Gene expression profile of antithrombotic protein c defines new mechanisms modulating inflammation and apoptosis.

Human protein C is a natural anticoagulant factor, and a recombinant activated form of the molecule (rhAPC) is completing clinical evaluation for treatment of severe sepsis. Because of the pathophysiologic role of endothelial dysfunction in severe inflammatory disease and sepsis, we explored the possibility that rhAPC might directly modulate endothelial function, independent of its anticoagulant activity. Using broad transcriptional profiling, we show that rhAPC directly modulates patterns of endothelial cell gene expression clustering into anti-inflammatory and cell survival pathways. rhAPC directly suppressed expression of p50 and p52 NFkappaB subunits, resulting in a functional decrease in NFkappaB binding at target sites. Further, rhAPC blocked expression of downstream NFkappaB regulated genes following tumor necrosis factor alpha induction, including dose-dependent suppression of cell adhesion expression and functional binding of intracellular adhesion molecule 1, vascular cell adhesion molecule 1, and E-selectin. Further, rhAPC modulated several genes in the endothelial apoptosis pathway, including the Bcl-2 homologue protein and inhibitor of apoptosis protein. These pathway changes resulted in the ability of rhAPC to inhibit the induction of apoptosis by the potent inducer, staurosporine. This new mechanistic understanding of endothelial regulation and the modulation of tumor necrosis factor-induced endothelial dysfunction creates a novel link between coagulation, inflammation, and cell death and provides insight into the molecular basis for the efficacy of APC in systemic inflammation and sepsis.

Endothelial dysfunction plays a critical role in uncontrolled inflammatory conditions such as sepsis and multiorgan dysfunction syndrome. The elucidation of cytokines following insult, particularly tumor necrosis factor ␣ (TNF), 1 initiates cell surface activation affecting a number of pathways (e.g. oxida-tion, adhesion, cytokine release, apoptosis, and nitric oxide production), as well as releasing a tissue factor that further contributes to inflammation though thrombin-induced activation of endothelium, platelets, and vascular smooth muscle (1). The overall result of this cellular activation is a dysregulation of endothelial function, leading to microvascular thrombosis, end organ damage, multiple organ dysfunction, and often death (2,3). In the last several years, blocking disseminated intravascular coagulation, and the microthrombi that may promote end-organ dysfunction, has been proposed as a new target for clinical treatment in sepsis (4).
Human protein C is a plasma serine protease that plays a key role in maintaining normal hemostasis (5,6). The thrombin-activated form of protein C (APC) acts as a feedback inhibitor of the coagulation cascade and has demonstrated antithrombotic activity in numerous model systems (7). APC has shown efficacy in models of lethal endotoxemia and has been reported to prevent microvascular coagulation in patients with meningococcal sepsis (8 -12). A recombinant version of human activated protein C (rhAPC) is completing clinical evaluation in patients with severe sepsis, targeting the effects of microvascular coagulation. However, there has been considerable speculation both historically and recently (13) as to the mechanism of the apparent efficacy of APC in the treatment of sepsis, with the implication that the beneficial activity must be more that just antithrombotic.
Although the critical role of APC in modulating microvascular coagulation through the inhibition of thrombin generation has been well studied, its direct effects on endothelial function have not been elucidated. We used a novel approach of broad transcriptional profiling and cluster analysis to define possible new molecular mechanisms for the APC pathway in vascular endothelial function. Transcript profiling, the simultaneous monitoring of gene expression for a significant portion of a genome, has emerged as a powerful tool in genetics and biology, because it allows for the analysis of many of the signal transduction pathways and other biological systems identified from the sequencing of the human genome (14). This profiling has been used to reveal novel targets and mechanisms of action using yeast as a model system (15)(16)(17). Using this novel approach, we demonstrate that rhAPC directly modulates cell signaling and alters gene expression in two major pathways of inflammation and apoptosis. rhAPC suppressed NFB-modulated genes by directly reducing NFB expression and functional activity. Further, rhAPC inhibited cytokine signaling, including TNF-induction of cell surface adhesion molecules (e.g. VCAM, ICAM, E-selectin, and fractalkine). rhAPC also modulated apoptosis pathways, including up-regulation of the endothelial Bcl-2 homolog (A1), eNOS, and the inhibitor of apoptosis (IAP), and suppression of the apoptosis-associated genes calreticulin and TRMP-2. Moreover, treatment of cells with rhAPC blocked the induction of apoptosis. These data * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ Contributed equally to the study. § To whom correspondence should be addressed. Tel.: 317-276-2293; Fax: 317-277-2934; E-mail: grinnell@lilly.com. 1 The abbreviations used are: TNF, tumor necrosis factor; APC, activated protein C; rhAPC, recombinant human activated protein C; NFB, nuclear transcription factor B; ICAM, intracellular adhesion molecule; VCAM, vascular cell adhesion molecule; A1, Bcl-2 homologue protein; IAP, inhibitor of apoptosis protein; eNOS, endothelial nitric oxide synthase; HUVEC, human umbilical vein endothelial cell; EMSA, electrophoretic mobility shift assay; PCNA, proliferating cell nuclear antigen; FBS, fetal bovine serum; PCR, polymerase chain reaction; TBE, Tris-buffered EDTA; FACS, fluorescence-activated cell sorter; FITC, fluorescein isothiocyanate. provide a novel view of how inflammation is modulated at the endothelial cell level and defines new relationships among hemostasis, inflammation, and cell death/apoptosis. Moreover, our results demonstrate the value of gene profiling in defining novel mechanisms of drug action and specifically in providing possible mechanistic answers to rhAPC efficacy in sepsis.

EXPERIMENTAL PROCEDURES
Cell Culture-Human umbilical vein endothelial cells (HUVEC) and pools HUVEC PO96, P146, or P150 were grown in endothelial growth medium with 2% FBS (Clonetics, San Diego, CA). Cells were prepared untreated or treated with rhAPC (Eli Lilly and Co., Indianapolis, IN) and/or TNF␣ (R & D Systems, Minneapolis, MN). For the Affymetrix™ experiment, cells were treated with APC for 16 h and TNF for the last 7 h. For the four conditions (untreated; vehicle of 0.1% bovine serum albumin in Tris/NaCl, 183 nM APC; 1 ng/ml TNF; and 1 ng/ml TNF plus 183 nM APC) cells were observed for a total of 16 h. EAhy926 cells (17), obtained from the University of North Carolina, were grown in T-75 flasks with Dulbecco's modified Eagle's medium/F-12 (3:1), 5% FBS, 50 M/liter gentamicin, and 20 mM/liter HEPES. Cells were grown to confluence (70 -90%) and were treated under various conditions as described in individual experiments. TNF concentrations were 1 ng/ml, and APC ranged from 8 nM to 180 nM. Cells were treated at various time points between 4 and 48 h. In control experiments, recombinant hirudin (lepirudin rDNA) was added to the cell culture medium at ϳ1 g/ml. There was no residual thrombin in any of the preparations as determined by amidolytic assay using S-2366 (Chromogenix) for thrombin proteolytic activity, compared with control thrombin at subnanomolar concentrations. During the course of the experiments, samples were tested for amidolytic activity in the presence and absence of hirudin, and no significant thrombin activity could be detected (Ͻ25 pM limit of detection).
Microarray-Cells were washed, and total RNA was isolated by Triazol™ (Life Technologies, Inc.) following the manufacturer's recommendations. RNA was stored at Ϫ80°C in diethyl pyrocarbonatetreated deionized water. Detailed methods for labeling the samples and subsequent hybridization to the arrays are available from Affymetrix (Santa Clara, CA). Briefly, 1.5 g of poly(A) ϩ RNA was converted to double-stranded cDNA (Superscript; Life Technologies, Inc.) priming the first-strand synthesis with a T7-(dT) 24 primer containing a T7 polymerase promoter. 1 g of double-stranded cDNA was subsequently used as a template to generate biotinylated cRNA using the incorporated T7 promoter sequence in an in vitro transcription system (Megascript kit; Ambion and Bio-11-CTP and Bio-16-UTP; Enzo). Control oligonucleotides and spikes were added to 15 g of cRNA, which was then hybridized to Hu6800 oligonucleotide arrays for 16 h at 45°C with constant rotation. The arrays were then washed and stained on an Affymetrix fluidics station using the EUKGE-WS1 protocol and scanned on an Affymetrix GeneArray scanner. Data analysis was performed using GeneChip 3.1 software. Internal controls of housekeeping genes and a test chip trial were run prior to test samples.
DNA Binding Assay-Reactions containing Buffer D (20 mM HEPES, pH 7.9, 100 mM KC1, 5 mM MgC1 2 , 0.2 mM EDTA, 0.5 mM dithiothreitol), poly(dI⅐dC), nuclear extract, and labeled oligonucleotide were incubated for 15-45 min. In some experiments antibodies to NFB subunits were also included in the reaction. Complexes were run on 5% (0.5ϫ TBE) polyacrylamide gels (Owl, Inc.). The probe was ␥-32 P-labeled (PerkinElmer Life Sciences), and nuclear extracts from Eahy926 and HUVEC were prepared as described (18). Samples were run for 90 min at 110 V, 2.0 A in 1/2ϫ TBE running buffer. Gels were dried on DE81 paper and developed overnight on Kodak film (Eastman Kodak Co., Rochester, NY), and monoclonal antibodies to NFBp65, NFBp50, and NFBp52 (NFB2) (Santa Cruz Biotechnology, Santa Cruz, CA) were used for specificity controls. Quantification of gel shift to a consensus NFB-C 5Ј-CAGTTGAGGGGACTCCAGGCC-3Љ site was performed by a scanning phosphorimager.
Flow Cytometry-Primary antibody at 1-2 g/ml in 100 l of FACS buffer (PBS, 5% albumin, 0.02% sodium azide) was applied at 4°C for 30 min. The secondary antibody, anti-mouse IgG-FITC, at 1 g/ml in 100 l of FACS buffer was applied at 4°C for 30 min. FACS analysis was done with a Coulter TM flow cytometer (Coulter). Primary antibodies were to adhesion markers ICAM-1, E-selectin, VCAM-1, and fractalkine (R & D Systems, Minneapolis, MN).
Apoptosis Assay-Cells were seeded at 3 ϫ 10 4 cells per well in a 96-well plate and treated with 1 g/ml staurosporine (Sigma) for 1 h or with staurosporine and APC (pretreatment for 16 h). Cells were prepared and stained according to APOPercentage TM apoptosis assay per the manufacturer's instructions (Biocolor Ltd., Belfast, Northern Ireland).

RESULTS
Transcript Profile of rhAPC Alone and in TNF-treated Endothelial Cells-The effect of rhAPC on TNF-activated or nonactivated human endothelial cells was assessed using transcript profiling with Affymetrix TM microarrays. Treatments were performed under conditions where no thrombin or other protease activity could be detected, and no detectable thrombin was being generated during the cell culture experiments as described under "Experimental Procedures." After treatment, cells were lysed and mRNA was isolated, followed by analysis on Affymetrix TM arrays, and results were confirmed by semiquantitative RT-PCR. We found that rhAPC and TNF treatments of HUVECs resulted in 10 genes that were regulated by rhAPC alone and 31 genes regulated by TNF alone. Shown in Table I are the genes whose expression was altered by rhAPC treatment alone and 13 genes of the 31 TNF-regulated genes that we found to be comodulated by rhAPC. In general, the genes clustered with respect to cellular function into those involved in inflammation/immune modulation and cell survival/apoptosis. We also observed that the TNFactivated genes, which primarily fell into the proinflammatory and apoptotic pathways, were counter modulated by the rhAPC treatment. There were several TNF-activated genes, associated with feedback pathways for controlling TNF responses, (19) that were further enhanced by rhAPC treatment. Notably, the TNFactivated A20 gene, which was enhanced by rhAPC, has recently been shown in knockout experiments to be critical in regulating TNF-induced effects and cell death responses (20). Overall, the pattern of gene expression modulated by rhAPC was consistent with an induction of anti-inflammatory and antiapoptotic pathways.
Effect of rhAPC on the NFB Pathway-A very notable change observed on the transcript profile following rhAPC treatment was a suppression of p52 (NFB2 subunit) on the array. Further analysis of the array results indicated that expression of this NFB subunit was suppressed both by rhAPC alone and in combination with TNF (Fig. 1A). The average difference from the 20 perfect and mismatch signals on the Affymetrix array demonstrated a reduction in NFB2 mRNA by rhAPC, with or without TNF cotreatment. Semiquantitative RT-PCR confirmed that the NFB2 mRNA was suppressed by rhAPC alone and that rhAPC attenuated its induction by TNF (data not shown). To further examine the functional effect of the suppression at the mRNA level, an electrophoretic mobility shift assay (EMSA) for NFB was used to determine the amount of NFB DNA binding in nuclear extracts from treated HUVECs using a 32 P-labeled oligonucleotide consensus probe to NFB. (Fig. 1B). In repeated experiments, rhAPC alone suppressed NFB nuclear extract DNA binding by ϳ60% (n ϭ 4), consistent with the reduced mRNA expression (lanes 1 and 2). Moreover, in repeated experiments, rhAPC also attenuated the amount of binding following TNF induction of NFB (lanes [3][4][5]. In addition to these results using HUVECs, we obtained the same results using the Eahy926 (21) human endothelial line (data not shown). The specificity of the EMSA binding assays was demonstrated with antibody supershifts to p65, p50 (NFB1), and p52 (NFB2), as well as competition with cold consensus sequence (data not shown). This showed that the effect was specific and that rhAPC a can directly inhibit the functional NFB pathway (both NFB1 and NFB2) in endothelial cells. Table I, a number of TNF-modulated genes were countermodulated by rhAPC, and the induction of many of these genes has previously been shown to be mediated through NFB (see Table I). We focused additional studies on the adhesion molecules as being of particular importance with regard to anti-inflammatory effects. Semiquantitative RT-PCR confirmed that rhAPC mediated suppression of several of the adhesion molecule mRNAs (induced by TNF) including CX3C (fractalkine) (22), ICAM-1, E-selectin, and VCAM-1 (23). By flow cytometry, we confirmed that ICAM-1 surface expression induced by TNF could be suppressed by rhAPC in endothelial cells ( Fig. 2A). The effect of APC on the reduction in ICAM-1 expression was concentrationdependent, as shown in Fig. 2B. Similarly, we have shown that rhAPC could inhibit the expression both E-selectin and VCAM-1, two other important leukocyte adhesion molecules. As shown in Fig. 2C, rhAPC reduced cell surface E-selectin expression, measured by flow cytometry, in a concentration-dependent manner. Further, using a direct cell binding assay, dependent on the interaction of endothelial VCAM-1 and very late antigen-4 on a target cell (U937), we demonstrate that rhAPC directly inhibits the cell-cell interaction up-regulated by TNF (Fig. 2D). Additional studies with VCAM-1 and fractalkine showed comparable dose-dependent inhibition by rhAPC.

Suppression of Cell Adhesion Molecules by rhAPC-As shown in
APC-dependent Modulation of Apoptosis-In addition to the described effect of rhAPC on NFB and its downstream mediation of inflammation, our profiling results also suggested a second clustering around pathways promoting antiapoptosis and cell survival. As shown in Fig. 3A, rhAPC suppress two proapoptotic genes, calreticulin, an endoplasmic reticulum luminal protein that when suppressed has been shown to decrease cell apoptosis (24), and TRMP-2, a marker of cell apoptosis (25). In contrast, rhAPC increased genes shown to be antiapoptotic or markers of cell survival such as A1 (19), IAP (26), cell cycle-related human Gu helicase (27), and proliferat-  ing cell nuclear antigen (PCNA) (28). Interestingly, rhAPC also significantly up-regulated eNOS, which is important in view of the significant role it plays as an endothelial survival factor (reviewed in Ref. 29).
Considering this pattern of gene expression, experiments were performed to determine whether rhAPC could actually block the induction of apoptosis. As shown in Fig. 3B, Eahy926 endothelial cells treated with staurosporine, a potent inducer of the cell death pathway, underwent apoptotic cell death by 3 h post-treatment. However, treatment with rhAPC significantly reduced the number of cells undergoing apoptosis. We observed a similar protective effect using kidney 293 cells and HUVECs, which were extremely sensitive to apoptotic induction. In repeated and quantified experiments, rhAPC significantly blocked the induction of apoptosis in each of these cells (Fig.  3C). The effect of rhAPC on the inhibition of apoptosis was dose-dependent with concentrations as low as ϳ2 nM effectively suppressing the induction of apoptosis. Further, these apoptosis results were confirmed by intracellular staining of caspase 3 (data not shown).
As indicated above, we observed no significant thrombin generation during the course of our cell culture experiments analyzing the effect of APC on endothelial function. However, to assure that the observed effects were independent of the ability of APC to inhibit the generation of thrombin, studies were conducted in the presence of hirudin, a potent thrombin inhibitor. In repeated experiments in which hirudin was included in the cell culture treatments, we found no effect of this inhibitor on the antiinflammatory and apoptosis responses with or without APC (data not shown). In the apoptosis analysis shown in Fig. 3, the addition of hirudin had no effect on the ability of APC to suppress staurosporine-induced apoptosis. The addition of hirudin also had no effect on the apoptosis suppression by APC as measured by caspase 3 staining. Moreover, we found no effect of thrombin inhibition on the ability of APC to suppress the TNF activation of the cell surface adhesion markers as in Fig. 2, measuring both VCAM and ICAM levels. Thus, the effect of APC on modulating endothelial cell pathways was independent of its ability to inhibit thrombin generation. DISCUSSION We have utilized a novel approach to define new molecular pathways for the subcellular action of the anticoagulant human protein C. Moreover, our data demonstrate the power of transcriptional profiling in defining drug mechanism and implications for disease therapy. In Fig. 4, we diagram the expanded paradigm of the physiological role of APC. We have found that beyond its indirect effect on inflammation, via inhibition of thrombin generation, APC has novel direct anti-inflammatory effects via suppression of the NFB pathway and an apparent ability to prevent apoptosis and modulate cell survival. This effect of rhAPC on inhibiting apoptosis appears to be separate and antithetical to most of the TNF/NFB pathway effects, yet consistent with an effect that would be beneficial in antithrombotic/anti-inflammatory situations. In fact, there is growing evidence for the role of apoptosis in systemic inflammatory response and sepsis (30 -32). The emerging data suggest that organs-specific cell death involving both parenchymal and microvascular endothelium underlies organ dysfunction, with increased apoptotic rates occurring in organ dysfunction. The concept of suppression of proinflammatory pathways and the switch to cellular survival mechanisms at the endothelial interface suggests a complex adaptive response at the vessel wall directly connected at the subcellular level to the action of APC. The protein C pathway in inflammatory states could protect the organism from vascular insult and possibly prolong endothelial, cellular, and organ survival.
The ability of APC to suppress the up-regulation of adhesion molecules demonstrates a significant anti-inflammatory activity and, in conjunction with antithrombotic activity, further supports its role in severe systemic inflammatory disorders, including sepsis. The latter is strongly supported by studies in the ICAM-1 knockout mouse with no endothelial surface ICAM-1 that is resistant to induction of sepsis (33) and protected from ischemic renal injury (34). Recently, diminished neutrophil myeloperoxidase activity in a renal ischemia reperfusion animal model suggested suppression of inflammation related to APC and adhesion, further supporting our proposed mechanism (35). Thus, failure of endothelial function in severe inflammatory processes without the immune modulatory influences of APC may help explain both the adverse outcomes of septic patients with protein C deficiency and the favorable outcomes seen in meningococcal purpura fulminans patients receiving protein C and APC treatment (9,12,36).
Further investigation of the broader effects of NFB suppression by APC on endothelium and immune leukocytes may help expand our understanding of the link between coagulation and inflammation and those disease states where protein C pathway regulation is important. APC may play a physiological role in other disease states involving endothelial inflammation and apoptosis e.g. arteriosclerosis, acute respiratory distress syndrome, systemic vasculitis syndromes, allograft vasculopathy (37), hepatic veno-occlusive disease (38), and sickle cell syndromes (39). Moreover, ongoing studies will help define the role of other possible inflammation/immune modulation pathways suggested from our broad profiling (e.g. chemoattractant B61 (Table I) and the suppression of Class I human lymphocyte antigen genes; data not shown). Recently, the endothelial protein C receptor (40) has been demonstrated to be important in survival in the baboon sepsis model (41). It will be important to determine whether this receptor is involved in the cell signaling we have observed with rhAPC.
At the time of this writing, enrollment in the PROWESS trial, a phase III, placebo-controlled, 28-day, all-cause mortality study, was stopped early based on interim analysis showing that trial results met the criteria for reduced mortality among rhAPC-treated patients. There has been considerable speculation, both historically and recently (13), as to the mechanism of the apparent efficacy of APC in the treatment of sepsis. Our data provide additional mechanistic understanding, not only for the physiologic role of APC, but for its potential therapeutic role in sepsis and systemic inflammatory responses.