Modulation of Cytokine-Induced Cyclooxygenase 2 Expression by PPARG Ligands Through NFκB Signal Disruption in Human WISH and Amnion Cells

Abstract

Cyclooxygenase (COX) activity increases in the human amnion in the settings of term and idiopathic preterm labor, contributing to the generation of uterotonic prostaglandins (PGs) known to participate in mammalian parturition. Augmented COX activity is highly correlated with increased COX2 (also known as prostaglandin-endoperoxide synthase 2, PTGS2) gene expression. We and others have demonstrated an essential role for nuclear factor κB (NFκB) in cytokine-driven COX2 expression. Peroxisome proliferator-activated receptor gamma (PPARG), a member of the nuclear hormone receptor superfamily, has been shown to antagonize NFκB activation and inflammatory gene expression, including COX2. We hypothesized that PPARG activation might suppress COX2 expression during pregnancy. Using primary amnion and WISH cells, we evaluated the effects of pharmacological (thiazolidinediones) and putative endogenous (15-deoxy-Δ¹²,¹⁴-prostaglandin J₂, 15d-PGJ₂) PPARG ligands on cytokine-induced NFκB activation, COX2 expression, and PGE₂ production. We observed that COX2 expression and PGE₂ production induced by tumor necrosis factor alpha (TNF) were significantly abrogated by 15d-PGJ₂. The thiazolidinediones rosiglitazone (ROSI) and troglitazone (TRO) had relatively little effect on cytokine-induced COX2 expression except at high concentrations, at which these agents tended to increase COX2 abundance relative to cells treated with TNF alone. Interestingly, treatment with ROSI, but not TRO, led to augmentation of TNF-stimulated PGE₂ production. Mechanistically, we observed that 15d-PGJ₂ markedly diminished cytokine-induced activity of the NFκB transcription factor, whereas thiazolidinediones had no discernable effect on this system. Our data suggest that pharmacological and endogenous PPARG ligands use both receptor-dependent and -independent mechanisms to influence COX2 expression.

INTRODUCTION

Prostaglandins (PGs) that elicit myometrial contractions and cervical maturation are pivotal to the onset and maintenance labor in humans and all other mammalian species that have been investigated [1]. The exact triggers for the onset of parturition at term are incompletely understood, but there is a growing consensus that cytokines, such as interleukin 1β (IL1B) and tumor necrosis factor alpha (TNF), are instrumental in unleashing vigorous biosynthesis of PGE₂ and PGF_2α within intrauterine tissues, leading to active labor and birth [2]. In addition, it appears that many of the fundamental biological forces governing labor at term are also evoked during episodes of preterm labor elicited by inflammatory insults to the mother [3].

While the majority of parturition research has focused on the feed-forward mechanisms through which uterine stimulants are generated, a relevant question remains unanswered: What biological inputs keep the uterus from contracting and the cervix from ripening during the vast majority of pregnancy? It is reasonable to hypothesize that tonically active signals prohibit myometrial activation before term and that these biomolecules are of fundamental importance for a thorough understanding of the mechanisms governing parturition. In this regard, it has recently been reported that PGs of the J₂ series may exert anti-inflammatory activity in a variety of circumstances [4]. Peroxisome proliferator-activated receptor gamma (PPARG) is a nuclear receptor whose ligands include PGs of the J₂ series (particularly 15-deoxy-Δ¹²,¹⁴-prostaglandin J₂, 15d-PGJ₂) in addition to polyunsaturated fatty acids, products of lipoxygenase metabolism, and oral hypoglycemic agents (thiazolidinediones) [5, 6]. Several groups have reported that PPARG antagonizes the expression of inflammatory signals in a broad array of experimental contexts [7, 8], including human gestational tissues [9]. Moreover, some PPARG ligands inhibit the expression of inflammatory gene products, such as inducible nitric oxide synthase (also known as nitric oxide synthase 2A, NOS2A), cyclooxygenase 2 (COX2, also known as prostaglandin-endoperoxide synthase 2, PTGS2), and microsomal prostaglandin E synthase-1 (also known as prostaglandin E synthase, PTGES) by mechanisms that may be PPARG-dependent and/or -independent [7, 8, 10, 11].

A common theme among the proinflammatory enzymes cited above is their regulation at the transcriptional level by nuclear factor κB (NFκB) [12, 13]. The genes encoding these inflammatory proteins harbor one or more copies of a κB-binding motif through which transcription may be enhanced [12]. For example, cytokines use NFκB within intrauterine cells to stimulate expression of the COX2 gene [14–17], whose product catalyzes the committing and rate-limiting step in uterotonic PG formation [18]. Recent evidence suggests that COX2-mediated synthesis of PGD₂ metabolites (including the PPARG ligand, 15d-PGJ₂) may provide a mechanism for feedback control of PG biosynthesis [4, 19]. Furthermore, we recently reported that a reciprocal relationship exists between the expression of COX2 and PPARG proteins in fetal membranes obtained from women before the onset of labor compared with tissues collected following delivery [20]. Thus, in the present study, we examined the mechanism by which known PPARG ligands govern COX2 expression in WISH cells and primary cultures of human amnion.

MATERIALS AND METHODS

Materials

Recombinant human TNF was purchased from R&D Systems (Minneapolis, MN). Antibodies against inhibitory factor κBα (IκBα, also known as nuclear factor of kappa light polypeptide gene enhancer in B-cells inhibitor alpha, NFKBIA), IκB kinase α (IKKα, also known as conserved helix-loop-helix ubiquitous kinase, CHUK), IκB kinase β (IKKβ, also known as inhibitor of kappa light polypeptide gene enhancer in B-cells kinase beta, IKBKB), COX2, and NFκB subunits p65 (also known as reticuloendotheliosis viral oncogene homolog A, RELA), p50 (also known as nuclear factor of kappa light polypeptide gene enhancer in B-cells 1, NFKB1), p52 (also known as nuclear factor of kappa light polypeptide gene enhancer in B-cells 2, NFKB2), cRel (also known as reticuloendotheliosis viral oncogene homolog, REL), and RelB (also known as reticuloendotheliosis viral oncogene homolog B, RELB) were obtained from Santa Cruz Biotechnology (Santa Cruz, CA). Mouse anti-rabbit glyceraldehyde-3-phosphate dehydrogenase (GAPD) antibody, which is cross-reactive with the human isoform, was purchased from Chemicon International (Temecula, CA). Antibodies recognizing phosphorylated IκBα (Ser³²) and IKKα (Ser¹⁸⁰)/IKKβ (Ser¹⁸¹) were from Cell Signaling Technology (Beverly, MA), as were rabbit anti-human PPARG antibodies. A second PPARG antibody was purchased from Affinity Bioreagents (Golden, CO). The 1.8-kilobase (kb) cDNA fragment used for COX2 mRNA Northern blotting was a kind gift from Dr. Timothy Hla (University of Connecticut, Farmington, CT). Arachidonic acid, all PGs, rosiglitazone, and PGE₂ ELISA kits were obtained from Cayman Chemical (Ann Arbor, MI). Troglitazone was from BIOMOL (Plymouth Meeting, PA). DIG Nucleic Acid Detection and DIG-High Prime kits were purchased from Roche Diagnostics (Indianapolis, IN). Assays-on-Demand gene expression target assay mix (Hs00153133 m1), 18S rRNA assay mix, and TaqMan Universal Master Mix were obtained from Applied Biosystems (Foster City, CA). SuperSignal chemiluminescent detection reagents were obtained from Pierce Biotechnology (Rockford, MA). Prolong antifade mounting reagent and Alexa Fluor-594-conjugated goat anti-rabbit antibodies were purchased from Molecular Probes (Eugene, OR). The NFκB consensus oligonucleotide was obtained from Promega (Madison, WI). All other reagents, unless otherwise specified, were obtained from Sigma (St. Louis, MO).

Cell Cultures

Human WISH cells were obtained from the American Type Culture Collection (CCL-25) and maintained in Ham F-12/Dulbecco modified Eagle medium (F-12/DMEM, Invitrogen, Carlsbad, CA) supplemented with 2 mM l-glutamine, 1 mM sodium pyruvate, and 10% (v/v) newborn calf serum. Cells were grown at 37°C in a humidified atmosphere of 95% air/5% CO₂ and used for experiments between the 3rd and 25th passages.

Primary cultures of human amnion cells were obtained at the time of uncomplicated scheduled cesarean delivery at term in the absence of labor or membrane rupture. There were no clinical signals of infection in any of these subjects. Informed consent was obtained in all cases following a protocol approved by the Institutional Review Board of The Ohio State University. Cultures were established according to the method of Okita et al. [21], with modifications. Briefly, amnion tissue was dissected from the choriodecidua within minutes of delivery, placed in sterile Hanks buffered salt solution (HBSS), and transported to our laboratory for aseptic processing. Amnion tissue was washed in ice-cold HBSS until free of blood, minced into ~1-cm² fragments, and digested for 30 min in HBSS containing 0.4% (w/v) trypsin at 37°C with constant rocking. Undigested material was recovered and fresh trypsin solution was added. Following an additional 30-min incubation, cells were strained through 60 mesh and the flow-through was collected. This process was repeated using the remaining undigested tissue. The flow-through from the second and third digestions was pooled and cells were collected by centrifugation at 600 × g for 10 min. Cells were then washed, plated into tissue culture dishes, and grown in F-12/DMEM supplemented with 2 mM l-glutamine, 1 mM sodium pyruvate, 50 μg/ml gentamicin sulfate, and 10% (v/v) fetal bovine serum. Within 7–10 days, cultures had attained confluence and were used for experiments.

Northern Blot Analysis

To assess COX2 mRNA expression in WISH cells following treatments, total RNA was extracted using TRIzol according to manufacturer’s specifications (Invitrogen). Total RNA (20 μg/lane) was fractionated on a 1% (w/v) agarose gel containing 2% (v/v) formaldehyde, transferred to a nylon membrane, and immobilized using ultraviolet irradiation. Membranes were then probed with digoxigenin (DIG)-containing cDNA probes labeled using the DIG-High Prime kit. A 1.8-kb cDNA fragment encoding human COX2 and a 0.6-kb cDNA fragment encoding human GAPD were used for detection. Blots were hybridized overnight at 42°C in 7% (w/v) SDS, 50% (v/v) formamide, 325 mM sodium chloride, 32.5 mM sodium citrate, 50 mM sodium phosphate (pH 7.0), 0.1% (w/v) N-lauroylsarcosine, 50 μg/ml sheared salmon sperm DNA, and 2% (w/v) DIG Blocking Reagent (Roche). Bound probes were identified using the DIG Nucleic Acid Detection kit (Roche). Chemiluminescent signals were detected using the VersaDoc Imaging System and analyzed using Quantity One software (Bio-Rad Laboratories, Hercules, CA).

Real-Time Reverse Transcriptase-Polymerase Chain Reactions

Following treatments, total RNA was extracted from amnion cell cultures using TRIzol (Invitrogen). Because total RNA yields from amnion cells were typically ~10% that of the WISH cells, quantities were insufficient to allow for detection by Northern blotting using the relatively insensitive DIG-labeled cDNA probes. Therefore, relative abundance of COX2 mRNA was performed by real-time reverse transcriptase-polymerase chain reactions (RT-PCR) detection using an ABI PRISM 7700 sequence detector (Applied Biosytems, Foster City, CA). Two μg/reaction of total RNA were reverse transcribed to cDNA using random hexamer primers and SuperScript II reverse transcriptase (Invitrogen), according to instructions provided by the manufacturer. Human COX2 expression was detected using the Assays-on-Demand gene expression target assay mix (Hs00153133 m1; Applied Biosystems), which contained a mixture of specific primers and a hydrolysis probe containing the fluorescent dye 6-FAM. As the reporter probe sequence was designed to detect a region spanning multiple intron-exon junctions, target cDNA was detected selectively, irrespective of the presence of contaminant genomic DNA. A primer/VIC-labeled probe set for detecting 18S rRNA (Applied Biosystems) was used as an internal control in each reaction. Amplification mixtures contained 2.5 μl 20× target assay mix (Applied Biosystems), 1 μl first-strand cDNA synthesis mixture (corresponding to 50 ng reverse transcribed RNA), 0.5 μl of 18S rRNA control mix (Applied Biosystems), 25 μl of 2× TaqMan Universal Master Mix (Applied Biosystems), and nuclease-free distilled deionized water to a total volume of 50 μl. Amplification was performed over 40 cycles of denaturation at 95°C for 15 sec and annealing/extension at 60°C for 1 min.

For each treatment condition, duplicate amplifications were performed. Relative abundance of COX2 mRNA was normalized to the 18S rRNA internal standard and calculated as 2^−ΔC_T, where ΔC_T denotes difference between the sample C_T (cycle threshold) and that of the reference message present in cells under nonstimulated conditions. Amplification curves were analyzed using software provided by the manufacturer (Applied Biosystems).

Electrophoretic Mobility Shift Assay

Cells were treated with the test substances indicated and nuclear protein extracts were prepared [22]. Nuclear protein extracts (10 μg/reaction) were then mixed with an oligonucleotide (5′-AGTTGAGGGGACTTTCC CAGGC; Promega) representing the consensus sequence of the κB-binding motif (underlined) labeled with [³²P]-ATP (approximately 10⁴ cpm/tube) as previously described [22]. Control incubations contained either no nuclear extracts, a 20-fold excess of an unlabeled, irrelevant oligonucleotide (5′-GGCGAAACTTCTGGAATATTCCCGAACTTTCAG, a sequence from the human heat shock 70-kDa protein 1 [HSPA1A] gene promoter [23]), or a 20-fold excess of unlabeled κB oligonucleotide. After a 20-min incubation at room temperature, the extracts were fractionated on nondenaturing 5% (w/v) polyacrylamide gels in 0.25 M Tris-boric acid-EDTA buffer. Gels were then dried and exposed to x-ray film at −80°C.

To conduct supershift assays, samples were prepared for electrophoretic mobility shift assay (EMSA) as described above and to each tube 2 μg of anti-p65, anti-p50, anti-p52, anti-RelB, or anti-cRel was added for 30 min at room temperature. Control incubations contained the same concentration of isotype-specific IgG. Samples were fractionated on nondenaturing gels and autoradiographed as described above.

Immunoblot Analysis

Cellular proteins were extracted as previously described [24]. Proteins (30 μg/lane) were resolved by SDS-PAGE and transferred to nitrocellulose. Immunoblotting was performed using antibodies directed against native and phosphorylated IκBα, native and phosphorylated IKKα/β, PPARG, COX2, and GAPD in Tris-buffered saline (pH 8.0) containing 0.05% (v/v) Tween-20 and 5% (w/v) BSA. Following exposure to horse-radish peroxidase-conjugated secondary antibodies, chemiluminescent signals were revealed using SuperSignal chemiluminescent detection reagents (Pierce). Immunoreactive proteins were visualized using the VersaDoc Imaging System and analyzed using Quantity One software (Bio-Rad Laboratories).

Immunofluorescence

Cells were seeded onto flame-sterilized glass coverslips placed in a 24-well tissue culture plate. Following treatments, cells were fixed for 1 h in 4% (w/v) paraformaldehyde/PBS, made permeable by the addition of 0.2% (v/v) Triton X-100/PBS for 15 min, and blocked in 5% (v/v) horse serum/PBS overnight at 4°C before the addition of antibodies. Monoclonal antibodies directed against p65 (sc-8008; Santa Cruz) were then applied. In control experiments, an equivalent concentration of isotype-specific normal mouse IgG was substituted for the primary antibody. After stringent washing in PBS, the coverslips were exposed to fluorochrome-conjugated secondary antibodies (Molecular Probes). Nuclei were counterstained with 5 μg/ml 4′,6-diamidino-2-phenylindole (DAPI, Sigma). Cells were mounted using the ProLong Antifade kit (Molecular Probes) and visualized using an epifluorescence microscope (Nikon Instruments, Melville, NY).

Prostaglandin E₂ ELISA

Cells were plated in 48-well tissue culture plates in serum-free culture medium supplemented with 5 μM arachidonic acid. Following 6-h incubations with test substances, the media were collected and frozen at −80°C until assays were performed. Media were analyzed for PGE₂ content using commercially available ELISA kits (Cayman Chemical). Cells were solubilized with 1 N NaOH, total protein was measured, and the data were expressed as nanograms of PGE ₂ produced/mg protein. In all ELISA experiments, 4–6 replicates were performed per assay and the experiments were repeated a minimum of twice. The intra- and interassay coefficients of variation of this assay were <10% and the limit of detection was 15 pg/ml (corresponding to approximately 12.5 pg/mg when adjusted for the average protein content/well).

Statistical Analysis

ELISA and densitometric data were assessed by one-way analysis of variance (ANOVA) followed by the Tukey-Kramer multiple comparisons post hoc test; P < 0.05 was considered statistically significant.

RESULTS

PPARG Ligands Modulate PGE₂ Biosynthesis

Among several potential endogenous ligands for PPARG, 15-deoxy-Δ¹²,¹⁴-PGJ₂ (15d-PGJ₂), a nonenzymatically-derived terminal metabolite of PGD₂, has been extensively studied [19, 23, 25–28]. We examined the effects of this cyclopentenone in relation to pharmacological PPARG agonists of the thiazolidinedione class (rosiglitazone [ROSI] and troglitazone [TRO]). WISH cells were preincubated for 30 min in the absence or presence of 15d-PGJ₂, ROSI, or TRO and then challenged with TNF (25 ng/ml) for 6 h. Control cultures were incubated with 0.1% (v/v) ethanol (vehicle for the thiazolidinediones) or 0.8% (v/v) methyl acetate (MeAc, vehicle for 15d-PGJ₂). To exclude the possibility of cross-reactivity between the antibodies and the reagents, in some cases, we spiked aliquots of test medium with the highest experimental concentrations of 15d-PGJ₂, ROSI, or TRO and subjected these to ELISA. None of these compounds were found to interfere with the assay system. Expression of the PPARG receptor protein in WISH and primary amnion cells was confirmed by immunoblotting (Fig. 1A).

FIG. 1.

Effects of PPARG ligands on TNF-induced PGE₂ production. A) Expression of PPARG (57 kDa) in WISH and primary amnion cells was confirmed by immunoblotting using polyclonal antibodies purchased from Cell Signaling Technology (2492, top panel) and Affinity Bioreagents (PA3-821, bottom panel). B) WISH cells were preincubated for 30 min with 15d-PGJ₂ (0.3–30 μM), ROSI (0.1–10 μM), TRO (0.1–10 μM), or vehicle (0.8% [v/v] MeAc) and then treated for 6 h in the absence or presence of 25 ng/ml of TNF. The media were collected and subjected to PGE₂ ELISA. PGE₂ content was normalized to the total protein content for each sample and has been represented graphically (mean ± SEM for quadruplicate determinations, n = 2 individual experiments at the concentration ranges shown, which were confirmatory of two additional experiments in which single concentrations [30 μM for 15-dPGJ₂, and 10 μM each for ROSI and TRO] were used). It should be noted that the effect of a second vehicle (0.1% [v/v] ethanol, not shown) on stimulated PGE₂ production was similar to that of MeAc. * P < 0.001 vs. TNF (ANOVA).

In the presence of 30 μM 15d-PGJ₂, TNF-stimulated PGE₂ production was diminished compared with cells preincubated with vehicle alone (Fig. 1B and data not shown). At 3 μM of 15d-PGJ₂, this response was less marked, and no effect was evident when 0.3 μM of 15d-PGJ₂ was used (Fig. 1B). By contrast, preincubation with ROSI or TRO did not abrogate PG generation (Fig. 1B). In fact, in preliminary experiments, it was found that preincubation with 10 μM ROSI led to a significant enhancement of cytokine-induced PGE₂ release, whereas 10 μM TRO had no discernable effect (data not shown). When tested at an expanded range of concentrations (0.1–10 μM), it was discovered that cytokine-mediated PGE₂ synthesis was consistently enhanced in the presence of ROSI at concentrations of 1 μM or greater (Fig. 1B). There was no enhancement of basal PGE₂ production when these cells were incubated with ROSI or TRO in the absence of TNF (data not shown). These data indicated that, while both 15d-PGJ₂ and thiazolidinediones are activators of PPARG, they exert differential effects on PGE₂ formation in this model system.

15-deoxy-Δ¹²,¹⁴-PGJ₂ Prohibits TNF-Induced COX2 Accumulation

The mechanisms through which 15d-PGJ₂ exerts its anti-inflammatory effects remain controversial, as both PPARG-dependent and -independent actions have been reported [10, 28, 29]. Like the other PGD₂ derivatives (PGJ₂ and Δ¹²-PGJ₂), 15d-PGJ₂ has a cyclopentenone ring structure that is characterized by one or two α,β-unsaturated carbonyl moieties with electrophilic carbons that are very susceptible to nucleophilic addition reactions (reviewed in [30], also see Fig. 2A). Many of the PPARG-independent effects of 15d-PGJ₂ appear to be similar to those seen with other cyclopentenone PGs, and may arise through the ability of these compounds to form thiol conjugates with cysteinyl residues in a variety of cellular proteins [30, 31].

FIG. 2.

Effects of cyclopentenone PGs and thiazolidinediones on TNF-elicited COX2 protein expression. A) Chemical structures of PGD₂ and the cyclopentenone PGs used in this study are shown in relation to that of the parent compound, cyclopentenone. Electrophilic carbons are denoted by gray asterisks. B, C) Representative immunoblots prepared from WISH cell lysates 4 h following challenge with TNF (25 ng/ml) alone or in combination with PGD₂ (30 μM), cyclopentenone PGs (PGA₁, PGJ₂, Δ¹²-PGJ₂, 15d-PGJ₂, 15d-PGA₁, 15d-PGA₂, and 15d-PGD₂, all at 30 μM) or thiazolidinediones (ROSI and TRO, 0.1–10 μM). Control cells received either 0.8% (v/v) MeAc (B) or 0.1% (v/v) ethanol (C). D, E) Immunoreactive COX2 (72 kDa) was quantified by densitometric scanning. Data were normalized to GAPD (34 kDa) protein levels for each treatment group and have been represented graphically as ratios of arbitrary densitometric units (mean ± SEM, n = 2 [D] or 3 [E] individual experiments). * P < 0.01 vs. TNF (AN-OVA). ** P < 0.05 vs. TNF (ANOVA).

To determine whether diminished PGE₂ biosynthesis was due to decreased COX2 protein following cytokine stimulation, we preincubated WISH cells either with 15d-PGJ₂ (30 μM), ROSI (0.1–10 μM), TRO (0.1–10 μM), or vehicles and then challenged the cells with TNF for 4 h. Additionally, to compare the effects of 15d-PGJ₂ with other cyclopentenone PGs, cells were also treated with equivalent concentrations (30 μM) of the following: PGA₁, PGD₂, PGJ₂, Δ¹²-PGJ₂, 15d-PGA₁, 15d-PGA₂, and 15d-PGD₂. Cell extracts were subjected to immunoblotting with anti-COX2 antibodies. In the absence of cytokine treatment, immunoreactive COX2 was barely detectable. By contrast, treatment with TNF elicited a dramatic increase in COX2 protein (Fig. 2, B and C, lane 2), which was consistent with an earlier report from our group [24]. When cells were preincubated with 15d-PGJ₂, TNF did not stimulate an increase in COX2 production (Fig. 2B, compare lanes 2 and 7). Interestingly, among cells pretreated with related cyclopentenones (PGA₁, PGJ₂, Δ¹²-PGJ₂) or their 15-deoxy derivatives (15d-PGA₁, 15d-PGA₂, 15d-PGD₂), only Δ¹²-PGJ₂ inhibited COX2 protein expression, and this effect was modest (Fig. 2D). Neither ROSI nor TRO inhibited TNF-induced COX2 protein expression (Fig. 2C, in which it should be noted that the apparent diminution of immunoreactive COX2 in lane 4 was not a consistent finding). Rather, for both ROSI and TRO at 10 μM, we observed a consistent trend toward increased COX2 protein (Fig. 2E).

15-Deoxy-Δ¹²,¹⁴-PGJ₂ and Thiazolidinediones Differentially Regulate COX2 mRNA Expression

After demonstrating that 15d-PGJ₂ inhibits TNF-stimulated COX2 protein expression in WISH cells, we determined whether 15d-PGJ₂ or thiazolidinediones attenuate COX2 mRNA upregulation following cytokine exposure. In Northern blots, COX2 mRNA was characteristically upregulated following a 60-min challenge with TNF (Fig. 3A, lane 2). When WISH cells were preincubated with 15d-PGJ₂ for 30 min before cytokine stimulation, we noted nearly complete abrogation of COX2 mRNA accumulation (Fig. 3A, compare lanes 1, 2, and 7). A similar, but less substantial, effect was noted for Δ¹²-PGJ₂, a second PGD₂ metabolite that has been shown to exhibit both PPARG-dependent and -independent activities [31, 32]. With respect to the other PGs, PGA₁, 15d-PGA₁, and 15d-PGA₂ all exhibited similar effects in attenuating TNF-elicited COX2 mRNA expression (Fig. 3, A and B). By contrast, neither PGD₂, PGJ₂, nor 15d-PGD₂ produced a discernable effect on COX2 mRNA levels in response to cytokine challenge, despite the presence of electrophilic carbons on PGJ₂ and 15d-PGD₂ (Fig. 3, A and B). RNA from WISH cells that had been pretreated with ROSI or TRO followed by TNF showed no diminution in COX2 mRNA (Fig. 3A). Instead, at thiazolidinedione concentrations of 10 μM, there was a modest but reproducible enhancement of cytokine-mediated COX2 mRNA accumulation (Fig. 3A, lanes 11 and 13).

FIG. 3.

Effects of cyclopentenone PGs and thiazolidinediones on cytokine-induced COX2 mRNA expression. A) Representative Northern blot demonstrating COX2 mRNA levels in WISH cells 1 h following treatment with TNF (25 ng/ml) in the absence or presence of PGD₂ (30 μM), cyclopentenone PGs (PGA₁, PGJ₂, Δ¹²-PGJ₂, 15d-PGJ₂, 15d-PGA₁, 15d-PGA₂, and 15d-PGD₂, all at 30 μM) or thiazolidinediones (ROSI and TRO, 1–10 μM). Control cells received either 0.8% (v/v) MeAc or 0.1% (v/v) ethanol, with similar effects. B) Chemiluminescent detection of the major COX2 mRNA band (4.5 kb) was quantified by densitometric scanning. Data were normalized to GAPD mRNA levels in each treatment group and have been represented graphically as a ratio of arbitrary densitometric units (mean ± SEM, n = 3 individual experiments). C) Real-time RT-PCR was used to assess COX2 mRNA expression in primary amnion cells following treatment with vehicle alone, vehicle with TNF (25 ng/ml), or TNF in the presence of 15d-PGJ₂ (30 μM), ROSI (10 μM), or TRO (10 μM). The threshold cycle (C_T) at which COX2 mRNA was detected was normalized to that of the 18S rRNA internal control. The ratios of COX2 mRNA/18S rRNA have been expressed graphically as the fold increase in relation to control cells, for which the level of COX2 mRNA expression was set to 1 (mean ± SEM for duplicate determinations, n = 2 individual experiments). * P < 0.01 vs. TNF (ANOVA). ** P < 0.05 vs. TNF (AN-OVA).

To examine whether the observations in WISH cells were also seen in amnion cells cultured from full-term fetal membranes, we preincubated amnion cell monolayers for 30 min with vehicle, 15d-PGJ₂, ROSI, or TRO and then challenged cells for 1 h with TNF. Total RNA was extracted and assayed by real-time RT-PCR using a human COX2 primer pair/probe set. Figure 3C demonstrates that TNF caused a nearly fivefold increase in COX2 mRNA expression following cytokine treatment. In contrast, amnion cells preincubated with 15d-PGJ₂ manifested an approximately 40% decrease in COX2 mRNA when challenged with cytokine (Fig. 3C). When amnion cells were preincubated with ROSI or TRO and then TNF, there was an enhancement of cytokine-induced COX2 mRNA accumulation (Fig. 3C). Thus, both WISH and primary amnion cells exhibited decreased expression of COX2 mRNA when preincubated with 15d-PGJ₂, while there was no such diminution when cells were pretreated with thiazolidinediones. In fact, thiazolidinedione pretreatment elicited a modest and consistent elevation in COX2 mRNA expression following cytokine stimulation in both cell types.

Cyclopentenones Inhibit TNF-Stimulated Activation of NFκB

Proinflammatory cytokines use the NFκB pathway as a principal means to elicit COX2 expression within intrauterine cells [15–17]. Under basal conditions, NFκB dimers (which can be composed of the subunits p65, p50, p52, cRel, and RelB) are bound to an inhibitor (IκB) [12]. In the presence of IκB, the net intracellular distribution of NFκB is cytoplasmic; however, when cells are stimulated, IκB is phosphorylated by IκB kinases (IKKs), which results in the destruction of this inhibitor by the 26 S proteasome complex [12]. Once this occurs, the NFκB complex translocates to the nucleus, binds to response elements within the promoters of target genes, and activates transcription.

To examine the biochemical mechanisms that underlie cyclopentenone inhibition of cytokine signaling via NFκB, WISH cells were preincubated either with PGD₂, cyclopentenones, thiazolidinediones, or vehicles and then stimulated with TNF (25 ng/ml) for 15 min (a time point at which the signaling events responsible for NFκB activation were at their peak, as established previously [17]). Immunoblots were probed with antibodies recognizing the native and phosphorylated forms of IKKα, IKKβ, and IκBα.

TNF elicited phosphorylation of IKKα and IKKβ, which was followed by phosphorylation and rapid degradation of IκBα (data not shown). This resulted in the loss of IκBα immunoreactivity observed by 15 min poststimulation (Fig. 4A, compare lanes 1 and 2). As previously demonstrated [17], this loss in IκBα immunoreactivity could be blocked with MG-132, an inhibitor of protein degradation by the 26 S proteasome complex (data not shown). Neither PGD₂, the cyclopentenones, nor the thiazolidinediones were found to affect signaling events leading to TNF-mediated phosphorylation of IKKα or IKKβ (Fig. 4, A and B). However, in the presence of 15d-PGJ₂, cytokine-induced phosphorylation of IκBα was completely blocked, and there was no evidence of lost IκBα immunoreactivity (Fig. 4A, lane 7). Several other cyclopentenones, including PGA₁, PGJ₂, Δ¹²-PGJ₂, 15d-PGA₁, and 15d-PGA₂, caused retention of phosphorylated IκBα following TNF stimulation, consistent with disruption or delay in the events leading to NFκB activation (Fig. 4A). The findings of inhibited IκBα phosphorylation and/or degradation were congruous with prior reports, in which the activities of both IKK and the proteasome complex were found to be directly inhibited by the cyclopentenone PGs [28, 33–35]. By contrast, neither PGD₂, 15d-PGD₂, nor the thiazolidinediones blocked the phosphorylation/degradation of IκBα.

FIG. 4.

Effects of cyclopentenone PGs and thiazolidinediones on cytokine-stimulated NFκB activation. A, B) WISH cells were challenged for 15 min with TNF (25 ng/ml) in the absence or presence of PGD₂ (30 μM), cyclopentenone PGs (PGA₁, PGJ₂, Δ¹²-PGJ₂, 15d-PGJ₂, 15d-PGA₁, 15d-PGA₂, and 15d-PGD₂, all at 30 μM) or thiazolidinediones (ROSI and TRO, 0.1–10 μM). Control cells received either 0.8% (v/v) MeAc (A) or 0.1% (v/v) ethanol (B). Activation of NFκB was assessed using antibodies detecting native and phosphorylated forms of IKKα (85 kDa), IKKβ (87 kDa), and IκBα (39 kDa). These blots are representative of the results of two independent experiments. C) Intracellular localization of NFκB subunit p65 was detected by immunofluorescence in WISH (panels 1–5) or primary amnion (panels 6–12) cells treated with or without 25 ng/ml of TNF for 15 min in the absence (vehicle only) or presence of ROSI (10 μM, panels 3 and 8), 15d-PGJ₂ (30 μM, panels 4 and 9), or PGA₁ (30 μM, panels 5 and 10). Arrows indicate cell nuclei. A representative control experiment is shown (panel 11) in which an equivalent amount of normal mouse IgG₁ was substituted for the primary antibody; DAPI staining within this same field demonstrates the location of cell nuclei (panel 12). Bars = 20 μm.

We next conducted immunolocalization studies to evaluate whether 15d-PGJ₂ attenuated cytosolic to nuclear translocation of the NFκB subunit p65. WISH cells were preincubated for 30 min either with vehicle, ROSI, or 15d-PGJ₂. To test the potential specificity of 15d-PGJ₂, we also preincubated some cultures with a second cyclopentenone, PGA₁. The cells were then challenged with TNF and fixed for immunostaining with anti-p65 at various intervals from 5 to 60 min. As previously observed with IL1B [17], TNF stimulation resulted in translocation of p65 from cytosol to nucleus within 15 min (Fig. 4C, arrows in panel 2). ROSI had no effect on the movement of p65 to the nuclear compartment (Fig. 4C, arrows in panel 3). 15d-PGJ₂, however, completely abrogated the ability of TNF to elicit p65 translocation to the nucleus, even through the 60-min time period tested (the effect of this agent at 15 min poststimulation is shown in Fig. 4C, panel 4). PGA₁, though somewhat less efficacious, was able to delay p65 nuclear localization (data not shown). By 15 min following TNF challenge in the presence of PGA₁, nuclear p65 immunoreactivity was observed in a minority of cells (Fig. 4C, panel 5). Similar to the data obtained in WISH cells, we also noted in primary amnion cells that 15d-PGJ₂, but not ROSI, inhibited TNF-stimulated p65 nuclear translocation, while the effects of PGA₁ were incomplete (Fig. 4C, panels 6–10).

We next examined whether 15d-PGJ₂ could prevent binding of NFκB to radiolabeled oligonucleotides representing the consensus κB motif, similar to those found in the 5′-promoter region of the human COX2 gene [36]. In initial EMSA experiments, it was observed that nuclear extracts prepared from WISH cells stimulated for 30 min with TNF led to the restricted mobility of two specific bands (I and II in Fig. 5, A and B). A third specific band migrating above band II was noted in some cases (Fig. 5B). Super-shifting with polyclonal antibodies directed against NFκB subunits p65 and p50 demonstrated that the DNA binding activity of band I likely consisted of p50/p50 homodimers, while that of band II likely consisted of p65/p50 heterodimers. We ruled out the likely participation of other NFκB members using antibodies to p52, cRel, and RelB (data not shown). Following a 30-min preincubation with 30 μM of 15d-PGJ₂ or 10 μM of ROSI, we observed that TNF failed to elicit protein binding to the κB motif in the presence of 15d-PGJ₂ only (Fig. 5C). Collectively, these data demonstrated that 15d-PGJ₂, like some of the other cyclopentenones (PGA₁, PGJ₂, Δ¹²-PGJ₂, 15d-PGA₁, and 15d-PGA₂), but in contrast with the pharmacological PPARG ligands, was able to attenuate cytokine-stimulated NFκB activation.

FIG. 5.

Modulation of cytokine-induced NFκB DNA binding by cyclopentenone PGs and thiazolidinediones. A) The conditions for EMSA were verified by incubating a [³²P]-labeled oligonucleotide probe (5′-AGTT-GAGGGGACTTTCCCAGGC) containing the κB consensus site (underlined) in the presence of nuclear extract prepared from TNF-stimulated WISH cells. As a control for probe specificity, companion extracts were prepared in the absence of nuclear extract (Probe Only) or in the presence of a 20-fold excess of an unlabeled competitor probe containing either an irrelevant sequence (HSF Comp) or the κB binding motif (κB Comp). B) Nuclear extracts from WISH cells treated for 30 min with media alone (Veh) or with TNF (25 ng/ml) were prepared for supershift EMSA using polyclonal antibodies directed against NFκB subunits p65 (αp65) or p50 (αp50). C) Nuclear extracts were prepared from WISH cells 30 min following stimulation with TNF (25 ng/ml) in the absence or presence of 15d-PGJ₂ (30 μM) or ROSI (10 μM) and subjected to EMSA. Control cells were treated with vehicle only (0.8% [v/v] MeAc). The blots are representative of two independent experiments. Arrows denote the locations of specific bands (I and II).

DISCUSSION

In this investigation, we evaluated the ability of PPARG ligands to downregulate the COX2 gene and subsequent formation of PGE ₂ in WISH and primary amnion cells. Our data demonstrated an interesting dichotomy between 15d-PGJ₂, a cyclopentenone PG, and the thiazolidinediones, which are pharmacological PPARG activators. Whereas 15d-PGJ₂ treatment of amnion cells in primary culture or WISH cells almost completely abrogated TNF-driven COX2 expression, the thiazolidinediones partially upregulated COX2 expression. These data render a straightforward mechanism through ligation of nuclear PPARG unlikely and instead invoke a more complex level of COX2 regulation by these agonists.

In this study, we employed WISH cells in many experiments, as these cells are easily manipulated genetically and manifest many phenotypic features of amnion epithelial cells [37–39]. Although WISH cells have been used as reliable and reproducible facsimile for amnion epithelial cells for many years [14, 17, 24, 27, 40], it must be acknowledged that these cells contain HeLa chromosome markers, as we have recently reported [41]. We verified the applicability of our data by repeating key experiments using primary amnion cell cultures, which in these cases showed similar results. Thus, to a degree, WISH cells remain a very realistic surrogate for molecular mechanistic studies in amnion cells.

In the case of 15d-PGJ₂, suppression of cytokine-elicited COX2 expression appears to have arisen to a large extent through interference with NFκB activation. In amnion cell models, the NFκB signaling cascade appears to be the predominant mechanism through which cytokines stimulate COX2 gene expression [14, 16, 17]. Moreover, our group has reported that NFκB is a prominent transcription factor governing COX2 gene expression in ED₂₇ cells [22]. Like other cyclopentenone PGs, 15d-PGJ₂ has been shown to impede NFκB signaling at multiple steps: it may interfere with the phosphorylation of IκB by suppressing IKK activity [33], it may interfere with IκB degradation by inhibiting activity of the 26 S proteasome complex [34], and it may interfere directly with the ability of NFκB subunits to bind to their cognate DNA response elements [28]. These effects are thought to occur independently of PPARG, through the formation of covalent adducts between 15d-PGJ₂ and critical cysteinyl residues within these proteins [28, 35]. Consistently, we observed that other cyclopentenone PGs bearing α,β-unsaturated ketone groups also interfered with cytokine-mediated NFκB activation, albeit far less potently, whereas thiazolidinediones did not alter this signaling pathway. Although prior reports have lead to the speculation that PPARG might interfere with NFκB activity through direct protein-protein interactions [8] or through competition for limiting amounts of coactivator proteins [42], our data suggest that, in WISH and in amnion cells cultured in vitro, PPARG may not be involved directly in the antagonism of NFκB activation. In support of this, Lindstrom and Bennett have recently reported that antagonism of cytokine-elicited NFκB activation by 15d-PGJ₂ occurs independently of PPARG in amnion and myometrial cells [43].

In addition to inhibiting NFκB, cyclopentenone PGs have also been shown to suppress the activity of other transcription factors involved in cytokine-stimulated COX2 expression, such as activator protein-1 (AP-1) [8, 44]. By impeding such signaling cascades, it is likely 15d-PGJ₂ and other cyclopentenones suppress TNF-stimulated COX2 expression. In general, the ability of the cyclopentenones to attenuate COX2 induction was correlated with the number of electrophilic β-carbons present within the molecules. Both 15d-PGJ₂ and Δ¹²-PGJ₂, with cross-conjugated α,β- unsaturated ketones, exhibited greater potency relative to the cyclopentenones having only a simple α,β-unsaturated carbonyl moiety, which is consistent with prior observations [35]. However, given that 15d-PGJ₂ suppressed cytokine-induced COX2 expression to a greater degree than did Δ¹²-PGJ₂, our results suggest that the number of reactive β-carbons alone is insufficient to explain the effects of these agents. It is therefore noteworthy that 15d-PGD₂, which bears an exocyclic α,β-unsaturated ketone, was unable to suppress either cytokine-mediated NFκB activation or COX2 expression. Although we do not yet have a suitable explanation for this observation, we speculate that endocyclic β-carbons might exhibit greater reactivity than exocyclic β-carbons for the formation of cyclopentenone protein conjugates.

Many previous studies have demonstrated proapoptotic actions of cyclopentenones such as 15d-PGJ₂ in several cell types [45, 46]. Keelan et al. [27] reported that 15d-PGJ₂ elicited programmed cell death in WISH cells. However, in the present study, we treated WISH cells and primary amnion cells for relatively brief periods (<12 h) and observed little evidence of apoptotic cell death (data not shown). Thus, it is unlikely that cytotoxicity is an explanation for the observed decreases in COX2 gene expression and subsequent PGE₂ biosynthesis reported here. We cannot exclude the possibility that longer incubations may lead to apoptosis, however.

In contrast with 15d-PGJ₂ and other cyclopentenones, we observed a tendency for thiazolidinediones, when given at high concentrations, to augment TNF-stimulated COX2 expression. This effect was particularly evident when low concentrations of TNF (5 ng/ml) were used (data not shown). While this finding was unexpected, there is precedence that PPARG might positively regulate COX2 gene expression in certain cell types [47]. The human COX2 gene contains a PPAR response element located approximately 3900 bp upstream of the translational start site, through which ligand-bound PPARG may promote transcriptional activation [47]. However, in numerous experiments, we failed to detect any stimulation in basal COX2 expression by thiazolidinediones (including troglitazone, rosiglitazone, and pioglitazone, data not shown) in the absence of cytokine stimulation. Furthermore, despite the fact that TRO and ROSI enhanced cytokine-elicited COX2 expression to similar degrees, only ROSI was found to augment TNF-stimulated PGE₂ production. In ongoing experiments, we are examining whether this effect might have occurred through differential regulation of an enzyme downstream of COX2 in the PGE₂ metabolic pathway (such as PTGES), or through a more global change in cellular arachidonic acid utilization. It is noteworthy that these effects were seen only when high concentrations of these agents were employed. While it is not unusual for thiazolidinediones to be used at concentrations of 10 μM (or higher) in cell culture experiments [9, 10, 48], it is nevertheless true that such amounts are often in excess of what is required for saturation of the PPARG receptor (which, in the case of ROSI, is below 1 μM [49]). Therefore, we speculate that some of the pharmacological effects observed may have occurred independently of PPARG. We are currently conducting experiments to discern the PPARG dependency of these effects in a comprehensive manner, but the present data provide a vital backdrop for these investigations.

The true physiological significance of these data awaits further experimental clarification. While it has been demonstrated that cells of the fetal membranes, including the amnion, express the PPARG receptor [20], the existence of endogenous PPARG ligands in these tissues remains unclear. It has been established that the placenta is the major source of PGD₂, the precursor for PGs of the J series (PGJ₂, Δ¹²-PGJ₂, and 15d-PGJ₂) within the uterus [50]. Furthermore, amniotic fluid contains PGD₂ [51], and there is some evidence to suggest that 15d-PGJ₂ also might be present [52], although this latter finding is controversial. Therefore, it is not unreasonable to suggest that substances such as the PGs of the J₂ might serve a physiological role during gestation. However, there is a growing body of literature suggesting that 15d-PGJ₂ is not the only potential endogenous PPARG ligand; polyunsaturated fatty acids (PUFAs) and products of lipoxygenase metabolism have also been shown to activate PPARG [6, 53], and a novel, as yet uncharacterized, endogenous PPARG ligand has recently been discovered [54]. These and potentially other unidentified ligands may comprise a mileau of anti-inflammatory biomolecules that are present endogenously throughout normal human gestation, as suggested by the recent findings of Waite and colleagues [55]. Members of our laboratory are currently developing screening systems to allow for the detection and identification of such PPARG ligands within amniotic fluid, maternal plasma, and gestational tissues.

In summary, we have demonstrated in WISH and primary amnion cells that reported ligands for PPARG (cyclopentenone PGs and thiazolidinediones) exert differential effects on COX2 gene expression. Inasmuch as these two classes of PPARG agonists do not act in precisely the same manner, we have proposed a working model to delineate the potential mechanisms through which these agents govern COX2 expression and PG biosynthesis (Fig. 6).

FIG. 6.

Working model for the effects of 15d-PGJ₂ and thiazolidinediones on TNF-stimulated NFκB activation and COX2 expression. Depicted in this schematic diagram is the NFκB signaling pathway, through which TNF and other cytokines stimulate COX2 gene expression in WISH and amnion cells [17]. Upon binding to its cellular receptor, TNF invokes a signaling cascade whereby the IκB kinase (IKK) complex becomes activated. IKK acts to phosphorylate IκB, leading to the destruction of this protein by the 26 S proteasome complex. Once this occurs, NFκB can migrate to the nucleus, bind to response elements within the 5′-promoter region of the COX2 gene, and activate transcription. Independently of PPARG, 15d-PGJ₂ may act at multiple steps in this pathway to impede NFκB activation, either by inhibiting of IKK activity (1), or by attenuating destruction of IκB by the 26 S proteasome complex (2). The effects of 15d-PGJ₂ through activation of the PPARG receptor are unclear (3); however, our current data indicate that pharmacological activation of PPARG by thiazolidinediones (TZDs) might lead to enhanced COX2 transcription.