Arachidonic Acid Influences Proinflammatory Gene Induction by Stabilizing the Inhibitor-κBα/Nuclear Factor-κB (NF-κB) Complex, thus Suppressing the Nuclear Translocation of NF-κB*

Arachidonic acid (AA), through its myriad metabolites, is involved in inflammation in a number of ways. AA is produced and released by several cell types, including endothelial cells (EC), and acts on a variety of cells. EC activation plays a key role in inflammation presumably by modulating the immune response through up- or down-regulation of several genes. We have previously shown that AA and its nonmetabolizable analogue, 5,8,11,14-eicosatetraynoic acid (ETYA), inhibit up-regulation of proinflammatory genes in EC. In the present study we identify a mechanism to explain the inhibitory effects: AA and ETYA both inhibit the translocation of nuclear factor-κB (NF-κB) to the nucleus by blocking the degradation of the inhibitor of NF-κB (IκB) and thus stabilizing the IκB/NF-κB complex. To investigate the mechanism whereby AA inhibits up-regulation of genes encoding proinflammatory mediators, we examined the ability of ETYA to inhibit tumor necrosis factor-α (TNF-α) mediated phosphorylation and degradation of IκBα. Western blot analysis revealed that preincubation of EC with ETYA for 40 min prior to stimulation with TNF-α inhibits the phosphorylation and degradation of IκBα. These findings establish a mechanism by which AA inhibits nuclear translocation of NF-κB and thereby explaining its modulatory role in the induction of proinflammatory genes.

Studies of arachidonic acid (AA) 1 have focused primarily on metabolites of AA and their role in inflammation, including thrombosis and clinical disorders such as asthma. While AA metabolites can have potent proinflammatory effects, prostacyclin and certain AA metabolites have also been shown to have antiinflammatory actions.
Little is reported on the effects mediated by AA itself. Endothelial cells (EC), which both release AA and serve as the possible targets of AA released by EC and other cells, play a prominent role in inflammation and hemostasis and organ rejection. The activation of EC results in the up-regulation in EC of genes encoding adhesion molecules, procoagulant factors, cytokines and other molecules, which can participate in the inflammatory-thrombotic responses characterizing a number of conditions including graft rejection. Stimulated by reports demonstrating a significant improvement in graft survival of animals treated with linoleic acid and other precursors of AA, we have previously demonstrated (1) that AA itself can inhibit the up-regulation in EC of several key genes/molecules involved in inflammation and rejection, including E-selectin, ICAM-1, and IL-8. Furthermore, using E-selectin as a prototype gene, we showed that inhibition is at the transcriptional level.
Transcriptional up-regulation of proinflammatory genes involved in EC activation is strongly dependent on activation of NF-B, which involves phosphorylation and degradation of IB leading to translocation of the NF-B heterodimer to the nucleus. The NF-B/IB␣ system has been shown to exert transcriptional regulation on proinflammatory genes involved in EC activation (2). Most genes encoding adhesion molecules, cytokines, and other proinflammatory genes have functional NF-B binding elements in their promoter regions. In the cytoplasm of quiescent EC, IB␣ forms a complex with NF-B and thereby prevents migration of NF-B into the nucleus (3). One mechanism by which an agent can inhibit NF-B activation is by preventing the phosphorylation and thus degradation of IB␣, referred to as "IB stabilization." We show in this paper that the inhibitory effect of AA on EC activation is due to interference at this level of transcriptional regulation.

EXPERIMENTAL PROCEDURES
Preparation of Porcine Aortic Endothelial Cells-Porcine EC were isolated and passaged in our laboratory as described previously (4). Briefly, EC were cultured in Dulbecco's modified Eagle's medium (DMEM) with 4.5 g/liter glucose and supplemented with 10% heatinactivated fetal bovine serum (Hyclone), L-glutamine and 50 units/ml penicillin/streptomycin. Experiments were performed in DMEM under serum-free conditions. Before incubating EC with polyunsaturated fatty acids (PUFA), cells were washed three times with prewarmed serum-free DMEM. Tumor necrosis factor-␣ (TNF) was diluted in serum-free medium and added to the EC; in cases where lipopolysaccharide (LPS) was used as a stimulus, fetal calf serum was added simultaneously with LPS to a final concentration of 5%. Human umbilical vein endothelial cells were a gift from Dr. B. Ewenstein, Brigham Women's Hospital, Boston, MA. Pig EC from passages 3 to 7 were used in these experiments.
Electrophoretic Mobility Shift Assay-Nuclear extracts from porcine and human EC were prepared as described (5). The double-stranded oligonucleotides used in all experiments were end-labeled using T4 polynucleotide kinase and [␥-32 P]ATP. After labeling, 5 g of nuclear extract was incubated with 100,000 cpm of labeled probe in the presence of 3 g of poly(dI-dC) at room temperature for 30 min followed by separation of this mixture on a 6% polyacrylamide gel in Tris/glycine/ EDTA buffer at pH 8.5. For specific competition, 7 pmol of unlabeled NF-B oligonucleotides was included; and for nonspecific competition, 7 pmol of the double-stranded mutant B oligonucleotides 5Ј-AGCTTA-GATTTTACTTTCCGGAGAGGA-3Ј and 7 pmol of pig E-selectin CRElike were used. For supershift assays, 1 l of the monoclonal anti-NF-B p56 subunit antibody (Boehringer Mannheim) was added to the nuclear extract simultaneously with the labeled probe.
Western Blot Analysis-Cytosolic extracts were prepared as described (5) except that further protease inhibitors, TPCK, TLCK, aprotinin, leupeptin, antipain, aprotinin, benzamidine, chymostatin, and pepstatin, were added. Equal amounts of protein were separated by SDS-polyacrylamide gel electrophoresis (12%), transferred to an Immobilon P polyvinylidene difluoride membrane using a semi-dry transfer cell (Bio-Rad) and probed with the rabbit polyclonal antibody C21 (Santa Cruz, Biotechnology) directed against IB␣ as described (23). Bands were visualized using horseradish peroxidase-conjugated donkey anti-rabbit IgG and the enhanced chemiluminescence assay (Amersham Life Science Inc., Arlington Heights, IL) according to the manufacturer instructions.
Statistic-All assays were done at least three times, and representative experiments are shown.

RESULTS AND DISCUSSION
We reported earlier (1) that AA and a nonmetabolizable AA analog, ETYA, inhibit the expression of several proinflammatory genes in EC. We studied E-selectin as an example of these genes and demonstrated by run-off analysis that the up-regulation of this gene is blocked at the transcriptional level. We now report that this inhibitory effect of AA and ETYA in EC is based on the stabilization of IB␣ by these 20:46 fatty acids, therefore blocking nuclear translocation of NF-B, a transcription factor known to be essential for the up-regulation of several genes that characterize EC activation (6). Fig. 1 represents an electrophoretic EMSA, showing a time course of TNF-induced NF-B translocation. EC were stimulated with TNF for 0, 10, 30, 60 and 120 min (Fig. 1, lanes TNF 10Ј, TNF 30Ј, TNF 60Ј, and TNF 120Ј, respectively), after which the nuclear extract was isolated as described under "Experimental Procedures." The EMSA shows that after only 10 min of stimulation with TNF, there is an easily detectable translocation of NF-B into the nucleus, which increases further with time. However, when EC were pre-treated with 55 M ETYA for 30 min followed by stimulation with 5 ng/ml of TNF for an additional 10, 30, 60, or 120 min (lanes E ϩ TNF 10Ј, E ϩ TNF 30Ј, E ϩ TNF 60Ј, and E ϩ TNF 120Ј, respectively), the TNF-induced translocation (activation) of NF-B into the nucleus was completely blocked. These results are not surprising as several of the AA metabolites are known to be involved in cell activation (7). In this and similar experiments on human umbilical vein cells (data not shown), 35 M AA consistently demonstrated a stronger net inhibitory effect on NF-B activation. Taken together, these data demonstrate that AA can inhibit the TNF-induced activation of NF-B and might explain the phenomenon, seen in other experiments, that higher doses of AA do not lead to a correspondingly greater inhibition of TNF-induced gene up-regulation, as higher amounts of AA result in increasing amounts of AA metabolites that activate NF-B. Thus the AA effect on NF-B activation reflects the net effect of inhibition and activation. The hypothesis that some AA metabolites might be responsible for activation, but that AA itself inhibits NF-B activation, is further supported by the fact that the stable and nonmetabolizable AA analog ETYA does not activate NF-B.
Also shown in Fig. 2   only leads to activation and translocation of NF-B.
We reported earlier (1) that the precursors of AA, linoleic acid and ␥-linolenic acid inhibited the TNF, LPS, or phorbol 12-myristate 13-acetate-induced up-regulation of several genes although, compared with AA or ETYA, higher doses of these unsaturated fatty acids had to be used to get comparable inhibition. We tested whether the inhibitory effect of the AA precursor, ␥-linolenic acid, is due to the same mechanism as that seen with AA. Fig. 3 15Ј-ETYA 285Ј).
The main objective of this work has been to expand our insight into the surprising effect of AA on gene regulation in endothelial cells as reported earlier (1). Nevertheless, to gain a better understanding of possible effects that polyunsaturated fatty acids might play in the regulation of inflammatory responses, we tested a panel of saturated, mono-, and polyunsaturated fatty acids for their effectiveness as inhibitors of NF-B translocation. Preincubation of EC with 65 M of ␥-linoleic acid (6,9,12-octadecatrienoic acid), dihomo-␥-linoleic acid (8,11,14eicosatrienoic acid), oleic acid (9-octadecenoic acid), 5,8,11,14,17-eicosapentaenoic acid, 4,7,10,13,16,19-docosahexaenoic acid, Mead acid (5,8,11-eicosatrienoic acid), 11,4eicosadienoic acid, and the saturated fatty acids arachidic acid (C20:0) and stearic acid (octadecanoic acid, C18:0) for 45 min prior stimulation with TNF (5 ng/ml) did not influence the TNF-induced translocation of NF-B (data not shown). Under our test conditions, none of these fatty acids lead to NF-B activation per se, as incubation of EC with these fatty acids, for a total of 135 min and without adding TNF, did not lead to activation of NF-B (data not shown).
Originally, ETYA has been synthesized as an AA analog, only later has it become clear that ETYA blocks the enzymes involved in metabolism of AA. Like AA, ETYA binds to cyclooxygenase and lipoxygenase. However, because of some structural differences, ETYA cannot be metabolized by these enzymes, and, therefore, bound ETYA blocks further binding of AA. Antioxidants have been shown to block NF-B activation (9 -11). To what degree AA metabolites might be involved in the effect seen with AA has been addressed in our previous work where we tested several AA metabolites for their effectiveness in EC gene suppression, none of them had any significant impact on TNF-induced E-selectin up-regulation. Several known cyclooxygenase and lipoxygenase inhibitors are also known to possess antioxidative properties, which excludes these reagents from further testing the unlikely hypothesis that ETYA acts through inhibition of prostaglandins or lipoxygenase metabolites. Besides, EC do not produce leukotriene, the lipoxygenase products of AA, without leukocyte-derived substrates (12). Furthermore, with the exemption of a few, such as prostaglandin A (11), most AA metabolites are known to be proinflammatory and therefore most likely leading to activation of NF-B (13). In the early events leading to NF-B activation by TNF, prostaglandin or lipoxygenase metabolites play essentially no role. In accordance with our conclusions are the results of experiments (data not shown) where NS-398, a potent inhibitor of cyclooxygenase (14) was used without any significant effect on TNF (5 ng/ml) induced NF-B activation. Similarly, treating EC with indomethacin (10 Ϫ4 and 10 Ϫ5 M, respectively) prior to stimulation with TNF has no influence on TNF (5 ng/ml) -induced NF-B translocation. The polyunsaturated fatty acid, 8,11-eicosadiynoic acid is another nontoxic acetylenic fatty acid that has been shown to inhibit eicosanoid biosynthesis at several stages (15). This unsaturated fatty acid failed to influence TNF-induced NF-B up-regulation in our experiments on EC. Lastly, we tested 9,12-octadecadiynoic acid a polyunsaturated fatty acid and irreversible inhibitor of cyclooxygenase and lipoxygenase, which is, compared on a mol to mol ratio, a stronger inhibitor of cyclooxygenase than ETYA (16,17). Our data indicate that incubation of EC with 10, 25, and 50 M of 9,12-octadecadiynoic acid prior to stimulation with TNF (5 ng/ml) for 1.5 h had significantly less suppressive effect on the TNF-induced NF-B activation than treatment with similar amounts of the "weaker" cyclooxygenase and lipoxygenase inhibitor ETYA (data not shown). These data support the concept that it is indeed the structural similarities between AA and ETYA that are responsible for the similarities seen on NF-B inhibition and that AA metabolites are not involved in the events leading to AA/ETYA-induced suppression on TNF-activated NF-B translocation.
Our working concept is based on the hypothesis that upon cell activation free AA is released intra-or extracellularly to keep inflammatory events localized or as a feedback mechanism to control the extent of inflammatory responses. For these reasons, the amount of free AA has to be tightly controlled. One method commonly used in experiments involving AA and other fatty acids is the binding of these compounds to carrier proteins such as albumin, where such binding seems to render AA inactive, as experiments performed in the presence of albumin demonstrated little or no influence on TNF-induced E-selectin up-regulation, a gene which has been shown to be NF-B dependent (data not shown). Due to their highly hydrophobic properties, fatty acids are difficult to keep evenly immersed in cell culture solutions. Binding of fatty acids to albumin facilitates the dispersion of AA in culture medium and therefore minimizes the loss due to binding of fatty acids to tubes as well as tissue culture material. For the above mentioned reason, this method was not appropriate in our experiments. To better understand the amount of AA being taken up by EC in our experiments performed under serum-free conditions, EC were incubated with 60 M AA pulsed with 3 H-AA for 45 min. Subsequently, the amount of 3 H-AA associated with EC was compared with 3 H-AA in culture medium. In these experiments, an equivalent of 9 M (15 Ϯ 4%) of the added 60 M AA was bound to EC.
An important mechanism in controlling NF-B activation is provided by a protein-protein interaction involving members of the IB family with NF-B. As demonstrated by others (18), IB binds to and forms a complex with the subunits (p65 and p50) of NF-B, thereby inhibiting transmigration of NF-B into the nucleus. Upon stimulation of cells with a wide array of reagents including TNF, IB is phosphorylated (19), ubiquitinated (20) and subsequently degraded. We tested the hypothesis that AA might influence the activation/translocation of NF-B by preventing the degradation of IB, which would result in stabilization of the IB-NF-B complex. Fig. 4 shows the result of a Western blot, where cytoplasmic extracts of porcine EC were separated on a 12% polyacrylamide gel, transferred to a polyvinylidene difluoride membrane and stained with an anti-IB␣ antibody as described under "Experimental Procedures." As shown in this figure and demonstrated by others (21), stimulation of EC with 5 ng/ml TNF leads to rapid proteolysis of IB␣, as virtually all the IB␣ present in the cytoplasm is degraded within 10 min. In contrast to the nearly nondetectable levels of IB␣ in the cytoplasm of cells stimulated with TNF only for 10 and 30 min (Fig. 4A, lanes TNF 10Ј and TNF 30Ј) PAEC pre-treated with 55 M ETYA for 30 min prior to addition of TNF for 10 or 30 min (Fig. 4A, lanes E ϩ TNF 10Ј and E ϩ TNF 30Ј), show no changes in the amount of IB␣ present in the cytoplasm. We achieved identical results when we used human umbilical vein endothelial cell instead of porcine EC to study the effect of ETYA on IB␣ stabilization (data not shown).
The first step leading to IB␣ degradation requires phosphorylation of IB␣ at serines 32 and 36 (22). The inducible phosphorylation of IB␣ is readily detected in Western blots as a slight decrease in protein mobility as deomonstrated by us (23) and others (24,25). As shown in Fig. 4A, treatment of EC with ETYA (ETYA 30Ј) only, or with ETYA followed by TNF (E ϩ TNF 10Ј), did not lead to changes in the mobility of IB␣. Gliotoxin has been shown to allow IB␣ phosphorylation but prevents its degradation (24). We performed identical control experiments using EC pretreated with gliotoxin (1 g/ml) followed by okadaic acid (OA) (0.5 or 1 M, respectively). The results of an experiment, shown in Fig. 4B, are consistent with the hypothesis that ETYA prevents the phosphorylation of IB␣ and therefore inhibits IB␣ degradation leading to the inhibition of NF-B translocation.
The controls for the EMSA are shown in Fig. 5. In the first lane (M), nuclear extracts of noninduced cells were separated and compared with nuclear extracts isolated from cells treated with 5 ng/ml TNF (lanes 2-5) for 1 h. The specificity of binding to the consensus NF-B element is demonstrated by the ability of unlabeled NF-B (cold) and the failure of a mutated NF-B element (mut) and a consensus CRE-element (CRE) to bind competitively to NF-B. The last line in this figure (p65)  antibody was incubated with the nuclear extract, it blocked the formation of the upper, slower migrating complex.
In summary, we have demonstrated that AA, independent of its metabolites, can play a key role in EC gene regulation. AA stabilizes the IB␣/NF-B complex and therefore suppresses NF-B translocation; given the key role of NF-B in the upregulation of proinflammatory genes with EC activation, this results in inhibition of gene induction. Despite the large amount of data collected on AA and its metabolites, the relationship of AA uptake and AA release to the production of metabolic compounds and their subsequent biological impact still remains obscure (26).
In the past, AA was merely seen as a precursor for bioactive eicosanoids. More recently it has emerged that AA itself can influence cellular communication systems at several levels (8,(27)(28)(29)(30)(31). Our findings strongly support and help to further establish the critical role of AA in intracellular signaling. Our results also bolster the development and use of AA analogs such as ETYA as antiinflammatory drugs. Whether our findings might provide further insight into the antiinflammatory mechanisms of actions of drugs such as aspirin, ibuprofen, and others, remains a matter of speculation. For a long time, aspirin, the most widely used nonsteroidal antiinflammatory drug, has been thought to act solely through inhibition of prostaglandins and thromboxanes. Only recently has it become clear that aspirin can block the translocation of NF-B, the key transcription factor in the activation of most proinflammatory genes (14). It seems likely, however, that these drugs, which block either cyclooxygenase, lipoxygenase, or both, may counteract inflammation by increasing intracellular levels of free AA, subsequently suppressing NF-B. This hypothesis, supported by our data, would offer yet another appealing explanation for the beneficial effect and the "mode of action" of these compounds.