Introduction: are natural anticoagulants candidates for modulating the inflammatory response to endotoxin?

Recent studies have focused on the potential role of natural anticoagulants in modulating host responses to endotoxin. The natural anticoagulants that have drawn the greatest attention are antithrombin, protein C, and tissue factor pathway inhibitor (TFPI). Interest in this area is spurred by clinical and basic observations. On the clinical side, there is a correlation between low levels of protein C and antithrombin and a negative clinical outcome. The situation with TFPI is much more difficult to interpret because the majority of TFPI is associated with the blood vessel1 and changes in plasma levels can reflect altered distribution between the vessel wall and plasma. On the basic side, animal models of septic shock employing either endotoxin or Escherichia coli, usually performed by intravenous infusion of these agents, have shown that antithrombin,2 protein C/activated protein C,3or TFPI4 can reduce the frequency of lethal responses in nonhuman primates and other animals. In general, these studies have found that, in addition to dampening the disseminated intravascular coagulation (DIC) associated with the endotoxin challenge, elevation of the natural anticoagulant levels also decreases the inflammatory response including IL-6 and IL-8 levels in the case of TFPI (see Creasey et al4) and antithrombin (see Minnema et al, this issue).

Contrasting these reports are studies of human volunteers administered low doses of endotoxin: there was an increase in their coagulation response, but the administration of TFPI failed to alter the IL-6 levels or those of other markers of inflammation (see de Jonge et al, this issue). This discrepancy between the response of human volunteers and experimental animals to endotoxin infusion raises several questions that warrant some discussion. For instance, are the differences simply a matter of species, or are there differences in experimental design that could account for the different outcomes? Before dealing with this issue, it is worth reviewing a rapidly growing literature on the role of coagulation factors in activating cells and eliciting inflammatory responses. It is in this context that the potential role of natural anticoagulants in modulating these responses is most easily understood. When considering any in vivo results, it is important to bear in mind that altering the concentration of antithrombin or TFPI will diminish thrombin formation in response to endotoxin and that this will diminish the activation of protein C. Thus if activated protein C has antiinflammatory functions not replicated by the other inhibitors, elevation of the other natural anticoagulants could have unexpected negative effects on the regulation of both the coagulation and inflammation processes.

How the natural anticoagulants might impact the inflammatory system remains an active area of investigation, but several observations help provide a framework by which this might be accomplished. In part, the answer may lie in the recent observations that factor VIIa, factor Xa, and thrombin can all activate cells directly, probably mediated in large part by the cleavage of cell surface protease activated receptors (PARs). The mechanism of activation of PARs is reviewed in Coughlin.5 Briefly, cleavage of these receptors generates a new N terminus that serves as a “tethered” ligand to activate these 7 transmembrane G protein-coupled receptors. Because each of the above coagulation enzymes appears to accomplish cell activation by different mechanisms, their mechanisms of cell signaling will be discussed separately.

Recently, several studies have shown that the tissue factor–factor VIIa complex can activate cells,6-8 causing a Ca²⁺ influx resulting in the activation of map kinase, c-Jun N terminal kinase, and the early growth response gene-1 (egr-1).9 This process potentially contributes to inflammatory mediator release from cells. Cell activation requires the factor VIIa to be proteolytically active;8 hence cell activation is not mediated by direct binding of factor VIIa to tissue factor. Instead, the cell activation appears to be mediated in part by activation of a protease activated receptor, either PAR 2 or a closely related receptor.9 

Other unidentified factors apparently play a role in signaling since cellular signaling could not be reconstituted by cotransfection of cells with tissue factor and PAR 2 alone. Signaling through PAR 2 has particular relevance to a potential role in inflammation because PAR 2 is induced by TNF α and IL-1 in endothelial cells,10potentially augmenting the ability of the tissue factor–factor VIIa complex to activate the endothelium either through endothelial cell tissue factor or tissue factor on adherent monocytes. The relevance of tissue-factor-mediated signaling to monocyte/macrophage activation and the potential for augmenting cytokine elaboration have recently been demonstrated by showing that the tissue factor–factor VIIa complex could elicit a variety of proinflammatory responses in macrophages, including reactive oxygen species, and induction of MHC class II and adhesion receptors.7 The macrophage activation required both the active site of factor VIIa and the cytoplasmic tail of tissue factor.

Inhibitors that block factor VIIa proteolytic activity, as TFPI does, would be expected to block cell activation in this system. Thus, by increasing the rate of tissue factor–factor VIIa inhibition on the cell surface, it may be possible to limit some of the cellular activation mechanisms that could contribute to the inflammatory response. It is worth noting that signaling requires relatively high levels of surface tissue factor expression and factor VIIa. These levels correspond to complete activation of the factor VII,^{7, 9} a point that may be important when considering the observation by de Jonge et al that they failed to see an impact of TFPI on cytokine elaboration in human volunteers challenged with low levels of endotoxin. Under these conditions, one would expect most of the factor VII to remain in the zymogen form, and hence the factor VIIa concentration might be too low to signal effectively. In the more severe cases of endotoxin challenge when overt DIC is occurring, it is likely that most of the factor VII at or near the endothelial or monocyte/macrophage cell surface may become activated.

Tissue-factor signaling seems to be more complex than simply facilitating factor VIIa cleavage of a PAR on the cell surface. As mentioned above, tissue-factor-mediated macrophage activation requires the presence of the cytoplasmic tail of tissue factor. Potential signaling involvement through the cytoplasmic tail is suggested by the observation that it is a substrate for phosphorylation on 3 serine residues.11 One aspect of tissue factor that has drawn considerable attention is its participation in tumor metastasis. Like the situation in macrophage activation, the cytoplasmic tail of tissue factor is required to potentiate metastasis.12 The cytoplasmic tail of tissue factor also interacts actin binding protein 280.13 The tissue factor–factor VIIa complex can interact with TFPI bound to the matrix or presumably to the endothelial cell surface proteoglycans where it can work in concert with intergrins to tighten cell-cell interaction and, in the case of monocytes, potentially augment the inflammatory response by facilitating cell extravisation.

In addition to these roles of tissue factor in cell activation and tumor-cell migration, reverse migration of monocytes from the basal to apical surface of the endothelium was shown recently to involve tissue factor or, specifically, to be blocked by antibodies to tissue factor.14 Whether this process is dependent on factor VIIa, requires the cytoplasmic tail of tissue factor, or is modulated by TFPI or other factor VIIa inhibitors is unclear. While the cytoplasmic tail of tissue factor seems to play an important role in some of these aspects of cell migration and signaling, it is not critical for survival because tissue factor with the cytoplasmic domain truncated can rescue tissue-factor-null mice.15 

Factor Xa is another candidate enzyme for augmenting inflammation. Importantly in the current context, both TFPI and antithrombin are effective inhibitors of factor Xa at the concentrations employed in the studies presented in this issue. Recently a receptor for factor Xa, effector protease receptor 1 (EPR-1), was identified and cloned.16 The receptor is expressed on a wide variety of cells, including leukocytes,17 endothelium,18and smooth muscle cells.19 Current evidence suggests that EPR-1 signals via 2 mechanisms, one mediated by factor Xa binding^{20, 21} and the other requiring the active site.^{18, 19} Factor Xa can also activate cells in an apparently EPR-1-independent fashion.9 On endothelium, factor Xa elicits synthesis and release of IL-6, IL-8, and monocyte chemotactic protein-1 by an active site-dependent reaction independent of EPR-1.

EPR-1-dependent inflammatory events that are apparently factor Xa-active-site independent, and hence presumably insensitive to factor Xa inhibitors, include augmentation of IL-1 mediated lymphocyte proliferation22 and edema.21 Edema could be blocked by peptides that prevent factor Xa binding to EPR-1. In vivo in mice, EPR-1 has been implicated in CD3/T-cell-receptor–dependent lymphocyte proliferation. Blocking receptor synthesis or function eliminated among other things graft-versus-host disease.23 It is unfortunate that, with regard to factor Xa mediated signaling involving EPR-1, it is unclear whether the natural protease inhibitors, antithrombin and TFPI, will prevent the factor Xa–EPR-1 complex from forming. In several cases of protease receptor interactions in the coagulation system, the complex can form with low-molecular-weight inhibitors in the active site, but antithrombin inhibition results in the loss of affinity for the receptor.24 

Thrombin is the most frequently recognized clotting enzyme to be implicated in cell activation and inflammation. Thrombin triggers cell activation through PAR 1, 3-4.25 Activation results in a Ca²⁺ flux and generation of a host of second messengers.5 In the case of endothelium, thrombin facilitates leukocyte adhesion through elaboration of adhesion molecules and stimulates platelet activating factor formation, a potent agonist for neutrophils.26 Thrombin induced IL-8 release in clotting blood and CD14⁺ monocytes. Thrombin also caused IL-6 and IL-8 production from endothelial cells in culture.27 Antithrombin would be anticipated to shorten the half-life of thrombin in the circulation and, hence, to decrease thrombin mediated augmentation of the inflammatory events. In addition, cells containing thrombomodulin at high concentrations, such as those in the endothelium, are less sensitive to thrombin than cells lacking thrombomodulin,28 probably because thrombomodulin binds thrombin with high affinity and masks the thrombin receptor recognition site.24 Because thrombomodulin can be down-regulated by cytokines29 and proteolytically released from endothelium by netrophil elastase,30 it is possible that the endothelial-cell PARs become more susceptible to thrombin activation as the inflammatory process in sepsis proceeds.

Not all coagulation inhibitors are equal in their ability to protect animals from endotoxin shock. Heparin blocks endotoxin initiated clotting but is ineffective in preventing organ failure and death in baboons.31 Likewise, factor Xa blocked with a small molecule in the active site effectively inhibits coagulation induced byE coli in baboons but does not protect from organ failure or prevent death.32 It is possible that failure of these compounds to protect may have to do with some of their nonanticoagulant functions. In the case of heparin, it can bind growth factors and facilitate cell signaling.33 In addition, heparin displaces TFPI from the endothelium,1 the impact of which is uncertain with regard to protection of the endothelium, especially the microvascular endothelium during sepsis. In the case of the active-site-blocked factor Xa, it is possible that failure to protect the animals was linked to the fact that the modified factor Xa would still interact with EPR-1, possibly eliciting some of the inflammatory responses reviewed above. The alternative interpretation is that the natural anticoagulants initiate other functions distinct from their role in coagulation.

The third major natural anticoagulant candidate being considered for modulation in sepsis is protein C/activated protein C. Protein C is activated when thrombin binds to thrombomodulin, primarily on the surface of the endothelium.24 Because activation is dependent on thrombin generation, potent anticoagulants can prevent activation. Activated protein C, however, has been shown to inhibit monocyte activation by LPS,34 and inhibition of the pathway has resulted in increased levels of inflammatory cytokines in baboons challenged with E coli.35 These effects are thought to be mediated by an as yet uncharacterized receptor for APC on the monocyte.36 In septic patients receiving activated protein C, the IL-6 levels were reduced substantially.37 Therefore, at least with this natural anticoagulant, in vivo studies in primates and humans indicate that modulating the activated-protein-C levels results in decreases in the inflammatory response. In considering how the system is actually regulated, however, it is important to recall the multiple sites of interaction. A telling example of the complexity are studies with α₁-antitrypsin Pittsburgh, a mutation of the normal inhibitor that results in a very potent thrombin inhibitor that is also a potent activated protein C inhibitor. Infusion of this inhibitor into baboons challenged with E coli actually decreased survival time relative to controls,38 suggesting that despite controlling coagulation there was an exacerbated inflammatory response.

Particularly relevant to the issue of coagulation factor induction of inflammatory mediators, in many settings the coagulation enzymes are either ineffective or require high concentrations to induce cytokine elaboration. In many of these settings, however, the coagulation factors work synergistically with other agonists, such as endotoxin, to stimulate cytokine elaboration. As an example, thrombin and factor Xa failed to elicit IL-1 production from macrophages alone but increased IL-1 production up to 200-fold in the presence of suboptimal levels of endotoxin.39 

Given the myriad of mechanisms by which the coagulation system can interact to modulate the inflammatory response, how can one explain the failure of TFPI to modulate inflammation in human volunteers when the same agent did modulate inflammation in baboons? There are several differences in the studies. In the baboons E coli were used rather than endotoxin. The bacteria may elicit greater complement activation and may interact in other ways distinct from endotoxin. The physiology of the response in the baboons was different with the E coli than endotoxin and the response to E coli was much more reproducible than to endotoxin (Dr. Lerner Hinshaw, personal communication, 1980). Second and perhaps more importantly, the nonhuman-primate and other sepsis models in which natural anticoagulants have been effective in preventing death and in dampening the inflammatory response have all used lethal or near lethal levels of endotoxin or E coli as the challenge to which the inflammatory mediators are compared in the presence or absence of natural anticoagulant supplementation. Under these conditions, vascular integrity is compromised, allowing coagulation factors to contact extravascular cells where receptors for coagulation factors are abundant.5 Activation of these cells could potentially augment the inflammatory response. Natural anticoagulant supplementation could, among other things, diminish loss of vascular integrity and thereby diminish the inflammatory response simply by maintaining the response to the intravascular space. In addition, in the in vitro systems, signaling by both factor Xa and factor VIIa requires relatively high levels of these enzymes.^{7, 22, 40}These levels may be reached in severe sepsis but are not obtained with the human volunteers given endotoxin. Although it remains possible that humans differ dramatically from nonhuman primates and other animals in the linkage between coagulation and inflammation, given that at least with some natural anticoagulants inhibition of inflammation has been observed in septic patients with these agents, the difference is more likely related to dose. In severely challenged animals, coagulation almost certainly amplifies to an acute inflammatory response, but at low levels of endotoxin the monocyte-driven inflammatory response is probably largely coagulation independent, in part because the coagulation factors may not reach the critical levels to further stimulate the cells. The dramatic differences between the models of sepsis can be seen by the remarkable differences in the levels of markers measured in the following 2 papers. In the human volunteers IL-6 reached 3-5 ng/mL, whereas in the baboon it was approximately 1000 ng/mL. Likewise, the levels of the thrombin-antithrombin complex, a surrogate for the circulating thrombin levels, were approximately 100 ng/mL in the human volunteers and 30 times that level in the baboons.

Because it is obviously not ethical to subject the human volunteers to lethal levels of endotoxin, it would be of interest to examine whether these natural anticoagulants fail to prevent the elevation of inflammatory mediators in experimental animals given the low doses of endotoxin. This would at least argue against the idea that the apparent discrepancies are related to species differences.

Clinical studies to date have suggested a trend toward improved survival of septic patients given antithrombin 41 or protein C/activated protein C.^{37, 40, 42} But either these were uncontrolled studies or the numbers of patients enrolled were too few to provide a definitive answer about the impact of these agents on mortality. The efficacy of these agents in treating sepsis/septic shock should be resolved in the near future with the completion of phase 3 clinical trials now in progress. Regardless of the outcome of these trials, it will be important to gain a better understanding of the unique mechanisms by which different natural anticoagulants function in the regulation of the inflammatory response. This information should prove useful in providing better guidelines for the use of natural anticoagulants or combinations of these agents in the treatment of sepsis or trauma patients.