Sepsis is a systemic inflammatory host response to microbial infection that is characterized by exaggerated and deregulated activation of the immune, inflammatory, and blood coagulation systems [1–4]. The rationale for the development of aPC as a therapeutic agent in septic patients was based on this multifactorial concept of pathogenesis, because aPC targets both the inflammatory and the procoagulant manifestations of sepsis [5–7]. Therapy with recombinant aPC reduces absolute 28-day mortality of adult patients with severe forms of sepsis by ∼6% . Yet, treatment also increases the risk of severe bleeding complications caused by aPC’s anticoagulant effect, and aPC infusion is not effective in children and patients with less than severe sepsis. Persisting concerns about the overall risk-to-benefit ratio of aPC therapy are currently addressed in phase IV clinical trials . Challenges that must be met for the further development and improvement of sepsis therapy with aPC are the enhancement of overall efficacy for absolute mortality reduction, decreasing unwarranted effects, such as bleeding to enhance the safety profile of the drug, improving efficacy in patients with less than severe sepsis and in children, and developing rapid bed-side assays for identifying patients most likely to benefit from aPC therapy. Since the early seminal demonstration by Drs Taylor and Esmon  that aPC infusion can reduce sepsis mortality in baboons, a large number of in vitro cellular studies and in vivo animal experiments have produced ample evidence that aPC is a powerful modulator of inflammation, cell survival, vascular integrity, organ damage, and overall morbidity and mortality in sepsis models. As summarized in recent reviews [5,6,11], the majority of such protective aPC effects was based on aPC’s ability to engage signaling-competent receptors on the cell surface, including the endothelial cell protein C receptor (EPCR) and protease-activated receptors for thrombin (PAR1) and other proteases (PAR2, PAR3). Importantly, whereas PAR1 activation by thrombin predominantly elicits inflammation-promoting cellular responses, engagement of the same receptor by the EPCR–aPC complex produces outcomes linked to the dampening and resolution of inflammation. The complex molecular mechanisms underlying this dichotomy of PAR signaling in the context of inflammation are discussed in other reviews in this issue of the journal. In aggregate, these insights raised important questions about the relative importance of aPC’s anticoagulant and signaling activities in the context of therapy (Fig. 1). The development and characterization of aPC variants with selectively altered anticoagulant or signaling bioactivity has generated an opportunity to address this problem by distinguishing the relative contributions of the anticoagulant potential of aPC, mediated via its receptor-independent interaction with the coagulation factor substrates fV and fVIII, and of the anti-inflammatory cell-signaling function of aPC, mediated via engagement of EPCR and PAR1.
Summary. Activated protein C (aPC) is the key effector protease of the natural protein C anticoagulant pathway and exerts anticoagulant, as well as anti-inflammatory activity. This dual mode of action has been thought to underlie the therapeutic efficacy of recombinant aPC in the treatment of patients suffering from severe forms of sepsis. The development and characterization of recombinant variants of aPC with altered bioactivity profiles has generated an opportunity to test this concept by dissecting the roles of aPC’s anticoagulant and cell-signaling functions in the treatment of sepsis. Animal studies suggest that aPC variants with near-normal signaling function, but with greatly diminished anticoagulant potential may exhibit a substantially improved risk-to-benefit ratio in sepsis therapy.
Recombinant aPC variants with selectively reduced cell-signaling or anticoagulant activity
A panel of recombinant variants of protein C zymogen and aPC is available that differ in their rate of activation by thrombin or the thrombomodulin–thrombin complex [12,13], exhibit altered dependence on Na+ or Ca2+ ions , show reduced activity towards the coagulation substrate fVa [15–18], altered interaction with protein S in the degradation of fVIII, as well as additional variants with extended plasma half-life, resistance against inactivation by the Serpins α1-antitrypsin and protein C inhibitor, or with enhanced affinity for phospholipids [19–21]. Testing such variants in cell assays measuring PAR1 and EPCR-dependent responses, such as inhibition of apoptosis or protection against thrombin-induced endothelial monolayer permeability revealed that it is feasible to disrupt to a large extent aPC’s ability to cleave fVa without severely diminishing its capacity to engage EPCR and PAR1. At least three such prototypic ‘non-anticoagulant’ or ‘signaling-only’ aPC variants have been described: 5A-aPC  and Cys67-Cys82-aPC  contain mutations in the fV-interacting regions of the catalytic domain, and aPC-L38D  with a single amino-acid substitution in the protein S-interacting Gla-domain (Fig. 2). These variants exhibit an almost complete suppression of anticoagulant function, but retain close to normal signaling activity in currently used assays measuring PAR1 and EPCR-dependent cellular read-outs. Conversely, mutations in the C-terminus of the PC light chain (E149A-aPC ) or in EPCR-interacting residues within the Gla-domain (L8W-aPC ) can produce ‘anticoagulant-only’ or ‘signaling-defective’ variants, such as with intact or even enhanced anticoagulant activity toward fVa, but greatly diminished ability to elicit EPCR- and PAR1-dependent responses (Fig. 2).
Cell-signaling selective aPC variants are effective in murine sepsis models
Two studies have yielded first insights into how two prototypic signaling- or anticoagulant-selective aPC variants, that is the ‘signaling-selective’ 5A-aPC and the ‘signaling-defective’ E149A-aPC, modify the outcome of sepsis and of LPS-induced endotoxic shock in mice. Kerschen et al.  first showed that normal aPC reduces mortality of endotoxemia in an EPCR- and PAR1-dependent manner, and then compared the efficacy of 5A-aPC and normal aPC. In this model, aPC was administered at the onset of endotoxemia, shortly after administration of LPS, and the dosing of 5A-aPC was titrated to minimize any residual in vivo anticoagulant effect of 5A-aPC. Under these conditions, 5A-aPC improved 7-day survival to the same extent as normal aPC. Niessen et al.  extended these findings to a ‘late intervention’ LPS model, in which 5A-aPC was administered 10 h after LPS challenge. We further documented that 5A-aPC was also effective in three different models of bacterial sepsis, even when administered up to 10 h after infection . The potential role of the anticoagulant activity of aPC is currently investigated in ongoing collaborations with Dr Laurent Mosnier and Dr Griffin (Scripps Institute, La Jolla). We find that treatment with the E149A-aPC variant, which exhibits somewhat increased anticoagulant activity, but greatly diminished signaling potential, does not confer a measurable survival benefit in LD50 or LD90 endotoxemia challenges, despite full in vivo anticoagulant activity. These findings imply that the cell-signaling of activity of aPC is not only necessary, but also sufficient to confer a therapeutic benefit.
An important caveat in the interpretation of the above experimental data is that LPS challenge triggers the host response to infection by providing a massive ‘danger’ signal via toll-like receptor 4 in the complete absence of actual pathogen, thus tricking the host into a war against a non-existing enemy. In fact, a different picture emerges when aPC variants are tested in models of infection with live bacteria. For example, when administered several hours after infection, 5A-aPC is much more effective than an equivalent dose of normal aPC (Fig. 3). Such data suggest that ‘non-anticoagulant’ aPC variants may not only reduce the bleeding risk, but also potentially improve the risk-to-benefit ratio through additional mechanisms.
Safety and dose escalation
APC variants with selectively diminished anticoagulant, but near normal cytoprotective activity may be predicted to improve the safety profile of standard aPC therapy, and enable dose escalation without a concomitant elevation of bleeding risk. The current clinical protocol for aPC administration consists of i.v. infusion of 24 μg aPC kg−1 h−1 over 96 h, producing a steady state plasma aPC concentration of 45–54 ng/mL [8,28]. Development of this protocol was in part based on the anticoagulant effect of aPC, which sets an upper limit on dosing due to its adverse effects on haemostasis. Maintaining the current mode of therapy and just substituting normal aPC for 5A-aPC or similar variants should therefore improve the risk-to-benefit ratio, but is unlikely to improve efficacy as measured by overall mortality reduction.
With few exceptions , animal studies documenting receptor-dependent protection by aPC employ 10- to 100-fold higher peak aPC concentrations than achieved by the clinical regimen. APC and the zymogen PC bind EPCR with similar affinity [30–34], and exogenous aPC must displace endogenous PC from EPCR to induce cell signaling. A major open question is therefore what the optimal dosing is that is required to fully engage the in vivo receptor-dependent cellular effects of aPC. Even if one considers that PC levels are markedly reduced in septic patients (30–50% of normal ), current doses of aPC in the clinic yield plasma concentrations of at least one order of magnitude lower than endogenous PC (typically ∼60 nM). These considerations suggest that more intense dosing might increase efficacy by enhancing the formation of signaling-competent aPC-EPCR complexes. Such dosing issues are the objective of ongoing clinical trials to optimize dosing of normal aPC[36–38]. For example, the RESPOND study explores the use of endogenous protein C levels, as a biomarker and steering parameter for administration of 2-fold higher doses of aPC (48 μg kg−1 h−1) and extended duration of infusion for an additional 24–144 h.
As human recombinant 5A-aPC exhibits less than 10% of normal anticoagulant activity, tenfold higher therapeutic dosing (240 μg kg−1 h−1) may be predicted to produce a similar or slightly lower risk of bleeding, as the current dosing of normal aPC. Pilot data from LPS studies in mice suggest that survival-promoting receptor interactions are already saturated by a single 2 μg bolus, yielding peak loading plasma levels of ∼500 ng/mL. On the basis of the projected anticoagulant effect of such high-dose 5A-aPC infusion, this concentration seems within the clinically achievable range. However, plasma levels of the aPC inhibitor PCI are much higher in humans than in mice, and such a substantial increase in dosing would also raise new questions about cost, unwarranted side-effects not observed with low-dose therapy (such as potential receptor desensitization), or mechanistic differences related to the mode and duration of infusion.
Existing evidence from preclinical studies already provides a strong incentive to test and further develop aPC variants for clinical use. However, realization of the full therapeutic potential of aPC will require a much-improved understanding of the mechanisms underlying aPC’s efficacy in sepsis treatment. Much more detailed insight is needed into the relevance of specific cellular and molecular targets of aPC for sepsis survival, how therapy affects the numerous interactions between vascular endothelium and leukocytes, and how aPC modifies the overall development and resolution of the innate and adaptive immune response to infection. The various effects of aPC in different cells types may not all depend on engagement of the EPCR-PAR1 signaling axis, but involve other and/or additional molecular targets. For example, aPC suppresses expression of tumor necrosis factor related apoptosis-inducing ligand (TRAIL) in endothelial cells in a PAR1-dependent, but EPCR-independent manner . In monocytes, aPC’s anti-inflammatory effects appear to involve suppression of signaling processes initiated through the frizzled receptor for Wnt5A . One of the early studies to document an anti-inflammatory effect of aPC on monocytic cells reported that specific binding of aPC to the cell surface could not be competed with protein C zymogen , suggesting the existence of an as yet to be identified aPC-selective receptor. More recent work has shown that aPC—in an EPCR-dependent manner—suppresses tissue factor activity in monocytic cells by engaging the apoER2/lrp8 receptor and subsequent activation of the reelin signaling pathway . In whole animal models, the overall effect of aPC on inflammation and sepsis, therefore in all likelihood involves more than a single receptor pathway on multiple cellular targets. The current knowledge about the individual contribution of each of these pathways to overall sepsis outcome is sketchy at best, and a much-improved understanding of the underlying biology is a prerequisite for evaluating the potential use of second-generation aPC variants in a clinical setting.
Space limitations prevent us from citing the majority of the critically published work, that is, the foundation of the work described in the current review, and we apologize to our colleagues for this limitation. We gratefully acknowledge the tremendous support from our collaborators L. Mosnier, J. Griffin, J. Fernandez, F. Castellino, W. Ruf, B. Cooley, N. Mackman and their laboratory staff. We are further indebted to C. and N. Esmon, A. Rezaie, J. Degen and all our colleagues at the BRI who have supported this work by providing reagents and through insightful discussions. Support was provided by funding from NHLBI HL093388 (HW).
Disclosure of Conflicts of Interests
HW receives funding from NIH to investigate the mechanism of aPC in sepsis therapy; has in the past received financial research support and consulting fees from Eli Lilly & Co.; and is co-applicant on a pending patent potentially related to the described work (WO 08/073603). EK declares no conflicts of interest.