which polymerizes
to form fibrin, the fibrous matrix of blood clots. Thrombin also has a host
of direct actions on cells6 (Fig. 1).
It triggers shape change in platelets and the release of the platelet activators
ADP, serotonin and thromboxane A2, as well as chemokines and growth factors.
It also mobilizes the adhesion molecule P-selectin and the CD40 ligand to
the platelet surface7, 8 and activates the integrin 
IIb/
3
(ref. 9). The latter binds fibrinogen and von
Willebrand factor (vWF) to mediate platelet aggregation1. Thrombin
also triggers expression of procoagulant activity on the platelet surface,
which supports the generation of additional thrombin10. In cultured
endothelial cells, thrombin causes release of vWF11, the appearance
of P-selectin at the plasma membrane11, and production of chemokines
— actions thought to trigger binding of platelets and leukocytes to
the endothelial surface in vivo12, 13. Endothelial cells
also change shape and endothelial monolayers show increased permeability in
response to thrombin14 — actions predicted to promote
local transudation of plasma proteins and oedema15. Thrombin
can also regulate blood vessel diameter by endothelium-dependent vasodilation;
in the absence of endothelium, thrombin acting on smooth muscle cells evokes
vasoconstriction. In cultures of fibroblast or vascular smooth muscle cells,
thrombin regulates cytokine production and is mitogenic, and in T lymphocytes
it triggers calcium signalling and other responses. These cellular actions
suggest that thrombin connects tissue damage to both haemostatic and inflammatory
responses and perhaps even to the decision to mount an immune response. They
also raise the possibility that regulation of endothelial and other cell types
by thrombin might have a role in leukocyte extravasation, vascular remodelling
and/or angiogenesis in contexts other than tissue injury. The recent characterization
of receptors that mediate thrombin signalling provides an opportunity to test
these ideas.
Figure 1: The actions of thrombin on blood cells and blood vessels.
Thrombin is a multifunctional serine protease generated at sites of vascular injury. It is arguably the most effective agonist for platelet activation. Thrombin also elicits a host of responses in the vascular endothelium, including shape and permeability changes, mobilization of adhesive molecules to the endothelial surface and stimulation of autocoid (small molecule mediators such as prostaglandins and platelet-activating factor) and cytokine production. Thrombin is chemotactic for monocytes and mitogenic for lymphocytes and mesenchymal cells.
High resolution image and legend (37K)How does thrombin talk to cells?
Thrombin signalling is mediated at least in part by a small family of G-protein-coupled PARs16. PAR1, the prototype of this family, is activated when thrombin cleaves its amino-terminal extracellular domain (exodomain) at a specific site17, 18. This cleavage unmasks a new N terminus that then serves as a tethered ligand, binding intramolecularly to the body of the receptor to effect transmembrane signalling17 (Fig. 2). Intermolecular ligation of PARs can occur but, not surprisingly, seems to be less efficient than intramolecular ligation19, 20. Synthetic peptides that mimic the tethered ligand of PAR1 activate the receptor independently of protease and receptor cleavage17. Thus PAR1 can be viewed as a peptide receptor that carries its own ligand. The latter remains silent until activated by cleavage of the PAR1 N-terminal exodomain. PAR1–thrombin interactions are accounted for by sequences surrounding the cleavage site within the N-terminal exodomain of the receptor, and cleavage at that site is both necessary and sufficient for PAR1 activation. Indeed, PAR1 mutants bearing enteropeptidase or trypsin cleavage sites in place of the thrombin cleavage site conferred the capacity for enteropeptidase or trypsin signalling, respectively, in heterologous expression systems. Thus the role of thrombin in PAR1 activation seems to be simply to unmask the receptor's tethered ligand6, 16.
Figure 2: Mechanism of PAR1 activation.
Thrombin (large green sphere) recognizes the N-terminal exodomain of the G-protein-coupled thrombin receptor PAR1. This interaction uses sites both N-terminal (small blue sphere) and C-terminal (small pink oval) to the thrombin cleavage site. The latter sequence resembles the C-terminal tail of the thrombin inhibitor hirudin and binds to thrombin in an analogous manner. Thrombin cleaves the peptide bond between receptor residues Arg 41 and Ser 42. This serves to unmask a new N terminus, beginning with the sequence SFLLRN (diamond) that functions as a tethered ligand, docking intramolecularly with the body of the receptor to effect transmembrane signalling. Synthetic SFLLRN peptide, which mimics the tethered ligand sequence, will function as an agonist independently of receptor cleavage. Thus PAR1 is, in essence, a peptide receptor that carries its own ligand, the latter being active only after receptor cleavage.
High resolution image and legend (23K)PAR1 can couple to members of the G12/13, Gq and Gi families and hence to a host of intracellular effectors (see Box 1). Such pluripotent signalling fits well with the known effects of thrombin on platelets, endothelial and other cells.
Irreversible activation and disposable receptors
The mechanism by which PAR1 is activated is striking in several ways. Cleavage of the receptor is irreversible, and the 'peptide agonist' unmasked by cleavage remains tethered to the receptor. Moreover, thrombin is an enzyme, implying that one thrombin molecule might cleave and activate several molecules of PAR1. This raises several important and related questions. Given the irreversibility of the activation mechanism, how is PAR1 signalling terminated? Given that thrombin is an enzyme, how does PAR1 mediate responses that are dependent on thrombin concentration? And, given tethering of ligand to receptor, will development of drugs that block PAR1 signalling be possible? There are strong hints of interesting answers16.
Like other G-protein-coupled receptors, activated PAR1 is rapidly uncoupled from signalling and internalized by phosphorylation-dependent mechanisms. Instead of recycling, it is then delivered to lysosomes for degradation with remarkable efficiency. Some PAR1 molecules that escape this fate appear to return to the cell surface with tethered ligand in an inactive state. Thus PAR1 is used once and then discarded. In fibroblasts and endothelial cells, responsiveness to thrombin is maintained by delivery of new PAR1 to the cell surface from a preformed intracellular pool. By contrast, in human megakaryocyte-like cell lines, recovery of PAR1 signalling requires new protein synthesis. Perhaps there is no need for a special resensitization mechanism in platelets. Once activated and incorporated into a clot, they are presumably not reused.
The rapid shut-off of activated PAR1 provides a plausible answer to how PAR1 mediates graded responses that vary with thrombin concentration21. Each cleaved receptor is active for a finite interval and therefore triggers production of some average 'unit' of second messenger (for example, inositol trisphosphate). Because the second messenger is itself cleared, the level of second messenger achieved is proportional to the rate at which receptors are cleaved and activated, and hence to thrombin concentration. Together with the relatively low avidity of the interaction between PAR1 and its tethered ligand, this mechanism makes us optimistic about the possibility of developing useful PAR1 antagonists. It suggests that, in order to attenuate cellular responses an antagonist need only delay PAR1 activation. Indeed, effective antagonists structurally related to the PAR1 tethered ligand have been generated22.
A family of PARs
Four PARs are known in mouse and human. Human PAR1 (refs 17, 18), PAR3 (ref. 23), and PAR4 (refs 24, 25) can be activated by thrombin. PAR2 is activated by trypsin26 and tryptase27 as well as by coagulation factors VIIa and Xa28, but not by thrombin. It is certainly possible that these receptors mediate responses to other proteases or even to peptide ligands in vivo. Indeed, cofactors that localize proteases to the cell surface and modulate their activity can help orchestrate PAR activation28, 29. Thus the full repertoire of proteases that signal through PARs remains to be defined.
It is worth noting that the N-terminal exodomains of PAR1 and PAR3 have thrombin-interacting sequences both N- and carboxy-terminal to the thrombin-cleavage site (Fig. 2). The C-terminal sequence resembles the C-terminal tail of the leech anticoagulant hirudin and, like the latter, binds to thrombin's fibrinogen-binding exosite; this interaction is important for receptor cleavage at low concentrations of thrombin6. The presence of such extended thrombin-interacting sequences in PAR1 and PAR3 is consistent with the notion that these receptors evolved to mediate responses to thrombin rather than to other proteases. A hirudin-like sequence is not evident in PAR4, and PAR4 indeed requires higher thrombin concentrations for activation than the other receptors24, 25.
PARs and platelet activation
Recent studies have provided a working model of thrombin signalling in human and mouse platelets and reveal both curious species differences and a variation on the paradigm for PAR activation. The model frames important questions regarding strategies for drug development and suggests that answers, at least in principle, can be derived from studies of PAR-knockout mice (Fig. 3).
Figure 3: Thrombin signalling in human and mouse platelets.
Human platelets express PAR1 and PAR4, and available data suggest that these receptors can independently mediate thrombin signalling — PAR1 at low and PAR4 at high thrombin concentrations. By contrast, mouse platelets express PAR3 and PAR4 and, surprisingly, it seems that mPAR3, rather than itself mediating transmembrane signalling, functions as a cofactor that supports cleavage and activation of mPAR4 at low thrombin concentrations.
High resolution image and legend (20K)Human platelets
Human platelets express PAR1 and PAR4, and activation of either is sufficient to trigger platelet secretion and aggregation17, 24, 30. Antibodies to the thrombin-interaction site in PAR1 blocked receptor cleavage and platelet activation at low, but not high, concentrations of thrombin30, 31, 32. By contrast, PAR4-blocking antibodies by themselves had no effect on platelet activation by thrombin, but when these were combined with PAR1 blockade, platelet activation was markedly inhibited, even at high concentrations of thrombin30. These results suggest that PAR1 mediates activation of human platelets at low thrombin concentrations and that, in the absence of PAR1 function, PAR4 can mediate platelet activation but only at high thrombin concentrations (Fig. 3). Given that PAR1 does normally function in human platelets, what does PAR4 contribute? It is possible that PAR4 simply provides 'back up' in an important system. It is equally possible that PAR4, which lacks a thrombin-binding hirudin-like sequence, mediates responses to proteases other than thrombin. In this regard, platelet activation by cathepsin G33, a granzyme released by activated neutrophils, seems to be mediated by PAR4 (ref. 34). PAR4 may make other unique contributions to platelet function. Indeed, PAR4 is activated and shut off more slowly than PAR1, and the tempo of calcium signalling in response to thrombin in human platelets appears to represent the sum contribution of both receptors35.
It is worth noting that thrombin binds to the platelet surface glycoprotein
GPIb36, part of a protein complex that also binds vWF
and P-selectin37. The role of this binding to GPIb is
unclear. It is possible that GPIb serves as a cofactor that modulates
thrombin's ability to cleave other platelet surface or plasma proteins, or
that GPIb has a more direct signalling role. Studies in knockout mice
will soon reveal whether the known PARs account for thrombin signalling in
platelets.
The presence of PAR1 and PAR4 in human platelets raises an important question regarding the development of antithrombotic drugs. Given that activation of PAR4 requires relatively high concentrations of thrombin, might inhibition of PAR1 be sufficient to prevent thrombosis? Or will inhibition of both PAR1 and PAR4 be required? In the absence of drugs that might be used to address this question in relevant animal models, answers in the near term are likely to come from PAR-deficient mice.
Mouse platelets
In contrast to human platelets, mouse platelets express PAR3 and PAR4 (ref. 25). Indeed, PAR1-activating peptides activate human but not murine platelets38, 39, 40, and knockout of mouse PAR1 (mPAR1) had no effect on thrombin signalling in mouse platelets but abolished thrombin signalling in fibroblasts40. These observations triggered a search for other thrombin receptors in mouse platelets and led to the identification of PAR3 (ref. 23). Expression of human PAR3 cDNA in COS cells or Xenopus oocytes conferred phosphoinositide hydrolysis in response to low concentrations of thrombin, and in situ hybridization using a mouse PAR3 probe detected mPAR3 mRNA in mouse megakaryocytes23. Knockout of mouse PAR3 revealed PAR3 to be necessary for activation of mouse platelets at low but not high concentrations of thrombin. Persistent thrombin signalling in PAR3-deficient mouse platelets was attributable to mPAR4 (ref. 25). On the face of it, these data conjured up a dual-receptor model analogous to that described for human platelets. In mouse platelets, PAR3 mediated activation at low thrombin concentrations and, in the absence of PAR3 function, PAR4 triggered activation at high thrombin concentrations25.
Subsequent characterization of the mouse homologue of PAR3 presented a
paradox. In spite of strong evidence that mPAR3 was necessary for mouse platelet
responses to low concentrations of thrombin, expression of mPAR3 cDNA
in heterologous expression systems failed to confer the property of thrombin
signalling. Resolution of this paradox came in the form of an interesting
variation on the mechanism of PAR activation29. Whereas expression
of mPAR3 in COS cells did not, by itself, confer thrombin signalling, co-expression
of mPAR3 with mPAR4 reliably enhanced both mPAR4 cleavage and signalling at
low concentrations of thrombin compared with mPAR4 alone. When tethered to
the plasma membrane, the N-terminal exodomain of mPAR3 was sufficient for
this activity, and the thrombin-interacting sequences within this domain were
necessary. Thus, it appears that mPAR3 does not by itself mediate transmembrane
signalling, but instead functions as a cofactor for cleavage and activation
of mPAR4 at low thrombin concentrations — a curious form of G-protein-coupled
receptor interaction in which one receptor acts as an accessory protein that
aids 'ligation' of another (Fig. 3)29. This model predicts that thrombin signalling in mouse platelets
is dependent on PAR4. A definitive test of this prediction will be possible
with platelets from PAR4-deficient mice, which should be unresponsive to thrombin
despite the presence of mPAR3 (Fig. 3). Whether mPAR3
and mPAR4 heterodimerize, and whether other similar PAR–PAR interactions
will be found, is not known. There is no evidence to suggest an analogous
interaction between human-PAR1 (hPAR1) and hPAR4 or between the thrombin-binding
site GPIb in human platelets and hPAR4.
Utility of mouse models
Despite species differences, mouse models may provide important hints on how to inhibit thrombin signalling in human platelets. The model in Fig. 3 makes several predictions. First, PAR3-deficient mouse platelets are analogous to PAR1-inhibited human platelets — both rely on PAR4 for thrombin signalling. If PAR3 deficiency protects against thrombosis in mouse models, PAR1 inhibition may be worth investigating as an antithrombotic strategy in humans. Second, PAR4-deficient mouse platelets may be analogous to human platelets in which both PAR1 and PAR4 function are blocked — thrombin signalling should be absent in both. Thus a PAR4-deficient mouse might provide an opportunity to define the importance of thrombin-triggered platelet activation in haemostasis and thrombosis.
PARs in endothelial activation
PAR1 seems to be the major mediator of thrombin signalling in vascular endothelial cells in both mice and humans, and most of the actions of thrombin on endothelial cells described above have been reproduced using PAR1 agonist peptide. Endothelial cells also express PAR2, which may mediate responses to tryptase released from mast cells27 or to coagulation factors VIIa or Xa28 in this setting.
What are the functions of endothelial PARs? One might imagine the following scenario. Tissue injury, whether by trauma, infection or metabolic or inflammatory mediators, triggers local generation of coagulation proteases and/or release of mast-cell tryptase which, by way of PARs, activate endothelial cells (Fig. 1). The activated endothelial surface in turn promotes adhesion and rolling of platelets and leukocytes as well as leakage of plasma proteins to the extravascular space. Thrombin also triggers endothelial production of platelet-activating factor, a potent neutrophil activator41, as well as the interleukins IL-6 and IL-8 (ref. 42). Thus PARs may link tissue injury to endothelial responses that recruit platelets, leukocytes and effector proteins to examine the locale for damage or infection.
The possibility of a positive feedback loop in which thrombin triggers endothelial responses that beget additional thrombin generation and endothelial activation is clear (Box 2). Undamped, such a system would trigger intravascular thrombosis and, perhaps, local tissue damage from leukocyte products. On a microscopic scale, this might be beneficial for walling off infection. However, disseminated intravascular coagulation with microvascular thrombosis and tissue infarction can occur in the setting of a strong systemic inflammatory stimulus (for example, sepsis) and/or deficiencies that disinhibit thrombin production (for example, protein C deficiency, protein S deficiency and the presence of factor V Leiden).
PAR-deficient mice provide an opportunity to test the role of endothelial PARs in inflammatory responses. The species differences in PAR expression between mouse and human may be fortunate in this regard. PAR1 appears to be the major thrombin receptor in endothelial cells in both species. Because PAR1 is expressed in human platelets but not in those of mice, however, PAR1-deficient mice offer an opportunity to abolish thrombin signalling in endothelial cells without perturbing platelet signalling. This can help define the contribution of endothelial activation by thrombin to thrombosis and inflammation. Intriguingly, PAR1 deficiency did protect against leukocyte infiltration and renal damage in a mouse model of antibody-mediated glomerulonephritis43.
Thrombin signalling in embryonic development
Approximately 50% of PAR1-deficient mouse embryos die at mid-gestation40. PAR1 does not act in mouse platelets, so this and other recent studies in knockout mice suggest that signalling by coagulation proteases and PARs may have an important role in embryonic development that is unrelated to haemostasis in any usual sense16, 44. The available data suggest that PAR1 and coagulation factors may contribute to normal blood vessel development. This is exciting, in that it may point to a new role for the 'coagulation' cascade — one of monitoring and regulating new blood vessel formation.
Future directions
The studies described above raise a host of questions regarding the molecular mechanisms of PAR activation and protease signalling. How general are PAR–PAR interactions and is receptor oligomerization involved? To what extent do cofactors increase the diversity of proteases to which cells can respond through PARs? Will the known PARs account completely for signalling by thrombin and other coagulation proteases, or will new PARs and/or other mechanisms be identified? Because PAR1 and PAR4 are the only PARs known to mediate transmembrane signalling in response to thrombin in the mouse, the presence or absence of residual thrombin signalling in cells from mice deficient in both PAR1 and PAR4 will be telling.
Important questions also remain regarding the roles of PARs in physiology and disease. For example, thrombin is a powerful activator of platelets and it is clear that both thrombin and platelets are important for haemostasis and thrombosis. But in addition to activating platelets, thrombin triggers fibrin formation, and platelets can be activated by a host of other mechanisms. Thus the relative importance of thrombin activation of platelets in haemostasis and thrombosis is unknown. As discussed above, the phenotype of a PAR4-knockout mouse may be enlightening in this respect. Similarly, a panoply of signalling systems and cell types orchestrates inflammatory responses, and efforts to define the relative contribution of PARs are just beginning.
The answers to these questions will influence decisions as to whether or not PARs are rational drug targets. Blockade of platelet activation by thrombin might well be a useful antithrombotic strategy. Attenuating inflammatory responses by blocking PAR signalling in endothelial cells is a more novel and untested notion, and affecting new blood vessel formation by the same route is more speculative still. Results from mouse models may stimulate the development of drugs to further explore these ideas.
