Activated protein C


  • J. H. GRIFFIN,

    1. Division of Translational Vascular Medicine, Department of Molecular and Experimental Medicine (MEM-180), The Scripps Research Institute, La Jolla, CA, USA
    Search for more papers by this author

    1. Division of Translational Vascular Medicine, Department of Molecular and Experimental Medicine (MEM-180), The Scripps Research Institute, La Jolla, CA, USA
    Search for more papers by this author
  • A. J. GALE,

    1. Division of Translational Vascular Medicine, Department of Molecular and Experimental Medicine (MEM-180), The Scripps Research Institute, La Jolla, CA, USA
    Search for more papers by this author

    1. Division of Translational Vascular Medicine, Department of Molecular and Experimental Medicine (MEM-180), The Scripps Research Institute, La Jolla, CA, USA
    Search for more papers by this author

John H. Griffin, Department of Molecular and Experimental Medicine (MEM-180), The Scripps Research Institute, 10550 North Torrey Pines Road, La Jolla, CA, USA.
Tel.: +1 858 784 2220; fax: +1 858 784 2243; e-mail:


Summary.  Protein C is a vitamin K-dependent plasma protein zymogen whose genetic mild or severe deficiencies are linked with risk for venous thrombosis or neonatal purpura fulminans, respectively. Studies over past decades showed that activated protein C (APC) inactivates factors (F) Va and VIIIa to down-regulate thrombin generation. More recent basic and preclinical research on APC has characterized the direct cytoprotective effects of APC that involve gene expression profile alterations, anti-inflammatory and anti-apoptotic activities and endothelial barrier stabilization. These actions generally require endothelial cell protein C receptor (EPCR) and protease activated receptor-1. Because of these direct cytoprotective actions, APC reduces mortality in murine endotoxemia and severe sepsis models and provides neuroprotective benefits in murine ischemic stroke models. Furthermore, APC reduces mortality in patients with severe sepsis (PROWESS clinical trial). Although much remains to be clarified about mechanisms for APC’s direct effects on various cell types, it is clear that APC’s molecular features that determine its antithrombotic action are partially distinct from those providing cytoprotective actions because we have engineered recombinant APC variants with selective reduction or retention of either anticoagulant or cytoprotective activities. Such APC variants can provide relatively enhanced levels of either cytoprotective or anticoagulant activities for various therapeutic applications. We speculate that APC variants with reduced anticoagulant action but normal cytoprotective actions hold the promise of reducing bleeding risk because of attenuated anticoagulant activity while reducing mortality based on direct cytoprotective effects on cells.


A vitamin K-dependent plasma protein was named ‘Protein C’ when Stenflo purified it from bovine plasma [1]. Activated protein C (APC) is generated from the protein C zymogen by proteolytic activation by thrombin (Fig. 1A); notably, APC, designated autoprothrombin II-A where ‘A’ indicated anticoagulant (Fig. 1B), had been studied by Seegers and colleagues, who thought it was a derivative of prothrombin.

Figure 1.

 Protein C activation and activated protein C (APC) activities. (A) The endothelial cell receptors, thrombomodulin (TM) and endothelial cell protein C receptor (EPCR), are required for efficient activation of protein C by thrombin (IIa). Dissociation of APC from EPCR allows expression of APC’s anticoagulant activity (B), whereas retention of APC bound to EPCR allows APC to express multiple direct cellular activities (C). (B) APC conveys its anticoagulant activity when bound to cell membrane surfaces by cleaving the activated cofactors Va (FVa) and VIIIa (FVIIIa) to yield the inactivated cofactors, FVi and FVIIIi. (C) Beneficial cytoprotective activities of activated protein C (APC) that involve direct effects of APC on cells require the cellular receptors EPCR and PAR-1. These activities include APC-mediated alteration of gene expression, profiles, anti-inflammatory activities, anti-apoptotic activities and protection of endothelial barrier functions. This figure is modified from a figure originally published in Blood: Mosnier, Zlokovic and Griffin. The cytoprotective protein C pathway, 207; 109: 3161–72. ©The American Society of Hematology.

Protein C concentration in plasma is 70 nm, while its activated form, APC, is 40 pm [2]. Mature human protein C contains 419 amino acids, and the human gene (PROC) on chromosome 2 (2p13-14) contains 9 exons. Key post-translational modifications include ß-hydroxylation at Asp71, N-linked glycosylation at residues 97, 248, 313 and 329 and gamma-carboxylation of nine glutamic acid residues giving rise to nine Gla residues in the amino terminus (the ‘Gla domain’). The domains of protein C comprise the Gla domain (residues 1–37), two epidermal growth factor (EGF)-like regions (EGF-1, residues 46–92 and EGF-2, residues 93–136), an activation peptide (residues 158–169), and the enzymatic serine protease domain (residues 170–419) (Fig. 2A). Thrombin cleavage of the zymogen at Arg169 removes the activation peptide and generates APC, which is a trypsin-like serine protease with a typical serine protease active site triad, His211, Asp257 and Ser360, corresponding to chymotrypsin’s His57, Asp102 and Ser195 active site triad. This active site triad is characteristic of all the plasma clotting enzymes.

Figure 2.

  Activated protein C (APC) exosite structures mediating APC activities. Interactions of APC with its macromolecular substrates and cofactors are dictated by exosite interactions, and APC protease domain surface loops specifically recognize various APC substrates. (A) The serine protease domain of APC is shown at the top of panel A, with the active site residues His211, Asp257 and Ser360 in green, the 37-loop residues Lys191, Lys192 and Lys193 and the calcium-binding loop residues Arg229 and Arg230 in blue. APC anticoagulant activity requires binding of the APC Gla-domain to negatively charged phospholipids on cell and/or platelet membranes or microparticles, whereas APC cytoprotective activity requires binding of the APC Gla-domain to endothelial cell protein C receptor (EPCR) (purple). (B) Exosite specificity is schematically illustrated by the juxtaposition of both 37-loop and calcium-binding loop positively charged residues (blue) of APC near negatively charged residues of FVa in this APC:FVa complex model (from Pellequer et al. [80]) and these positively charged APC amino acids are required for normal cleavage of FVa at Arg506 and thus for normal anticoagulant activity. (C) and (D) The surface contours with positive (blue) and negative (red) side-chains shown for wt-APC (C) and for the 5A-APC variant (containing 5 Ala substitutions at Lys191, Lys192, Lys193, Arg229 and Arg230). Comparison of the protease domain for 5A-APC (D) with that in (C) shows the loss of five positively charged surface residues from an exosite that is required for normal recognition of FVa. The model of full length APC [80] is based on the serine protease domain structure of APC (Protein Data Bank entry 1AUT [42]). The Gla-domain of the model of full length APC was aligned with the protein C Gla-domain peptide crystallized in complex with soluble EPCR (1LQV) [80,81]. The model of 5A-APC was generated from the serine protease domain structure of APC 1AUT [42] using Modeller [82].

In contrast to the various procoagulant plasma factors, APC has anticoagulant and cytoprotective activities (Fig. 1). The Gla domain mediates binding to lipids to express anticoagulant activity and binding to endothelial protein C receptor (EPCR) to express cytoprotective activities (Figs 1 and 2A) [3–5]. Amino acid sequences on APC’s surface that are remote from the active site, termed ‘exosites’, mediate APC’s interactions with cofactors and substrates (Fig. 2). Studies identified positively charged exosites in the protease domain that are critical for factor (F) Va inactivation [6–11], and genetic engineering of APC variants has provided molecules with selective reduction of either anticoagulant or cytoprotective activities (see below).

Protein C hereditary deficiency

Physiologic roles of protein C are evident because massive, usually lethal, thrombotic complications arise in infants with severe homozygous protein C deficiency and because heterozygous-deficient adults present with increased risk for venous thrombosis [12,13]. Related to the latter risk, an APC-cleavage site mutation (Arg506Gln, FV Leiden) is widely appreciated [14–16]. Protein C knockout mice studies show perinatal lethality [17,18], while protein C levels of 1–18% in mice permit growth and birth. Mice with low levels of protein C suffer early onset thrombosis and inflammation [19].

Anticoagulant and cytoprotective protein C pathways

Biochemical reactions on cell surfaces include protein C activation, expression of APC anticoagulant activity, and initiation of APC’s cytoprotective actions (Fig. 1A–C). Normal activation of protein C by thrombin requires two membrane receptors, thrombomodulin and EPCR [20,21].

APC’s anticoagulant actions involve the irreversible proteolytic inactivation of FVa and FVIIIa (Fig. 1B). Multiple APC-cofactors, comprising both proteins and lipids such as protein S, FV, high density lipoprotein, anionic phospholipids (e.g. phosphatidylserine, cardiolipin, etc.) and glycosphingolipids (e.g. glucosylceramide), mediate anticoagulation as previously reviewed [20,22,23].

APC’s multiple cytoprotective effects include: (i) alteration of gene expression profiles; (ii) anti-inflammatory activities; (iii) anti-apoptotic activity, and (iv) protection of endothelial barrier function (Fig. 1C) [24–28]. These cytoprotective effects often appear to require EPCR [20] and the G-protein coupled receptor, protease activated receptor-1 (PAR-1 [29]).

Protease activated receptors (PARs) are derived from four genes in the human genome. PAR-1 is a G-protein coupled receptor that can be activated by proteolysis [30]. In reactions requiring EPCR, APC can activate endothelial PAR-1 [25–28,31] and inhibits hypoxia/hypoglycemia or staurosporine-induced apoptosis in various endothelial cell types [24,25,27]. These same two receptors mediate multiple in vivo anti-inflammatory and neuroprotective effects of pharmacologic doses of APC in various murine injury models [25,32–34]. PAR-1 seems a promiscuous receptor as it can be either proinflammatory or anti-inflammatory, depending on activating protease, the dose of protease, and the cell type [35]. Notably, gene expression profile studies showed certain PAR-1-dependent effects of APC differed from those of thrombin (e.g. down-regulation of p53 and thrombospondin-1 expression) [31]. Although clear evidence for a requirement for PAR-1 in the beneficial actions of pharmacologic doses of APC has been provided, the importance of PAR-1 for physiologic APC’s in vivo effects has been debated [36–38].

Molecular distinction of APC’s anticoagulant and cytoprotective activities

The accumulation of insights into APC’s cytoprotective actions in recent years raises fundamental questions about the relative importance of the enzyme’s anticoagulant actions vs. its cytoprotective actions. For example, for APC’s protective effects in humans with severe sepsis and in various animal injury models, what molecular mechanisms explain APC’s in vivo actions? Is the anticoagulant activity of APC both essential for success in the PROWESS trial and also for the increased risk for serious bleeding in severe sepsis patients, especially during the 4-day period of APC infusion [39,40]? Alternatively, does one or more of APC’s cytoprotective actions reduce mortality in sepsis? And in various kinds of animal injury models, which activities of APC are critical, which are dispensable and which are deleterious? To help answer some of these questions, initial efforts in our laboratory used site-directed mutagenesis to engineer APC variants with decreased anticoagulant activity and normal cytoprotective activity [41], while later efforts were made to decrease cytoprotective activity while retaining anticoagulant activity.

The basic initial idea was to make potentially safer APC variants with reduced risk of bleeding because of reduced anticoagulant activity that retained normal cytoprotection; thus, we sought to alter FVa binding exosites in APC without affecting exosites that recognize PAR-1. APC’s anticoagulant activity involves cleavage at Arg506 in FVa (Fig. 1B); this cleavage depends on positively charged residues in surface loops on APC’s protease domain, including loop 37 (protein C residues 190–193, equivalent to chymotrypsin (CHT) residues 36–39), the Ca++-binding loop (residues 225–235, CHT residues 70–80) and the autolysis loop (residues 301–316, CHT residues 142–153) (Fig. 2C) [6,8–11,42,43]. We made two APC variants with Ala substitutions for Arg229 and Arg230 and Lys191, Lys192 and Lys193 (Fig. 2A) designated 229/230-APC and 3K3A-APC, and these had greatly attenuated anticoagulant activity (5 to 13% compared with wild-type APC) while each retained normal anti-apoptotic activity which still required PAR-1 and EPCR (Fig. 3) [41]. Calculation of the cytoprotective to anticoagulant activity ratio for these APC variants gave 7/1 and 25/1 (Fig. 4). When another APC variant containing the five Ala substitutions of these two variants (designated 5A-APC) was studied, we found that human 5A-APC had < 3% anticoagulant activity of wild-type APC but normal anti-apoptotic activity (see Mosnier et al., ISTH 2007 Congress) and the ratio was 36/1 (Fig. 4). Thus, we show that APC variants can be engineered to reduce anticoagulant activity while preserving the enzyme’s direct cytoprotective effects, which require EPCR and PAR-1.

Figure 3.

 Anticoagulantly impaired APC variants retain normal anti-apoptotic activity. The dose–response for activated protein C (APC)’s anti-apoptotic activity was determined in staurosporine (STS)-induced apoptosis assays using human endothelial cells. APC variants with Ala substitutions for Arg229 and Arg230 and Lys191, Lys192 and Lys193 (see Fig. 2A), designated 229/230-APC and 3K3A-APC, had normal activity compared with wild-type (wt) APC. The active site APC mutant, S360A-APC, had no activity. This shows that APC variants with reduced anticoagulant activity retain normal anti-apoptotic activity. This figure is taken with permission from Mosnier et al., Blood. 2004;104:1740–1744. ©The American Society of Hematology.

Figure 4.

 Activated protein C (APC) variants with greatly reduced anticoagulant activity relative to cytoprotective activity and vice versa. For wild-type-APC (wt-APC) and APC variants, the values for anti-apoptotic activity and anticoagulant activity are shown and the ratio of these activities is given numerically on the left side of the figure. The values for each activity are shown on the abscissa where the values for wt-APC are defined as 100%. These data highlight important distinctions between structural requirements for APCs anticoagulant and cytoprotective functions. The availability of APC variants with greatly reduced anticoagulant activity relative to cytoprotective activity (e.g. 1/36 ratio) and vice versa (e.g. 72/1 ratio) permits proof of principle studies in animal models to distinguish the relative importance of APCs cytoprotective vs. anticoagulant activities for APC’s in vivo benefits in sepsis, stroke or other injuries.

Our second and complementary goal in engineering APC variants with selectively altered activities was to make a molecule that had reduced cytoprotective activity but normal anticoagulant activity. For this purpose, we found that a single mutation of Glu149 to Ala (E149A-APC) gives an APC variant with increased anticoagulant activity but greatly reduced anti-apoptotic potency (see Mosnier et al., ISTH 2007 Congress). As indicated in Fig. 4, E149A-APC has a ratio of 1/72 for its cytoprotective to anticoagulant activity ratio. Thus, we show that APC variants can be engineered to reduce selectively the enzyme’s cytoprotective activity while preserving or actually increasing the enzyme’s anticoagulant activity.

Overall, our data show that mutagenesis of protein C can effectively create APC variants with decreased anticoagulant activity and normal cytoprotective activity or vice versa. Furthermore, these data (Fig. 4) highlight important distinctions between structural requirements for APCs anticoagulant and cytoprotective functions. The availability of APC variants with greatly reduced anticoagulant activity relative to cytoprotective activity (e.g. 1:36 ratio) and vice versa (e.g. 72:1 ratio) (Fig. 4) will permit proof of principle studies in animal models to distinguish the relative importance of APC’s cytoprotective vs. anticoagulant activities for APC’s in vivo benefits in sepsis, stroke or other injuries. Such studies may ultimately help contribute to safer or better therapeutic APC variants. For example, safer APC variants with reduced bleeding risk may permit therapies using higher APC doses for shorter times. Higher doses of APC may stabilize stressed cells at risk for excessive inflammation or apoptosis and prevent organ failure.

In recent collaborative studies, Dr Harmut Weiler and colleagues showed that recombinant murine 3K3A-APC and 5A-APC variants seem as active as wild-type murine APC in preventing death from endotoxemia in mice, and both EPCR and PAR-1 were required for mortality reduction by APC in the murine endotoxemia model [44,45]. These findings strongly imply a primary role for APC’s cytoprotective in vivo activities in this endotoxemia model. Additional studies in various murine or other animal injury models using a selection of APC variants with altered properties (Fig. 4) should help clarify the extent to which either anticoagulant or cytoprotective activities of APC play key roles for APC’s beneficial in vivo actions in reducing mortality in severe sepsis or in other injuries.

In vivo role for APC cytoprotection

Both anticoagulant and cytoprotective activities may contribute to APC’s beneficial in vivo actions. For some time, researchers posited that APC’s systemic anticoagulant action explained APC’s anti-inflammatory effects via down-regulation of thrombin generation because thrombin is pro-inflammatory. But recent findings dictate otherwise. In contrast to the PROWESS trial showing reduction of mortality in severe sepsis by APC, other anticoagulants such as antithrombin and tissue factor pathway inhibitor failed to reduce mortality [39,46,47]. Murine ischemic stroke model studies show that APC’s neuroprotective effects require EPCR and PAR-1 and are, at least in part, independent of APC’s anticoagulant activity [25,32–34]. Moreover, both EPCR and PAR-1 are required for mortality reduction in murine endotoxemia, and APC variants with < 10% anticoagulant activity but normal cytoprotective action are normally active in reducing death in murine endotoxemia (see below) [41,44,45]. Thus, we believe that APC’s direct effects on cells are established as crucial for the life-saving effects of pharmacologic doses of APC in at least two animal injury models, namely sepsis and stroke.

APC cytoprotection mechanistic considerations

APC’s direct effects on cells include: (i) alteration of gene expression profiles; (ii) anti-inflammatory activities; (iii) anti-apoptotic activity; and (iv) endothelial barrier stabilization. One or more of these effects are termed APC’s cytoprotective effects. Each of these activities is biologically distinct and may or may not involve the same intracellular mechanisms, and much remains to be clarified. However, mechanistic studies show that EPCR is essential for most if not all of APC’s effects on cells while PAR-1 often appears essential.

Gene expression profiles.

Gene expression profiling pattern studies revealed modulation of gene expression by APC for some major genes of inflammation and apoptosis, with the general effect of down-regulation of proinflammatory and proapoptotic pathways and up-regulation of anti-inflammatory and anti-apoptotic pathways [24,25,28,31,32,48,49]. Notably, APC decreases proapoptotic p53 and Bax expression and increases expression of anti-apoptotic genes such as Bcl-2.

For stressed endothelial cells, APC and thrombin, although both working via PAR-1 activation, differ significantly in their transcriptome alteration effects. Thus, APC is capable of modulating gene expression by mechanisms that are, at least partially, different from those used by thrombin.

APC anti-inflammatory activity

APC’s anti-inflammatory effects on endothelial cells involve inhibition of inflammatory mediator release and of expression of vascular adhesion molecules with the net result of inhibiting leukocyte adhesion and leukocyte tissue infiltration. In addition, by helping to maintain endothelial barriers, APC reduces extravascular inflammatory processes [50–55]. Inflammatory mediator release by leukocytes or by endothelial cells is inhibited by APC [56–59].

EPCR exists on the surface of monocytes, CD56 +  natural killer cells, neutrophils and eosinophils [55,60–62], and EPCR is critical for APC’s anti-inflammatory effects on leukocytes [63–66]. Soluble EPCR binds the integrin CD11b/CD18 (Mac-1) (αMβ2) (CR3) on leukocytes, as does the neutrophil molecule, proteinase-3 [67]. To date, no evidence for a role of the EPCR:PR3:CD11b/CD18 has appeared, although this complex might help mediate APC cellular effects.

Anti-apoptotic activity

Apoptosis is a widely known and key process that is based on efficient cell death programs [68–70]. For the intrinsic pathway of apoptosis that can be triggered by a variety of cellular stresses, the release of cytochrome c from mitochondria and subsequent procaspase 3 activation are central events and certain molecules such as the tumor suppressor protein, p53, and the Bcl-2 family of proteins, help to balance apoptotic signals and survival signals. The extrinsic apoptotic pathway, involving activation of initiator caspases, such as procaspase-8, requires membrane-bound death receptors and sensors of the extracellular environment. The extrinsic apoptotic pathway may also co-opt members of the intrinsic pathway via activation of the pro-apoptotic members of the Bcl-2 family or procaspase-3, which can be activated by either pathway, yielding caspase-3 that cleaves multiple substrates to advance apoptotic cell death [68–70].

APC exhibits anti-apoptotic activity both in vitro and in vivo via reactions requiring APC’s enzymatic activity and EPCR and PAR-1 [24,25,27,33]. In murine injury models, APC inhibits apoptosis in the brain, and the intrinsic apoptosis pathway is blunted with decreases in p53 and Bax and with increases in Bcl-2. APC also appears capable of blunting extrinsic pathway apoptosis because it counteracts the neurotoxicity of tissue plasminogen activator (tPA) that exerts caspase 8-dependent, proapoptotic activity via the extrinsic pathway [33]. In murine endotoxemia models, apoptosis is reduced by APC in association with improved survival in sepsis [41,44,45]. In severe sepsis patients, recombinant APC (XIGRIS) reduced apoptosis of circulating blood cells [71]. Thus, APC’s anti-apoptotic effects are documented both in vitro and in vivo, and both murine injury models and clinical findings suggest that APC’s anti-apoptotic activity may be very important for APC’s pharmacologically beneficial effects in reducing mortality in sepsis [72,73].

Endothelial barrier stabilization

Endothelial barrier breakdown with infiltration of extravascular space by cells and inflammatory mediators contributes to the pathogenesis of inflammation. Specifically, increased endothelial permeability promotes edema and hypotension and thereby promotes inflammation, acute lung injury and organ failure (see [74–76]). Protective endothelial barrier effects are caused by APC acting via EPCR and PAR-1; in this context, APC can induce sphingosine kinase-1 and up-regulate sphingosine-1-phosphate (S1P) formation via sphingosine kinase [50,53]. The sphingolipid S1P acts to reduce endothelial permeability through the S1P receptor-1 (S1P1), a G-protein coupled receptor belonging to the endothelial differentiation gene (Edg) family. The actions of S1P1 that stabilize the cellular cytoskeleton are dependent on Rho family GTPases and mitogen-activated protein kinases (MAPK) (see [74,76,77]). Endothelial barrier protection by APC requires PAR-1. In contrast to APC, high concentrations of thrombin, acting via PAR-1, cause destabilization of endothelial barriers, whereas PAR-1 activation by APC stabilizes barriers [50]. It appears that endothelial barrier effects of APC and thrombin may depend on Rac (protective) and Rho (destabilizing) signaling, respectively [77], and, for APC’s effects, there may be direct or indirect interactions between EPCR and S1P1 [53].

APC’s barrier stabilization effects may be more effectively accomplished by endogenously generated APC than by exogenously added APC [78], supporting the idea that endogenous generation of APC might promote vascular integrity and limit bleeding when thrombin is locally generated.

Strikingly, although APC is appropriately recognized as a potent anticoagulant, APC can also potently prevent bleeding in the brain! In murine ischemic stroke models where tPA administration induces brain hemorrhage, co-administration of APC reduces tPA-induced bleeding [25,32–34]. These antihemorrhagic effects of APC in stroke models require PAR-1 and may be partially based on both direct and indirect stabilization of endothelial barriers. APC’s direct and immediate effects may involve its ability to stabilize the endothelial cytoskeleton while its indirect effects, in this setting, involve attenuation of tPA-induced up-regulation of MMP9, which itself promotes breakdown of the blood–brain barrier [25,32–34].

APC’s in vivo effects

Multiple beneficial in vivo effects of APC have been published and a detailed summary of them is beyond the scope of this review, although some key studies are cited in the text above. The multiple in vivo beneficial effects are no doubt derived from both anticoagulant and cytoprotective actions of APC, depending on the injury and the context of the study. Current and potential therapeutic applications for APC include severe sepsis, ischemic stroke, lung injury and inflammation, and angiogenesis and wound healing, and these indications have been discussed elsewhere in detail [79].


Inflammation, apoptosis and thrombosis are inextricably intertwined in host defense protective reactions. Protein C provides physiologic homeostasis via its antithrombotic and anti-inflammatory actions, while pharmacologically administered APC reduces mortality in severe sepsis in humans and in murine and animal injury models, including ischemic stroke. In animal injury models and in vitro experiments, APC’s protective actions often require EPCR and PAR-1. Thus, basic preclinical and clinical studies heighten interest in the cytoprotective protein C pathway, which is distinct from the anticoagulant protein C pathway, suggesting the cytoprotective actions of APC are clinically very interesting. APC variants have been made with decreased anticoagulant activity but normal cytoprotective activity and vice versa. These APC variants are valuable not only for dissecting in vivo modes of action of APC but also for potentially reducing serious bleeding risks while retaining APC’s beneficial therapeutic properties.


Because of space limitations, we were unable to cite all of the interesting and relevant papers on APC, and we apologize to our colleagues for this limitation. We gratefully acknowledge stimulating discussions with Scripps colleagues X. Yang, S. Yegneswaran, W. Ruf, Z. Ruggeri and M. Riewald and with collaborators B. Zlokovic and H. Weiler. Support was provided in part by NIH grants HL21544, HL31950, HL52246 and HL63290 (J.H.G), HL082588 (A.J.G.) and HL087618 (L.O.M.) and a Basic Research Scholar Award from the American Society of Hematology (L.O.M.).

Disclosure of Conflict of Interests

JHG has carried out work on behalf of Socratech LLC. JAF, AJG and LOM state that they have no conflict of interests.