15-Hydroxyprostaglandin Dehydrogenase-derived 15-Keto-prostaglandin E2 Inhibits Cholangiocarcinoma Cell Growth through Interaction with Peroxisome Proliferator-activated Receptor-γ, SMAD2/3, and TAP63 Proteins*

Background: 15-PGDH catalyzes PGE2 oxidation to form 15-keto-PGE2. Results: 15-PGDH-derived 15-keto-PGE2 is a PPAR-γ ligand that causes Smad2/3 association with TGFBRI/SARA and induces formation of pSmad2/3-TAP63-p53 ternary complex. Conclusion: 15-PGDH-mediated 15-keto-PGE2 signaling cascade interacts with PPAR-γ, Smad2/3, and TAP63. Significance: Induction of 15-PGDH expression or administration of 15-keto-PGE2 may represent a promising anti-cancer therapeutic strategy. Prostaglandin E2 (PGE2) is a potent lipid mediator that plays a key role in inflammation and carcinogenesis. NAD+-dependent 15-hydroxyprostaglandin dehydrogenase (15-PGDH) catalyzes the oxidation of the 15(S)-hydroxyl group of PGE2, which leads to PGE2 biotransformation. In this study, we showed that the 15-PGDH-derived 15-keto-PGE2 is an endogenous peroxisome proliferator-activated receptor-γ (PPAR-γ) ligand that causes PPAR-γ dissociation from Smad2/3, allowing Smad2/3 association with the TGF-β receptor I and Smad anchor for receptor activation and subsequent Smad2/3 phosphorylation and transcription activation in human cholangiocarcinoma cells. The 15-PGDH/15-keto-PGE2-induced Smad2/3 phosphorylation resulted in the formation of the pSmad2/3-TAP63-p53 ternary complex and their binding to the TAP63 promoter, inducing TAP63 autotranscription. The role of TAP63 in 15-PGDH/15-keto-PGE2-induced inhibition of tumor growth was further supported by the observation that knockdown of TAP63 prevented 15-PGDH-induced inhibition of tumor cell proliferation, colony formation, and migration. These findings disclose a novel 15-PGDH-mediated 15-keto-PGE2 signaling cascade that interacts with PPAR-γ, Smad2/3, and TAP63.

associated with colorectal cancer risk (12). In mouse models of carcinogenesis, overexpression of 15-PGDH decreases cancer cell growth or delays tumor formation, whereas deletion of 15-PGDH increases the susceptibility to chemically or genetically induced colon tumors (8). Targeted adenovirus-mediated delivery of 15-PGDH gene inhibited colon cancer growth in a mouse xenograft model (21). The hepatocyte growth factor and its receptor c-Met signaling promotes PGE 2 biogenesis in colorectal cancer cells via up-regulation of COX-2 and down-regulation of 15-PGDH (22). 3-Polyunsaturated fatty acids reduce the level of PGE 2 in cholangiocarcinoma cells through downregulation of COX-2 and induction of 15-PGDH (23). Indeed, reciprocal regulation between COX-2 and 15-PGDH expression has been documented in several cancers (24). Anti-cancer therapeutics, such as transforming growth factor (TGF)-␤1, glucocorticoids, and histone deacetylase inhibitors, have been shown to exert their anti-carcinogenic activity in part through induction of 15-PGDH expression (17,25). All of these findings point toward an important tumor suppressive function of 15-PGDH.
To date, the action of 15-PGDH is largely attributable to its conversion of biologically active PGE 2 , with 15-keto-PGE 2 being considered as largely inactive. This study provides novel evidence for an active role of 15-keto-PGE 2 in 15-PGDH-mediated anti-tumor effect. Our data reveal a novel 15-PGDH/15keto-PGE 2 -mediated signaling cascade that interacts with peroxisome proliferator-activated receptor-␥ (PPAR-␥), Smad2/3, and TAP63 in human cholangiocarcinoma cells. We have shown that 15-keto-PGE 2 is a natural ligand that binds to PPAR-␥ and causes its dissociation from Smad2/3, which allows subsequent Smad2/3 phosphorylation and activation of TAP63. Given that 15-PGDH converts the pro-inflammatory and pro-tumorigenic PGE 2 to the anti-inflammatory and tumor-suppressive 15-keto-PGE 2 , induction of endogenous 15-PGDH expression or delivery of exogenous 15-PGDH/ 15-keto-PGE 2 may represent promising future therapeutic interventions.
Stable Cell Lines with Double Transfections-The CCLP1 cells were first transfected with pCMV6-AV-GFP-15PGDH vector using Lipofectamine 2000 (Invitrogen), and the stably transfected cells were selected by using 1 mg/ml G418 (Invitrogen). pGFP-V-TAP63 was then co-transfected into the above stable cell line, and the transfected cells were grown in presence of 1-2 g/ml puromycin (Invitrogen) for selection. Alternatively, the CCLP1 cells were first transfected with pGFP-V-RS-15PGDH vector using Lipofectamine 2000, and the stably transfected cells were selected by using 1-2 g/ml puromycin; pcDNA3-TAP63 was then co-transfected into the above stable cell line, and the transfected cells were grown in presence of 1 mg/ml G418 for selection. The transfection efficiency was verified by Western blotting analysis.
Reversed-phase Electrospray Ionization Mass Spectrometry-An analytical LC-MS/MS method was utilized to determine the 15-keto-PGE 2 levels in CCLP1 cells. For prostaglandin extraction, CCLP1 cells in 50:50 hexane/ethyl acetate were vortexmixed, and the samples (4 ml) were dried under nitrogen and reconstituted in 180 l of 50:50 methanol, 10 mM ammonium acetate. 20 l of PGE 2 -d 4 was added to each sample and vortexed for ϳ15 s. 3 l of each sample was injected for LC-MS/MS analysis using the AB Sciex API 4000 mass spectrometer system. For HPLC, prostaglandins were chromatographically resolved using a 2.6-m, Phenomenex Kinetex Phenyl Hexyl, 100 ϫ 2.1-mm analytical column, and a linear gradient with mobile phase A (10 mM ammonium acetate in water) and mobile phase B (methanol). Individual analytes were detected using electrospray negative ionization and monitoring the transitions m/z 349 3 331 for 15-keto-PGE 2 and m/z 355 3 275 for PGE 2 -d 4 . The identification of the compounds was confirmed by comparison with reference standards. A calibration curve for 15-keto-PGE 2 was constructed and used to determine the concentration of 15-keto-PGE 2 in each sample. Standards of PGE 3 and PGD 3 were analyzed during method development to ensure chromatographic separation because they were structural isomers of 15-keto-PGE 2 as well as having similar product ion spectra.
Nuclear Extract-Approximately 10 9 cells were homogenized in a Dounce homogenizer with 15 gentle strokes of a tight-fitting pestle in 3 ml of cell lysis buffer (250 mM sucrose, 30 mM KCl, 6 mM MgCl 2 , 20 mM HEPES, pH 7.9, 0.5 mM EDTA, and 0.1% Triton X-100 with protease inhibitor mixtures from Roche Applied Science). The extract was centrifuged at 1,500 ϫ g for 10 min at 4°C. The supernatant was used as the cytosolic fraction. The pellet was resuspended in nuclear lysis buffer (50 mM Tris, pH 8, 150 mM NaCl, 0.1% Nonidet P-40, and 0.5% Triton X-100 with protease inhibitor mixtures) while rotating slowly for at least 40 min at 4°C and then sonicated to lyse the nuclei and thoroughly shear the genomic DNA. The resulting extract was centrifuged at 13,000 ϫ g for 15 min at 4°C, and the supernatant was used as the nuclear fraction.
Co-immunoprecipitation (IP) and Repeat IP-For co-immunoprecipitation, cells were transfected using Lipofectamine 2000 (Invitrogen) in a 100-mm diameter dish. At the end of each treatment, the cells were lysed in 1 ml of the whole-cell extract buffer A (50 mM Tris-Cl, pH 7.6, 150 mM NaCl, 0.5-1% Nonidet P-40, 0.1 mM EDTA, and 1.0 mM DTT) with protease inhibitor mixtures. In brief, 500-l cell lysates were pre-cleared with 30 l of protein G/A-plus agarose beads (Santa Cruz Biotechnology) by rotation for 1 h at room temperature, and the supernatant was obtained after centrifugation (1000 ϫ g) for 3 min at 4°C. Precleared supernatants were incubated with 2 g of antibody by rotation for 4 h at 4°C. The immunoprecipitates were incubated with 30 l of protein A/G-plus agarose beads by rotation overnight at 4°C and then centrifuged at 5000 rpm for 5 min at 4°C. The precipitates were washed five times for 10 min with beads wash solution (50 mM Tris-Cl, pH 7.6, 150 mM NaCl, 0.1% Nonidet P-40, 1 mM EDTA) and then resuspended in 40 l of 2ϫ SDS-PAGE sample loading buffer to incubate for 5 min at 100°C. Then Western blot was performed with another related antibody indicated in Western blotting. Repeat IP was conducted by cleansing the first precipitates with elution buffer (0.1% Triton X-100, 0.1% SDS, 0.5% BSA in PBS). Specifically, a 40-l aliquot of immunoprecipitates was eluted with 750 l of elution buffer and incubated for 50 min at room temperature, and the samples were subjected to an additional round of IP as described above.
EMSA-Cells were washed and scraped in ice-cold PBS to prepare nuclei for electrophoretic gel mobility shift assay with the use of the gel shift assay system (Promega) modified according to the manufacturer's instructions. In brief, consensus oligonucleotides for TAP63-binding site were biotin-labeled (hot probe) as follows: 5Ј-biotin forward, GATGGATTGG-ACAGGTAAAG-3Ј, and reverse CTTTACCTGTCCAATCC-ATC-3Ј; no-biotin-labeled (cold probe) forward, 5Ј-GATGGA-TTGGACAGGTAAAG-3Ј, and reverse, 5Ј-CTTTACCTGTC-CAATCCATC-3Ј (cold probe). Each binding reaction was carried out with 1 g of biotinylated dsDNA probe and 200 g of purified nuclear protein in 20 l of binding buffer containing 0.5 mg/ml poly(dI-dC) (25 mM HEPES, pH 8.0, with 50 mM KCl, 0.1% Triton X-100, 2 mM MgCl 2 , 3 mM DTT, and 5% glycerol). Twenty five pmol of unlabeled cold DNA motifs (a 250-fold excess) were added in the competition assays. Reactions were carried out for 30 min of incubation at room temperature, followed by overnight incubation at 4°C. Reaction mixtures were loaded onto 6% Tris/borate/EDTA (TBE) polyacrylamide gels and separated in 0.5% TBE at 100 V on ice until the dye front migrated two-thirds of the way to nitrocellulose membranes and Western blotting for anti-biotin.
Immunohistochemistry-Tissues were fixed with 4% paraformaldehyde, dehydrated, embedded in paraffin, and sectioned at 4 m. Sections were immunohistochemically stained using primary mouse anti-human Ki67, PCNA, and TAP63 antibodies (Santa Cruz Biotechnology). The primary antibodies were appropriately diluted. As the secondary antibody, anti-mouse IgG (horseradish peroxidase linked whole antibody from sheep, GE Healthcare) was used at 200ϫ dilution. In brief, deparaffinized sections were treated with 0.3% H 2 O 2 in methanol for 30 min to abolish endogenous peroxidase activity and then microwaved in antigen unmasking solution for antigen retrieval. Sections were blocked with 10% goat serum in PBS for 1 h at 37°C and then incubated with 2 g/ml anti-PCNA antibody (or appropriate diluted other antibodies) at 4°C overnight. Then the sections were incubated with anti-mouse IgG (horseradish peroxidase-linked whole antibody) at 37°C for 1 h. Staining was performed using 3,3-diaminobenzidine substrate kit for peroxidase according to the manufacturer's instructions (Vector Laboratories) and counterstained with hematoxylin. As a negative control, duplicate sections were immunostained without exposure to the primary antibodies. The frequency of positive cells was determined by counting the total number of cells and total positively stained cells in randomly selected ϫ200 magnification fields, including at least 1000 cells. Average numbers from the field sets were then determined and reported as the percentage of positively stained cells to the total numbers of cells.
Immunofluorescence-Following treatment, cells on a coverslip were washed twice with cold PBS and were then fixed in 4% paraformaldehyde for 10 min. The fixed cells were treated with 0.3% H 2 O 2 in methanol for 30 min to abolish endogenous peroxidase activity and then treated with 0.1% Triton and 5% DMSO. After washing three times with PBS, the samples were blocked with 5% BSA at 37°C for 1 h and then were incubated with 2 g/ml primary antibody at RT for 1 h or at 4°C overnight. Following three washings with PBS, the sections were incubated with TRITC-linked IgG or FITC-linked IgG. The samples were counterstained with DAPI (diluted 1:1000 in double distilled H 2 O) for 15 min. Coverslips were applied using Mounting Medium for fluorescence, and images were captured using Olympus FV1000II laser scanning confocal microscope (Olympus) by FV10-ASW1.7 software or fluorescence microscope (Olympus).
BrdU Staining-80% confluent cells were cultured for 24 h before treatment with 10 l of BrdU (Roche Applied Science) for 4 h. Immunofluorescent staining with an anti-BrdU antibody was performed according to the manufacturer's instructions (BD Biosciences). BrdU-positive cells from 10 random chosen fields of at least three independent samples were counted.
Wound Healing Assay-Cells were cultured to Ͼ90% confluence in 10-cm dishes. The cells were rinsed with PBS and starved in low serum media (1.5 ml; 0.5-0.1% serum in DMEM) overnight. A sterile 200-l pipette tip was used to scratch wounds through the cells. The cells were then rinsed gently with PBS. Photographs were taken using phase contrast at 10ϫ at 4, 8, and 24 h (the media were changed after each measurement).
Culture Plate Colonization Assay-1 ϫ 10 3 cells were plated in a 10-cm dish and allowed to grow for 14 days. The colonies were stained with crystal violet (Amresco) or stained with 0.25% Coomassie Brilliant Blue R-250 (Sigma) in 50% methanol and 10% glacial acetic acid. The colonies were counted, and the data were obtained from three independent experiments.
Soft Agar Colony Formation Assay-2 ϫ 10 2 cells were plated on a 6-well plate containing 0.5% (lower) and 0.35% (upper) double layer soft agar. To form lower layer soft agar, 1% agar was melted in a microwave and cooled to 40°C in a water bath, and 2ϫ DMEM containing 20% FCS was warmed to 40°C. After allowing for the temperature (40°C) to equilibrate at least 30 min, the two solutions were mixed in equal volumes to give 0.5% lower layer soft agar (0.5% agar ϩ 1ϫ DMEM ϩ 10% FCS). Then 1 ml of 0.5% lower layer soft agar was rapidly added into each well of a 6-well plate and set aside for 5 min or more until solidification at RT. To form the upper layer soft agar, 0.7% agar was melted in a microwave and cooled to 40°C in a water bath, and 2ϫ DMEM containing 20% FCS was warmed to 40°C. After allowing for the temperature (40°C) to equilibrate at least 30 min, 3 ml of the each solution (0.7% agarose, 2ϫ DMEM containing 20% FCS) was added into a 10-ml tube containing 0.1 ml of 12,000 cells/ml of cell suspension. After mixing gently by swirling, the 1-ml mixture was added into each well of the three replicate 6-well plates. Then the 6-well plates were incubated at 37°C in humidified incubator for 21 days. The cells were fed 1-2 times per week with cell culture media (DMEM).
Soft agar colonies on the 6-well plates were stained with 0.5 ml of 0.005% crystal violet for more than 1 h, and the colonies were counted.
Dual-Luciferase Reporter Assay-Cells (1 ϫ 10 5 /well of a 6-well plate) were transiently transfected with 1 g of luciferase construct (pGL3/PPRE-luc, p3TP-luc, pGL3/TAP63 promoter-luc, and pGL3-Luc) and 0.1 g of pRL-Tk (Promega) together with the indicated plasmids using Lipofectamine/plus reagent (Invitrogen). After transfection for 36 h, the cells were harvested with the lysis buffer, and luciferase activities of cell extracts were measured by DLReady TM Centro XS 3 LB960 with the use of the Dual-luciferase assay system (Promega) according to the manufacturer's instructions. Luciferase activity was normalized for transfection efficiency with Renilla luciferase activity.
Cell Proliferation WST-1 Assay-To describe growth curves, cells were synchronized in G 0 phase by serum deprivation and then released from growth arrest by re-exposure to complete medium with serum. Cell proliferation was detected by reagent WST-1 kit (Roche Applied Science) according to the manufacturer's instructions. Cell growth curve was based on the normalized values of OD 450 , and each point represents the mean of three independent samples.
DNA Pulldown-Cells were lysed by sonication in HKMG buffer (10 mM HEPES, pH 7.9, 100 mM KCl, 5 mM MgCl 2 , 100% glycerol, 1 mM DTT, and 0.5% Nonidet P-40) containing protease and phosphatase inhibitors for the preparation of nuclear exact. Equal amounts of cell nuclear extracts were precleared with streptavidin-agarose resin (Thermo) for 1 h and then were incubated with 1 g of biotinylated double-stranded oligonucleotides (TAP63-binding site), 5Ј-biotin forward, GATGGATTGGACAGGTAAAG-3Ј, and reverse, CTTTA-CCTGTCCAATCCATC-3Ј) (synthesized by Integrated DNA Technologies), together with 10 g of poly(dI-dC) at 4°C for 24 h. DNA-bound proteins were collected with incubation with streptavidin-agarose resin for 1 h with gentle shaking to prevent precipitation in solution. Following five washings of the resin-bound complex with 0.5-1.0 ml of binding buffer, the samples were boiled and subjected to SDS-PAGE and Western blot analysis.
Chromatin Immunoprecipitation-Formaldehyde crosslinking and chromatin immunoprecipitation assays were performed as described by Shang et al. (52) or according to the protocol provided by Upstate Biotechnology with modifications. In brief, cells with 90% confluence in a 150-mm dish were cross-linked by adding formaldehyde to a final concentration of 1% formaldehyde (0.68 ml of 37% formaldehyde in 25 ml of media) and rocked for 10 min at room temperature. The DNA was purified using a QIAquick spin column and eluted in 50 l/column of 10 mM Tris, pH 8.0. PCR conditions were 60 s at 94°C, followed by 35 cycles of 30 s at 94°C, 30 s at 55°C, and 30 s at 72°C, followed by 10 min at 72°C. The TAP63 promoter PCR primer sequences are as follows: P1, 5Ј-ACTTAAT-GAGATGGGAGAGG-3Ј; P2, 5Ј-GATAACAGAACTCAA-GTCCC-3Ј.
Real Time Quantitative PCR-Q-PCR was used to analyze ChIP DNA and INPUT DNA in triplicate using Fast SYBR Green PCR Master Mix (Qiagen) according to the manufactur-er's instructions. Q-PCR conditions were 20 s at 94°C, followed by 40 cycles of 10 s at 94°C, 30 s at 60°C. The input of the TAP63 promoter was normalized to the input of the IP (after subtraction of the rabbit IgG control IP). The PCR primer sequences are as follows: P1, 5Ј-ACTTAATGAGATGGGAG-AGG-3Ј; P2, 5Ј-GATAACAGAACTCAAGTCCC-3Ј.
Xenograft Tumor Study in SCID Mice-Four-week-old male athymic NOD CB17-prkdc/SCID (severe combined immunodeficiency) mice were purchased from The Jackson Laboratory and maintained in the animal facilities according to the protocol approved by the American Association for Accreditation of Laboratory Animal Care. Six athymic SCID mice per group were injected subcutaneously at the axillary area with human CCLP1 cells (stably transfected with pCMV6-AC-GFP, pCMV6-AV-GFP-15PGDH, pGFP-V-RS, and pGFP-V-RS-15PGDH, respectively) (1 ϫ 10 7 cells in 100 l of PBS). The mice were observed over 4 weeks for tumor formation. Animals were stratified so that the mean tumor sizes in all groups were nearly identical. The mice were then sacrificed, and the tumors were recovered. The wet weight of each tumor was determined for each mouse. A portion of the tissue from each tumor was snap-frozen. An additional portion of each tumor was fixed in 4% paraformaldehyde and embedded in paraffin for histological examination. 4-m sections were made and stained with hematoxylin and eosin (H&E) or anti-PCNA, anti-Ki67, anti-TAP63 antibodies.
Injection of 15-PGDH Adenoviral Vector into Xenograft Tumors-Human CCLP1 cell suspensions of 1 ϫ 10 8 (in 0.2 ml of PBS) were injected subcutaneously at the axillary area of SCID mice. The mice were divided into two groups based on the appearance of tumors and treated by injection of pAd and pAd-15-PGDH (10 10 pfu), and the diameter was determined for each mouse every 3 days, respectively. The mice were then sacrificed and the tumors recovered. The wet weight of each tumor was determined for each mouse. The tumors diameters were taken by caliper ruler measurement in two dimensions. Tumor volume was calculated using the formula: V ϭ L/2⅐w 2 .
Statistical Analysis-The significant differences between mean values were obtained from at least three independent experiments. Each value was presented as mean Ϯ S.E. unless otherwise noted, with a minimum of three replicates. The results were evaluated by SPSS12.0 statistical software (SPSS Inc., Chicago) and Student's t test was used for comparisons, with p Ͻ 0.05 considered significant. (23,26) were transfected with the GFP control vector (pCMV6-AC-GFP), the 15-PGDH expression vector, the RNAi control vector (pGFP-V-RS), and the 15-PGDH RNAi vector (pGFP-V-RS-15PGDH), respectively. Successful alteration of 15-PGDH in the stably transfected cell lines were confirmed by immunofluorescence and Western blotting (Fig. 1A). Immunofluorescence staining showed that the level of 15-PGDH was increased in cells transfected with the 15-PGDH expression vector compared with transfection with the corresponding GFP control vector; the 15-PGDH staining intensity in 15-PGDH knock-down cells was significantly lower than the corresponding RNAi control vector cells. Western blotting analysis revealed that the expression of 15-PGDH was increased in 15-PGDHoverexpressed cells (56 kDa, GFP-15PGDH fusion protein) and decreased in 15-PGDH knockdown cells (29 or 58 kDa). The 15-PGDH-overexpressed cells had reduced PGE 2 and high levels of 15-keto-PGE 2 , whereas the 15-PGDH knockdown cells show accumulation of PGE 2 and reduction of 15-keto-PGE 2 (Fig. 1B). Because 15-keto-PGE 2 is known to activate PPAR-␥ (27), we examined the PPAR response element (PPRE) reporter activity in CCLP1 cells with altered 15-PGDH expression. As shown in Fig. 1B, 15-PGDH overexpression significantly increased the PPRE reporter activity, whereas 15-PGDH knockdown reduced it. The effect of 15-PGDH on PPRE reporter activity was abolished when the cells were co-transfected with a vector expressing 15-ox prostaglandin-⌬13-reductase (PGR-2) (PGR-2 overexpression reduces the level of 15-keto-PGE 2 , as the enzyme catalyzes the reaction converting 15-keto-PGE 2 to 13,14-dihydro-15-keto-PGE 2 (27)). Furthermore, treatment of wild type CCLP1 cells with the 15-PGDH metabolite, 15-keto-PGE 2 , also increased the PPRE-luciferase activity in comparison with treatment with the DMSO vehicle control (Fig. 1B). Accordingly, EMSA analysis showed that 15-PGDH overexpression increased PPAR-␥ binding to PPRE, whereas 15-PGDH knockdown reduced it (see supporting information for Fig. 1, part I). Treatment of wild type CCLP1 cells with 15-keto-PGE 2 also increased PPAR-␥ binding to PPRE. The specificity of PPAR-␥ binding was confirmed by the observation that PPRE probe mutation abolished PPAR-␥ binding. Consistent with these observations, 15-PGDH overexpression or 15-keto-PGE 2 treatment increased the expression of the PPAR-␥ target genes, p21 WAF1/CIP1 , PTEN, and PON1, as determined by real time RT-PCR, RT-PCR, and Western blotting (supporting information for Fig. 1, part II).

15-PGDH Inhibits Cholangiocarcinoma Growth in Vivo-To
further examine the effect of 15-PGDH on tumor growth in vivo, CCLP1 cells with 15-PGDH overexpression or knockdown were inoculated subcutaneously into SCID mice, and the animals were closely monitored for tumor development. As shown in Fig. 2A  The lower panel shows PPRE-luciferase reporter activity in CCLP1 cells with indicated transfection or treatment (the data are presented as mean Ϯ S.E., **, p Ͻ 0.01). C, cell proliferation assay. Cells were synchronized in G 0 phase by serum deprivation and then released from growth arrest by re-exposure to serum (the cells were then grown in complete medium). WST-1 assay was performed in 96-well plates. Each sample was assayed in triplicates for 6 days consecutively. Cell growth curve was based on the corresponding relative values of OD 450 (the data represent mean Ϯ S.E. from three independent experiments; *, p Ͻ 0.05; **, p Ͻ 0.01). D, cell colony formation efficiency assay. 1 ϫ 10 3 cells (the numbers of cells were determined by trypsinization and counting in a hemocytometer) were plated in a 10-cm dish and incubated in a humidified atmosphere of 5% CO 2 incubator at 37°C for 14 days. For visualization, colonies were stained with crystal violet or with 0.25% Coomassie Brilliant Blue R-250 in 50% methanol and 10% glacial acetic acid. The left panel shows representative photographs of colony formation from different stable cell lines. The right panel shows colony formation rate, as determined by calculating the percentage of colonies from the 1 ϫ 10 3 cells (the data were presented as mean Ϯ S.E. from three independent experiments).
was approximately one-fifth of the control weight (0.52 Ϯ 0.09g, p Ͻ 0.01); the weight of the 15-PGDH-depleted tumors (1.45 Ϯ 0.18g) was ϳ2.5 times of the control tumor weight (0.59 Ϯ 0.08g, p Ͻ 0.01) and 12 times of the 15-PGDH overexpressed tumor weight. The time of the tumor appearance in the 15-PGDH overexpressed group was significantly longer compared with the control group (18.65 Ϯ 3.12 days versus 9.12 Ϯ 2.14 days, p Ͻ 0.01). Conversely, the tumor appearance time in the 15-PGDH knockdown group was significantly shortened compared with the control group (6.01 Ϯ 1.27 days versus 9.37 Ϯ 2.12 days, p Ͻ 0.01). As shown in Fig. 2B,  We next employed an additional tumor xenograft model in which an adenoviral vector expressing 15-PGDH (pAd-15-PGDH) was directly injected into the tumors grown in SCID mice (at 3-day intervals, starting 11 days after inoculation until the end of the experiment). As shown in Fig. 2C, pAd-15-PGDH injection significantly inhibited the growth of the xenograft
Furthermore, we observed that TGF-␤ treatment increased the expression of 15-PGDH in three human cholangiocarcinoma cell lines (CCLP1, SG231, and HuCCT1) (see supporting information for Fig. 3). Our results suggest a positive feedback loop between TGF-␤ and 15-PGDH/15-keto-PGE 2 signaling pathways in cholangiocarcinoma cells. This finding is consistent with the previous reports that TGF-␤ induces 15-PGDH in human gastrointestinal and lung cancers (30,31).
To further examine the effect of 15-PGDH on Smad2/3 activation, we performed immunofluorescence staining and Western blotting analysis to detect the phosphorylation of Smad2/3 in CCLP1 stable cell lines. As shown in Fig. 4A, Western blotting and immunofluorescence staining showed that the level of pSmad2/3 increased when 15-PGDH was overexpressed but decreased when 15-PGDH was knocked down. The effect of 15-PGDH on Smad2/3 phosphorylation was not mediated through TGF-␤ production, as 15-PGDH overexpression or knockdown or 15-keto-PGE 2 treatment did not alter the level of TGF-␤ in these cells (Fig. 4A). However, the 15-PGDH-mediated Smad2/3 phosphorylation and transcriptional activity was influenced by the addition of exogenous TGF-␤ or anti-TGF-␤ antibody. As shown in Fig. 4B, TGF-␤ treatment enhanced 15-PGDH-mediated increase in Smad2/3 transcription activity, whereas inhibition of TGF-␤ by its antibody prevented 15-PGDHinduced Smad2/3 phosphorylation and transcription activation. These findings further suggest the cross-talk between 15-PGDHderived 15-keto-PGE 2 and TGF-␤ for Smad2/3 activation.
We carried out further experiments to examine the direct effect of 15-keto-PGE 2 on Smad2/3 phosphorylation and activation. As shown in Fig. 4D, 15-keto-PGE 2 treatment enhanced Smad2/3 phosphorylation and transcription activity; these effects were inhibited when the cells were treated with the anti-TGF-␤ antibody. The 15-keto-PGE 2 -induced Smad2/3 phosphorylation and transcription activity were inhibited by GW9662, a PPAR-␥ antagonist. Although PPAR-␥ overexpression alone inhibited Smad2/3 phosphorylation and transcription activity, the presence of 15-keto-PGE 2 reversed the inhibitory effect of PPAR-␥ and enhanced Smad2/3 phosphorylation and transcription activity. Whereas PPAR-␥ associates with Smad2/3 and inhibits the interaction between Smad2 and SARA, treatment with 15-keto-PGE 2 causes Smad2/3 dissociation from PPAR-␥ and subsequent association with SARA.
Given that pharmacological PPAR-␥ ligands are known to induce the expression of p53 and its downstream genes p21 WAF1/CIP1 and GADD45 in cholangiocarcinoma (37), we performed further experiments to determine whether 15-PGDH and 15-keto-PGE 2 might influence p21 WAF1/CIP1 and GADD45 expression. As shown in Fig. 5, the levels of p21 WAF1/CIP1 and GADD45 were increased in cells with 15-PGDH overexpression but decreased in cells with 15-PGDH depletion. Accordingly, 15-keto-PGE 2 treatment also induced the expression of p21 WAF1/CIP1 and GADD45; this effect was abolished by the PPAR-␥ antagonist GW9662. These findings suggest that 15-PGDH and 15-keto-PGE 2 can activate p53 downstream genes in cholangiocarcinoma cells.
The 15-PGDH RNAi-induced reduction of TAP63, p53, and pSmad2/3 association with the TAP63 promoter was partially reversed by 15-keto-PGE 2 , but not by 15-keto-PGF 2 ␣, 5-oxo-ETE, 12-oxo-ETE, 15-oxo-ETE, or 12-oxo-LTB 4 (see supporting information for Fig. 6). This phenomenon was observed in intact cells as well as in a cell-free system with isolated cell lysates. However, the 15-PGDH-induced association of TAP63, p53, and pSmad2/3 with the TAP63 promoter was not altered by LXA 4 or LXB 4 , and this phenomenon was also observed in intact cells as well as in isolated cell lysates. These results fur-ther support the role of 15-keto-PGE 2 in mediating 15-PGDH actions in cholangiocarcinoma cells.
The 15-PGDH-induced TAP63 binding to its promoter was partially inhibited when Smad2/3 were knocked down (Fig. 7, A  and B). No TAP63 binding to its promoter was detected when both 15-PGDH and Smad2/3 were knocked down. However, overexpression of Smad2/3 enhanced 15-PGDH-induced TAP63 binding to its promoter. The 15-PGDH-induced binding of TAP63, p53, and pSmad2 to the TAP63 promoter consensus sequence was inhibited by pretreatment with the anti-TGF-␤ antibody (Fig. 7C), by overexpression of PGR-2 (Fig.  7D), or by pretreatment with the PPAR-␥ antagonist GW9662 (Fig. 7E). Consistent with these observations, 15-keto-PGE 2 treatment also increased the binding of TAP63, p53, and pSmad2 to the TAP63 promoter consensus sequence, and this effect was inhibited by pretreatment with the anti-TGF-␤ antibody (Fig. 7F) or by the PPAR-␥ antagonist FIGURE 6. 15-PGDH facilitates the assembly of pSmad2/3, p53, and TAP63 complex on the TAP63 DNA consensus sequence. A, IP, repeat IP, and repeat rIP assays in CCLP1 cells with stable 15-PGDH overexpression or knockdown. B, TAP63-Super-EMSA in CCLP1 cells with stable 15-PGDH overexpression or knockdown. The TAP63 site sequence probes include the following: 5Ј-GATGGATTGGACAGGTAAAG-3Ј (cold probe) and 5Ј-biotin-GATGGATTGGACAGGTAAAG-3Ј (hot probe). The binding of p63 protein to its DNA site was increased in 15-PGDH-overexpressed cells but decreased in 15-PGDH knockdown cells. C, p53-Super-EMSA in CCLP1 cells with stable 15-PGDH overexpression or knockdown using the TAP63 DNA-binding site sequence as indicated above. The binding of p53 to the TAP63 DNA site was increased in 15-PGDH-overexpressed cells but decreased in 15-PGDH knockdown cells. D, pSmad3-Super-EMSA in CCLP1 cells with stable 15-PGDH overexpression or knockdown using the TAP63 DNA-binding site sequence as indicated above. The binding of pSmad3 to the TAP63 DNA site was increased in 15-PGDH-overexpressed cells but decreased in 15-PGDH knockdown cells. E, TAP63 site DNA pulldown combined with TAP63 co-immunoprecipitation and Western blotting analysis using indicated antibodies. The TAP63 site DNA probe used was 5Ј-biotin-GATGGATTGGACAGGTAAAG-3Ј. Ternary complex of pSmad2, p53, and TAP63 was detected on the TAP63 DNA consensus site. The binding ability was increased in 15-PGDH-overexpressed cells but decreased in 15-PGDH knockdown cells. Biotin and histone were used as loading controls. F, TAP63 site DNA pulldown using the TAP63 site DNA probe (as described above), in combination with the pSmad2 co-immunoprecipitation and Western blotting (WB) analysis using the indicated antibodies. GW9662 (Fig. 7G). These observations demonstrate the involvement of PPAR-␥ in 15-PGDH/15-keto-PGE 2 -induced TAP63 autotranscription.
We further performed chromatin immunoprecipitation assays to determine the binding of TAP63, p53, and pSmad2/3 to the TAP63 promoter consensus site. As shown in Fig. 8A, the FIGURE 7. Effect of TGF-␤ signaling, PPAR-␥, and PGR-2 on 15-PGDH-and 15-keto-PGE 2 -induced binding of the pSmad2/3, p53, and TAP63 tertiary complex to the TAP63 consensus sequence. A, TAP63-Super-EMSA using the aforementioned TAP63 site DNA probe in cells transfected with indicated plasmids. Knockdown of Smad2/3 prevents 15-PGDH-induced TAP63 binding to its consensus DNA site. B, TAP63 site DNA pulldown using TAP63 site DNA probe (as described above) and the indicated antibodies. Smad2/3 knockdown prevented 15-PGDH-induced binding of pSmad2/3, p53, and TAP63 to the TAP63 consensus sequence on the TAP63 promoter, whereas Smad2/3 overexpression enhanced the 15-PGDH induced binding ability. Histone and biotin were used as loading controls. C, TAP63 site DNA pulldown in combination with co-immunoprecipitation and Western blotting (WB) using the indicated antibodies. Incubation with the anti-TGF-␤ antibody (10 ng/ml) prevented 15-PGDH-induced binding of pSmad2/3, p53, and TAP63 to the TAP63 consensus sequence. D, TAP63 site DNA pulldown in combination with co-immunoprecipitation and Western blotting using the indicated antibodies. Overexpression of PGR-2 prevented 15-PGDH-induced binding of pSmad2/3, p53, and TAP63 to the TAP63 consensus sequence. E, TAP63 site DNA pulldown in combination with co-immunoprecipitation and Western blotting using indicated antibodies. The PPAR-␥ antagonist GW9662 prevented 15-PGDH-induced binding of pSmad2/3, p53, and TAP63 to the TAP63 consensus sequence. F, TAP63 site DNA pulldown in combination with co-immunoprecipitation and Western blotting using indicated antibodies. 15-Keto-PGE 2 treatment increased the binding of pSmad2, p53, and TAP63 to the TAP63 DNA consensus site, and this effect was prevented by anti-TGF-␤ antibody treatment (10 ng/ml). G, TAP63 site DNA pulldown in combination with co-immunoprecipitation and Western blotting using indicated antibodies. GW9662 prevented 15-keto-PGE 2 -induced binding of pSmad2/3, p53, and TAP63 to the TAP63 consensus sequence.

DISCUSSION
To date, the action of 15-PGDH is largely attributable to its conversion of biologically active PGE 2 , with its enzymatic products being considered as largely inactive. This study provides novel evidence for an active role of 15-keto-PGE 2 in 15-PGDH-mediated anti-tumor effect. We have shown that the 15-PGDH-derived 15-keto-PGE 2 is an endogenous PPAR-␥ ligand that causes PPAR-␥ dissociation from Smad2/3, allowing Smad2/3 binding to TGFBRI and SARA and subsequent Smad2/3 phosphorylation and activation. By using chromatin immunoprecipitation assay as well as immunoprecipitation and Western blotting analysis, we were able to detect the PPAR-␥ and Smad2/3 binding complex in cell lysates and nuclear extracts. In the absence of 15-keto-PGE 2 , Smad2/3 is associated with PPAR-␥ and unable to associate with TGFBRI and SARA. Binding of 15-keto-PGE 2 to PPAR-␥ leads to dissociation of PPAR-␥ from Smad2/3, allowing Smad2/3 association with TGFBRI and SARA and subsequent Smad2/3 phosphorylation and activation. These findings suggest that 15-keto-PGE 2 is a PPAR-␥ ligand that reverses PPAR-␥ inhibition of Smad2/3. In contrast, other PPAR-␥ agonists, such as ciglitazone, rosiglitazone, or 15-d-PGJ 2 , had no effect on Smad2/3 activation. Thus, the concentration of 15-keto-PGE 2 in cells and tissues represents a key factor that determines Smad2/3 activity.
Although reduction of PGE 2 may partly contribute to 15-PGDH-mediated inhibition of tumor cell growth, our findings in this study provide novel evidence for an active role of Panel b, WST cell proliferation assay. Each sample was assayed in triplicates for 6 consecutive days, and the data represent means Ϯ S.E. from three independent experiments (**, p Ͻ 0.01; *, p Ͻ 0.05). Panel c, soft agar colony formation assay. The cells were incubated at 37°C for 21 days (the culture medium was changed 1-2 times per week). The culture plates were stained with 0.5 ml of 0.005% Crystal Violet (Sigma) for at least 1 h, and the numbers of colonies were counted. The data are expressed as means Ϯ S.E. from three independent experiment (**, p Ͻ 0.01). Panel d, wound healing assay at 0 and 24 h in indicated CCLP1 stable cell lines. The data are expressed as means Ϯ S.E. from three independent experiment (**, p Ͻ 0.01). B, overexpression of TAP63 prevents 15-PGDH knockdown effects. Panel a, Western blotting for 15-PGDH and TAP63 in CCLP1 cells stably transfected with 15-PGDH RNAi vector and/or TAP63 expression vector. Panel b, WST cell proliferation assay. Each sample was assayed in triplicates for 6 consecutive days, and the data represent means Ϯ S.E. from three independent experiments (**, p Ͻ 0.01; *, p Ͻ 0.05). Panel c, soft agar colony formation assay. The cells were incubated at 37°C for 21 days (the culture medium was changed 1-2 times per week). The culture plates were stained with 0.5 ml of 0.005% crystal violet (Sigma) for at least 1 h, and the numbers of colonies were counted. The data are expressed as means Ϯ S.E. from three independent experiment (**, p Ͻ 0.01). Panel d, wound healing assay at 0 and 24 h in indicated CCLP1 stable cell lines. The data are expressed as means Ϯ S.E. from three independent experiment (**, p Ͻ 0.01).
15-keto-PGE 2 in 15-PGDH-mediated inhibition of tumor cell growth. This assertion is based on several observations. 1) 15-Keto-PGE 2 treatment and 15-PGDH overexpression were found to have comparable effects on PPAR-␥ activation and Smad2/3 phosphorylation/activation. 2) 15-Keto-PGE 2 treatment and 15-PGDH overexpression had comparable effects on Smad2/3 association with TAP63 and p53, their DNA binding ability, and induction of TAP63 expression. 3) 15-PGDH-induced PPAR-␥ activation and Smad2/3 phosphorylation/activation were inhibited by overexpression of PGR-2, which converts 15-keto-PGE 2 to 13,14-dihydro-15-keto-PGE 2 . 4) 15-PGDH-induced Smad2/3 association with TAP63 and p53, their DNA binding ability, and induction TAP63 expression were all inhibited by PGR-2. 5) The effects induced by 15-PGDH depletion (e.g. binding of PPAR-␥ to Smad2, dissociation of Smad2 from SARA and TGFBRI, inhibition of Smad2/3 activity, and inhibition of TAP63 expression) were partially reversed by 15-keto-PGE 2 but not by the other 15-PGDH metabolites. 6) Two other 15-PGDH substrates (LXA 4 and LXB 4 ) did not significantly alter Smad2/3 activation and TAP63 expression in 15-PGDH-overexpressed cells. All of these observations support an active role of 15-keto-PGE 2 in 15-PGDHmediated inhibition of cholangiocarcinoma cell growth. In our system, we observed that 15-PGDH overexpression led to ϳ2-fold reduction of PGE 2 and a 5-fold increase of 15-keto-PGE 2 . The difference in the fold changes may relate to the low basal level of 15-keto-PGE 2 in the cells, although it is possible that other factors that regulate PGE 2 and 15-keto-PGE 2 biosynthesis and degradation may also be implicated. Given the relatively high concentrations of 15-PGDH metabolites utilized in the experiments in vitro (at micromolar concentrations), the physiological implication of these findings in vivo remains to be further defined.
In this study, the effect of 15-keto-PGE 2 was compared with several other PPAR-␥ agonists, including ciglitazone, rosiglitazone, and 15-d-PGJ 2 . We found that only 15-keto-PGE 2 was able to cause PPAR-␥ dissociation from Smad2/3 (thus removing PPAR-␥ inhibition on Smad2/3), whereas ciglitazone, rosiglitazone, and 15-d-PGJ 2 (15-deoxy-⌬ 12,14 -prostaglandin J 2 ) were unable to dissociate PPAR-␥ from the Smad2/3 binding complex, despite their ability to enhance PPAR-␥ transcription activity in the PPRE reporter activity assays. These findings suggest differential regulation of Smad2/3 by PPAR-␥ via specific ligands. Although 15-keto-PGE 2 enhances PPAR-␥-mediated PPRE reporter activity (as an agonist), our data showed that 15-keto-PGE 2 counteracted PPAR-␥ action in the regulation of Smad2/3 activity (as an antagonist). Thus, 15-keto-PGE 2 differentially regulates PPAR-␥ activity depending on specific downstream targets. To our knowledge, this is the first study describing activation of Smad2/3 by a PPAR-␥ ligand. The importance of 15-keto-PGE 2 in regulation of Smad2/3 is underscored by the diverse functions of TGF-␤ and Smad signaling and by the fact that 15-keto-PGE 2 is an endogenous PPAR-␥ ligand produced by 15-PGDH, an enzyme that is abundantly expressed in a vari- 15-Keto-PGE 2 is an endogenous PPAR-␥ ligand that causes PPAR-␥ dissociation from Smad2/3, allowing Smad2/3 association with TGFBRI and SARA and subsequent Smad2/3 phosphorylation and transcription activation. The 15-PGDH/15-keto-PGE 2 -induced Smad2/3 phosphorylation facilitates the formation of the pSmad2/3-TAP63-p53 ternary complex and their binding to the TAP63 promoter, which leads to induction of TAP63 and inhibition of cancer growth. ety of non-neoplastic human tissues and cells. This is in contrast with 15-d-PGJ 2 , a PPAR-␥ ligand with limited biological significance, as it is the dehydration end product of PGD 2 with extremely low or undetectable concentration in vivo (51). In our study, 15-d-PGJ 2 as well as ciglitazone and rosiglitazone, the two most commonly used thiazolidinedione anti-diabetic drugs, showed no significant effect on Smad2/3 activation.
The exact mechanism for the different actions between 15-keto-PGE 2 and other PPAR-␥ ligands is not clear and remains speculative. It is possible that this process might involve conformational change of the Smad2/3-binding site of the PPAR-␥ triggered by 15-keto-PGE 2 but not thiazolidinediones. However, it is also possible that some other yet to be identified factors might be involved.
In summary, this study depicts a novel 15-PGDH-mediated 15-keto-PGE 2 signaling cascade that interacts with PPAR-␥, Smad2/3, and TAP63 to inhibit cholangiocarcinoma cell growth. 15-Keto-PGE 2 is identified as a natural ligand that binds to PPAR-␥ and causes its dissociation from Smad2/3, which allows subsequent Smad2/3 phosphorylation and activation of TAP63. Given that 15-PGDH converts the pro-inflammatory and pro-tumorigenic PGE 2 to the anti-inflammatory and tumor-suppressive 15-keto-PGE 2 , induction of 15-PGDH expression or administration of 15-keto-PGE 2 may represent a promising anti-cancer therapeutic strategy that warrants further investigation.