Up-regulation of Cyclooxygenase-2 Expression and Prostaglandin Synthesis in Endometrial Stromal Cells by Malignant Endometrial Epithelial Cells A PARACRINE EFFECT MEDIATED BY PROSTAGLANDIN E 2 AND NUCLEAR FACTOR- (cid:1) B*

, We investigated the regulation of prostaglandin production in normal endometrial stromal cells (ESC) by malignant endometrial epithelial cells. We found that cyclooxygenase (COX)-2 mRNA and protein levels and prostaglandin (PG)E 2 production in ESC were signifi- cantly increased by Ishikawa malignant endometrial epithelial cell conditioned medium (MECM). By using transient transfection assays, we found that the (cid:1) 360/ (cid:1) 218-bp region of the COX-2 promoter gene was critical for MECM induction of promoter activity. This MECM-responsive region contained a variant nuclear factor (NF)- (cid:2) B site at (cid:1) 222 to (cid:1) 213 that, when mutated, completely abolished COX-2 promoter activation by MECM. Employing electrophoretic mobility shift assays, we further demonstrated that binding of NF- (cid:2) B p65 to this NF- (cid:2) significantly the new control medium treated with Act D was 3.0 h. Relative levels of COX-2 mRNA expression were determined by densitometric scanning of the bands and normalized to the GAPDH signal. expressed as a at control time point are three MECM D

Supported by a plethora of experimental evidence, prostaglandin (PG) 1 production emerged as a highly promising therapeutic target not only in the treatment of many inflammatory diseases but also several types of human cancers. The proposal that PGs contribute to carcinogenesis is supported further by compelling evidence that inhibitors of cyclooxygenase (COX) activity (and thereby of PGs formation) protect against colon, mammary, esophageal, and lung cancer in humans (1). The increased amounts of PGs in tumors reflect enhanced synthesis, which occurs by COX-catalyzed metabolism of arachidonic acid. PGs are synthesized from arachidonic acid by two different isoforms of COX, referred to as COX-1 and COX-2. They share ϳ60% identity at the amino acid level and have similar enzymatic activities, but although they catalyze the same reaction, these two isoforms have been suggested to have distinct biological functions (2)(3)(4)(5). COX-1 is constitutively expressed in most mammalian tissues and is thought to carry out housekeeping functions such as cytoprotection of the gastric mucosa, regulation of renal blood flow, and control of platelet aggregation. In contrast, COX-2 mRNA and protein are normally undetectable in most tissues but can be rapidly induced in response to proinflammatory or mitogenic stimuli, which included various cytokines, growth factors, oncogenes, endotoxins, and chemicals (6 -12). Enhanced expression of COX-2, but not COX-1, has been found in colon, pancreatic, and gastric cancer tissues (13)(14)(15). Previous studies (16 -19) have shown that overexpression of COX-2 reduces the rate of apoptosis, increases the invasiveness of malignant cells, and promotes angiogenesis. Therefore, it is believed that increased production of PGs (especially PGE 2 ) in tumors is a result of enhanced COX-2 gene expression (13).
We and others (20 -23) 2 have examined the expression and the distribution of the COX-2 by immunohistochemistry in human normal endometrium and endometrial cancer. Specific staining for COX-2 could be found only in the surface and glandular epithelium in normal endometrium but not in the endometrial stroma. On the other hand, not only the COX-2 expression in the surface and glandular epithelia of the endometrial cancer was increased as compared with normal endometrium, an increase in the COX-2 immunostaining in the stroma of the endometrial cancer relative to that of normal endometrium was also noted. Similar results have been described for colon carcinoma, esophageal carcinoma, and malignant melanoma (24 -26). These results were suggestive of a cross-talk between malignant epithelial cells and surrounding stromal cells to favor COX-2 expression in the endometrial tumors. To test this hypothesis, the present investigation was designed to examine the direct effect of conditioned medium of a cancerous endometrial cell line (Ishikawa cells) on COX-2 expression in normal stromal cells of the human endometrium. We hypothesized that malignant epithelial cells secreting factor(s) that act in a paracrine fashion might be responsible for increased COX-2 expression in the stromal cells. In addition, we attempted to characterize the critical cis-acting elements that mediate induction the human COX-2 gene expression in normal endometrial stromal cells by malignant endometrial epithelial cell conditioned medium.
Cell Culture-Human normal endometrial tissues were obtained at the time of surgery from reproductive aged women (n ϭ 12) who were undergoing hysterectomy for advanced cervical dysplasia after obtaining informed consent following a protocol approved by the Office for Protection of Research Subjects of the University of Illinois, Chicago. These patients did not receive hormonal treatments and not take antiinflammatory drugs (especially COX inhibitors) before surgery. Six specimens were in the proliferative phase, whereas the other six were secretory. No differences in experimental results were noted with respect to the cycle phase. Normal endometrial stromal cells (ESC) were cultured using a protocol reported previously (27). The cells were studied at passage 4 -6. Confluent ESC were serum-deprived for 16 h in serum-free, phenol red-free Dulbecco's modified Eagle's medium/Ham's F-12 (DMEM/F-12) before subjected to the following two treatments: (i) serum-free, phenol red-free DMEM/F-12 as the base-line control; (ii) serum-free, phenol red-free DMEM/F-12 conditioned with Ishikawa human malignant endometrial epithelial cells (malignant epithelial cell conditioned medium (MECM)). Treated ESC were then used to isolate total RNA for reverse transcriptase-PCR, whole cell protein extracts for Western blot analysis, and nuclear extracts for electrophoretic mobility shift assay (EMSA). The conditioned medium was generated in the following fashion. Ishikawa epithelial cells were initially grown in 75cm 2 flasks in DMEM/F-12 with 10% fetal bovine serum, penicillin (100 units/ml), streptomycin (100 g/ml), and amphotericin B (250 ng/ml) (growth medium). After Ishikawa cells were grown to confluence, culture medium was switched to serum-free, phenol red-free DMEM/F-12 containing antibiotics for a 16-h washout period to collect MECM. Then the cells were incubated in new serum-free, phenol red-free DMEM/ F-12 containing antibiotics for 72 h to allow accumulation of secreted factors in the medium. We collected MECM and centrifuged it to remove the cell debris. The supernatant were transferred to a clean tube and immediately frozen and kept at Ϫ80°C for the future use. As an additional control, serum-free, phenol red-free DMEM/F-12 incubated with HES human benign endometrial epithelial cell line (benign epithelial cell conditioned medium (BECM)) was prepared following a similar procedure described above. We also generated the conditioned media from ECC-1 human malignant endometrial epithelial cell line (ECC-1CM), HEC-1A human malignant endometrial epithelial cell line (HEC-1ACM), T47D human malignant mammary epithelial cell line (T47DCM), MCF-7 human malignant mammary epithelial cell line (MCF-7CM) and LNCaP human malignant prostate epithelial cell line (LN-CaPCM) following a similar procedure described above. For assessing the concentration dependence, MECM was concentrated 5-or 10-fold in an Ultrafree PF-60 Ultrafiltration Device using 5,000 molecular weight cut-off membranes (Millipore, Bedford, MA).
Semi-quantitative RT-PCR Amplification-ESC were cultured in 100-mm dishes until confluent in the growth medium as described above and switched to serum-free, phenol red-free media for 16 h. These cells were then incubated under various conditions, i.e. control or MECM for 8 h. Total RNA was isolated from ESC using the RNeasy mini kit (Qiagen, Valencia, CA), following the protocol suggested by the manufacturer. The integrity of the RNA was confirmed by agarose gel electrophoresis. For RT-PCR analysis of COX-2 mRNA, the SuperScript First-Strand Synthesis System (Invitrogen, Carlsbad, CA) was used to synthesize the first strand cDNA as instructed by the supplier. Briefly, 5 g of total RNA isolated from ESC was treated with DNase I (1 units/l). One l of this was reserved for PCR amplification with primers specific for glyceraldehyde-3-phosphate dehydrogenase (GAPDH), providing a control for equal starting amounts of total RNA in samples and PCR efficiency. The remainder of the DNase-treated RNA was directly reverse-transcribed. One l of the reverse transcriptase reaction mix was used for PCR with oligonucleotide pairs specific for COX-2 and GAPDH. The nucleotide sequences of the primer pairs employed and PCR conditions were reported previously (27). The PCR cycle numbers were 38 for COX-2 and 30 for GAPDH. PCR performed with the original RNA sample after DNase I digestion (see above) did not yield any products, confirming that amplified products were dependent on the presence of template generated by reverse transcription and not the result of contamination with extraneous DNA. Aliquots of the reaction products were analyzed by electrophoresis in an agarose gel and ethidium bromide staining. Intensity of PCR products was quantified using the Quantity One 1-D Analysis Software (Bio-Rad). We assert that these data are semi-quantitative (relative to control GAPDH) based on the following test performed prior to data analysis. Both products were assayed in the linear response range of the RT-PCR amplification process; the cycle number used in this assay was determined by finding the midpoint of linear amplification on a sigmoidal curve for both amplification products with cycle numbers 25-42 plotted against band density (see Fig. 1A).
Prostaglandin E 2 Measurements-ESC were plated in 6-well tissue culture plates in the growth medium as described above and allowed to become established as confluent monolayers for 24 h. All of the cells were serum-depleted for at least 16 h and treated with control or MECM (2 ml/well). Stimulations were performed for 24 h, and the supernatants were transferred to clean microcentrifuge tubes. Two 100-l aliquots of supernatant/sample were assayed by prostaglandin E 2 Immunoassay Kit (R & D Systems, Minneapolis, MN), according to the manufacturer's instructions. The concentration of PGE 2 was determined for competitive binding enzyme-linked immunosorbent assay (ELISA) using the Microplate Reader, model 550 (Bio-Rad). These measurements were made in duplicate and repeated in three separate experiments.
Western Blotting-ESC were cultured in 100-mm dishes until confluent in the growth medium as described above and switched to serumfree, phenol red-free media for 16 h. These cells were then incubated under various conditions, i.e. control, BECM, or MECM for 8 h. Total protein was extracted from whole cells using M-PER mammalian protein extraction reagent (Pierce), following the protocol suggested by the manufacturer. Protein concentration was measured using a BCA Protein Assay Kit (Pierce), according to the manufacturer's instructions. The lysate (20 g of total protein) was mixed with 6ϫ standard electrophoresis sample buffer and fractionated on an SDS-PAGE (4% stacking gel, 7.5% resolving gel) at 25 mA for 4 h. Protein samples were then electroblotted to the Trans-Blot nitrocellulose membrane (0.2 m) (Bio-Rad). The membrane was then incubated with an anti-COX-2 polyclonal antibody at 1:5,000 dilution (0.04 g/ml) (Santa Cruz Biotechnology, Santa Cruz, CA) in blocking buffer for 1 h at room temperature and then incubated similarly with horseradish peroxidase-conjugated anti-goat IgG (Santa Cruz Biotechnology, Santa Cruz, CA), diluted in blocking buffer 1:100,000, for 1 h at room temperature. The signal was detected using SuperSignal West Femto Maximum Sensitivity Substrate Chemiluminescence Kit (Pierce) according to the manufacturer's protocol and exposed to BioMax ML x-ray film (Eastman Kodak) for less than 2 min. Band intensity of protein expression was quantitated using the Molecular Analyst version 1.5 software (Bio-Rad).
Plasmid Construction-Construction of the deletion mutants containing specific regions of the human COX-2 gene promoter in the luciferase reporter vector pGL3 Basic (Promega, Madison, WI) was accomplished using PCR amplification of the desired region using the recombinant plasmid containing a 7-kb promoter region of the human COX-2 gene (kindly provided Dr. Stephen M. Prescott, University of Utah, Salt Lake City) as the template. For each PCR, primer pairs used to amplify specified regions of the COX-2 promoter region had suitable flanking restriction sites that forced cloning of the fragment in the desired orientation into the pGL3 basic vector. Orientation and se-quence of all constructs were verified by direct sequencing using the ABI PRISM 377 DNA Sequencer (Applied Biosystems, Foster City, CA).
Site-directed Mutagenesis-Mutant construct, phCOX2(KBM), with a mutation at the NF-B site was constructed as described previously (28). Briefly, the sequence was changed from GGGGACTACC to GGc-cACTACC, the lowercase nucleotides indicate the mutations. The mutations and the orientation of insert were confirmed by direct sequencing. Plasmid used in the transfection experiment was purified using an EndoFree Plasmid Isolation Kit (Qiagen, Valencia, CA), and purity was verified by spectrophotometry and agarose gel electrophoresis.
Transient Transfections and Luciferase Assays-The day before transfection, ESC were plated into 6-well tissue culture plates at a density such that the cells reached 70 -80% confluence by the time of transfection. Transfections were performed using the LipofectAMINE PLUS reagent (Invitrogen), following the protocol provided by the manufacturer. Each transfection was done using 0.4 g of firefly luciferase reporter construct DNA that contains serial deletion and site-specific mutants of COX-2 promoter gene and 1 ng of an internal control Renilla luciferase reporter plasmid pRL-TK (Promega, Madison, WI). Three hours after transfection, the transfection medium was removed by aspiration; 2 ml of DMEM/F-12 containing 10% fetal bovine serum and antibiotics was added, and the plates were returned to the incubator for 16 h. Cells received serum-free DMEM/F-12 for an additional 16 h and were then switched to control or MECM for another 24 h. Then medium was removed, and wells were rinsed with phosphate-buffered saline to remove detached cells and residual growth medium. Then 250 l of 1ϫ passive lysis buffer, provided in the Dual-Luciferase Reporter Assay System (Promega, Madison, WI), was added per well. Ten l of supernatant was used for assay of luciferase activities. Luciferase activities were determined using the LUMAT LB9507 luminometer (Berthold Technologies GmbH & Co. KG, Bad Wildbad, Germany). Firefly luciferase activities were normalized based on the Renilla luciferase activity in each well. These measurements were performed in triplicate and repeated in three independent experiments.
EMSA-ESC were cultured in 100-mm dishes until confluent in the growth medium as described above and switched to serum-free media for 16 h. These cells were then incubated with control or MECM for 16 h. Nuclear protein was extracted from whole cells using NE-PER nuclear and cytoplasmic extraction reagents (Pierce), following the protocol suggested by the manufacturer. Protein concentration was measured with a BCA Protein Assay Kit (Pierce), according to the manufacturer's instructions. Double-stranded oligonucleotide probes (see below) were end-labeled with [␥-32 P]ATP (3,000 Ci/mmol at 10 mCi/ml) using T4 polynucleotide kinase (10 units/l) (Promega, Madison, WI). Approximately 20,000 cpm of labeled probe and 0.5 g of nuclear extract were incubated for 10 min at 30°C in a reaction mix (total volume, 20 l) containing 4% (v/v) glycerol, 10 mM Tris-HCl, pH 7.5, 50 mM NaCl, 1 mM MgCl 2 , 0.5 mM dithiothreitol, and 2 g of poly(dI-dC). For supershift experiments, 2 l of antibody was added, and incubation was continued for an additional 1 h on ice. Samples were then analyzed on nondenaturing 6% acrylamide gel. Dried gels were exposed to Biomax MR x-ray film (Kodak). We used the following double-stranded probes. NF-B site wild type probe (5Ј-CAGGAGAGT-GGGGACTACCCCCTCTGCT-3Ј) was designed to represent a 28-bplong sequence (Ϫ232/Ϫ205 bp). NF-B site mutant probe (5Ј-CAGGAGAGTGGccACTACCCCCTCTGCT-3Ј) was designed to represent a 28-bp-long sequence (Ϫ232/Ϫ205 bp). The underlined nucleotides indicate the transcription factor binding domain, and the lowercase nucleotides indicate the mutations.
Statistical Analysis-Statistical analysis for comparison between treatment groups was performed by one-way analysis of variances followed by Tukey's multiple comparison test using the StatView 5.0 statistical software package (SAS Institute, Cary, NC). p Ͻ 0.05 was considered significant. All values are given as the mean, with the bars (in all figures) showing S.E.

Effect of MECM on COX-2 mRNA and Protein Levels in
Normal ESC-We initially carried out experiments to evaluate the optimal conditions for determining the effects of MECM on COX-2 mRNA levels in ESC. The time course of COX-2 mRNA abundance as examined by RT-PCR showed an increase following treatment at 4 h and peaked at 8 h (data not shown). Based on this observation, the ESC were treated with control or MECM for 8 h (Fig. 1, A and B). To determine where PCR amplification for COX-2 mRNA was in the logarithmic phase, total RNA isolated from ESC treated with MECM was reversetranscribed and was amplified under different cycle numbers. Single PCR products were obtained for COX-2. A linear relationship between PCR products and amplification cycles was observed for COX-2 treated with MECM in ESC (Fig. 1A). Consequently, 38 cycles for COX-2 were employed for quantification. Compared with the control, MECM treatment significantly increased the COX-2 mRNA level in ESC (Fig. 1B). PCR was also performed using an aliquot of the same RT products for the housekeeping gene GAPDH mRNA to control for the RT reaction, PCR efficiency, and equal starting amounts of total RNA. There was no apparent change in the GAPDH mRNA abundance with MECM treatment. It should be pointed out that COX-2 seems to be constitutively expressed in Ishikawa malignant epithelial cells in relatively high levels (Fig. 1B, 1st  lane). Quantitative densitometry for three independent experiments confirmed these results. We also performed a time course experiment for COX-2 protein levels in ESC treated with MECM employing Western analysis. COX-2 levels started to increase at 4 h, and peak level was observed at the 8-h treatment time (Fig. 1C, part of data not shown). To examine the specificity of effects of MECM on ESC, we used BECM to treat ESC and compared with MECM by immunoblot analysis (Fig. 1C). Incubation with BECM did not increase COX-2 levels demonstrating that the stimulatory effect was specific for malignant endometrial epithelial cells. These results demonstrate that malignant endometrial epithelial cells in culture produce specific factor(s), which stimulate COX-2 expression.
MECM Caused an Increase in PGE 2 Synthesis in ESC-To determine whether the induction of COX-2 mRNA and protein levels was correlated with comparable changes in PGE 2 production, PGE 2 concentrations were measured in culture media of ESC after MECM treatment. The effect of MECM on the PGE 2 synthesis in ESC is shown in Fig. 2. PGE 2 synthesis by ESC was measured after incubation in control or MECM. Incubation of ESC with MECM for 24 h caused a significant increase in the PGE 2 concentration in the medium by 4.2-fold compared with base-line PGE 2 levels in MECM itself or by 3.2-fold compared with incubation in control medium.

Activation of the COX-2 Promoter by MECM Requires a Critical Regulatory Region Containing an NF-B-binding Site-
Deletion and site-specific mutants of COX-2 promoter-driven luciferase reporter gene constructs were transiently transfected to ESC and treated with MECM (24 h). As shown in Fig.  3A, an induction in promoter activity upon MECM treatment was observed only in the reporter construct containing the COX-2 promoter region Ϫ360/Ϫ218 bp, indicating that a critical regulatory region at Ϫ360/Ϫ218 bp was responsible for this induction. Sequence analysis of this region and the literature review (5) revealed the existence of an NF-B-binding site at Ϫ222/Ϫ213 bp. Site-directed mutation of this NF-B-binding site significantly reduced MECM-induced COX-2 promoter activity in ESC (Fig. 3A). Thus, the presence of this NF-B element (Ϫ222/Ϫ213 bp) was required, at least in part, for MECM-mediated induction of COX-2 promoter in ESC. The concentration-dependent effect of MECM on COX-2 gene induction in ESC was evidenced by the transient transfection experiments using 5-and 10-fold concentrated MECM (Fig. 3B). Compared with a consistently observed 1.5-fold induction in Ϫ360/ϩ56-bp construct by 1ϫ MECM, treatments with 5 and 10ϫ concentrated MECM elicited Ͼ2-fold inductions of COX-2 promoter activity in ESC.
Identification of Protein(s) That Bind to the NF-B cis-Acting Element (Ϫ222/Ϫ213 bp)-EMSA was performed using nuclear proteins from ESC treated with control or MECM to determine the protein/DNA binding activities at the NF-B site. Nuclear extract from ESC incubated with control medium was composed of two faint but specific bands, upper and lower. Interestingly, nuclear extract prepared from MECM-treated cells showed strikingly more intense upper and lower complexes, suggesting increased protein/DNA binding activity at NF-B site upon MECM treatment (Fig. 4A). Preincubation with a cold wild type Ϫ232/Ϫ205-bp probe completely abolished both shifted bands. Conversely, the cold Ϫ232/Ϫ205-bp probe containing a mutation in NF-B (Ϫ222/Ϫ213-bp) element had no effect on the complex formation, confirming the specificity of the reaction. Furthermore, a generic probe containing a consensus NF-B element also abolished both complexes.
To determine which proteins were responsible for the formation of the two inducible nuclear complexes, supershift experiments using antibody directed against members of the NF-B/ Rel family (p65, p50, p52, RelB, and c-Rel) were performed (Fig.  4B). Only the antibody against the NF-B subunit p65 completely eliminated the formation of the lower complex. Antibodies against the p50, p52, RelB, or c-Rel, however, did not affect the formation of DNA-protein complexes. Comparable results were obtained using nuclear extracts derived from HeLa cells (data not shown). From these experiments, we concluded that p65 composed the lower complex. It should be noted that inability of any of the other members of the NF-B family to compete or supershift the upper complex indicates involvement of additional transcription factor(s) in this complex formation.
Effect of MECM on COX-2 mRNA Stability-Because mRNA stabilization has been demonstrated as a major mechanism of regulation of COX-2 gene expression, we therefore investigated this possibility in MECM-mediated induction of COX-2 gene expression in ESC. To examine the stability of COX-2 mRNA in ESC, 10 g/ml Act D was added with or without MECM (Fig.  5A). First, ESC in culture were treated with MECM for 8 h (maximum mRNA level). At this time point (control, 0 h), we distinguished four conditions. In the first two conditions, MECM was retained for another 4 h (upper portion of Fig. 5A and closed and open circles in Fig. 5B). Closed circles represent control cells with no additions. Open circles represent the addition of Act D to MECM. The COX-2 mRNA half-life value (t1 ⁄2 ) was 5.7 h for MECM plus Act D treatment. On the other hand, when MECM was removed and replaced by the DMEM/F-12 serum-free control medium, COX-2 mRNA levels declined significantly (lower portion of Fig. 5A and open and closed inverted triangles in Fig. 5B). The absence or presence of Act D did not cause a significant difference in mRNA levels between these two new control medium conditions. The t1 ⁄2 for the new control medium treated with Act D was 3.0 h, and the difference of t1 ⁄2 between the treatment of MECM plus Act D (5. suggest that MECM significantly increased COX-2 mRNA stability. (29 -31) have demonstrated up-regulation of COX-2 expression by PGE 2 via a positive feedback stimulation. Significantly, semi-quantitative RT-PCR analysis showed a relatively high expression level of COX-2 in Ishikawa malignant epithelial cells (Fig. 1B, 1st lane), suggesting PGE 2 present in MECM may contribute to the increased COX-2 expression in ESC by MECM. In order to investigate this possibility, experiments using PGE 2 -deprived MECM and exogenous PGE 2 at the same concentration present in MECM were conducted. To generate PGE 2 -deprived MECM, we followed the same procedure used to prepare MECM (see "Experimental Procedures") except that the incubation medium contained 40 M indomethacin (non-selective COX-1 and 2 inhibitor), and the indomethacin treatment was repeated every 24 h. No apparent morphological changes were noted on Ishikawa cells after indomethacin treatment (data not shown). We attempted to measure the concentration of PGE 2 in indomethacin-treated MECM by ELISA (assay sensitivity, 8 pg/ml), and no detectable amount of PGE 2 was evident (data not shown). We performed semi-quantitative RT-PCR to evaluate the effect of indomethacin on the COX-2 mRNA levels in Ishikawa cells. As shown in Fig. 6A, treatment with indomethacin did not affect COX-2 or GAPDH mRNA levels in Ishikawa cells. In addition, we reported previously (27) that treatment with indomethacin did not affect COX-2 mRNA levels in ESC. In contrast to MECM, incubation with PGE 2 -deprived MECM (MECM prepared in the presence of indomethacin) did not increase COX-2 mRNA levels in ESC (Fig. 6B). Moreover, incubation with the same concentration of PGE 2 present in MECM (40 pg/ml, as determined by ELISA, see Fig. 2) resulted in a substantial increase in the density of the COX-2 mRNA band (Fig. 6B). Taken together, these results indicate that PGE 2 present in MECM may in part be responsible for the up-regulation of the COX-2 expression in ESC.

Induction of COX-2 mRNA Expression by MECM Is Mediated by PGE 2 -Several recent studies
Recent demonstrations of the ability of PGE 2 to not only up-regulate the COX-2 expression by itself but also potentiate the interleukin (IL)-1␤-mediated COX-2 gene induction in many cell types prompted us to examine this possibility in ESC (32,33). Compared with the IL-1␤ (1 ng/ml) treatment alone, co-incubation with exogenous PGE 2 (40 pg/ml) significantly increased the COX-2 mRNA levels in ESC (Fig. 6C). The optimal concentration and time course of the IL-1␤ treatment used in the experiment were determined in a recent study (27) from our laboratory. IL-1␤ concentration in MECM was below the assay detection level by ELISA (data not shown).
To determine whether elevated PGE 2 synthesis in Ishikawa malignant endometrial epithelial cells was typical of malignant endometrial epithelial cells or idiotypic for this cell line, PGE 2 concentrations were measured in conditioned media of other malignant endometrial epithelial cell lines by ELISA. As shown in Fig. 6D, similar to the case of MECM, as compared with BECM, significantly elevated levels of PGE 2 concentration were detected in conditioned media prepared from ECC-1 and HEC-1A. Interestingly, elevated levels of PGE 2 concentration were also noted in conditioned media of malignant epithelial cells of mammary or prostatic origin. These results suggested that increased PGE 2 production might be a common property of many malignant epithelial cells. In summary, our results suggest that modulation of COX-2 expression in stromal cells by malignant epithelial cells is achieved via a combination of paracrine/autocrine factors and/or signaling and that PGE 2 is a key factor in this stimulatory mixture. DISCUSSION We showed here a potential paracrine interaction between malignant endometrial epithelial cells and adjacent stromal cells on endometrial cancer. This interaction favors increased prostaglandin synthesis in stromal cells and involves COX-2 and NF-B. Interestingly, PGE 2 can autoregulate its own synthesis through a positive feedback loop in endometrial cancer. Evidence for expression of all four PGE 2 receptors (EP 1 , EP 2 , EP 3 , and EP 4 ) in the endometrium (including stroma) further supports this possibility (34 -36). 3 A self-amplifying loop, based on increased PGE 2 production leading to increased COX-2 expression and concomitant PGE 2 production by ESC surrounding malignant epithelial cells, may be critical for the pathophysiology of the endometrial cancer growth. Recently, enhanced PGE 2 synthesis has been shown to promote cell growth in some cancer models (1,10). PGE 2 can cause decreased programmed cell death in HCA-7 human colonic cancer cells and increased growth and motility of the human colorectal carcinoma cell line, LS-174 (37,38). In this study, we also determined that PGE 2 production was commonly elevated in malignant epithelial cells irrespective of tissue origin and might contribute to the pathophysiology of tumor growth.
We have shown previously (27) that ESC express COX-2 in response to IL-1␤ stimulation and synthesize PGE 2 . The identification of the stimulatory effect of PGE 2 on IL-1␤-dependent COX-2 expression in ESC is likely to be of physiologic relevance. In fact, other investigators (6 -8) showed that cytokines such as tumor necrosis factor-␣ and IL-1␤ increased the binding activity of NF-B to the COX-2 promoter and up-regulated its activity in other systems. Interestingly, it was also proposed that NF-B is important for oncogenic transformation, at least partly through its ability to block apoptosis (39 -41). Because apoptosis is the primary mechanism of tumor cell killing by radiation and by chemotherapy, the speculation that the activation of NF-B suppresses apoptotic potential generated interest in the role of NF-B in cancer therapies. Indeed, suppression of NF-B activation significantly enhances cell killing in culture in response to these treatments (42).
NF-B is a dimeric DNA-binding protein composed of members of the NF-B/Rel family of proteins including the mammalian forms, p65, p50, p52, RelB, and c-Rel (43,44). NF-B proteins are capable of forming numerous homodimers and heterodimers with other family members, and this adds another level of complexity to the interaction of NF-B with specific target genes. In this study, we found that NF-B p65 subunit binds to the Ϫ222/Ϫ213-bp element in the CDX-2 promoter in response to MECM treatment. We continue to search for other partners in these DNA-protein complexes observed by EMSA. Supershift experiments showed that p50, p52, RelB, or c-Rel were not part of these complexes. Other investigators (45,46) showed that members of NF-B/Rel family interacted with other proteins, particularly members of the C/EBP family. In order to investigate the possibility, antibodies to various C/EBP proteins were tested in EMSAs. None of the antibodies against C/EBP␣, C/EBP␤, and CEBP␦ had any effect on the inducible complex formation (data not shown).
There are two NF-B consensus sites in the promoter region of the human COX-2 gene (47): the NF-B-5Ј site (Ϫ447 to Ϫ438) and the NF-B-3Ј site (Ϫ222 to Ϫ213). NF-B-5Ј has been shown to have a role in the mechanism of COX-2 induction by tumor necrosis factor-␣ in a murine osteoblast cell line (6). NF-B-3Ј may play a role in facilitating the induction of COX-2 by lipopolysaccharide and phorbol ester in concert with the nuclear factor-interleukin-6 expression site and a cAMPresponse element site in bovine aortic endothelial cells (12). We discovered that the NF-B-3Ј is necessary for MECM-mediated COX-2 transcription. In addition, because the reporter construct containing the COX-2 promoter region Ϫ828/ϩ56 was unresponsive to MECM, it appears that for the optimal induction of COX-2 by MECM, inhibitory site(s) are included between the Ϫ828 and Ϫ360-bp region of COX-2 promoter.
It now seems clear from published evidence (47-50) that the COX-2 gene is regulated through both 5Ј (transcriptional) and 3Ј (post-transcriptional) regulatory elements. We identified the critical cis-acting element, i.e. the NF-B site in the COX-2 gene promoter required for the MECM-mediated COX-2 transcriptional increase. However, MECM did not affect the transcription of COX-2 reporter constructs so much. Early studies by Raz et al. (51) demonstrated that inducible COX-2 synthesis could be divided into early transcriptional and late post-transcriptional phases. High levels of encoded protein products from COX-2 genes are usually required for only a short period and must be expressed in a burst (50). The entire 3Ј-untranslated region (2.5 kb) of the human COX-2 gene is encoded by exon 10, which contains three canonical (AAUAAA) polyadenylation sequences and 22 copies of AUUUA " . EMSA to identify the proteins that bind NF-B. A, competitive EMSAs were performed with an NF-B-radiolabeled probe and nuclear extract from control (CON) or MECM-treated ESC. A 100-fold molar excess of unlabeled wild type Ϫ232/Ϫ205-bp probe, mutant Ϫ232/ Ϫ205-bp probe, or consensus NF-B probe was used in the competition reactions. Nuclear extract from ESC incubated with control medium was composed of two faint but specific bands, upper and lower. Interestingly, nuclear extract prepared from MECM-treated cells showed strikingly more intense upper and lower complexes, suggesting increased protein/DNA binding activity at NF-B site upon MECM treatment. Preincubation with a cold wild type Ϫ232/Ϫ205-bp probe or consensus NF-B probe completely abolished both shifted bands. On the other hand, the cold Ϫ232/Ϫ205-bp probe containing a mutation in NF-B element had no effect on the complex formation, confirming the specificity of the reaction. B, supershift experiments were performed with an NF-B radiolabeled probe and nuclear extract from MECMtreated ESC and antibodies against transcription factors that should bind to the NF-B site. Only the antibody against the NF-B subunit p65 completely eliminated the formation of the lower complex. Antibodies against the p50, p52, RelB, or c-Rel, on the other hand, did not affect the formation of DNA-protein complexes. Comparable results were obtained using nuclear extracts derived from HeLa cells (data not shown). These results were reproduced in three other experiments.  A and inverted triangles in B). The t1 ⁄2 for the new control medium treated with Act D was 3.0 h. Relative levels of COX-2 mRNA expression were determined by densitometric scanning of the bands and normalized to the GAPDH signal. Values were depicted for mRNA abundance expressed as a percentage at control time point 0 h. Results are expressed as the mean Ϯ S.E. of three independent experiments. The difference of t1 ⁄2 between the treatment of MECM plus Act D (5.7 h) and the treatment of new control medium plus Act D (3.0 h) was significant (p Ͻ 0.05).
Kamen" motifs (47)(48)(49)(50). The latter sequences are believed to be associated with message instability, translational efficiency, and rapid turnover (52)(53)(54). Because COX-2 mRNA is highly unstable, and because MECM stabilizes COX-2 mRNA in the absence of transcription, we suggest that post-transcriptional mRNA stability is an important consequence of MECM action as well as the transcription step.
Thus, the three key molecules, PGE 2 , COX-2, and NF-B, are closely linked together with the common thread of oncogenesis. This observation promotes new insights into the paracrine interactions in cancer development and may lead to new therapeutic strategies capable of interrupting the oncogenetic cascade at key points.