Inhibition of Tumor Necrosis Factor α-mediated NFκB Activation and Leukocyte Adhesion, with Enhanced Endothelial Apoptosis, by G Protein-linked Receptor (TP) Ligands*

Tumor necrosis factor (TNF) α is a critical mediator of inflammation; however, TNFα is rarely released alone and the “cross-talk” between different classes of inflammatory mediators is largely unexplored. Thromboxane A2(TXA2) is released during I/R injury and exerts its effects via a G protein-linked receptor (TP). In this study, we found that TXA2 mimetics stimulate leukocyte adhesion molecule (LAM) expression on endothelium via TPβ. The potential interaction between TXA2 and TNFα in altering endothelial survival and LAM expression was examined. IBOP, a TXA2 mimetic, attenuated TNFα-induced LAM expression in vitro, in a concentration-dependent manner, by preventing TNFα-enhanced gene expression, and also reduced TNFα-induced leukocyte adhesion to endothelium both in vitro and in vivo. IBOP abrogated TNFα−induced NFκB activation in endothelial cells, as determined by reduced IκB phosphorylation and NFκB nuclear translocation, by inhibiting the assembly of signaling intermediates with the intracellular domain of TNF receptors 1 and 2 in response to TNFα. This inhibition resulted from the Gαq-mediated enhancement of STAT1 activation and was reversed by anti-STAT1 antisense oligonucleotides. TNFα-mediated TNFR1-FADD association and caspase 8 activation were not inhibited by IBOP co-stimulation, however, resulting in a 2.6-fold increase in endothelial cell apoptosis. By stimulating the vessel wall and inducing endothelial cell apoptosis, TXA2, in combination with TNFα, may hamper the angiogenic response during inflammation or ischemia, thus reducing revascularization and tissue viability.

Tumor necrosis factor (TNF) ␣ is a critical mediator of inflammation; however, TNF␣ is rarely released alone and the "cross-talk" between different classes of inflammatory mediators is largely unexplored. Thromboxane A 2 (TXA 2 ) is released during I/R injury and exerts its effects via a G protein-linked receptor (TP). In this study, we found that TXA 2 mimetics stimulate leukocyte adhesion molecule (LAM) expression on endothelium via TP␤. The potential interaction between TXA 2 and TNF␣ in altering endothelial survival and LAM expression was examined. IBOP, a TXA 2 mimetic, attenuated TNF␣-induced LAM expression in vitro, in a concentration-dependent manner, by preventing TNF␣-enhanced gene expression, and also reduced TNF␣-induced leukocyte adhesion to endothelium both in vitro and in vivo. IBOP abrogated TNF␣؊induced NFB activation in endothelial cells, as determined by reduced IB phosphorylation and NFB nuclear translocation, by inhibiting the assembly of signaling intermediates with the intracellular domain of TNF receptors 1 and 2 in response to TNF␣. This inhibition resulted from the G␣ q -mediated enhancement of STAT1 activation and was reversed by anti-STAT1 antisense oligonucleotides. TNF␣-mediated TNFR1-FADD association and caspase 8 activation were not inhibited by IBOP co-stimulation, however, resulting in a 2.6-fold increase in endothelial cell apoptosis. By stimulating the vessel wall and inducing endothelial cell apoptosis, TXA 2 , in combination with TNF␣, may hamper the angiogenic response during inflammation or ischemia, thus reducing revascularization and tissue viability.
A critical mediator of several inflammatory and ischemic conditions (1), tumor necrosis factor ␣ (TNF␣) 1 promotes leu-kocyte adhesion to vascular endothelium both in vitro and in vivo; such adhesion is mediated by the complex interplay of adhesion receptors on both leukocytes and endothelial cells. TNF␣ induces the transcription of the leukocyte adhesion molecules ICAM, VCAM, and E-selectin by activating the transcription factor NFB. Multiple vascular cell types, including macrophages, cardiac myocytes, mast cells, and neutrophils release TNF␣ during ischemia and inflammation (1)(2)(3). In addition to its proinflammatory effect, TNF␣ can cause cell death by apoptosis and promote cell survival (4). These diverse effects can result in apparently contradictory experimental findings, as experimental data support both deleterious and beneficial roles for TNF␣ in, for example, preservation of myocardial function following ischemia and reperfusion (5)(6)(7)(8).
The complexity of the physiologic effects of TNF␣ is mimicked by the mechanisms by which it generates signals to produce those effects. TNF␣ interacts with two cell surface receptors, TNFR1 and TNFR2, ubiquitously expressed on cells of the cardiovascular system. TNFR1 is dominant in most cell types; an independent role for TNFR2 is yet to be established (4). After ligand binding, a conformational change in the pretrimerized TNF receptor complex causes dissociation of the SODD protein and, in a series of protein-protein interactions, intracellular mediators are recruited to the cytosolic portion of the receptor (9). TNFR1 activates two broad signaling pathways, both initiated by the recruitment of TRADD to TNFR1. One pathway, mediated by the subsequent recruitment of FADD, results in caspase 8 activation and initiation of apoptosis. Conversely, the survival pathway involves association of RIP and TRAF-2 with TRADD. The TRADD⅐RIP⅐TRAF-2 complex recruits the IKK signalosome to TNFR1 thus activating NFB. In addition, the TRADD⅐RIP⅐TRAF-2 complex also activates MAPK pathways that lead to AP-1 activation (4,10).
During inflammation, multiple cytokines and stimulants of vascular endothelium are released. Thus, any pathophysiological effect produced by TNF␣ must take place in the context of "cross-talk" with those mediators. The most common class of receptor for these mediators are G protein-linked receptors, multiple forms of which are expressed on vascular cells. One mediator of particular importance in ischemia and inflammation is the eicosanoid thromboxane A 2 (TXA 2 ), released from activated platelets and damaged vessel walls. Local and systemic elevations in TXA 2 are reported in several thrombotic and vascular diseases (11). TXA 2 mediates vascular damage by inducing platelet activation, vasoconstriction, vascular smooth * This work was supported by National Institutes of Health Grants HL47032, HL51043, and HL55552 and American Heart Association postdoctoral fellowship 0020186T. 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.
¶ muscle hypertrophy/hyperplasia, and by converting the endothelial surface to a "prothrombotic" state. These biological actions of TXA 2 are mediated by a G protein-linked receptor (TP) of which two alternatively spliced subtypes, TP␣ and TP␤, exist in humans (12,13). Although the intracellular tail of TP␤ differs from that of TP␣, these isoforms couple to 80% of the same G proteins (14). Differences in signaling reported so far mediate receptor trafficking and desensitization (15)(16)(17).
In this study, we sought to investigate whether stimulation of TP, a prototypical G protein-coupled receptor that is activated during ischemia and inflammation, could alter TNF␣induced endothelial cell apoptosis and leukocyte adhesion in vivo. Furthermore, we tested the possibility that such an effect resulted from a selective effect on specific signal transduction pathways employed by TNF␣. The data present a currently unrecognized interaction between TNF␣ and ligands for G protein-coupled receptors that may provide a mechanism by which endothelial cell death occurs during inflammation.

EXPERIMENTAL PROCEDURES
Chemicals-The G protein inhibitors NF023 (G␣ i ), NF449 (G␣ s ), GP antagonist 2A (G␣ q ), pertussis toxin, and GDP were all purchased from Calbiochem-Novabiochem. All other chemicals were of suitable grade and purchased from Sigma unless otherwise stated.
Leukocyte Adhesion Assays-Confluent monolayers of first passage HEC were stimulated with TNF␣, IBOP, or the two in combination for 8 h and were washed extensively before the addition of leukocytes. U937 cells, a leukocyte cell line, were labeled with 2Ј,7Ј-bis(2-carboxyethyl)-5(6)-carboxyfluorescein-AM (2 M) for 30 min and 1 ϫ 10 6 cells were added to stimulated HEC monolayers for 2 h. At the conclusion, cells were washed with phosphate-buffered saline and incubated with lysis buffer (50 mM HEPES, pH 7.4, 150 mM NaCl, 1% (v/v) Nonidet P-40, 0.1 mM phenylmethylsulfonyl fluoride) for 30 min. U937 attachment was quantitated in a fluorometric plate reader using fluorescein isothiocyanate filters.
Analysis of Leukocyte Adhesion and Extravasation in Vivo-The modulation of TNF␣ by TXA 2 in vivo was analyzed by monitoring the passage of leukocytes along postcapillary venules in the externalized cremaster muscle as previously described (20). C57-black mice were injected intraperitoneally with murine TNF␣ (150 pM), the TXA 2 mimetic IBOP (50 nM), or the two together and prepared for intravital microscopy after 4 h. Adherent leukocytes were quantified per 100 m of vessel length. Extravasated leukocytes were determined as the number of interstitial leukocytes adjacent (within 30 m) to venules.
Immunoprecipitation and Immunoblotting-Confluent HEC were treated with IBOP, IL-1␤, or TNF␣ alone or in combination for 5 min. Monolayers were washed twice in phosphate-buffered saline containing 1 mM phenylmethylsulfonyl fluoride and scraped into immunoprecipitation lysis buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 1% (v/v) Nonidet P-40, 0.5% (w/v) sodium deoxycholate, 0.1 mM phenylmethylsulfonyl fluoride, 1 mg/ml aprotinin, 1 mg/ml leupeptin). Cell suspensions were sonicated, incubated on ice for 30 min, and clarified by centrifugation. For immunoprecipitation, protein content was determined and 300 g of total protein was incubated for 16 h at 4°C with protein G-agarose beads coated with saturating amounts of antibodies to TNFR1, TNFR2, NFB (p65 RelA ) (Santa Cruz Biotechnology, Santa Cruz, CA), RIP, or TRADD (Transduction Laboratories). The resulting immune complexes were recovered after centrifugation by boiling for 10 min in SDS-PAGE loading buffer. Nuclear proteins were isolated as previously reported (21).
Annexin V Staining-For annexin V staining, HEC stimulated with IBOP, TNF␣, or the two in combination (16 h) were harvested using trypsin/EDTA and stained for annexin V using the Pharmingen annexin-fluorescein isothiocyanate staining kit (Pharmingen). Annexin V staining was quantitated using flow cytometry.
Caspase Activity Assays-Following treatment with TNF␣, IBOP, or the two in combination for 16 h, HEC were scraped into lysis buffer and incubated on ice for 30 min. Lysates were clarified by centrifugation and protein concentration was estimated. Protease reactions utilized cleavage of the chromogenic peptide substrates Z-DEVD-p-nitroanilide and Ac-IETD-p-nitroanilide for caspase 3-and 8-like activity, respectively (Biomol Research Laboratories, Plymouth Meeting, PA). Caspase activity assays were carried out in reaction buffer (100 mM HEPES, pH 7.5, 5 mM dithiothreitol, 20% (v/v) glycerol, 0.5 mM EDTA, 0.1% (w/v) bovine serum albumin, 10 mM caspase substrate) using 300 g of total protein.
Reactions were incubated at 30°C for 30 min and cleavage of colorimetric substrates quantitated at 415 nm.
Cell Transfection and Luciferase Assays-HEC were transfected with the NFB-dependent luciferase reporter construct pBII-Luc(2ϫ Igk B-fos-luciferase) (22) or STAT1 oligonucleotides using the Gene-PORTER transfection reagent (Gene Therapy Systems, San Diego, CA). Transfected HEC were stimulated with TNF␣ or IL-1␤ alone or in combination with IBOP for 16 h and relative luciferase activity was determined using the luciferase gene reporter kit (Roche Molecular Biochemicals). STAT1 sense (5Ј-ggTggCAggATgTCTCAgTgG 3Ј) and antisense (5Ј-CCACTgAgACATCCTgCCACC-3Ј) oligonucleotides were transfected into cells 24 h prior to assay; preliminary experiments established that this time was sufficient for the ablation of STAT1 expression.
RNA Isolation and Northern Blot Analysis-Confluent HEC were exposed to TNF␣ or IBOP or both agents for up to 8 h. HEC were washed twice with phosphate-buffered saline and total mRNA was isolated using TRIzol reagent (Invitrogen). RNA (5 g of each sample) was separated on a denaturing 1% agarose gel, transferred to nylon membrane (Amersham Biosciences), and blotted for leukocyte adhesion molecule (LAM) expression as previously described (23). Full-length human cDNA probes to ICAM-1, VCAM-1, E-selectin, and GAPDH mRNAs were made from all cDNAs by random priming using the Megaprime Labeling kit (Amersham Biosciences). After blotting, membranes were dried and autoradiographed for the appropriate period of time with changes in mRNA levels quantitated by scanning densitometry.
Statistical Analysis-Data were pooled and statistical analysis was performed using the Mann-Whitney U test.

TXA 2 Mimetics Stimulate LAM Expression on HEC-Stimu-
lation with the TXA 2 mimetic IBOP (50 nM) increased expression of LAM molecules on HEC, with maximal induction after 16 h for ICAM-1 and 6 -8 h for VCAM-1 and E-selectin (p Յ 0.001, Fig. 1A). Induction of LAM expression by IBOP was concentration-dependent. ICAM-1 and VCAM-1 expression on HEC were maximally induced at IBOP concentrations Ն25 nM (IC 50 12 nM, p Յ 0.005) at 16 and 6 h, respectively. E-selectin was maximally induced at IBOP concentrations Ն50 nM (IC 50 25 nM) after 6 h (p Յ 0.01, Fig. 1B). U46619, another TP agonist, also stimulated LAM expression on HEC with maximal induction at 400 nM and time courses similar to those shown for IBOP (data not shown). Inclusion of the TP antagonist SQ29548 (5 M) abrogated IBOP-induced LAM expression (data not shown), indicating the effect was mediated by the TXA 2 receptor (TP). LAM expression is regulated by increased transcription, which in turn is mediated by NFB (24). HEC express both TP isoforms, TP␣ and TP␤. In cell lines overexpressing each of the two TP isoforms on a null background, NFB nuclear translocation (Fig. 1C) and LAM expression (data not shown) were found to follow ligation of TP␤, but not of TP␣. NFB nuclear translocation in response to IBOP in TP␤ expressing cells was inhibited by loading the cells with GDP, indicating the process was mediated by G proteins, and by two pharmacological inhibitors of the heterotrimeric G protein G␣ i (Fig. 1C). Antagonists of G␣ q and G␣ s were ineffective, consistent with the concept that this process was mediated by G␣ i signaling.
The inhibition of LAM expression by IBOP in TNF␣-stimulated HEC was also concentration-dependent (Fig. 2C). TNF␣ treatment induced a 4.5-fold increase in ICAM-1 expression (Fig. 2C), above that seen with untreated HEC. Increasing concentrations of the TXA 2 mimetic IBOP attenuated the induction of ICAM-1 by TNF␣ with inhibition greatest at IBOP concentrations Ն100 nM (p Յ 0.01, IC 50 ϭ 50 nM) (Fig. 2C). The induction of E-selectin and VCAM-1 expression were similarly inhibited by IBOP in TNF␣-stimulated HEC (data not shown). Inhibition of LAM expression on TNF␣-stimulated HEC was not because of alterations in the time course for maximal expression, as cells stimulated with both agents displayed only basal expression of ICAM-1 (Fig. 2D), VCAM-1, and E-selectin (data not shown) at all time points. Thus, the prevention of TNF␣-induced LAM expression by TXA 2 mimetics did not result from cross-talk between the two agents in which the time to peak expression was altered.
Consistent with these observations, IBOP-mediated inhibition of LAM expression prevented leukocyte adhesion to TNF␣stimulated HEC. Stimulation with either the TXA 2 mimetic IBOP or TNF␣ induced a 3-4-fold increase in adhesion of the monocytic cell line U937 to HEC (p Յ 0.005, Fig. 3A). Simultaneous treatment with both agents resulted in U937 adhesion similar to that in untreated controls (p Ն 0.4, Fig. 3A). Similar results were observed when the adhesion of leukocytes to the vessel wall and their extravasation from the vasculature into surrounding tissues was examined in vivo. Injection of recombinant murine TNF␣ (150 pM) into mice resulted in a 7-and 11-fold increase in leukocyte adhesion to and extravasation from, respectively (p Յ 0.0001), the venous network of cremaster muscle preparations after 4 h (Fig. 3B). Co-stimulation with IBOP (50 nM) resulted in an attenuation of TNF␣-mediated leukocyte adhesion and extravasation (p Յ 0.005 versus TNF␣ alone), with leukocyte rolling increased 3-fold (data not shown). Consistent with the observation above that ligation of TP␤ induces LAM expression in EC, IBOP stimulation did not cause an increase in leukocyte trafficking in vivo (Fig. 3B), as a mouse homologue for TP␤ has not been identified, and the murine TP resembles the human TP␣ isoform.
Regulation of ICAM-1, VCAM-1, and E-selectin by TNF␣ depends upon increased transcription (24). To determine how TP inhibits LAM expression in response to TNF␣, we examined the pattern of LAM mRNA expression. Fig. 3C shows that only ICAM-1 mRNA was detectable in untreated endothelial cells by Northern blot analysis. TNF␣ and IBOP both induced ICAM-1 expression 9-fold over controls after stimulation for 8 h (Fig.  3C). Both agents also induced robust expression of VCAM-1 and E-selectin after 3 h. The equivalent induction of ICAM-1, VCAM-1, and E-selectin mRNA by TNF␣ and TP stimulation is consistent with the pattern of protein expression determined by flow cytometry (Fig. 2A). Consistent with the results in panels A and B, incubation with both TNF␣ and IBOP prevented the increased mRNA expression for ICAM-1, VCAM-1, and E-selectin observed with either agent alone (Fig. 3C).
TP Stimulation Prevents Activation of NFB by TNF␣-Comparison of the ICAM-1, VCAM-1 promoters, and E-selectin promoters indicated that NFB and AP1 were factors that possibly indicated transcriptional regulation by TNF␣. To examine whether TNF␣-stimulated NFB activity is regulated by TXA 2 mimetics, we transfected HEC with a NFB-sensitive luciferase reporter construct. TNF␣, IL-1␤, and IBOP treatment induced luciferase activity in transfected HEC to similar levels when used individually (Fig. 4A, p Յ 0.005). In contrast, IBOP abrogated the transcriptional activity of NFB when used in combination with TNF␣, but not IL-1␤ (Fig. 4A).
Nuclear translocation and full transcriptional activity of NFB require phosphorylation of the NFB inhibitor protein, IB (4). Treatment with TNF␣, IL-1␤, or IBOP for 20 min increased IB␣ phosphorylation in HEC lysates 10 -12-fold and induced nuclear translocation of p65 RelA , the most common NFB subunit. Untreated endothelial cells had little or no phosphorylated IB or nuclear p65 RelA as assessed by immunoblotting (Fig. 4B). In accordance with the luciferase data (showing a lack of NFB-mediated transcriptional activity), co-stimulation with IBOP prevented the nuclear translocation of p65 RelA and the enhanced IB␣ phosphorylation observed with TNF␣, but not IL-1␤, alone (Fig. 4B). The inhibition of TNF␣-induced NFB activation by TP was prevented both in GDP-loaded cells, showing that the effect was G protein-mediated, and was also blocked by an antagonist of G␣ q signaling, but not that of G␣ s or G␣ i (Fig. 4C). IBOP stimulation also abrogated TNF␣-induced NFB activation in cell lines expressing TP␣ or TP␤ alone (data not shown), consistent with the previously reported data that both TP isoforms are coupled to G␣ q (14).
Interestingly, IBOP not only inhibited TNF␣-induced NFB nuclear translocation but also reversed the process (Fig. 4D). Simultaneous addition of TNF and IBOP (30 min) inhibited the nuclear accumulation of NFB observed with either agent alone. TNF␣ stimulation for 15 min induced nuclear NFB staining and IB phosphorylation similar to that obtained after 30 min (Fig. 4D, lane 3 versus lane 5). Addition of IBOP (to TNF␣-stimulated HEC) for a further 15 min resulted in an absence of IB phosphorylation and nuclear NFB staining (lane 6). In addition, NFB was found to be reassociated with IB in HEC treated with TNF␣ for 15 min prior to the addition of the TP agonist for a further 15 min, indicating that the levels of free NFB were returned to baseline (Fig. 4D, lane 6 versus  lane 5). The reciprocal was also true, as addition of TNF␣ after IBOP stimulation reversed the activation of NFB, decreased IB phosphorylation, and increased IB⅐NFB complexes (Fig.  4D, lane 8 versus lane 7). These data indicate that TP agonists both inhibit the activation of NFB by TNF␣ and also reverse previously activated NFB.

TP Specifically Inhibits the Activation of NFB, but Not Other TNF␣ Pathways, by Disrupting Early Events in Receptor
Signaling-NFB activation by IL-1␤ and TNF␣ converge at the IKK signalosome and share all distal elements. The inability of TXA 2 mimetics to prevent IL-1␤-mediated NFB activation indicates that the pathways involved may be specific to TNF␣. Stimulation with either IBOP or TNF␣ alone, or together, did not alter the cellular expression of any of the proteins investigated (data not shown). Thus, we immunoprecipitated TNFR1 and TNFR2 and blotted for the expected proteins to investigate whether TP stimulation altered the protein-protein interactions evoked by TNF␣ signaling. None of the proteins tested were recruited to the cytoplasmic tail of TNFR1 or TNFR2 in the absence of TNF␣ (Fig. 5, lanes 1 and 2). TNF␣ stimulation induced robust association of TRADD, RIP, TRAF-2, and IKK␥ with TNFR1. TRADD associated with TNFR1 to a similar degree in the presence or absence of IBOP. In contrast, IBOP co-stimulation prevented association of RIP, TRAF-2, and IKK␥ with TNFR1 (Fig. 5A, lane 4). Furthermore, neither TNFR1 nor TRADD co-immunoprecipitated with either anti-RIP or TRAF-2 antibodies in the presence of TNF␣ and IBOP (data not shown). This indicates that the interaction of RIP and TRAF-2 with TRADD, which occur independently, were both inhibited by IBOP. STAT-1 phosphorylation at Tyr 701 was recently reported to be part of the TNFR1⅐TRADD complex, and inhibited TRADD-TRAF-2/RIP interactions without influencing TRADD-FADD interactions (25). Interestingly, the TNF␣-induced STAT1-TNFR1 interaction was increased 3-fold by co-stimulation with the TXA 2 mimetic IBOP (Fig. 5A).
TNF␣ was found to increase STAT1 phosphorylation (Fig. 5B), which was augmented 3-fold by co-stimulation with IBOP. IBOP alone did not stimulate STAT1 phosphorylation (Fig.  5B), however, suggesting that the enhancement of TNF2-induced STAT1 phosphorylation by IBOP and TNF␣ is likely a result of another mechanism, such as inhibition of phosphatase activity. As is the case with inhibition of NFB, the IBOPenhanced phosphorylation of STAT1 was reduced by inhibition of G␣ q signaling (Fig. 5B). Thus, the abrogation of TNFR1 NFB signaling by IBOP may result from a TP-mediated enhancement of STAT1 activation by TNF␣.
When potential alterations in TNFR2 signaling were examined, we observed that TRAF-2 bound to TNFR2 and associated with NIK in TNF␣-stimulated HEC (Fig. 5C). Co-stimulation with IBOP and TNF␣ prevented association of NIK and TRAF-2, but not TRAF-2 and TNFR2. Furthermore, TP stimulation induced the association of RIP with TNFR2, which does not occur following TNF␣ alone, in contrast to the case with TNFR1 (Fig. 5C). Thus, TXA 2 mimetics may enhance those aspects of TNFR2 signaling that result in apoptosis, as well as inhibit NFB activation, by redirecting RIP to associate with TNFR2.
In this model, the effects of TP stimulation on the recruitment of signaling intermediates to TNFR1 and TNFR2 depend upon the increased association of STAT1 with TNFR1. To explore further the causative role of STAT1, we used antisense oligonucleotides to antagonize STAT1 expression. Transfection of HEC with the STAT1 antisense oligonucleotides abrogated STAT1 expression over a period of 16 h, and did not affect HEC viability. STAT1 sense oligonucleotides did not affect STAT1 expression during the same time course (data not shown). When STAT1 oligonucleotide-treated HEC were stimulated, those HEC treated with sense oligonucleotides were not found to differ from untransfected cells (Fig. 5, A and C). In contrast, STAT1 antisense oligonucleotides ablated the interfering effects of TP stimulation, thus allowing RIP, TRAF-2, and IKK␥ to be recruited to TNFR1. Similarly, the recruitment of NIK to TNFR2 was also restored and the recruitment of RIP did not occur. Thus, the enhanced phosphorylation and recruitment of STAT1 to TNFR1 in the presence of IBOP is a crucial step in the antagonism of TNF␣-induced NFB activation by TP stimulation.
The importance of MAPK activation to LAM transcription has recently become apparent (26 -29). To determine whether the effects of TXA 2 were specific for NFB, we probed for ). * and # indicate significance (p Յ 0.01) from control and TNF␣, respectively. B, IB phosphorylation and NFB nuclear translocation in response to TNF␣ and IL-1␤, with or without IBOP co-stimulation, after 30 min, was determined using specific antibodies for p65 RelA and phosphorylated IB. In panels B and C, histone H1 is shown as a loading control. C, NFB in nuclear extracts of HEC stimulated with TNF␣ and IBOP was measured. To determine the G protein(s) involved in the inhibition of TNF␣ induced activation of NFB by IBOP, cells were incubated with GDP (50 M), or antagonists of specific G proteins (10 M, G␣ q , G␣ i , and G␣ s ) for 30 min prior to the addition of stimulants. D, IBOP reverses the activation of NFB by TNF␣. HEC were incubated with IBOP or TNF alone or in combination for 15 or 30 min. In addition, HEC were incubated with IBOP 15 min after TNF␣ treatment had begun (lane 6); in the reverse experiment, TNF␣ was added 15 min after treatment with IBOP (lane 8). The activation of NFB was examined by nuclear translocation of NFB, phosphorylation of IB, and complex formation between NFB and IB. In the bottom two subpanels, immunoprecipitation (IP) of NFB is followed by immunoblotting (WB) with IB or NFB, the latter as a control. Histone H1 blotting was used to control for loading on nuclear protein preparations. The blots shown are representative of four individual experiments.
FIG. 5. IBOP-mediated prevention of the association of proteins responsible for NFB signaling with TNFR1 and TNFR2 upon TNF␣ stimulation. The binding of downstream effectors to TNFR1 (A) and TNFR2 (C) was analyzed after stimulation for 5 min with IBOP, TNF␣, or the two in combination. TNFR1 or TNFR2 were immunoprecipitated and immune complexes were probed with antibodies to the proteins as indicated in the figure. In addition, lysates from stimulated HEC were probed using antibodies against STAT1 phosphorylated at Tyr 701 (B). Blots were reprobed with a monoclonal STAT1 antibody to assess loading (n ϭ four experiments). The role of STAT1 in regulating the association of proteins with TNFR1 and TNFR2 was assessed using an antisense oligonucleotide to deplete HEC of STAT1 with the sense oligonucleotides used as a control (panels A and C). The blots shown are representative of three individual experiments.
MAPK activation in extracts of HEC stimulated with IBOP, TNF␣, or both agents with phosphospecific antibodies. TNF␣ treatment induced a 4-, 8.5-, and 13-fold increase in p38, ERK1/2, and JNK phosphorylation, respectively (Fig. 6A). The increased MAPK phosphorylation was not diminished by IBOP co-stimulation. Because IBOP disrupts TNF␣-mediated TRAF-2-TNFR1 interactions, we posited that MAPK activation following TNF␣ was a response to either MADD/TNFR1 or TRAF-2/TNFR2 signaling. TRAF-2 associated with Ask1 and MAKK1 in TNF␣-stimulated HEC (Fig. 6B) and co-stimulation with IBOP did not influence their recruitment. Similarly, the association of MADD with TNFR1 in response to TNF␣ was unaffected by co-stimulation with IBOP (Fig. 6C). Thus, TP stimulation induces a functional bifurcation in TNFR2 signaling that maintains MAPK activation but inhibits NFB activation.
TP Does Not Inhibit Caspase Activation by TNF␣ and Enhances Endothelial Cell Apoptosis-Concurrent TP stimulation inhibits TNF␣-induced activation of a major cell survival pathway. TNF␣ also promotes cell death by activating caspases. Thus, we examined the effect of TXA 2 mimetics on TNF␣mediated caspase activation and cell survival. TNF␣-mediated caspase activation causes the association of TRADD and FADD (4). Immunoprecipitation of TNFR1 showed that FADD associated with the receptor when TNF␣ was present; this association was not altered by co-stimulation with IBOP (Fig. 7A). Association of FADD and TNFR1 activates caspase 8 and subsequently caspase 3. Thus, we next examined caspase activity in stimulated HEC lysates. Control cells contained little caspase 3-or 8-like enzyme activity; however, robust caspase 3 and 8 activation were observed in TNF␣-treated cells, which was not altered by IBOP (Fig. 7B). Thus, proapoptotic signaling from TNFR1 remained intact in the presence of TXA 2 . We next examined the effect of simultaneous TNF␣ and TP stimulation on endothelial cell survival. Stimulation with IBOP or TNF␣ alone increased annexin V staining by 2-fold in untreated HEC after 16 h (p Յ 0.01, Fig. 7C). In contrast, the use of both agents induced a 5-fold increase in endothelial cell apoptosis (p Յ 0.001). This synergistic increase in endothelial cell death is likely the result of the preservation of the proapoptotic cascade of TNFR1, with the inhibition of prosurvival pathways of both TNFR1 and TNFR2, such as NFB activation. Indeed, when HEC that had been transfected with oligonucleotides with the NFB consensus binding site were treated with TNF␣, endothelial cell apoptosis was increased 2-fold and ICAM-1, VCAM-1, and E-selectin expression was ablated (data not shown). Thus, the inhibition of TNF␣-induced NFB activity by TXA 2 stimulation appears to be sufficient to promote endothelial cell apoptosis.

DISCUSSION
In this study we found that ligands for TP, the G proteincoupled receptor for TXA 2 , abrogated the enhanced expression of proinflammatory markers on TNF␣-stimulated endothelial cells and inhibited leukocyte adhesion to TNF␣-treated endothelium in vitro and in vivo. Furthermore, in endothelial cells incubated with both agents, apoptosis was markedly increased, as TXA 2 mimetics prevented TNF␣ from initiating pathways linked to cell survival, but not those linked to apoptosis.
TP stimulation prevented NFB activation by both TNF␣ receptors. The cross-talk between the TXA 2 and TNF␣ receptors is unlikely to occur at a point downstream of the convergence of the TNFR1 and TNFR2 signaling pathways, as activation of NFB by IL-1␤, which shares these pathways, was unaffected by TP stimulation. Thus, mechanisms that universally inhibit NFB activation, such as induction of A20 and inhibition of IB kinase (4), do not appear to be involved. In addition, TRAF-2 bound to TNFR2, but not TNFR1, in HEC stimulated with TNF␣ and IBOP. Together these data indicate that TP-mediated inhibition of both TNFR1 and TNFR2 is facilitated via novel mechanisms.
Association of RIP and TRAF-2 with TNFR1 are necessary for NFB activation and cells deficient in either protein do not activate NFB fully in response to TNF␣ (30,31). Overexpression of STAT1 inhibits the association of RIP and TRAF-2 with TNFR1⅐TRADD complexes, but does not affect FADD⅐TRADD complex formation, after TNF␣ stimulation (25). Concurrent stimulation with TNF␣ and IBOP also interfered with TRADD-RIP/TRAF-2 interactions but not FADD binding or caspase activation. Furthermore, stimulation with TNF␣ and IBOP increased STAT1 phosphorylation on Tyr 701 and the associa- In panel A, lysates from stimulated HEC were probed using monoclonal antibodies to the phosphorylated forms of p38, ERK1/2, and JNK following stimulation for 5 min with IBOP, TNF␣, or the two in combination. Blots were reprobed with a monoclonal ␣-tubulin antibody to assess loading (n ϭ four experiments). The interaction of TNFR2 (B) and TNFR1 (C) with downstream effectors that influence MAPK activation were determined after TNFR1 or TNFR2 were immunoprecipitated and immune complexes were probed with antibodies against MADD, Ask1, or MEKK1. tion of TNFR1 and STAT1. Thus, TP enhanced TNF␣-mediated STAT1 activation; this may inhibit NFB activation by TNFR1, most likely by competitively inhibiting RIP/TRAF-2 binding to a TNFR1⅐TRADD complex. The TXA 2 -and G␣ q -mediated enhancement of TNF␣-stimulated STAT1 activation is another novel finding of the present study. This model of antagonism by TP ligands is an extension of the findings of Wang et al. (25), who showed that STAT1 overexpression prevented the recruitment of RIP to TNFR1 in a concentration dependent fashion. Other G protein-linked receptor ligands, such as angiotensin II and ␣ 1 adrenergic receptors, have also been reported to stimulate STAT1 (32,33). Whereas the pathways of STAT1 activation may differ, all these receptors are linked to G␣ q , which suggests the possibility of a common mechanism of G proteinmediated inhibition of TNF-stimulated NFB activation during inflammation and reperfusion.
TP stimulation also inhibited TNFR2 activation of NFB. TP stimulation did not inhibit TRAF-2-TNFR2 interactions, however; TP induced a functional bifurcation in TNFR2 signaling, leading to activation of MAPK, but not NFB. This division of TNFR2 signaling was mediated by selective inhibition of NIK-TRAF-2 interactions after TRAF-2 binds to TNFR2. We propose that recruitment of RIP to TNFR2 in the presence of IBOP mediates the selective inhibition of NIK binding. RIP overexpression alters the signaling pathway of TNFR2, "switching" the downstream effects from NFB activation to a cascade inducing programmed cell death (34). TP signaling prevents the RIP-TRADD interaction in response to TNF␣ by inducing translocation of STAT1 to TNFR1. In this model, RIP displaced from TNFR1 by STAT1 binds to TNFR2. Thus, the enhanced association of RIP with TNFR2 may aid the induction of apoptosis and prevent NFB activation by TNF␣ in the presence of TP stimulation. Indeed, the importance of STAT1 to this pathway of inhibition was highlighted by the experiments with antisense STAT1 oligonucleotides, indicating that the recruitment of RIP to TNFR2 was abrogated in the absence of STAT1.
These data present a model for the inhibition of TNF␣mediated NFB activation by IBOP. The effects of IBOP are more complicated than just mere inhibition of the NFB pathway, however. Initiating IBOP treatment after TNF␣ stimulation had achieved maximal expression reduced NFB activation to baseline. This response is characterized by the movement of NFB of the nucleus, decreased IB phosphorylation, and reassociation of the two proteins. Whereas the inhibition of proximal events in TNF␣ signaling by IBOP explain why no further NFB activation occurs, it does not fully explain the rapid reversion to baseline for NFB activated during the previous period. One possibility, proposed by Ghosh and Karin (35), is that IB may be recruited to the nucleus to recomplex to NFB and cause its redistribution to the cytoplasm. In the case of the present experiments, TP might be suspected of triggering a phosphatase that leads to dephosphorylation of IB. The ability to inhibit previously activated NFB is a powerful proapoptotic effect of TP stimulation; regardless of the order that HEC receive the stimuli, the result is the inhibition of NFB and the induction of apoptosis.
We found that TP stimulation does not affect TNF␣-induced activation of JNK, p38, and ERK1/2. Activation of MAPK by TNF␣ is dependent upon TRAF-2 (36,37). As the TNFR2-TRAF-2-Ask1, but not the TNFR1-TRADD-TRAF-2, pathway is preserved in the presence of TP stimulation, it is most likely the pathway that leads to MAPK activation. Induction of LAM requires MAPK activation and p38 is required for full transcriptional activity of NFB (26,38). The inhibition of TNF␣stimulated LAM expression, despite activation of all three MAPKs, indicates either that AP-1 and NFB activation are both required for LAM expression, or instead that NFB alone is necessary and sufficient. Interestingly, ERK, JNK, and p38 all promote EC apoptosis (39) and have direct links to caspase activation as well as other cell death pathways, such as the functional suppression of Bcl-2 (40). Thus, p38, JNK, and ERK activation in the presence of TNF␣ and TXA 2 may enhance the caspase activation mediated by the TRADD-FADD pathway, thus leading to the enhanced apoptosis observed in endothelial cells treated with both agents.
Although not emphasized in this report, the inhibition of NFB activation in HEC treated with TNF␣ and TXA 2 mimetics is reciprocal. The ability of TXA 2 mimetics to induce NFB activation is specific for a single TP isoform; TP␤, through a G␣ i -coupled signaling pathway not used by TP␣, activates NFB in response to IBOP and U46619. In contrast, the inhibition of TNF␣ signaling is potentially mediated through either isoform, as both couple to G␣ q . TNF␣ stimulation prevented the activation of NFB in TP-stimulated HEC. Furthermore, TNF␣ could also reverse the activation of NFB if cells were stimulated with IBOP prior to TNF␣. The mechanism by which TNF␣ controls G␣ i -mediated NFB activation is not yet identified.
Circulating levels of TXA 2 and TNF␣ are increased during ischemia. In addition, ischemic episodes increase the number of circulating endothelial cells by stimulating apoptosis (41). These endothelial cells were macrovascular in origin and did not express markers of "activation" such as LAM. We have found that co-stimulation with TXA 2 and TNF␣ produces a similar phenotype in macrovascular endothelial cells in vitro. Although TXA 2 does not inhibit NFB induction by all cytokines, e.g. IL-1␤, similar mechanisms might prevent activation of endothelial cell NFB by such cytokines. The reciprocal inhibition of NFB activation, such as that which occurs with TXA 2 and TNF␣, is sufficient for the induction of endothelial cell apoptosis and thus may increase the number of circulating endothelial cells during ischemia. Consistent with this hypothesis is the recent observation that blocking TNF␣ decreases the number of circulating endothelial cells (42).
Apoptosis mediates inflammation and leukocyte extravasation associated with renal injury by cleavage of the monocyteactivating polypeptide II protein (43). We speculate that apoptotic endothelial cells, exposed to TNF␣ and TXA 2 , could mediate leukocyte infiltration into ischemic tissues by a similar mechanism. Co-stimulation by TXA 2 and TNF␣ may propagate an inflammatory response by perturbing the vessel wall, inducing injury to the vasculature, and exposing the subendothelial matrix following loss of endothelial cells because of apoptosis. In addition, the induction of endothelial apoptosis may attenuate an angiogenic response, thus reducing revascularization in inflamed or ischemic tissues.
Taken together, these data describe a previously unrecognized interaction between two mediators of inflammation and reperfusion injury that may provide a mechanism by which endothelial cell injury occurs during these conditions. In addition to elucidating this mechanism, a rationale is provided for the use of a TP antagonist to preserve the cell survival properties of TNF␣ during inflammation and reperfusion.