Selective Inhibition of Cyclooxygenase-2 Expression by 15-Deoxy-Δ^12,1412,14-prostaglandin J₂ in Activated Human Astrocytes, But Not in Human Brain Macrophages¹ Free

One of the hallmarks of neurodegenerative and inflammatory pathologies is the increased number of activated astrocytes and microglia in response to the pathological stimulus (1, 2). Microglial cells express some of the markers normally associated with hemopoietic monocytes and are the resident macrophages of the brain. They are believed to play a central role in pathological damages of cerebral tissue (3, 4). Astrocytes, which are responsible for maintaining the homeostasis of the brain tissue, also participate to a large extent in the neuroimmune responses (5, 6). During the inflammatory process, the increase in proinflammatory cytokines such as IL-1β activates microglia and astrocytes to secrete several potentially toxic products, including lipid mediators and free radicals (3, 4, 5). Among these mediators, PGs and NO are the products of the inducible isoforms of cyclooxygenase (COX-2)³ and NO synthase (iNOS) enzymes, respectively (7). Overexpression of these proinflammatory enzymes has emerged as an important determinant of the cytotoxicity associated with inflammation in pathologies such as Alzheimer’s disease, cerebral ischemia, multiple sclerosis, and HIV-1 encephalitis (8, 9, 10, 11, 12). Important differences exist in the molecular regulation of these enzymes between rodent and human glial cells. For example, in rodents, microglia are capable of high levels of NO production, particularly after LPS stimulation in vitro (13). However, the ability of human microglial cells to produce high levels of NO after LPS or proinflammatory cytokine stimulation in vitro remains controversial (14, 15, 16, 17, 18, 19). In contrast, human astrocytes can produce NO after IL-1β and IFN-γ stimulation, but not after LPS treatment (14, 15, 16, 17). Therefore, to study the therapeutic potential of different anti-inflammatory drugs against human neuropathologies, we should consider the relevance of results obtained in rodent systems, compared with human glial cells.

The peroxisome proliferator-activated receptors (PPARs) are ligand-activated transcription factors belonging to the nuclear receptor family. Three subtypes of PPARs have been described: PPARα, β, and γ (20). Several clinically important compounds, such as the fibrate class of hypolipidemic drugs, the thiazolidinediones, some nonsteroidal anti-inflammatory drugs, and fatty acids, can bind to and activate these receptors (21, 22, 23, 24). Recent studies suggest that some PPAR ligands may be important anti-inflammatory agents, including several PGs, of which 15-deoxy-Δ^12,14-PGJ₂ (15d-PGJ₂) is the most potent (21, 22). In stimulated murine and human peripheral blood macrophages, 15d-PGJ₂ binds to and activates PPARγ, resulting in various anti-inflammatory events such as inhibition of iNOS, COX-2, gelatinase b expression, and inflammatory cytokine production (21, 22, 23). More recently, this has been extended to the study of the microglial activation in rodents. In rodent microglial cells activated with LPS, 15d-PGJ₂ has been proposed as a potent suppressor of brain macrophage activation because it down-regulates iNOS expression (25, 26). Because the regulation of iNOS expression differs between human and rodent cells (27), we examined the anti-inflammatory effects of 15d-PGJ₂ and other PPAR agonists on primary cultures of human microglia and astrocytes using COX-2 induction as an index of cell activation. Our results demonstrated that 15d-PGJ₂ inhibits COX-2 expression in activated astrocytes by interfering with the NF-κB pathway through a novel mechanism, whereas human brain macrophages were resistant to the inhibitory effect of this compound. These data suggest that different pathways exist for the anti-inflammatory effects of 15d-PGJ₂ in comparison with human astrocytes, rodent brain macrophages or peripheral macrophages, and human brain macrophages.

The preparation of cells from human fetal brain has previously been described (14, 28). Purified cultures of microglia and astrocytes were obtained from CNS cells 10–14 days after plating, as described (14). Briefly, microglial cells were released by circular shaking and selected by a 20-min adhesion cycle (>95% CD68/KiM-7-positive cells). After the release of microglial cells, the adherent cells remaining were trypsinized, plated, and passaged three to four times to obtain purified cultures of astrocytes (>95% glial fibrillary acidic protein-positive astrocytes). Cells were maintained in Eagle’s MEM supplemented with 10% FCS, 2 mM l-glutamine, and antibiotics. Cell viability was assessed using thiazolyl blue (MTT) assay, as described (29).

To stimulate cultures, cells were refed with serum-free medium containing the tested inducer, inhibitor, or agonist. The following reagents were purchased from the designated companies: IL-1β and IFN-γ (Roche Molecular Biochemicals, Gaithersburg, MD); clofibrate and ciglitazone (Biomol Research Laboratories, Plymouth Meeting, PA); 15d-PGJ₂ (catalog no. 18570), WY14643, carbaprostacyclin, oleic acid, 9-hydroxyoctadecadieroic acid, and rabbit polyclonal anti-COX-2 and anti-iNOS Abs (1:1000), mouse anti-COX-1 mAb (1:500), ovine COX-2, COX-1, and mouse iNOS standards, and human COX-2 cDNA (Cayman Chemicals, Ann Arbor, MI); mouse GADPH cDNA (Ambion, Austin, TX); actinomycin D, protease inhibitors, and LPS from Escherichia coli serotype O26:B6 (Sigma-Aldrich, St. Louis, MO); and Abs to NF-κBp50, NF-κBp65, STAT-1, c-Jun, c-Fos, PPARα, and PPARγ (Santa Cruz Biotechnology, Santa Cruz, CA). Concentrations of PPAR ligands were chosen from previous published studies (30, 31, 32, 33).

After the treatment of microglial cells and astrocytes, cells were lysed in Laemmli buffer containing 5% 2-ME and sonicated for 1 min. The protein concentration was measured by a noninterfering protein assay (Geno-Tech, St. Louis, MO). Cell lysate proteins (15–50 μg/lane) were loaded onto a denaturing SDS-polyacrylamide gel, electrophoresed, and transferred onto nitrocellulose membrane. The nitrocellulose membrane was then blocked in 5% nonfat dry milk in PBS containing 0.1% Tween 20. The membrane was incubated with the primary Ab 1 h at room temperature, followed by incubation with the secondary Ab conjugated to HRP. The blots were probed with the ECL Western blot detection system from Amersham Pharmacia Biotech (Piscataway, NJ).

Northern blots were performed on total cellular RNA isolated from cells using an RNA isolation kit from Qiagen (Chatsworth, CA). Ten micrograms of total RNA per lane were electrophoresed on a formaldehyde-containing 1.2% agarose gel and transferred to nylon-supported membranes. Hybridizations were conducted at 42°C with cDNA probes and labeled with [³²P]CTP by random priming in Ultrahyb (Ambion). After hybridization, membranes were washed twice for 20 min in 2× SSC phosphate/EDTA and 0.1% SDS at room temperature, twice for 20 min in the same solution at 55°C, and twice for 20 min in 0.1× SSC phosphate/EDTA and 0.1% SDS at 55°C. To verify equivalency of RNA loading in the different lanes, the blots were rehybridized to determine the levels of GAPDH mRNA. For the quantification of COX-2 and GAPDH mRNA signals, densitometry of bands was performed using ImageQuant (Molecular Dynamics, Sunnyvale, CA), and relative COX-2 mRNA were determined as the ratio COX-2/GAPDH.

Protein extraction from cells was conducted by a modification of the Andrews and Faller procedure (34) in ice-cold buffer C (20 mM Tris-HCl (pH 7.9), 1.5 mM MgCl₂, 420 mM NaCl, 0.2 mM EDTA, 1 mM dithiotreitol, 25% glycerol) containing protease inhibitors, 1 mM PMSF, antipain, leupeptin, aprotinin, pepstatin A, and chymostatin (10 μg/ml each). Protein concentration was determined by the method of Bradford (Bio-Rad, Richmond, CA). AP-1 consensus and mutant oligonucleotides were from Santa Cruz Biotechnology. Oligomers were designed and synthesized with their complementary strands from the distal and proximal NF-κB sites within −455 to −428 and −232 to −205 from the transcriptional start site (+1) on the human COX-2 promoter (35). The NF-κB sequences used were as follows: distal NF-κB (5′-CGGCGGCGGGAGAGGGGATTCCCTGCGCCC-3′); distal NF-κB mutant (5′-CGGCGGCGGGAGAGCTCATTCCCTGCGCCC-3′); proximal NF-κB (5′-AGACAGGAGAGTGGGGACTACCCCCTCTGC-3′); and proximal NF-κB mutant (5′-AGACAGGAGAGTGGCCACTACCCCCTCTGC-3′). The NF-κB binding sequence is underlined. Sense and antisense oligomers were annealed in 20 mM Tris-HCl (pH 7.6), 50 mM NaCl, 10 mM MgCl₂, and 1 mM dithiotreitol. The annealed oligonucleotides were phosphorylated at the 5′ end with [γ-³²P]ATP and T4 polynucleotide kinase. For EMSA, the DNA-binding assay (20 μl) contained 10 mM Tris-HCl (pH 7.9), 50 mM NaCl, 5 mM MgCl₂, 0.5 mM EDTA, 1 mM DTT, 10% glycerol (v/v), 4 μg poly(dI-dC) (Amersham Pharmacia Biotech), 5 μg nuclear extract, and 0.8 ng ³²P-labeled DNA fragment. The reaction mixture was incubated at room temperature for 30 min. In the supershift assays, the Abs to NF-κBp50, NF-κBp65, c-Fos, or c-Jun (2 μg/reaction) were incubated with the reaction mixture for 2 h at 4°C before the addition of ³²P-labeled DNA fragment. In cold competition assays, 100-fold molar excess of cold wild-type or mutant oligomers was used. The protein-DNA product was run onto a 4 or 6% nondenaturing polyacrylamide gel. The gel was then dried and subjected to autoradiography.

The rate of GST or glutathione reductase (GR) enzymatic activities was assessed by measuring the rate of conjugation of 1-chloro-2,4-dinitrobenzene with reduced glutathione (GSH) or the rate of NADPH oxidation in the presence of oxidized glutathione, respectively. Whole cell extracts were prepared in cold buffer (50 mM potassium phosphate (pH 7.5), 1 mM EDTA). Protein concentration was determined, and GST and GR activities were measured at 25°C, using enzymatic assay kits, according to the manufacturer’s instructions (Cayman Chemicals). GST and GR sp. act. were expressed as nanomoles of substrate converted per minute per milligram of cellular protein. Comparisons of means were conducted by Student’s t test; differences with a value of p ≤ 0.05 were considered statistically significant.

Previous studies have demonstrated that both human microglial cells and astrocytes activated by IL-1β are capable of enhanced production of PGs, whereas only human astrocytes can be induced by IL-1β in combination with IFN-γ to produce detectable levels of NO, under our conditions of culture (14, 15, 16, 28, 36). To investigate the effect of PPAR agonists on these inflammatory pathways in human cells, we first monitored COX-2 and iNOS protein expression in microglia and astrocytes after cytokine stimulation. As expected, COX-2 protein was induced by IL-1β in both microglia and astrocytes (Fig. 1,A). The induction was detectable from 4 to 6 h and peaked at 12 h after stimulation (data not shown). Under these culture conditions, IL-1β in combination with IFN-γ (but not LPS) also induced iNOS protein expression in human astrocytes, but not in human microglia (Fig. 1 B).

The effect of different natural and synthetic agonists for each class of PPAR was subsequently tested by Western blot analysis on cytokine-induced COX-2 and iNOS protein expression in glial cells. Both synthetic agonists of PPARα, clofibrate and WY14643, enhanced IL-1β-induced COX-2 expression in astrocytes and microglia (Fig. 2, A and B). Among PPARγ agonists tested, only 15d-PGJ₂ displayed a dose-dependent inhibitory effect on the induced COX-2 expression in astrocytes, but not in human microglia (Fig. 2, A and B). A specific and high-affinity agonist of PPARγ, ciglitazone, had no significant effect on COX-2 induction in astrocytes, suggesting that the effect of 15d-PGJ₂ was independent of PPARγ activation. The concentrations of 15d-PGJ₂ that inhibited COX-2 induction by IL-1β had no significant toxicity to astrocytes, as judged by the MTT assay (data not shown). The only semisynthetic agonist available for PPARβ, carbaprostacyclin, had no effect on the induced COX-2 expression in human astrocytes or microglia (Fig. 2, A and B). The various agonists tested had a slight or no effect on the constitutive expression of COX-1 isoform (Fig. 2, A and B) and did not affect COX-2 expression in unstimulated cells (data not shown). 15d-PGJ₂ has been shown to inhibit cytokine-induced iNOS expression in rodent microglial cells (25, 26). However, in human glial cells tested in this study, IL-1β plus IFN-γ induced iNOS expression only in astrocytes, but not in microglia (Fig. 1,B). Similar to the observation in rodent microglia, 15d-PGJ₂, but not ciglitazone, inhibited the induced iNOS expression in human astrocytes (Fig. 2,C). PPARα agonists (clofibrate and WY14643), which enhanced COX-2 expression in astrocytes and microglia, decreased cytokine-induced iNOS expression in astrocytes (Fig. 2 C).

Western blot analysis demonstrated that astrocytes and microglia expressed both PPARα and γ. IL-1β slightly increased PPARα expression in astrocytes and microglia (Fig. 3). In contrast, PPARγ expression was not affected by IL-1β treatment alone or in combination with 15d-PGJ₂ or ciglitazone (Fig. 3).

The anti-inflammatory effect of 15d-PGJ₂ has been described on murine and human peripheral macrophages (21, 22). We found that in glial cells, the inhibitory effect of 15d-PGJ₂ was not observed on COX-2 induction in human brain macrophages (microglia), but in astrocytes. This observation prompted us to further examine the mechanism of the differential effect of 15d-PGJ₂ on the two cell types. Northern blot analysis demonstrated a strong inhibition of IL-1β-induced COX-2 mRNA accumulation in human astrocytes in the presence of 15d-PGJ₂ (10-fold inhibition, Fig. 4). In human microglial cells, 15d-PGJ₂ only partially inhibited COX-2 mRNA induction by IL-1β (2-fold inhibition; Fig. 4), which would explain the absence of a significant decrease in COX-2 protein level in these cells (Fig. 2 B). COX-2 message stability is an important factor in regulation of its abundance due to the presence of large number of instability motifs in its 3′ untranslated region (37). We observed no effect of 15d-PGJ₂ on the stability of IL-1β-induced COX-2 mRNA in astrocytes and microglia (data not shown), indicating that 15d-PGJ₂ would mainly act as a transcriptional inhibitor of COX-2 expression.

The activation of the transcriptional factor NF-κB is crucial to the induction of many inflammatory response genes, including COX-2 (35, 38). The human COX-2 promoter contains two putative binding sites for NF-κB (35, 38). We investigated the effect of 15d-PGJ₂ on the binding activity of NF-κB to these sites, using EMSA. Low levels of constitutive NF-κB/DNA-binding activity were found in nuclear extracts from unstimulated astrocytes and microglial cells (Fig. 5). The NF-κB binding to the distal, but not the proximal site in the COX-2 promoter was enhanced after IL-1β treatment in both astrocytes and microglia and generated a fast migrating specific gel-shifted complex (Fig. 5). A 100-fold molar excess of cold wild-type NF-κB oligomer, but not of the cold mutant oligomer, completely abolished the formation of this complex, demonstrating the specificity of the complex (Fig. 5). The treatment with 15d-PGJ₂ prevented the formation of the NF-κB/DNA complex in IL-1β-stimulated astrocytes (Fig. 6, upper panel), whereas in microglia only a slight decrease in this complex was observed (Fig. 6, lower panel). This complex was formed by both p50 and p65 NF-κB subunits, because it was supershifted with either anti-p50 or anti-p65 Abs, resulting in high-m.w. NF-κB bands migrating near the origin (Fig. 6). These results suggest 15d-PGJ₂ acts to inhibit NF-κB binding in astrocytes.

In contrast to 15d-PGJ₂, PPARα ligands enhanced IL-1β-induced COX-2 expression in both astrocytes and microglia (Fig. 2,C). Therefore, we investigated whether the effect of these ligands could also involve the NF-κB activity. To address this question, we tested the effect of PPARα ligand, WY14643, on the IL-1β-induced NF-κB activity in astrocytes and microglia. In contrast to 15d-PGJ₂, the treatment of cells with WY14643 had no effect on the activation of NF-κB binding in IL-1β-stimulated astrocytes and microglia (Fig. 7).

The proteins that comprise NF-κB are subject to complex modes of regulation, including cytoplasmic compartmentalization and retention by inhibitory proteins (I-κB), phosphorylation, and various associations of active subunits (39). Phosphorylation of I-κB and the consequent nuclear import of NF-κB complex upon cellular stimulation have been the most studied in the regulation of the NF-κB activity. We next investigated whether the differential effect of 15d-PGJ₂ on NF-κB transcriptional activity in human astrocytes and microglia was due to a differential alteration of its nuclear translocation. We monitored the levels of NF-κBp65 and p50 expression by Western blotting in both total and nuclear extracts from IL-1β-stimulated astrocytes and microglia in presence or absence of 15d-PGJ₂. IL-1β induced NF-κBp50 expression in astrocytes and microglia, while the level of NF-κBp65 was not significantly modulated by IL-1β. 15d-PGJ₂ had no effect on the amount of NF-κBp65 in total extracts from astrocytes or microglia, whereas in nuclear extracts it slightly decreased NF-κBp65 in astrocytes, but not in microglia (Fig. 8,A). In contrast, treatment with 15d-PGJ₂ strongly prevented the increase in NF-κBp50 protein induced by IL-1β in total extracts from human astrocytes, but not from microglia (Fig. 8,A). Consistently, the nuclear levels of NF-κBp50 were also decreased in IL-1β-activated astrocytes after treatment with 15d-PGJ₂ (Fig. 8 A). These results demonstrate that 15d-PGJ₂ modulates NF-κB transcriptional activity in human astrocytes at least through repression of the NF-κBp50 induction by IL-1β. In human microglia, no significant effect of 15d-PGJ₂ on NF-κBp65/p50 could be evidenced.

Cyclopentenone PGs are actively transported into cells via a membrane transporter (40). The absence of effect of 15d-PGJ₂ on the induction of COX-2 and NF-κB-binding activity in human brain macrophages could be due to a less efficient transport of the PG, as compared with astrocytes. In peripheral macrophages, 15d-PGJ₂ has been shown to interfere with other transcription factors such as AP-1 family (c-Fos, c-Jun) and STAT-1 that are also important in various proinflammatory pathways (21). To investigate whether 15d-PGJ₂ could affect inflammatory pathways other than NF-κB in human astrocytes and microglia, we studied the expression of c-Fos, c-Jun, and STAT-1 in activated astrocytes and microglia after treatment with 15d-PGJ₂. IL-1β-induced STAT-1 protein accumulation was reduced in both astrocytes and microglial cells, whereas c-Jun protein accumulation was enhanced in both cells after treatment with 15d-PGJ₂ (Fig. 8,B). In contrast, 15d-PGJ₂ induced c-Fos protein accumulation only in astrocytes, but not in microglia (Fig. 8,B). To verify that the increase in c-Fos and c-Jun corresponded to the active form of these proteins, EMSA experiments using an AP-1 consensus oligonucleotide were conducted (Fig. 8,C). The data demonstrated that 15d-PGJ₂ enhanced the binding activity of AP-1 complex in nuclear extracts from both astrocytes and microglia. Consistent with the Western blotting results, supershift experiments in presence of anti-c-Fos and anti-c-Jun Abs demonstrated that the AP-1 complex formed in astrocytes comprised c-Fos/c-Jun heterodimers after treatment with 15d-PGJ₂, whereas it contained mainly c-Jun homodimers in microglia (Fig. 8 C). These data indicate that 1) the entry of 15d-PGJ₂ also occurs in human brain macrophages; and 2) 15d-PGJ₂ does not act only as an inhibitor of inflammatory response through NF-κB inhibition in human astrocytes, but can also activate other inflammatory pathways through the increase in c-Fos and c-Jun expression in astrocytes and microglia.

Our data presented an important difference in the ability of 15d-PGJ₂ to inhibit NF-κB-binding activity in astrocytes as compared with human microglial cells. Cyclopentenone PGs are highly reactive with nucleophilic agents such as the thiol groups of nuclear proteins (40). In contrast, the transported PG in the cytosol may also conjugate with intracellular GSH, inhibiting the binding of the PG to the nuclear proteins (41). This event can be modulated by two enzymatic activities: 1) the cytoplasmic GST that is responsible for the metabolic elimination of cyclopentenone PGs by conjugation to GSH (41), and 2) the GR that catalyzes the reduction of oxidized glutathione to GSH and therefore maintains adequate levels of reduced cellular GSH. Thus, we compared GST and GR activities in human astrocytes and microglial cells. No significant difference in the GST sp. act. was observed between untreated or IL-1β-treated astrocytes or microglial cells (data not shown). However, a 35% decrease in the GR sp. act. was detected after IL-1β treatment only in human astrocytes, but not in microglia (448 ± 13 vs 293 ± 50 nmol/min/mg protein, untreated vs IL-1β-treated astrocytes, n = 3, p < 0.05). This suggests that intracellular GSH levels in astrocytes might decrease after IL-1β treatment, and thereby could increase the ability of the transported 15d-PGJ₂ to interfere with NF-κB pathway.

The activation of brain macrophages (microglia) is a key event associated with the neurodegeneration seen in many neuroinflammatory pathologies such as Alzheimer’s disease and HIV-1 encephalitis. The cyclopentenone PG, 15d-PGJ₂, is considered as a potent negative regulator of macrophage activation on murine and human peripheral blood macrophages (21, 22). This compound has been shown to also exert anti-inflammatory effects on rodent brain macrophages, based on the study of LPS-induced iNOS down-regulation by this PG (25, 26). However, in human microglial cells, studies of iNOS induction by proinflammatory stimuli in vitro led to apparently controversial results (14, 15, 16, 17, 18, 19). Several reports, including the observations reported in this work, failed to detect iNOS in purified cultures of human microglial cells (14, 15, 16), whereas other studies have shown that human microglia are capable of synthesizing iNOS in response to various stimuli in vitro (17, 18, 19). In latter studies, microglial cells were probably preactivated by culture conditions such as the use of a cytokine mixture (17) or up to 48–50 cell passages (18, 19) to expand the microglia before performing the experiments. In support of this possibility is the fact that iNOS mRNA and protein were also detected in unstimulated microglial cells in these studies (17, 18, 19). In addition, the presence of iNOS protein in human microglial cells has been shown in situ, under certain pathological conditions (12, 42). Altogether, these data suggest that a high degree of cell activation is necessary for human microglia to express iNOS protein, as compared with rodent microglia. Thus, to study in vitro the anti-inflammatory potential of compounds in human pathologies, using iNOS induction as an index of microglial activation in rodent cultures does not appear appropriate. In this study, we investigated anti-inflammatory effects of 15d-PGJ₂ on human astrocytes and microglia, using COX-2 induction as an index of cell activation. Our results demonstrated that in contrast to peripheral macrophages and rodent brain macrophages, human brain macrophages are not subject to the anti-inflammatory effect of 15d-PGJ₂ in terms of COX-2 inhibition. However, 15d-PGJ₂ inhibited COX-2 protein induction in human astrocytes by suppressing IL-1β-induced COX-2 mRNA accumulation. In human brain macrophages, the inhibition level of IL-1β-induced COX-2 mRNA was much lower than in astrocytes and was not sufficient to inhibit COX-2 protein accumulation. 15d-PGJ₂ is a ligand for PPARγ, but both PPARγ-dependent and independent mechanisms have been suggested for its anti-inflammatory effects among different cell types studied (21, 22, 23, 25, 26). Although human astrocytes expressed detectable levels of PPARγ, our results suggest that the effect of 15d-PGJ₂ on COX-2 induction in astrocytes is independent of PPARγ activation, because 1) the inhibition of COX-2 did not occur in human microglia that also express PPARγ; 2) a selective synthetic PPARγ agonist, ciglitazone, was not able to inhibit COX-2 in astrocytes; and 3) the inhibition of COX-2 in astrocytes was observed at higher 15d-PGJ₂ concentrations than those required for the PPARγ activation.

We subsequently investigated the effect of 15d-PGJ₂ on NF-κB-binding activity in human astrocytes and microglia. We found that IL-1β induced the NF-κB binding to the distal, but not the proximal, site in the COX-2 promoter in both human astrocytes and microglia. Interestingly, in human vascular endothelial cells, hypoxia has been shown to induce COX-2 expression via an enhanced NF-κB binding to the proximal site of the promoter but not the distal site (35). This suggests that the differential use of the NF-κB binding sites in human COX-2 promoter depends on the inducer and/or tissue-specific factors. Therefore, it is conceivable that other factor(s) might be involved in the NF-κB-binding activity to the human COX-2 promoter. 15d-PGJ₂ suppressed the NF-κB binding to the distal site in COX-2 promoter specifically in astrocytes. In human microglia, 15d-PGJ₂ had almost no effect on NF-κB-binding activity, which was consistent with the low level of inhibition of COX-2 mRNA accumulation and the absence of effect on COX-2 protein accumulation. In a murine macrophage cell line, 15d-PGJ₂ was shown to directly prevent the NF-κB-binding activity through multiple mechanisms, including modification of I-κB kinase activity, which reduces the NF-κBp65 nuclear translocation, and by direct modification of DNA binding domain of NF-κBp65 (43). Interestingly, 15d-PGJ₂ could affect NF-κB-binding activity via either mechanism depending on the cell type (43). Consistent with the absence of effect of 15d-PGJ₂ on COX-2 induction and NF-κB/DNA-binding activity in human brain macrophages, we found no changes in the total or nuclear accumulation of both NF-κB subunits in these cells after treatment with 15d-PGJ₂. In human astrocytes, although 15d-PGJ₂ slightly decreased the nuclear amount of NF-κBp65, it mainly acted by reducing the IL-1β induction of NF-κBp50 subunit. This suggests a new mode of action of 15d-PGJ₂ on NF-κB activation pathway in human astrocytes different from that described in murine peripheral macrophages (43). In contrast, we found that the expression and the DNA-binding activity of AP-1 proteins (c-Fos and c-Jun) were induced in human astrocytes and microglia in presence of 15d-PGJ₂. These data demonstrate that 15d-PGJ₂ can differentially affect transcriptional factors among different cell types because it has been shown to down-regulate AP-1-binding activity in peripheral macrophages (21). Regulatory sequences corresponding to the activating transcription factor (ATF)/CRE site have been reported in the 5′-flanking region of the human COX-2 gene (for a review, see Ref. 44). This site is typically activated by hetero- and homodimers of the c-Fos, c-Jun, ATF, and CREB. The ATF/CRE site has been shown to be critical for the COX-2 gene induction in several cell types, such as fibroblasts or epithelial cells (44). However, the relative importance of one transcription factor in the induction of the COX-2 gene over the others seems to be dependent on the stimulus and the cell type (44). In the case of human astrocytes, the increase in AP-1-binding activity by 15d-PGJ₂ could suggest an antagonistic role of these factors in COX-2 expression. In human microglia, 15d-PGJ₂ also increases IL-1β-induced AP-1-binding activity with no effect on COX-2 expression, perhaps due to the maintained activation of NF-κB. Our data also suggest that, besides the negative feedback on COX-2 expression, 15d-PGJ₂ could lead to the activation of other inflammatory response genes involving AP-1 protein activation.

The unresponsiveness of human brain macrophages to the anti-inflammatory effect of 15d-PGJ₂ in terms of COX-2 and NF-κB inhibition is surprising and is apparently specific to this macrophage population, because inflammatory pathways have been shown to be down-regulated by this compound in human peripheral blood macrophages and peritoneal macrophages (21, 22). We observed that 15d-PGJ₂ could affect the expression of other transcription factors, such as c-Jun and STAT-1, in human microglia. This indicates that the transport of 15d-PGJ₂ occurs in microglia. The effect of 15d-PGJ₂ on AP-1 and STAT-1 factors might likely occur via PPARγ activation, as suggested by other studies (21). Reactive oxygen species are believed to be important in the activation of NF-κB by a mechanism that is not completely understood (45). Among free radicals, superoxide anion production is induced upon inflammatory activation in microglia (14). However, it is not clear whether this production of superoxide anions could be in part responsible for antagonizing the repression of NF-κB activation by 15d-PGJ₂ in these cells. Another biological factor that interferes with 15d-PGJ₂ is the glutathione (GSH) that can conjugate with the cyclopentenone ring of the PG (40, 41). High levels of GSH could result in the quenching of 15d-PGJ₂, thereby inhibiting its binding to nuclear proteins (41, 43). Thus, a difference in the intracellular levels of GSH could also explain the difference in the effect of 15d-PGJ₂ observed between human microglia and astrocytes. In support of this hypothesis, we observed decreased activity of GR in IL-1β-stimulated astrocytes, which could lead to a decrease in the intracellular levels of GSH in these cells.

Under pathological conditions, COX-2 and iNOS are up-regulated within a single cell type or tissue (7). Most of the anti-inflammatory drugs, such as dexamethasone, down-regulate both enzymes. The general inhibition of inflammatory product secretion by such drugs may explain the inconsistency of their effects on the neurological lesions in different in vivo experimental systems for neuroinflammatory disorders (46, 47, 48). In addition to 15d-PGJ₂, we also tested the effect of synthetic activators of PPARα, clofibrate and WY14643, on astrocyte and microglial activation. We demonstrated that PPARα agonists increased COX-2 expression in human astrocytes and microglia stimulated with IL-1β, while they inhibited iNOS induction by IL-1β plus IFN-γ in human astrocytes. This finding provides, for the first time, the possibility of an inhibition of iNOS uncoupled from that of COX-2 in human astrocytes. A PPARα-dependent effect of clofibrate and WY14643 in activated glial cells would be suggested by our following observations: 1) both agonists exerted similar effects; 2) they exerted their effect only on IL-1β-activated glial cells, but not on unstimulated cells, and IL-1β stimulation of glial cells enhanced the amount of PPARα protein; and 3) the effect of WY14643 did not occur through modulation of IL-1β-induced NF-κB activity.

In conclusion, 15d-PGJ₂ and PPARα agonists exert differential effects on the inflammatory response among human astrocytes and microglial cells, in comparison with peripheral macrophages. Further studies on the molecular modes of action of these molecules will reinforce our understanding of the mechanisms that regulate inflammation in brain and should provide new insights in the treatment of neuroinflammatory diseases in humans.

I thank Dr. Eugene Major for giving me the privilege to accomplish this study and for his full support throughout this work. I also thank Blanche Curfman and Peter Jensen for careful reading of the manuscript.