Feedback Control of Cyclooxygenase-2 Expression through PPAR γ

Cyclooxygenase-2 (COX-2), a rate-limiting enzyme for prostaglandins (PG), plays a key role in inflammation, tumorigenesis, development, and circulatory homeostasis. The PGD(2) metabolite 15-deoxy-Delta(12, 14) PGJ(2) (15d-PGJ(2)) was identified as a potent natural ligand for the peroxisome proliferator-activated receptor-gamma (PPARgamma). PPARgamma expressed in macrophages has been postulated as a negative regulator of inflammation and a positive regulator of differentiation into foam cell associated with atherogenesis. Here, we show that 15d-PGJ(2) suppresses the lipopolysaccharide (LPS)-induced expression of COX-2 in the macrophage-like differentiated U937 cells but not in vascular endothelial cells. PPARgamma mRNA abundantly expressed in the U937 cells, not in the endothelial cells, is down-regulated by LPS. In contrast, LPS up-regulates mRNA for the glucocorticoid receptor which ligand anti-inflammatory steroid dexamethasone (DEX) strongly suppresses the LPS-induced expression of COX-2, although both 15d-PGJ(2) and DEX suppressed COX-2 promoter activity by interfering with the NF-kappaB signaling pathway. Transfection of a PPARgamma expression vector into the endothelial cells acquires this suppressive regulation of COX-2 gene by 15d-PGJ(2) but not by DEX. A selective COX-2 inhibitor, NS-398, inhibits production of PGD(2) in the U937 cells. Taking these findings together, we propose that expression of COX-2 is regulated by a negative feedback loop mediated through PPARgamma, which makes possible a dynamic production of PG, especially in macrophages, and may be attributed to various expression patterns and physiological functions of COX-2.

different regulation of COX-2 expression remain to be elucidated.
In the present study, we investigated the different effect of 15d-PGJ 2 on expression of COX-2 gene between macrophage-like differentiated U937 cells and BAEC. We provide evidence that a unique expression pattern of PPARγ is involved in this different effect.
Especially in the U937 cells, LPS down-regulates PPARγ mRNA but up-regulates GR mRNA although both 15d-PGJ 2 and DEX suppressed COX-2 expression, at least, by interfering the NF-κB signaling pathway. With additional evidences, we propose that expression of COX-2 will be regulated by a negative feedback loop mediated through PPARγ. This makes possible a dynamic production of PG especially in macrophages.

MATERIALS AND METHODS
Cell culture---U937 cells (10) and BAEC (11) were grown in RPMI 1640 and DMEM media, respectively, supplemented with 10% fetal calf serum (Flow), 50 µM 2mercaptoethanol, 100 units/ml penicillin and 100 µg/ml streptomycin. For differentiation into monocyte/macrophage, U937 cells were treated with 100 nM TPA and allowed to adhere for 48 h, after which they were fed with TPA-free medium and cultured for 24 h prior to use.
Determination of PG synthesis---The TPA-differentiated U937 cells (5 x 10 5 cells/well) were cultivated on 12-well tissue culture plates with 1 ml of the culture medium. After a further 24-h of incubation, the relevant reagents were added to the medium. After 12 h of levels of mRNA were calculated on the basis of hybridization signals as measured by an imaging analyzer BAS 5000 (Fuji Photo Film Co., Tokyo). RT-PCR analysis was performed using KOD DNA polymerase (Toyobo, Osaka) as described previously (16). The primer pair for PPARγ amplification was designed to anneal to both human (33) and bovine (34) sequences had the following sequence, 5'-CCAAAGTGCAATCAAAGTGGAGCC-3' and Western Blot Analysis---Cell lysates (10 5 cell equivalents) were subjected to SDSpolyacrylamide gel electrophoresis on 10% gels. The separated proteins were electroblotted onto a polyvinylidene difluoride membrane (Millipore). The membranes were probed with the human COX-2 antisera (IBL, Gunma, Japan) and visualized using the ECL Western blot analysis system (Amersham Pharmacia Biotech) according to the manufacturer's instructions.

Down-regulation of PPARγ by LPS in U937
Cells---DEX-mediated suppression of COX-2 expression is modulated by GR, which will be attributed to distinct effect of DEX on COX-2 expression between macrophages and endothelial cells (16). Similarly, we determined whether expression of PPARγ accounts for the different effects of 15d-PGJ 2 . Expression of PPARγ mRNA was observed in the differentiated U937 cells (33) as well as monocytes and macrophages (27) and that, was down-regulated by treatment of LPS in a time-dependent manner ( Fig. 3A and 3C). This down-regulation was not observed by treatment with 15d-PGJ 2 alone (data not shown). In contrast, no PPARγ mRNA was detected in BAEC ( Fig. 3A and 3B) and HUVEC although PPARδ mRNA was constitutively expressed in both cells as well as the U937 cells (data not shown). In aortic smooth-muscle cells (36), the PPARα activators inhibit the inflammatory response. However, in the U937 cells as well as activated macrophages (24), no expression of PPARα was observed by northern blot analysis, and that, a selective PPARα activator Wy-14643 (100 µM) showed no effect on COX-2 mRNA expression in the U937 cells (data not shown).

Inverse expression of PPARγ and GR by LPS in U937 cells---As described previously,
suppressive effect of 15d-PGJ 2 on COX-2 expression was milder than that of DEX in the U937 cells. To address this question, we examined expression levels of GR after various treatments (Fig. 4). LPS increased GR mRNA about two-fold, which shows inverse expression pattern between GR and PPARγ. Moreover, DEX partly restored the suppressive expression of PPARγ by LPS. This inverse expression pattern between GR and PPARγ will be partly explained the milder suppressive effect of 15d-PGJ 2 than that of DEX, and that, suggesting that different roles of GR and PPARγ on COX-2 expression. (also known as NF-IL6β) activates COX-2 transcription mainly through CRE whereas the NF-κB and NF-IL6 sites also contribute to the COX-2 expression (11). Transient transfection assay using the COX-2 promoter (-327/+59) showed that 15d-PGJ 2 did not suppress the COX-2 promoter activity in BAEC (Fig. 6), which is consistent with no suppression of COX-2 mRNA by 15d-PGJ 2 (Fig. 2). However, by coexpression of PPARγ, BAEC acquired the suppressive regulation of COX-2 gene by 15d-PGJ 2 but not by DEX, whereas by coexpression of GR (16), acquires more suppressive regulation by DEX than that by 15d-PGJ 2 ( Fig. 6), indicating the involvement of PPARγ in regulation of COX-2 expression by 15d-PGJ 2 .
Suppression of PGD 2 Production by NS398---In the presence of albumin or serum, PGD 2 is metabolized to PGJ 2 and ∆ 12 -PGJ 2 , natural ligands for PPARγ (22, 23), and that, these PGD 2 metabolites actively transport to cellular nuclei (37). Therefore, there is a possibility that COX-2 expression is self-regulated by PGD 2 metabolites, which is produced in a COX-2 dependent manner. To examine this possibility, we investigated whether U937 cells produce PGD 2 in a COX-2 dependent manner. EIA assay showed that U937 cells produced PGD 2 , and a COX-2 selective inhibitor NS398 suppressed this production (Fig. 7). Moreover, PGD 2 synthase mRNA was detected by RT-PCR analysis of the U937 cells (data not shown). These results are also consistent with previous reports in bone-marrow derived macrophages (19) and specialized antigen-presenting cells (38).

DISCUSSION
The present study has shown that 15d-PGJ 2 suppressed LPS-induced COX-2 mRNA in macrophage-like differentiated U937 cells, but not in vascular endothelial cells. This difference comes from different expression pattern of PPARγ, that is, much higher expression in the U937 cells ( Fig. 3) and acquisition of the 15d-PGJ 2 -sensitivity on the COX-2 expression by coexpression of PPARγ into BAEC (Fig. 6). Moreover, LPS down-regulates PPARγ mRNA but up-regulates GR although both suppressed COX-2 promoter activity by interfering with the NF-κB signaling pathway (13) (Fig. 5). On the other hand, U937 cells as well as macrophages (19) produces PGD 2 in a COX-2 dependent manner, and PGD 2 is spontaneously converted to PGJ 2 derivatives by non-enzymatic dehydration (37). Taken together, we propose that PGD 2 metabolites such as 15d-PGJ 2 work an intracellular signaling mediator which retains the low expression level of COX-2 by negative feedback loop meditated through PPARγ in macrophages (Fig. 8). After LPS-treatment, up-regulation of COX-2 was coincident with down-regulation of PPARγ (Fig. 3), which canceled the negative feedback loop. Simultaneously rapid increase of PGE 2 ( Fig. 1) was observed and cAMP enhanced the COX-2 transcription by LPS in the U937 cells (13) suggesting that COX-2 expression will be enhanced by a positive feedback loop (20) mediated through PG receptors.
In fact, existence of PGE receptor subtypes EP2 and EP4 increasing an intracellular cAMP level were reported in murine macrophage-like cell line, J774.1 (39). This positive feedback loop can be suppressed by DEX since LPS up-regulates GR mRNA and will increase the sensitivity to DEX (Fig. 4). Moreover, the possibility that COX-2 has anti-inflammatory properties at the later phase of a carrageenin-induced pleurisy was recently reported (40), which will be also explained by the negative feedback regulation of COX-2 by PPARγ.
PPARγ is activated by a range of synthetic and naturally occurring substances, including antidiabetic thiazolidinediones, polyunsaturated fatty acids, PGD 2 metabolites, components of oxidized LDL and 12/15-lipoxygenase products (41). Among them, rosiglitazone (BRL49653), the most potent synthetic ligand for PPARγ, did not suppressed LPS-induced expression of COX-2 mRNA in U937 cells (data not shown). Interestingly, in spite of the stronger binding activity of rosiglitazone in vitro, several reports point out the higher biological activity of 15d-PGJ 2 compared to rosiglitazone (42). In this context, 15d-PGJ 2 but not PPARγ agonists such as troglitazone is recently reported as a direct inhibitor of IκB kinase which is responsible for the NF-κB activation (43) TPA-differentiated U937 cells would be assumed as responsive macrophages because of similar expression patterns of COX-2 and TXA 2 synthase mRNAs in casein-elicited peritoneal macrophage (19). However, expression of PPARγ but not PPARα is observed in both undifferentiated and differentiated U937 cells, which is different from the report that PPARγ is induced upon differentiation into macrophages whereas PPARα is already present in undifferentiated monocytes (45). This discrepancy may be attributed to heterogeneity of macrophages (19).
COX-2 expression is regulated not only in cell-type specific but also species-specific manner. In fact, the delayed induction of COX-2 by gonadotropin was reported in bovine granulosa cells but not in the rat cells, however; the induction was observed in both species by guest on July 23, 2018 http://www.jbc.org/ Downloaded from (46). Similarity of nucleotide sequence of the COX-2 promoter region between bovine and human was higher than that between bovine and rat, and that, cis-acting elements for NF-κB, NF-IL6 sites, and CRE are conserved among human, bovine, rat and mouse COX-2 promoter regions. No suppression of 15d-PGJ 2 on the LPS-induced COX-2 mRNA and no detectable level of PPARγ mRNA were observed in HUVEC as well as BAEC. Therefore, there will be not so much difference on the regulation of COX-2 expression at least between human and bovine endothelial cells.
PPARγ and GR mRNAs are inversely regulated by LPS in the U937 cells (Fig. 4) although both 15d-PGJ 2 and DEX suppressed COX-2 promoter activity by interfering with the NF-κB signaling pathway (Fig. 5). Ligands for PPARs and DEX are reported to enhance COX-2 expression in some carcinoma cells (28, 29) and amnion cells (47), respectively.
These different effects on COX-2 expression may explain by different regulated expression of PPARs, steroid hormone receptors and C/EBPs. In this context, estrogen-induced production of a PPAR ligand was reported in a PPARγ-expressing tissue, in which induced conversion of PGD 2 to a metabolite was observed (48). Interestingly, a precise transcriptional network among these transcription factors is important for adipocyte differentiation. Therefore, it will be interesting to find each relationship between COX-2 and the transcriptional network in physiological and pathophysiological functions.

Fig. 2. Different effect of 15d-PGJ 2 between differentiated U937 cells and BAEC.
Macrophage-like differentiated U937 cells and BAEC were treated with LPS for 5 h in the presence or absence of 10 µM 15d-PGJ 2 . Isolated total RNA (10 µg) was examined by Northern blot analysis for expression of COX-2 mRNA. Relative amount of COX-2 mRNA was measured by an image-analyzer after normalization by that of GAPDH, and indicated as 100% without 15d-PGJ 2 since expression of COX-2 mRNA was very low in both cells without the LPS-treatment. Results represent the mean + standard deviations of three separate dishes. GAPDH expression level in the U937 cells is higher than that in BAEC although ethidium bromide-staining intensities of 28S RNA were equal between them as measured by an imaging analyzer FLA2000. Similar result was also obtained when using a bovine COX-2 cDNA probe instead of the human probe.

Fig. 3. Different expression patterns of PPARγ mRNA between U937 cells and BAEC.
Macrophage-like differentiated U937 cells and BAEC were treated with LPS for 5 h in the presence or absence of 10 µM 15d-PGJ 2 . A, isolated total RNA (10 µg) was examined by Northern blot analysis using radiolabeled probes for COX-2, PPARγ, COX-1 and GAPDH, respectively after stripping each probe in this order. B, RNA samples (1 µg each) extracted from U937 cells and BAEC were subjected to RT-PCR to confirm the relative expression levels of PPARγ, as described under "Materials and methods." C, U937 cells were treated with LPS (10 µg/ml), and at the times indicated total RNA was isolated and examined by Northern blot analysis using radiolabeled probes for PPARγ and GAPDH. Relative amount of  Macrophage-like differentiated U937 cells were treated with the indicated reagents for 5 h.
Isolated total RNA (10 µg) was examined by Northern blot analysis using radiolabeled probes for COX-2, PPARγ, GR and GAPDH, respectively after stripping each probe in this order.