Peroxisome Proliferators Enhance Cyclooxygenase-2 Expression in Epithelial Cells*

The formation of prostaglandins requires the catalytic activity of cyclooxygenase (COX) which converts arachidonic acid to the prostaglandin endoperoxide PGH2, from which all other prostaglandins are formed. COX-2 is the highly inducible isozyme of COX which is responsible for much of the prostaglandin production in inflammation and is a key factor in colon carcinogenesis. Because COX-2 activity can be rate-limiting in prostaglandin formation, COX-2 expression must be regulated tightly. Numerous factors, including mitogens, tumor promoters, and cytokines have been found to stimulate the transcription of COX-2. We show that fatty acids, prostaglandins, and non-steroidal anti-inflammatory drugs, compounds that are substrates, products, and inhibitors, respectively, of COX enzymatic activity, also increase its expression. These compounds are members of a heterogeneous group of compounds known as peroxisome proliferators, and the prototypical peroxisome proliferator, WY-14,643, also enhanced COX-2 expression. We demonstrate that these compounds increase COX-2 transcription, and we identify a region of the COX-2 promoter containing a peroxisome proliferator response element that is responsible for the enhancement of COX-2 expression seen with these compounds.

When these mice were crossed with COX-2 knockout mice, the offspring had only a small fraction of the intestinal polyps found in their COX-2-expressing peers. Mice that were heterozygous for expression of COX-2 had intermediate levels of colon polyps. Given the correlation between increased COX-2 expression and colon cancer, the effect of NSAIDs is most certainly mediated, at least in part, by COX inhibition. Interestingly, some NSAIDs have recently been demonstrated to act as peroxisome proliferators (24), suggesting that they may also regulate gene expression as part of their chemopreventative mechanism.
In the experiments described below, we examined the ability of fatty acids, prostaglandins, and NSAIDs to regulate COX-2 expression. We demonstrate that these compounds enhance COX-2 expression and that the regulation occurs transcriptionally. Further, we define a region of the promoter responsible for at least part of the regulation of COX-2 expression by these compounds.

EXPERIMENTAL PROCEDURES
Materials-Fatty acids, prostaglandins, NS-398, and several other NSAIDs were purchased from Cayman Chemical Company. WY-14,643, sulindac sulfide, and sulindac sulfone were purchased from BioMol. Other NSAIDs and common reagents were purchased from Sigma. Fatty acids and prostaglandins were diluted in ethanol, WY-14,643 and NSAIDs in dimethyl sulfoxide. Medium for the culture of human mammary epithelial cells was prepared in house, media for culture of colon cell lines were obtained from Life Technologies, Inc. Mouse anti-human monoclonal antibodies to COX-1 and COX-2 were the kind gift of J. Maclouf (INSERM, Paris). Mouse anti-human ␤-actin was purchased from ICN. Goat anti-mouse secondary antibody was from BIOSOURCE.
Clones for Human PPARs and RXR␣-We screened a human skeletal muscle 5Ј-STRETCH PLUS cDNA library (CLONTECH) to isolate clones for the known human isoforms of PPAR and for RXR␣. PPAR␣ (25), ␦ (26), and RXR␣ (27) were obtained by screening with 30-mer oligonucleotides corresponding to the region surrounding the translational start site, according to established procedures. PPAR␥ (28) was obtained by screening with a random prime-labeled cDNA probe (Cayman). Full-length cDNA clones were obtained for each. The clones were sequenced to ensure correctness and subcloned into pcDNAI/Amp (Invitrogen) for mammalian expression.
Isolation and Subcloning the 7-Kilobase COX-2 Promoter-Previous work in this laboratory isolated a 1.8-kb human COX-2 promoter clone (9). Because many PPREs are located further upstream from the translational start site (29), we felt it was necessary to obtain a larger promoter clone. We designed polymerase chain reaction primers corresponding to the 5Ј-most end of the COX-2 promoter, which were used to isolate a P1 clone (Genome Systems). We performed a series of restriction digestions and Southern blots to identify regions of the clone corresponding to further 5Ј-regions of the COX-2 promoter. We chose pieces from two restriction digestions, KpnI (5,710 bases) and SacI (2,971 bases), which were subcloned into pBluescript (Stratagene) for sequencing and then into pGL3-basic (see below) attached to our existing COX-2 promoter to allow us to search for potential PPREs in this larger promoter. The result of the subcloning gave us a total of approximately 7,270 bases of promoter for the KpnI construct and 3,920 bases of promoter for the SacI construct. The 7,270-base pair COX-2 promoter has been deposited with GenBank, Accession Number AF044206.
Epithelial Cell Culture-184B5 mammary epithelial cells (obtained from Martha Stampfer) (32) were grown in modified MCDB 170, a serum-free defined medium. Before treatment with fatty acids, prostaglandins, or NSAIDs, the cells were made quiescent by treating for 48 h with an antibody to the epidermal growth factor (EGF) receptor (monoclonal antibody 225) at 10 g/ml (33). HCT-116, DLD-1, CaCo-2, and HT-29 colonic epithelial cells were obtained from ATCC and grown according to recommended procedures. HCT-116 and HT-29 were cultured in McCoy's 5A with 10% fetal bovine serum, DLD-1 were grown in RPMI 1640 with 10% fetal bovine serum, and CaCo-2 were grown in minimal essential medium with Earle's salts, supplemented with sodium pyruvate and 20% fetal bovine serum. For 24 h before treatment with NSAIDs, the colonocytes were serum starved by replacing the normal growth medium with serum-free medium. To examine COX-2 expression in both mammary epithelial cells and colonocytes, the medium was changed to fresh serum-free medium containing fatty acid, prostaglandin, WY-14,643, or NSAID. For 184B5 cells, the medium also contained the EGF receptor antibody. Cells were incubated with agonist for 4 -8 h for mRNA expression or for 10 -12 h for protein expression before harvest. For time-course experiments, samples were harvested at the times indicated in the figures. For experiments utilizing actinomycin D, it was added at 5 g/ml simultaneously with agonist. For each fatty acid, prostaglandin, and NSAID used, a dose-response curve was performed. Only the maximally effective concentration of each fatty acid, prostaglandin, or NSAID is shown in the figures. Fatty acids including arachidonic acid, eicosapentaenoic acid, eicosatetraynoic acid, linoelaidic acid, and docosahexaenoic acid were used at 0. Transfection Experiments-Quiescent 50% confluent 184B5 cells were transfected using 2.5 l/P35 dish LipofectAMINE (Life Technologies, Inc.) and 0.25 g/P35 dish each of luciferase reporter plasmid, PPAR␣ (or ␦, or ␥, or vector only), and ␤-galactosidase and were performed in the presence of monoclonal antibody 225. At 24 h after transfection, the medium was changed to fresh MCDB 170 containing monoclonal antibody 225 with or without 100 M WY-14,643. The cells were harvested 72 h later, and luciferase and ␤-galactosidase assays were performed. Luciferase activity was measured using the Promega kit, and ␤-galactosidase was measured using Galactolight (Tropix). Triplicate wells were assayed in duplicate for each sample.
Western Blot Analysis-Cells were washed with ice-cold phosphatebuffered saline and then lysed with buffer consisting of 20 mM Tris, pH 7.5, 16 mM CHAPS, 0.5 mM dithiothreitol, 1 mM EDTA, 1 mM benzamidine hydrochloride, 1 g/ml leupeptin, and 10 g/ml soybean trypsin inhibitor. Cells were scraped and placed on ice for 30 min. Unlysed cells and debris were removed by centrifugation, and protein assays were performed (BCA protein assay, Pierce) to ensure equal loading on SDS-polyacrylamide gel electrophoresis. Proteins were separated through 10% denaturing polyacrylamide gels and transferred to polyvinylidene difluoride membranes. COX-2 and COX-1 monoclonal antibodies were used according to the procedures of Habib et al. (34). Monoclonal anti-human ␤-actin was also used in some experiments as a control for protein integrity. The secondary antibody was horseradish peroxidase-labeled goat anti-mouse (BIOSOURCE), and ECL (Amersham) was used as a detection method. Quantitation of Western blots was performed by scanning the blots into Photoshop and performing densitometry with NIH Image.
RNase Protection Assay-RNA was isolated using TRIzol reagent (Life Technologies, Inc.) and processed according to the manufacturer's instructions. Samples were resuspended in RNase-free water and concentrations determined spectrophotometrically. 10 g of each sample was used for RNase protection assay with probes for COX-2 and GAPDH. The COX-2 probe corresponds to the region spanning Ϫ90 to ϩ332 relative to the COX-2 translational start site. 2 The GAPDH probe is commercially available (Ambion). Radioactive transcripts were prepared using [␣-32 P]UTP and the Ambion MAXIscript in vitro transcription kit. 80,000 cpm/sample of the COX-2 probe and 8,000 cpm/sample of the GAPDH probe were used in the RNase protection assay (RPA II, Ambion). RNA and probes were co-precipitated, resuspended in hybrid-ization buffer, and incubated overnight at 43°C. The following day, unbound RNA was digested with RNase, and the remaining radiolabeled RNA was precipitated, resuspended in loading buffer, and separated on a 6% polyacrylamide gel. Gels were dried and exposed to BioMax MS film at Ϫ80°C for up to 48 h. After autoradiography, the protection assay was quantitated by densitometry as above.
Nuclear Run-on Assays-Nuclear run-on assays were used to assess absolute levels of transcription of COX-2 after treatment with EGF, phorbol 12-myristate 13-acetate, or LA. Assays were performed essentially as described by DeWitt and Meade (3). For nuclear isolation, 1 ϫ 10 7 cells were treated with vehicle, EGF (100 ng/ml), phorbol 12-myristate 13-acetate (60 ng/ml), or LA (18 M) until the time when the maximal change in transcription was expected based on changes in mRNA levels (30 min for EGF, 1 h for phorbol 12-myristate 13-acetate, and 2 h for LA and vehicle). Cells were scraped into cold phosphatebuffered saline, centrifuged and rinsed, and placed in buffer (50 mM Tris, pH 8.3, 40% glycerol, 5 mM MgCl 2 , and 0.1 mM EDTA) for storage in liquid nitrogen until assayed. mRNA and protein were harvested from identical plates at later time points to ensure that both COX-2 protein and mRNA expression were altered as expected by these treatments. For nuclear run-on assays, the following plasmids were used as probes: a cDNA fragment for COX-2 which corresponded to the entire open reading frame but did not contain the 3Ј-untranslated region and a full-length cDNA for COX-1. (Previous work demonstrated that COX-1 levels are unaltered by the treatments used in these experiments and that it could serve as a control in these experiments.) Both plasmids were in the same base vector, eliminating the need for a vector control. Plasmids were linearized, denatured, and applied to Hybond N membrane using a Minifold II Slot Blotter (Schleicher & Schuell). To begin run-on assays, nuclei were thawed and combined with a buffered nucleotide mixture containing [␣-32 P]UTP to label transcripts radioactively. Transcription was allowed to proceed for 30 min at 30°C, and then RNase-free DNase (200 units) was added and the incubation continued a further 10 min. Next, proteins were digested by the addition of proteinase K (400 g/ml) for 45 min at 37°C. The samples were extracted with phenol-chloroform and precipitated with NH 4 OAc and EtOH at Ϫ80°C. After repeating the DNase and proteinase K digestions, the samples were precipitated overnight at Ϫ80°C. The following day, the samples were centrifuged, and the RNA was resuspended in TE and denatured with NaOH for 10 min on ice. The samples were precipitated a third time and resuspended in TE. Incorporation of radiolabel into the nuclei was ascertained by scintillation counting, and an equal amount of radioactivity was added to individual slot blots for each sample. The slot blots that contained COX-1 and COX-2 probes had been prehybridized for 2 h before the addition of radiolabeled nuclei. After the addition of the nuclei, the blots were hybridized at 42°C for 48 h. The blots were then washed as described and exposed to BioMax MS at Ϫ80°C for up to 48 h. After autoradiography, blots were quantitated as described above.
Gel Shift Assay-Nuclear extracts were prepared from COS-7 cells transfected with PPAR␣ (or ␦, or ␥, or vector only) and with RXR␣ (35). Oligonucleotide cassettes corresponding to the ACO PPRE (30,31) and the COX-2 Ϫ3900 PPRE were either radiolabeled using [␥-32 P]ATP and T4 kinase or used unlabeled for competition experiments. The following oligonucleotides were used: ACO PPRE F, ACO PPRE R, COX-2 Ϫ3900 PPRE F, COX-2 Ϫ3900 PPRE R, whose sequences were defined earlier, and AP2 F, 5Ј-GATCGAACTGACCGCCCGCGGCCCGT-3Ј; and AP2 R, 5Ј-ACGGGCCGCGGGCGGTCAGTTCGATC-3Ј. Binding reactions were performed by preincubating 20 g of nuclear lysate in binding buffer with or without competitor oligonucleotide (50-fold excess compared with radiolabeled oligonucleotide) in a volume of 18 l for 10 min at room temperature. Binding buffer consisted of 10% glycerol, 20 mM Tris, pH 8.0, 80 mM KCl, 10 mM MgCl 2 , 2 mM dithiothreitol, and 5 g of poly(dI⅐dC). 2 l of radiolabeled oligonucleotide cassette was then added, and the reactions were incubated 20 min at room temperature. The products were separated by loading on a prerun 4% non-denaturing polyacrylamide gel containing 2.5% glycerol and separated in 1 ϫ TBE buffer at 150 volts at room temperature until the bromphenol blue marker dye was approximately two-thirds of the way down the gel. Controls, including AP2 oligonucleotides, were from the Promega gel shift assay system. Gels were dried and exposed to BioMax MR autoradiography film at Ϫ80°C for up to 48 h.

RESULTS
Polyunsaturated fatty acids, including LA, have diverse effects on cells, and we asked whether one component of the response was to alter COX-2 expression. We exposed mammary epithelial cells to LA and found that it strongly enhanced COX-2 expression (Fig. 1A). Many fatty acids are able to stimulate cell responses, so we next tested other fatty acids for their effect on COX-2 expression. We found that all of the unsaturated fatty acids that we tested enhanced COX-2 expression, including arachidonic acid, the normal substrate for COX-2, as well as fatty acids that are abundant in fish oils, eicosapentaenoic acid and docosahexaenoic acid. To determine whether metabolism of the fatty acid was necessary for stimulation of COX-2 expression, we exposed cells to linoelaidic acid, an analog of LA which has only trans-double bonds, and eicosatetraynoic acid, an analog of arachidonic acid with triple bonds, neither of which is a substrate for oxidases such as COX. These compounds also increased the level of COX-2, demonstrating that metabolism of fatty acid is not required for this effect.
The pattern of fatty acids that enhanced COX-2 expression is consistent with fatty acids that have been identified recently (14,15) to act through PPARs. Although their physiological ligands remain uncertain, it has been demonstrated that prostaglandins and fatty acids can act as ligands. We next tested the ability of a panel of potential COX-2 products to enhance its expression. We found that many of them enhanced COX-2 expression, particularly the cyclopentanone prostaglandins PGA 2 and 15⌬PGJ 2 , which are most potent in other PPARmediated responses (14,36,37) (Fig. 2A). Of all the compounds that we tested, 15⌬PGJ 2 was the most potent, strongly enhancing COX-2 expression at levels as low as 0.4 M (Fig. 2B). We also found that 15-HETE ( Fig. 2A), 8-HETE, and 9-and 13-HODE stimulated COX-2 expression (not shown). PGE 2 , which is produced at relatively high levels by mammary epithelial cells, did not alter COX-2 expression in these cells. This pattern of COX-2 expression in response to prostaglandins and HETEs supported the hypothesis that it occurred via a PPAR. As an additional test of this, we treated cells with the PPAR agonist WY-14,643 and found that it strongly stimulated COX-2 expression (Fig. 2C).
NSAIDs function by inhibiting COX-1 and COX-2 and recently have been identified as peroxisome proliferators (24). We asked if inhibitors of COX would also induce its expression. We tested a panel of NSAIDs on both mammary epithelial cells and colonic epithelial cells and found that many of them strongly enhanced COX-2 expression in both cell types (Fig. 3). Particularly potent were the members of the fenamate family, meclofenamate and mefenamic acid. Ibuprofen, a commonly used NSAID, enhanced COX-2 expression as did NS-398, the first well characterized inhibitor that is highly selective for COX-2. Sulindac sulfide, which has been used in the prevention of colon cancer, strongly enhanced COX-2 expression in mammary epithelial cells, with much lesser effects in colonic epithelial cells. Although only CaCo-2 cells are shown, similar results were obtained in all four colonic epithelial cell lines that we examined. COX-1 expression was unaffected by NSAID treatment. Treatment with NSAIDs at these concentrations did not alter cell viability (n ϭ 2, data not shown).
Peroxisome proliferators act by increasing the transcription of the genes that they regulate. If fatty acids, prostaglandins, and NSAIDs alter COX-2 expression through a PPAR, changes in COX-2 mRNA levels should correspond with changes in COX-2 protein levels. We examined the COX-2 mRNA over time in 184B5 cells treated with LA (Fig. 4A) and found that the levels rose markedly, reaching a maximum by 4 h after treatment. We also measured the change in COX-2 mRNA in CaCo-2 cells that had been treated with the panel of NSAIDs used for Western blot analysis (Fig. 4B). The same pattern of NSAIDs was found to enhance COX-2 mRNA expression as had increased protein levels, suggesting that the mechanism responsible for the increased expression was transcriptional. Actinomycin D inhibited COX-2 expression when used in combination with fatty acids, prostaglandins, and NSAIDs (Fig. 5, A  and B), providing further evidence that their effects are transcriptional. Nuclear run-on assays validated this conclusion; in two separate experiments 18 M LA enhanced COX-2 transcription in 184B5 cells 2.1-and 1.5-fold (COX-2/COX-1), whereas phorbol 12-myristate 13-acetate, which strongly stimulates COX-2 expression in these cells, enhanced transcription 4.9-and 2.1-fold.
We took two approaches to identify the portion of the COX-2 promoter responsive to peroxisome proliferators. First, we isolated new, longer regions of the COX-2 promoter than previously published, and we sequenced approximately 7 kb. These clones contained a PPAR consensus site (PPRE) at approxi- mately 3900 bases upstream of the translational start site (Ϫ3900). We made reporter constructs that contained between 7 kb and 200 base pairs of the COX-2 promoter, the Ϫ3900 PPRE, or with a reporter that contained the rat fatty ACO PPRE sequence (30,31). Cells were co-transfected with an expression plasmid for one of the PPARs. The transfected cells then were stimulated with WY-14,643 or vehicle for 72 h. Luciferase activity was measured and normalized to co-transfected ␤-galactosidase. In cells containing the Ϫ3900 PPRE construct, luciferase activity increased 1.4-fold on treatment with WY-14,643 (n ϭ 6) (Fig. 6A). This response depended on co-transfection with a PPAR, and, although modest, was highly consistent between experiments. The experiments shown used an expression plasmid for PPAR␣, but essentially equivalent results were obtained with either ␦ or ␥ (data not shown). In the second approach, we performed gel shift experiments using nuclear lysate obtained from COS-7 cells transfected with a PPAR and RXR␣ (Fig. 6B). As a positive control we used an oligonucleotide cassette for the ACO PPRE. Both the COX-2 Ϫ3900 PPRE and the ACO PPRE produced bands on gel shift. The ACO PPRE could be competed with cold ACO PPRE but not with cold AP2. The COX-2 Ϫ3900 PPRE could be competed with cold COX-2 PPRE or ACO PPRE but not cold AP2, supporting the hypothesis that this region of the COX-2 promoter is responsible for binding PPAR/RXR heterodimers and therefore is responsible for peroxisome proliferator-stimulated transcription of COX-2. To provide further support that the shifted band was truly a PPAR⅐COX-2 PPRE complex, we performed supershift experiments using both commercially available antibodies as well as an anti-PPAR␣ antibody that we prepared in house especially for this purpose. Although these antibodies recognized the corresponding PPAR when used for immunoblotting, none would supershift the PPAR⅐ACO PPRE complex used as a positive control, and none was therefore useful for characterizing the interaction between PPARs and the COX-2 PPRE. DISCUSSION Much debate has surrounded the role of dietary fat as a risk factor for the development of cancer. Although diet is a rela- tively well established risk for colon cancer (5), the mechanism behind this risk has remained undefined. Further, studies examining the impact of dietary fat on breast cancer have been inconclusive. The studies described in this manuscript may help guide future studies addressing this long running debate by identifying a potential mechanism for the increased risk of cancer development due to dietary fat. Our studies show that components of dietary fat induce the expression of COX-2, a protein that has been closely tied to colon cancer (8 -10), in mammary epithelial cells. The induction of the COX-2 gene by lipids was quite vigorous and occurred in response to a variety of fatty acids, many of which are found in a typical Western diet.
The regulation of COX-2 expression not only by fatty acids, but also by its enzymatic products, demonstrates a rare feedforward mechanism that may be key in the pathogenesis of cancer. Dietary fat may act through a PPAR to enhance COX-2 expression in one cell type, resulting in the release of prostaglandins to the intercellular milieu. There, the prostaglandins affect neighboring cells, resulting in enhanced COX-2 expression in these cells. Support for such a feed-forward mechanism is found in the work of Oshima et al. (10) who examined expression of COX-2 in the interstitial cells of mice harboring a mutant APC (APC ⌬716 ). These mice develop large numbers of intestinal polyps. They found COX-2, surprisingly, to be localized in the interstitial cells of the murine intestine rather than in the polyp epithelium where it is expressed in human colon. Thus, expression measured in the polyp epithelium may actually represent a second wave of COX-2 expression due to the paracrine effect of prostaglandins released by the interstitial cells.
NSAIDs are among the most prescribed drugs in the United States. Although these drugs have many well known, longstanding uses, one of the more recent uses is as a colon cancer prophylactic. As demonstrated in this manuscript, NSAIDS have the dual role of inhibiting COX-2 activity while increasing its expression. No in vivo role has currently been defined for this increased COX-2 expression stimulated by NSAIDS. It is likely that NSAIDS would block the activity of the induced COX-2 enzyme as long as the treatment continued, and in such a case, there presumably would be no net effect. However, if NSAID treatment were discontinued, or were intermittent, the effect might be deleterious. For example, if the level of the COX-2 enzyme had been increased markedly by the drug and then it was discontinued, the result might be the exact opposite of the intended therapeutic effect. However, estimating the likelihood of this unfavorable outcome is complex because it depends on the rate at which the NSAID dissociates from COX-2, the t 1/2 of the protein, and probably other variables. In vivo experiments will be necessary to assess whether the induction of COX-2 by NSAIDS can result in increased prostaglandin synthetic capacity under some circumstances.
It is interesting that COX-1 levels are not altered by treatment with NSAIDS, suggesting that the increase in COX-2 expression seen on treatment with NSAIDs is not a feedback mechanism designed to return prostaglandin production to a base-line level. Further, the different NSAIDS did not enhance COX-2 expression equally, although all were tested at concentrations sufficient to abolish prostaglandin formation. This result supports the hypothesis that enhanced COX-2 expression on treatment with NSAIDS is a direct effect mediated through a PPAR.
While this manuscript was in preparation, Staels et al. (38) reported that activators of PPAR␣ can inhibit the expression of a variety of genes involved in the inflammatory response, including COX-2, in smooth muscle cells. These experiments used a protocol in which the cells were treated with PPAR␣ activators and then with interleukin-1; the PPAR activators had no effects by themselves. One likely explanation for some of the differences between our results and theirs is that our studies used epithelial cells, mammary and colon, whereas theirs were on vascular smooth muscle cells. Moreover, their studies used our previously described 1.8-kb COX-2 promoter/reporter construct (9), which does not have a PPRE and which we found to be unresponsive to the agonists described here. Staels et al. (38) concluded that their results reflected a novel mechanism in FIG. 6. Fatty acids stimulate transcription of COX-2 through a PPAR. Panel A, peroxisome proliferators stimulate the transcription of COX-2. 184B5 cells were transfected with reporter constructs from the COX-2 promoter, the ACO PPRE, or a control promoter, driving luciferase expression. Cells were co-transfected with both PPAR␣ and ␤-galactosidase. Luciferase and ␤-galactosidase activities were measured 72 h after the addition of fresh medium with or without 100 M WY-14,643. The results are expressed as the fold increase in luciferase activity on treatment with WY-14,643 (n ϭ 6). Panel B, gel shift analysis. Nuclear lysates were prepared from COS-7 cells transfected with PPAR␦ and RXR␣. After the binding of oligonucleotide cassettes corresponding to the ACO PPRE or the COX-2 Ϫ3900 PPRE to these lysates, DNA-protein complexes were detected by separation on non-denaturing polyacrylamide gels. For competition experiments, cold oligonucleotide cassette for ACO, the COX-2 Ϫ3900 PPRE, or AP2 was incubated with the lysates before the addition of radiolabeled ACO or COX-2 PPRE. This result was obtained in two independent experiments. which PPAR␣ activation interfered with transcription mediated by nuclear factor-B and not through a typical PPRE; our work shows that the fatty acid, prostaglandin, and NSAID effects are through a traditional PPAR/PPRE mechanism. It is feasible that both mechanisms are involved in the regulation of COX-2 expression in response to peroxisome proliferators, and the predominant response may be tissue-specific. Several studies examining the role of PPAR␥ in colon cancer were published at the time that this manuscript was submitted (39 -41). In two of them, it was shown that selective activators of PPAR␥ increased the number and size of colon, but not small intestinal, polyps in Min mice, which have a mutation in one allele of the APC gene (39,40). The other study reported the opposite effect of PPAR␥ on the growth of human colon cancer cells transplanted into mice (41). In addition to measuring the size and number of tumors, Lefebvre et al. (39) examined molecular events including whether COX-2 was induced in the mouse adenomas. They also tested the effect of a PPAR␥ agonist on COX-2 expression in HT-29 cells, a human colon carcinoma cell line that we also examined. The activation of PPAR␥ did not increase COX-2 expression in their experiments, either in mouse adenomas in vivo or the human cells in culture. In contrast, we found that COX-2 was induced in HT-29 and other colon carcinoma cells in response to NSAIDS, which can activate PPAR␣. Taken together, we interpret these studies as showing two results: activators of PPAR␥ can alter colon epithelial cell function in a COX-2-independent manner, and activators of PPAR␣ (or ␦) induce transcription of COX-2 in epithelial cells, which may be relevant to carcinogenesis pathways in several organs. We found that the promoter constructs of COX-2 which contained a presumed PPRE were stimulated when a plasmid encoding any of the three human isoforms was cotransfected. Thus, our studies do not establish whether all of them mediate this action in vivo or whether it is a more restricted response. The work of others (as above) suggests that PPAR␥ is unlikely to serve as a trans-activator of the COX-2 gene in the colonic epithelium.