Critical roles for thrombin in acute and chronic inflammation


Anthony Dorling, Department of Immunology, Imperial College London, Hammersmith Hospital, Du Cane Road, London W12 0NN, UK.
Tel.: +44 208 383 1707; fax: +44 208 383 1712.


Summary.  Thrombin can amplify inflammation induced by other stimuli, either through ischemia (consequent upon thrombosis), indirectly through generation of downstream mediators such as activated protein C, or directly via signals through protease activated receptors (PAR). This paper will summarize recent data from our laboratory indicating that thrombin is required to initiate CCR2-dependent leukocyte recruitment and that it is the principal determinant of the outcome after vascular injury, via PAR-1 activation of a distinct subset of smooth muscle cell progenitors. In both, tissue factor (TF) initiates thrombin generation and the thrombin acts locally, exemplifying that the initiation phase can generate autocrine or paracrine signalling molecules. Thrombin is an important constituent of innate immunity, able to amplify and modify responses to invading pathogens or tissue damage. With novel anti-thrombin therapeutics and agents to target PAR, a new understanding of the importance of thrombin may allow the development of innovative anti-inflammatory strategies.


Tissue factor (TF) is expressed constitutively in the adventitia of vessels to initiate the serine protease clotting cascade after vessel injury, generating trace amounts of thrombin. For effective hemostasis, the cascade is propagated on platelet membranes to generate large amounts of thrombin to promote fibrin polymerization [1]. Additionally, after infection or inflammation, TF is found on numerous other cells including monocytes/macrophages (MØ), dendritic cells, platelets, endothelial cells (EC) and vascular smooth muscle cells (VSMC), so each of these is capable of initiating thrombin generation.

This is relevant as these cells also express specific G-protein-coupled protease activated receptors (PARs), designated PAR 1–4 [2–5], a family of receptors activated by proteases. Thrombin can cleave PAR-1, -3 and -4 [6], FXa induces signaling via PAR-1 and -2[7], TF/FVIIa can signal through PAR-2 and activated protein C, bound to EC protein C receptor can signal through PAR-1. PAR activation on EC leads to induction of IL-6, IL-8, TGFβ, MCP-1, PDGF, ICAM-1 and P-selectin. On MØ, PAR-1 signaling by thrombin enhances adhesiveness and increases production of IL-1, TNF, IL-6, MCP-1 and IL-10, with downregulation of IL-12 secretion, whereas PAR-2 signaling additionally enhances the production of IL-1, IL-6 and IL-8.

Thus, autocrine or paracrine signaling through PAR by trace amounts of proteases may be more important than cleavage of fibrinogen and thrombosis as a mechanism to modify inflammation. Exploring the role and importance of this mechanism has been challenging until recently, for a variety of reasons, including the lack of specific reagents to target PAR signaling, the redundancy between different PARs, the fact that expression can be cell-specific, the difference between rodent and human PAR expression and the confounding problems induced by concomitant thrombosis in many model systems.

Novel reagents to inhibit coagulation proteases on specific cell membranes

Several years ago, driven by a desire to protect transplanted organs from the thrombosis that occurs in acute antibody-mediated rejection (AMR), we generated genetic constructs encoding chimeric anticoagulant proteins for targeted expression on cell membranes. The chimeras were based on human tissue factor pathway inhibitor (TFPI) and the leech anticoagulant hirudin, which act to inhibit the initiation and propagation phases of coagulation, respectively [8,9]. Artificial cDNAs were generated using PCR-by-overlap-extension and a cassette cloning strategy. For stable, constitutive expression, the extracellular, transmembrane and intracytoplasmic domains from human CD4 were included. After transfection, the fusion proteins were expressed on cell surfaces where they had a normal conformation, retained the ability to bind appropriate coagulation factors and delayed TF-initiated clotting of recalcified human plasma [10]. In a further modification, the cytoplasmic portion of CD4 was replaced with that from P-selectin, which contains a short sequence to target proteins to secretory granules. These fusion proteins were expressed exclusively within granules in resting cells, but translocation to the surface was induced by activation with appropriate stimuli [11].

Both chimeras were incorporated into transgenes along with either truncated CD31 or α-smooth muscle actin (α-SMA) promoters. Four types of founder mice were generated by microinjection and high copy number individuals used to establish breeding colonies. Four lines were generated after backcrossing onto C57BL/6, which we have called CD31-TFPI-Tg, CD31-Hir-Tg, α-TFPI-Tg and α-Hir-Tg. All four lines have a normal phenotype with no evidence of a bleeding diathesis. Northern analysis showed expression of the transgene in all tissues examined and immunohistology of organs confirmed protein expression limited to EC (CD31 promoter) or VSMC in arteries and veins (α-SMA promoter) [12,13]. Resting EC purified from the CD31-Tg mice expressed no cell surface human TFPI or hirudin fusion proteins, but showed rapid expression after translocation of Weibel–Palade bodies following exposure to inflammatory cytokines. Platelets and MØ showed the same pattern of expression, as expected from the knowledge that these also express CD31.

Thrombin in acute vascular inflammation – models using CD31-Tg mice

To modify AMR, it was important to define whether manipulating the donor organ endothelium would influence intravascular thrombosis. As TF and thrombin were known to play a role in severe sepsis, we investigated the contribution of EC, platelets and MØ in triggering and propagating intravascular thrombosis in murine lipopolysaccharide (LPS) endotoxaemia as a prelude to the transplantation experiments.

Using a protocol that delivered a thrombotic phenotype, we confirmed that wild type (WT) mice developed severe thrombocytopenia, consumptive coagulopathy and extensive thrombosis in all tissues within 6 h of LPS injection. In contrast, CD31-TFPI-Tg and CD31-Hir-Tg mice developed mild clotting derangements and showed little intravascular thrombosis. Bone marrow (BM) reconstitution experiments to generate mice expressing fusion proteins on either EC or platelets and leukocytes revealed that EC expression significantly protected against the coagulopathy and thrombosis, whereas expression on platelets and leukocytes did not. Antibody inhibition experiments in the CD31-Hir-Tg mice confirmed that fusion protein expression on EC was equivalent to that on platelets and leukocytes, indicating a genuine qualitative advantage to expressing anticoagulants on EC. We concluded that activated EC, rather than circulating elements, determined the amount of thrombin generated during inflammation. These data suggested that targeting anticoagulants to the endothelium of donor organs might prevent intravascular thrombosis in AMR.

We tested this by transplanting mouse hearts into rats. This is a well-characterized model of xenoegeneic AMR, in which intragraft thrombosis is particularly prominent. WT mouse hearts were rejected within three days whereas hearts from either of the two CD31-Tg lines were resistant to thrombosis and survived for 7 days (P < 0.0001 c.f. WT) [14], before they were rejected by T cells; administration of ciclosporin A gave long-term (>100 days) survival without evidence of chronic rejection.

These results were very surprising, because the rat recipients quickly generated new antibodies against mouse hearts and these were deposited, with complement, on the graft EC. Our knowledge of AMR suggested that these hearts should have been severely damaged, but instead they were completely protected by the fusion proteins. This suggested that thrombin was playing a critical, hitherto unrecognized role in the pathophysiology of AMR.

To address how much of the survival advantage of the Tg hearts was due to inhibition of thrombosis, we transplanted WT hearts into rats depleted of ≥95% of their plasma fibrinogen (using the snake venom protein ANCROD). Despite no evidence of thrombosis or systemic coagulopathy, these hearts survived only slightly longer than in unmanipulated rats (mean survival of 3.8 vs. 2.8 days in rats with normal fibrinogen levels; P = 0.017). This suggested the survival advantage of CD31-Tg hearts was not due solely to inhibition of thrombosis.

We observed that AMR after ANCROD was accompanied by NK cell and MØ infiltration and showed that NK cells were causing rejection under these conditions [15]. In contrast, rejected Tg hearts contained few NK cells or MØ [14]. To investigate further, we determined that NK and MØ infiltration was dependent on the production of mouse MCP-1 (i.e. by the heart tissue), as confirmed by the lack of NK or MØ infiltration into hearts from MCP-1 deficient mice [16]. Using a combination of agonists or antagonists for PAR, we then showed that mouse MCP-1 generation was dependent on activation of PAR-1, specifically that expressed on the heart graft, as hearts from PAR-1-deficient mice were rejected as slowly as the CD31-Tg hearts. This led us to conclude that thrombin was needed to establish tissue chemokine gradients to initiate leukocyte recruitment. Although it was already known that PAR-1 could cause secretion of MCP-1, this was the first demonstration that it was necessary for MCP-1 secretion in vivo.

We confirmed that thrombin and PAR-1 played similar roles in a non-transplant model of thioglycollate-induced inflammation, characterized by a MCP-1-dependent, MØ-rich peritonitis. Importantly, in this model, thrombin was also necessary for MCP-3 and -5 generation, both of which are additional ligands for CCR2.

Thrombin in chronic vascular inflammation – models using α-Tg mice

As transplant immunologists, our interest in chronic vascular inflammation stems from a desire to understand the development of transplant arteriosclerosis, a form of chronic rejection. This is characterized by intimal hyperplasia (IH) and remodeling, leading to progressive obliteration of the vessel lumen and chronic ischemia. Because the initiating injury is the alloimmune response, strategies to suppress immune activation or interfere with immune effector mechanisms have been pursued with variable success. However, clinically arteriosclerosis remains a problem despite ‘optimized’ immunosuppression.

TF and PAR play a role in the development of blood vessels during vasculogenesis and angiogenesis [17], as illustrated by TF-deficient mice, which develop fragile capillaries, fail to recruit pericytes and die in utero due to widespread haemorrhage [18]. Fifty percent of mice lacking PAR-1 suffer a similar fate [19]. TF and thrombin are also involved in the development of IH, though the precise mechanisms by which they act in vivo remain ill defined. Recent studies in mice and humans have shown that significant numbers of VSMC in IH lesions are BM-derived, so there is an increasing appreciation that IH may involve dysfunctional repair by vascular progenitors (VP) recruited to sites of injury.

VPs are cells mobilized into the circulation and recruited to sites of vascular damage that are capable of differentiating into EC or VSMC. Characteristically they express vascular endothelial growth factor (VEGF) receptor-2 (VEGFR-2). Human EC progenitor cells (EPC) are the best characterized. They can be outgrown from peripheral blood mononuclear cells under appropriate conditions and comprise two main types [20], an ‘early’ EPC which derive from CD34 CD14+ monocytes and ‘late’ EPC which develop from a CD34+ CD133+ CD14 cell that is present at low frequency in peripheral blood [21,22]. Recently, ‘early’ and ‘late’ types of VSMC progenitors (VSMPC) have been identified after culture in platelet derived growth factor (PDGF) [23,24]. In mice, these phenotypes are ill defined, but EC and VSMC derive from a common progenitor [25], with the latter developing through an ‘intermediate’ mixed phenotype stage defined by co-expression of VEGFR-2, EC markers and VSMC-associated genes [26].

This is relevant to IH, as VPs recruited into the neointima do not differentiate into quiescent EC and VSMC. Rather, they have a ‘modulated’ phenotype, characterised by increased proliferation and synthetic functions [27] and expression of proteins normally restricted to EC, such as vWF or P-selectin [28,29]. These cells sometimes form a luminal ‘pseudoendothelium’ [30], implying a problem with recruitment of EPC.

Working in a model of wire-induced carotid artery injury, we confirmed that neointimal cells in WT mice were wholly derived from circulating CD34+ cells and that they had a mixed phenotype. Additionally, the cells were highly inflammatory, expressing markers such as TF and E-selectin [13]. We were struck by the resemblance between these cells and the ‘intermediate’ progenitors mentioned above.

In contrast injury in α-TFPI-Tg and α-Hir-Tg mice was followed by repair back to a pre-injured state, which included regeneration of a new, quiescent endothelium. This type of repair was also seen when WT mice were reconstituted with BM from Tg mice or injected with Tg CD34+ VPs at the time of injury. Conversely, Tg mice reconstituted with WT BM showed a WT-like neointima. Responses to injury were the same in both strains, indicating that thrombin generated by TF caused the changes seen in WT mice. Follow-up studies have confirmed that the outcome after wire-induced injury is entirely dependent on the actions of thrombin on CD34+ cells illustrated best by the fact that injured WT arteries show regenerative repair if injected, at the time of injury, with CD34+ cells that have been pre-treated with a PAR-1 antagonist [31].

A review of the literature gave little insight into what thrombin was doing to the VP. The few studies that have looked at the influence of thrombin on VP have showed that it promotes differentiation and enhances the pro-angiogenic properties of EPC, neither of which appeared directly relevant to our findings [32–35]. Therefore, we studied mouse CD34+ cells mobilized post-injury. All expressed PAR-1, PAR-2 and PAR-4 along with procoagulant TF, so that in vitro they promoted accelerated clotting of re-calcified plasma. We characterized EPC and VSMPC subpopulations within VP, based on the expression of CD31 and α-SMA, respectively, and found that approximately one-third of the VSMPC (approx. 3% of circulating CD34+ cells) expressed CD31. As they resembled the intermediate stage VP described in the embryonic stem cell literature (see above), we called them mixed phenotype intermediates (MPI). Our most striking finding was that in vitro, thrombin caused the survival and outgrowth of MPI cells from CD34+ VP, even when growth factors such as VEGF or PDGF were present [31].

Subsequently, we have compared the mobilization of VP post-injury in WT, α-TFPI-Tg and CD31-TFPI-Tg mice and this has been highly informative (manuscript submitted). VSMPC made up approximately 9% of circulating CD34+ cells in each strain. As in WT, the CD31-TFPI-Tg had a minority population of MPI and these cells expressed cell-surface TFPI. In α-TFPI-Tg mice, the most important finding was that there were no circulating MPI, but all VSMPC expressed cell-surface TFPI. However, when CD34+ cells from these mice were cultured in thrombin, outgrowth of MPI was dominant, as in WT mice. This suggested that the TFPI in these mice was preventing the appearance of MPI. This was confirmed when α-TFPI-Tg mice were given anti-TFPI antibody post-injury, as MPI appeared in the circulation at the expected frequency. The same inhibitory anti-TFPI antibody had no effect on the proportion of MPI in CD31-TFPI-Tg mice. Therefore, this was clear evidence that the presence of MPI in the circulation was dependent on TF expression by VSMPC.

To explain our findings, we have proposed that the circulating MPI exist in a dynamic equilibrium with more mature VEGFR-2-negative VSMPC. Our working hypothesis is that in WT and CD31-TFPI-Tg mice, thrombin generated on the surface of VSMPC (emerging from MPI during the differentiation process), signals through PAR-1 to sustain the expression of VEGFR-2 (and CD31), thereby transiently stabilizing the MPI phenotype. In contrast, in α-TFPI-Tg mice, emerging VSMPC undergo cellular differentiation, in the absence of a signal through PAR-1, so that MPI are undetectable.

We hypothesise that this equilibrium is important for the evolution of IH vs. regenerative repair. All our data indicates that IH occurs only when MPI are in the circulation, that the cells proliferating in the neointima are MPI and that thrombin is driving neointimal MPI proliferation. Observations in WT indicate that the first VP recruited to the injured vessel wall, visible on day 5 post-injury, have the MPI phenotype. This may indicate that these cells are preferentially recruited, in preference to EPC or more mature VSMPC, or it may indicate that VEGFR-2-negative VSMPC (which exist at higher frequency than MPI in the circulation) adopt the MPI phenotype under the influence of thrombin after recruitment. Implicit in this hypothesis is the idea that thrombin exerts the dominant influence on these cells, overriding differentiation cues provided by growth factors in the vessel wall. We have observed that the first VP recruited to the injured vessel wall in α-TFPI-Tg mice are mature VSMPC and these cells, incapable of making thrombin by virtue of their TFPI expression, appear responsive to the differentiation cues promoting complete regenerative repair.

Summary and conclusions

Our work in acute vascular inflammation has two implications. First, this pathway could be targeted therapeutically to limit leukocyte recruitment, especially in autoimmune diseases or after transplantation. Second, it may point to the selection pressures that have helped maintain, during mammalian evolution, the switch from an anticoagulant to a procoagulant endothelium during sepsis, a change that is potentially hazardous for the survival of individuals and hence for the species. There is already evidence that fibrin deposition is important to isolate and limit the spread of some pathogens [36], the importance of which is illustrated by the fact that plasminogen activators act as pathogenicity factors for some bacteria [37]. From our data, we propose that efficient chemokine production, efficient recruitment of leukocytes and effective clearance or containment of infection may be additional benefits that have helped maintain this response.

Our work in chronic inflammation and progenitor cell-mediated repair sheds new light on the long-established link between TF, thrombin and IH. The demonstration that thrombin maintains an equilibrium between two smooth muscle progenitor subsets is entirely novel and lays the foundation for further work to determine the influence that other inflammatory mediators have on these cells. This is potentially important for those investigators pursuing therapeutic strategies based on injecting large number of progenitors to regenerate or repair diseased tissues, as it will be vital to understand how the environment, into which the cells are recruited, influences phenotype and function.


Dr Chen performed the majority of laboratory work alluded to in this paper. Dr Dorling planned the work and wrote the manuscript.

Disclosure of Conflict of Interests

Imperial College London has a patent covering the use of membrane-targeted anticoagulants in xenotransplantation.