Polyoma Enhancer Activator 3, an Ets Transcription Factor, Mediates the Induction of Cyclooxygenase-2 by Nitric Oxide in Colorectal Cancer Cells*

Abundant evidence supports the role of cyclooxygen-ase-2 (COX-2) in colorectal cancer. Nitric oxide (NO), a pro-inflammatory signaling factor, may regulate COX-2 expression and activity thereby linking hyper-inflam-matory states to cancer susceptibility. Previously we showed that NO induced COX-2 expression. Although NO also activated the (cid:1) -catenin (cid:1) T-cell factor/lymphocyte enhancing factor transcriptional pathway, a direct causal link between this pathway and COX-2 expression was not demonstrated. In this current study, we focused on NO-induced transcriptional activity and elucidated its role in COX-2 expression. NO donors stimulated the expression of peroxisome proliferator-activated receptor- (cid:2) and c- myc , both downstream genes of (cid:1) -catenin. They also induced the expression of polyoma enhancer activator 3 (PEA3) and increased its DNA-binding activity. To establish a role for PEA3 to (cid:1) -catenin-induced COX-2, we transfected RKO cells with (cid:1) -catenin and found that (cid:1) -catenin increased PEA3 expression. Also, there was higher PEA3 in immortal mouse colon epithe-lium cells ( Apc Min/ (cid:3) ) compared with the nucleus, forms the complexes with TCF/LEF, and activates the (cid:1) -catenin (cid:1) TCF/LEF signaling pathway. (cid:1) -Catenin (cid:1) TCF/LEF activates PEA3 transcription factor, which stimulates the COX-2 activity. Moreover, NO, generated from iNOS transfection or from outside NO donors, also stimulates the interaction between PEA3 and transcription coactivator CBP/p300, which facilitates the induction of COX-2. We suggest that subscribers photocopy these corrections and insert the photocopies at the appropriate places where the article to be corrected originally appeared. Authors are urged to introduce these corrections into any reprints they distribute. Secondary (abstract) services are urged to carry notice of these corrections as prominently as they carried the original abstracts.

Multiple lines of evidence suggest that prostaglandins, metabolites of arachidonic acid, play an important role in colon cancer development. Elevated prostaglandin levels are found in colon cancers and in their precursor lesions, adenomatous polyps. It is now understood that at least two forms of the enzyme, COX-1 and COX-2, 1 are responsible for the metabolism of arachidonic acid. COX-2 is inducible and increased in inflammatory states as well as in numerous cancers and their associated pre-malignant lesions. It may be a critical step in colon carcinogenesis in that inhibitors of COX-2 can block carcinogenesis (1,2) NO is a pleiotropic regulator, critical to numerous biological processes, including vasodilatation, neurotransmission, and macrophage-mediated immunity. It also modulates the etiology and phenotype of cancer cells (3)(4)(5). Previous work from our group as well as others established a connection between NO and the induction of COX-2 (4,6). NO is produced from Larginine by nitric-oxide synthases (NOSs). NO treatment activates matrix metalloproteinases, which cause the degradation of E-cadherin, the relocalization of ␤-catenin to cytosol and nucleus, and the formation of ␤-catenin⅐TCF/LEF complexes. The ␤-catenin⅐TCF/LEF transcription complex, probably indirectly, stimulates COX-2 expression (6,7). Because other investigations showed that ␤-catenin may transactivate the expression of PEA3, a transcription factor of the Ets family (8), we considered whether ␤-catenin⅐TCF/LEF may stimulate COX-2 through activation of PEA-3. Thus, activation of COX-2 by NO may be by a sequential pathway: NO 3 ␤-catenin⅐TCF/ LEF 3 PEA3 3 COX-2.
PEA3 belongs to the Ets family of transcription factors. The signature of the Ets family is the ETS domain, a region of ϳ85 amino acids, which has been widely conserved during evolution. Most of the Ets family members bind to the core sequence 5Ј-GGA(A/T)-3Ј. A wide series of promoters have been characterized as containing active ETS binding sites (9,10). The specificity of each member of the Ets family to the regulation of a specific gene has not yet been established. It has been demonstrated that PEA3 is able to transactivate the promoters of collagenase types I and IV as well as stromelysin type I (11). Recently, a specific glutathione peroxidase gene gpx5 has been characterized as a putative target gene of the PEA3 DNAbinding protein in the mouse epididymis (12).
In this study, we employed conditionally immortal colonic cell lines. One is designated YAMC cell (Apc ϩ/ϩ ) derived from a SV40LT antigen parental mouse; IMCE cells (Apc Min/ϩ ) are derived from the F1 hybrids resulting from the mating of Apc Min/ϩ and SV40LT antigen transgenic mice (13,14). This pair of non-transformed murine colon epithelial cell lines of similar genetic background includes one that carries the Apc-Min/ϩ mutation and consequently expresses higher ␤-catenin levels. Because both YAMC and IMCE cells express the heatlabile SV40LT antigen that allows them to proliferate at 33°C, they revert to a non-transformed phenotype at the restrictive temperature of 39°C at which the proliferation of these cells ceases. RKO is a human colorectal cancer cell line with both wild type APC and wild type ␤-catenin. From studies in these cell lines, we found that NO treatment stimulated PEA3 expression that might be mediated by ␤-catenin⅐TCF/LEF signaling. PEA3 and some other transcription factors induced COX-2 promoter activity. One interesting finding was that NO, either added exogenously or produced endogenously by transfected iNOS, stimulated the interaction between PEA3 and transcription co-activator CBP/p300 in the induction of COX-2. These findings indicate the unique role of PEA3 in the regulation of COX-2. These data support our hypothesis that PEA3 mediates NO-induced COX-2.

EXPERIMENTAL PROCEDURES
Reagents and Tissue Culture-NO donors, SIN-1, NOR-1, SNAP, and iNOS inhibitor SMT were purchased from Calbiochem (San Diego, CA). Experiments were carried out using the conditionally immortalized murine colonic epithelial cell lines YAMC and IMCE (13,14). Both YAMC and IMCE cells express the heat-labile SV40 large T antigen with an IFN-␥-inducible promoter. The temperature-sensitive SV40 large T antigen with IFN-␥-inducible promoter is active only at 33°C. It becomes inactivated and non-functional once cells are transferred to 39°C. All cells were grown on a 75-cm 2 tissue culture flask coated with type I collagen (5 g/cm 2 ) in RPMI 1640 media supplemented with 5% neonatal calf serum, ITSϩ (insulin, 6.25 g/ml; transferrin, 6.25 g/ml; selenious acid, 6.25 ng/ml; linoleic acid, 5.35 mg/ml; and bovine serum albumin, 1.25 mg/ml), 5 IU/ml murine IFN-␥, 100,000 IU/liter penicillin, and 100 mg/liter streptomycin. They were cultured under transforming (permissive) conditions in a 33°C incubator with 5% CO 2 plus all the aforementioned supplements in the media. RKO colorectal cancer cells were kindly supplied by Dr. M. G. Brattain, Roswell Park Cancer Institute, Buffalo, NY, and maintained in Dulbecco's modified Eagle's medium with 10% of fetal bovine serum.
DNA and Plasmid Constructs-The following cDNA constructs have been used in this study. The PEA3 cDNA was kindly provided by Dr. J. A. Hassell of McMaster University, Canada. The Cox-2-Luc was kindly supplied by Dr. Lee-Ho Wang of the University of Texas-Houston Medical School. The ␤-catenin-S37A expression vector was kindly provided by Dr. Stephen Byers of Georgetown University, Washington, D. C. The pcDNA/TCF-4 was kindly provided by Dr. Bert Vogelstein of The Johns Hopkins University. The C/EBP-␣, -␤, and -␦ cDNAs were kindly provided by Dr. Peter Johnson of NCI, National Institutes of Health. The pcDNA3/iNOS was kindly provided by Dr. David Geller of the University of Pittsburgh. The p300 expression vector was from Upstate Biotech, Inc. The series of COX-2 luciferase reporter constructs in which deletions or site-specific mutations had been introduced, Ϫ1432/ϩ59, Ϫ327/ϩ59, Ϫ124/ϩ59, Ϫ52/ϩ59, KBM (NF-B, Ϫ223/ Ϫ214), ILM (NF-IL6, Ϫ132/Ϫ124), and CRM (CRE, Ϫ59/Ϫ53) were kindly provided by Dr. Tadashi Tanabe of the National Cardiovascular Center Research Institute, Osaka, Japan (21). Using a QuikChange II site-directed mutagenesis kit (Stratagene), 4 ETS binding sites (Ϫ75/ Ϫ72, Ϫ104/Ϫ101, Ϫ189/Ϫ186, and Ϫ287/Ϫ284) in the COX-2 promoter were mutagenized, respectively, where Ϫ327/ϩ59 construct was used as backbone. They were designated as ETSM-1, -2, -3, and -4.
Luciferase Assay-COX-2 transcriptional activity in RKO cells was measured using the Dual Luciferase TM reporter assay (Promega, Madison, WI) according to the manufacturer's protocol. The cells were co-transfected with the Cox-2-Luc reporter construct, or different truncation or mutation constructs, and pRL-TK, a Renilla construct for normalizing of transfection efficiency. To determine the effects of transcription factors on COX-2 activity, equivalent amounts of cDNA of various transcription factors, p300, or iNOS, or vector plasmid were transfected using LipofectAMINE 2000 (Invitrogen). Some cells were treated with 50 M of SIN-1 overnight. Transfected cells were lysed 36 h after transfection, and luciferase activity was measured with equal amounts of cell extract using a TD-20/20 luminometer (Turner Designs, Sunnyvale, CA) and normalized with the Renilla activity.

NO Treatment Increased Expression of ␤-Catenin⅐TCF/LEF Downstream Genes and Stimulated the Expression and DNA
Binding of PEA3-Our previously published studies showed that, in YAMC cells, NO treatment stimulated the expression of COX-2 and increased the production and activity of prostaglandin E 2 , a product of COX-2. Furthermore, NO induced the relocation of ␤-catenin from membrane to the cytosol and nucleus and stimulated the ␤-catenin⅐LEF-1 DNA complex formation. But a direct NO induction of COX-2 activity by ␤-catenin⅐TCF/LEF complex has not been shown (6).
To confirm that NO stimulated the ␤-catenin⅐TCF/LEF transactivation activity, we treated cells with NO donors and performed Western blot analysis on the expression for PPAR-␦ and c-myc, both downstream genes of ␤-catenin⅐TCF/LEF. The YAMC cells were treated with two different concentrations of SIN-1 (5.0 and 50 M), SNAP (10 and 100 M), and NOR-1 (5.0 and 50 M), respectively. NO treatment dramatically increased the expression of PPAR-␦ and c-myc (Fig. 1A), which is consistent with our previous interpretation that NO stimulated the ␤-catenin⅐TCF/LEF transactivation activity.
Previously published work by others suggested that PEA3 regulated COX-2 (8,16) which prompted us to consider that PEA3 might mediate NO-induced COX-2 in our model system. We first showed that NO stimulated PEA3 expression. Performing Western blot analysis for PEA3, we found that SIN-1 (5.0 and 50 M), SNAP (10 and 100 M), and NOR-1 (5.0 and 50 M) all induced the expression of PEA3 in YAMC cells (Fig. 1A).

PEA3 Mediates COX-2 Induction by Nitric Oxide
To further explore the effect of NO on PEA3 activity, we performed electrophoretic mobility shift assays, which showed that the NO donor SIN-1 increased the DNA-binding activity of PEA3 in both YAMC and IMCE cells (Fig. 1B).
The Status of ␤-Catenin Affected the Expression of PEA3-These initial studies showed that NO stimulated the activity of both PEA3 and ␤-catenin⅐TCF/LEF. Whereas others have suggested PEA3 stimulation of COX-2, no direct stimulation of COX-2 expression by ␤-catenin⅐TCF/LEF has been reported. Therefore, we considered that PEA3 might be a downstream target gene of ␤-catenin⅐TCF/LEF. To demonstrate this, we transfected RKO cells with a vector expressing an activated form of ␤-catenin, ␤-catenin-S37A, and obtained stable transfectants. Western blot analysis showed that ␤-catenin-S37A markedly increased the level of PEA3 as compared with vectortransfected control cells ( Fig. 2A). We also tested this hypothesis in YAMC (Apc ϩ/ϩ ) and IMCE (Apc Min/ϩ ) cells, which differ only in their Apc genotype. In IMCE cells, mutation in Apc leads to higher levels of cytosolic and nuclear ␤-catenin. There-fore, we compared the expression levels of PEA3 in these two cell lines. Western blot analysis showed that IMCE cells contained higher levels of cytosolic ␤-catenin when compared with its counterpart YAMC cells (Fig. 2B). Consistent with this finding, IMCE cells also had higher levels of c-myc, a target gene of ␤-catenin⅐TCF/LEF (Fig. 2B). Importantly, PEA3 levels were present at higher levels in IMCE cells (Fig. 2B), presumably due to higher ␤-catenin⅐TCF/LEF activity and supporting the interpretation that ␤-catenin⅐TCF/LEF regulates the expression of PEA3.
COX-2 Promoter Activity Was Regulated by PEA3, CBP/ p300, and Other Transcription Factors-The aforementioned findings led us to propose that PEA3 mediated NO induction of COX-2 activity. We performed studies assaying promoter activity in cells transfected with a COX-2-Luc construct to directly address this hypothesis. To determine whether PEA3 or various transcription factors and co-activators, alone or in various specified combinations, stimulated COX-2 promoter activity, we co-transfected vectors expressing c-Jun, ␤-catenin-S37A, TCF-4, cEBP-␣, -␤, -␦, and p300. We included these transcription factors, because they were reported either to reg- . Equal amounts of cell lysates were electrophoresed on acrylamide-denaturing gels and transferred by electroblotting onto nitrocellulose membranes. Western blots were performed using anti-PEA3 (Santa Cruz Biotechnology, 1:100), anti-PPAR-␦ (Santa Cruz Biotechnology, 1:100), and anti-c-myc (Chemicon, 1:200). Western blot to actin (Sigma, 1:800) was used as loading control. Blots were developed using the enhanced chemiluminescence kit (Amersham Biosciences, Piscataway, NJ). B, DNA⅐PEA3 complex formation induced by SIN-1. YAMC and IMCE cells were treated with or without 10 M SIN-1 for 4 h. An electrophoretic mobility shift assay was performed using double-stranded PEA3 oligonucleotide 5Ј-GGCTAGTGAGGCAG-GAAGTAGGGAGAG-3Ј. 100ϫ unlabeled PEA3 probe was used as a competitor and effectively competed away the DNA⅐PEA3 complex. ulate COX-2 activity in a cell system, and/or they were found to have binding sites in the COX-2 promoter (Fig. 3A) (8, 16 -20). Each set of transfectants was treated with or without the NO donor, SIN-1, to examine NO modulation of one or more specific transcription factors alone or in combination. As shown in Fig.  3B, PEA3 acting alone was the most potent among those tested with a COX-2 promoter activity increase of 7-fold over control. A number of other transcription factors, including c-Jun, cEBP-␣, cEBP-␦, and p300 increased COX-2 promoter activity 1-to 2-fold, whereas ␤-catenin and TCF-4 had little effect. In fact, cEBP-␤ showed significant inhibition of COX-2 promoter activity. The pattern of promoter activity seen with combinations of transcriptional factors was of special interest. As a co-activator, p300 significantly increased the stimulatory activity of c-Jun (ϳ4-fold) and modestly increased the already high activity level of PEA3. The synergistic interaction of PEA3 with c-Jun was dramatic. In combination these two stimulated COX-2 activity 17-fold as compared with the effect of PEA3 and c-Jun at ϳ7-fold and ϳ2-fold, respectively, when transfected alone. A modest synergism was seen between c-Jun and ␤-catenin. In contrast to the increased activity of p300 with PEA3, the cEBP species appeared to decrease the stimulatory effect of PEA3.
With our focus on the mediator of the NO effect on COX-2 expression, it was striking that NO modulated the promoter activity of only one set of transcriptional factors among those that we studied. With the use of SIN-1 as NO donor, we found an increase in stimulation of promoter activity only with the PEA3/p300 combination. Treatment with SIN-1 increased this activity almost 2-fold to the highest level seen in these studies.
At the concentration used, SIN-1 had no inhibitory effect on any of the COX-2 promoter activities.
Endogenous NO Also Stimulated PEA3/p300-induced COX-2 Activity-NO, a potent second messenger and regulator, is generated from L-arginine by several nitric-oxide synthases (NOSs). The expression of the inducible isoform (iNOS) is stimulated by cytokines and/or bacterial products in a wide range of cell types and tissues. Additionally, tumor cells may express high levels constitutively (3). Although our studies showed marked effects of exogenous NO from NO donors on the transcriptional activity of PEA3/p300 (Fig. 3B), the demonstration that endogenous NO generated from iNOS functioned like ex- ogenous NO would greatly strengthen the physiologic relevance of this finding. We, therefore, co-transfected RKO cells with a human iNOS cDNA expression construct as well as with PEA3/p300. The luciferase reporter assay showed that endogenous NO generated from transfected iNOS-stimulated COX-2 promoter activity over that of PEA3/p300 (Fig. 4). To support the interpretation that this effect was caused by NO produced by iNOS, we treated these cells with S-methylisothiourea sulfate (SMT), a specific inhibitor of iNOS catalytic activity. The results showed that SMT blocked the effect of iNOS in augmenting PEA3/p300-induced COX-2 promoter activity (Fig. 4). These studies clearly showed that endogenously generated NO as well as exogenous NO from NO donors stimulated PEA3/ p300-induced COX-2 promoter activity.
NF-IL6 and ETS-1 Sites Were Responsible for COX-2 Induction by PEA3 and PEA3/p300 Stimulated by NO-To determine the regulatory elements in the COX-2 promoter that were responsible for NO stimulation of PEA3/p300-induced COX-2 activity, a series of promoter reporter constructs were used in which deletions had been introduced (21). As shown in Fig. 5A, the NO-dependent PEA3/p300-induced, as well as NO-independent PEA3-and PEA3/p300-induced, promoter activity was totally abolished when using Ϫ124/ϩ59 or Ϫ52/ϩ59, whereas Ϫ327/ϩ59 retained most of its activity, suggesting that the sequences critical for the promoter induction probably lie between base pairs Ϫ327 and Ϫ125. Then we employed constructs with site-specific mutation to the transcription factor binding sites that locate between Ϫ327 and Ϫ125, including KBM, CRM, ILM, ETSM-3, and -4 (Fig. 3A). Because the ETS-1 site has been reported important for PEA3-induced COX-2 activity (20), ETSM-1 and -2 also were used. All these reporter constructs had Ϫ327/ϩ59 as their backbone. Although the promoter activity was retained using KBM, CRM, and ETSM-2, -3, -4, it was abrogated using ILM and ETSM-1 ( Fig. 5B; data for ETSM-3 and -4 not shown). These data agreed with the results from Subbaramaiah et al. (20) and Howe et al. (8), who suggested that ETS-1 and NF-IL6 sites, respectively, are important for PEA3-induced COX-2. The apparently contradictory observations that mutation of ETS-1 reduces COX-2 promoter transcription even though the native ETS-1 is inactive in the Ϫ124/54 promoter construct suggest that cooperation with other transcription factors that bind upstream of position Ϫ124 is necessary for ETS-1 activity. Thus, the data strongly suggested that NF-IL6 and ETS-1 sites are critical for COX-2 activity by PEA3, as well as PEA3/p300 stimulated by NO.
PEA3 Physically Interacted with p300 -As transcription coactivators, CBP/p300 binds to numerous transcription factors and nuclear receptors to modulate their activity. The binding of CBP/p300 to Ets-1 and Ets-2, both members of the Ets family, has been described (21). In our current studies, the finding that both exogenous and endogenous NO stimulated PEA3/p300 mediated COX-2 promoter activity suggested to us that PEA3 and p300 might bind to form a complex. The stimulation of their interaction by NO became an intriguing possibility. To test this idea, we performed co-immunoprecipitation studies using RKO cells transiently transfected with pcDNA-, PEA3-, p300-, or PEA3/p300-expressing vectors. After immunoprecipitation with an anti-p300 antibody, a Western blot to PEA3 was carried out. The results clearly demonstrated the physical binding between PEA3 and p300 (Fig. 6). DISCUSSION Two different cyclooxygenases are functional in mammals: COX-1 and COX-2. COX-2 is mainly an inducible isoform that shares significant features with iNOS in tissue distribution, regulatory function, and participation in pathophysiological phenomena. NO, the product of iNOS catalysis, is an important regulator of COX-2 expression and activity. There are various pathways proposed for NO-induced regulation of COX-2 expression and for modulation of cancer development. Among these, modulation of transcription factors such as nuclear factor-B and AP-1 appears important (4). We and others have FIG. 5. Mapping the regulatory elements in the Cox-2 promoter responsible for PEA3/p300 regulation in response to NO. RKO cells were transfected with equivalent amounts of cDNA of PEA3, p300, or vector plasmid as control. All cells were transfected with pRL-TK. A, the cells were also transfected with Ϫ1432/ϩ59 or with the truncated promoter constructs (Ϫ327/ϩ59, Ϫ124/ϩ59, and Ϫ52/ϩ59). B, the cells were also transfected with Ϫ327/ϩ59 or various mutation constructs (ILM, CRM, KBM, ETSM-1, and -2; data not shown for ETSM-3 and -4). Some cells were treated with 50 M SIN-1 overnight. 36 h after transfection, the cell lysates were harvested, and the COX-2 luciferase activity was determined using the Dual Luciferase kit (Promega). The results are expressed as mean Ϯ S.E. for three separate experiments. Pϩp, PEA3 plus p300.
FIG. 6. Physical interaction between PEA3 and CBP/p300. RKO cells were transiently transfected by pcDNA, PEA3, p300, or PEA3/ p300, respectively. 800 g of each sample was first pre-cleaned with Protein G-agarose beads and then immunoprecipitated with 2 g of anti-p300 polyclonal antibody (Santa Cruz Biotechnology). The protein was separated in SDS-PAGE. A Western blot to PEA3 was performed.
proposed that ␤-catenin⅐TCF/LEF signaling also plays a critical role in NO induction of COX-2, especially in colon epithelial cells (6,22). The data presented here are consistent with our previous results and extended our findings by demonstrating that PEA3 is a critical mediator of ␤-catenin⅐TCF/LEF signaling in NO-induced COX-2 in colorectal cancer. Our previous studies showed that NO treatment caused the activation of matrix metalloproteinases, which led to the degradation of E-cadherin and its subsequent dissociation from ␤-catenin, thereby contributing to the cytosolic accumulation of ␤-catenin and nuclear formation of ␤-catenin⅐TCF/LEF complex. Combining these data, we proposed a model that elucidates the mechanism of induction of COX-2 by NO (Fig. 7).
PEA3 expression is very weak in normal human tissue, although it can be activated by growth factors and stress-activated signaling pathways through the Ras-Raf-MAPK cascade (23). The Ets family of transcription factors plays key regulatory roles in development. Recent studies also suggest that it is important for oncogenesis. For example, Ewing's sarcoma is commonly associated with translocations that juxtapose a segment of the EWS gene to one of five different Ets genes, including PEA3 (24,25). PEA3 is overexpressed in tumors from human breast, ovary, and stomach (26,27). In a transgenic mouse model, dominant-negative PEA3 dramatically delayed the onset and reduced the number and size of mammary tumors in murine mammary tumor virus-neu mice (28). Moreover, it has been shown that human PEA3/E1AF confers an invasive phenotype on MCF-7 cells (29). Particularly, PEA3 is expressed frequently in mouse intestinal tumors and in every human colon tumor cell line examined (30).
The importance of the ␤-catenin/wnt pathway in the early stage of colorectal cancer development has been emphasized (31,32). In most cases, the accumulation of ␤-catenin comes from the mutations of the Apc gene or the ␤-catenin gene, itself. Both prevent the formation of ␤-catenin⅐APC⅐GSK-3␤⅐Axin complex, which phosphorylates and targets ␤-catenin for degradation by the ubiquitin-proteasome system (33,34). On the other hand, the accumulation of ␤-catenin may be augmented by NO stimulation in colorectal cells. NO is generated from high expression levels of iNOS, which is a common phenomenon in chronic inflammatory diseases and adenomatous polyps, both precursor lesions for colorectal cancer, or produced by immune cells in hyper-inflammatory states (35)(36)(37)(38)(39). Here we demonstrated that NO treatment not only caused relocation of ␤-catenin to cytosol and its nuclear DNA complex formation, but also it stimulated the expression of c-myc and PPAR-␦, both downstream genes of ␤-catenin⅐TCF/LEF signaling, indicating the activation of this pathway by NO.
Even if NO causes ␤-catenin relocation to cytosol, the nuclear formation of DNA⅐␤-catenin complex, and the activation of ␤-catenin⅐TCF/LEF, no evidence suggests the direct induction of COX-2 expression and activity. It is likely that a mediator links ␤-catenin⅐TCF/LEF with COX-2. Others have suggested that PEA3 is induced by ␤-catenin (8). Thus, we considered that this could also be important in our system. To test this, we performed Western blots for PEA3 for NO-treated YAMC cells. Surprisingly, all three NO donors used, SIN-1, NOR-1, and SNAP, increased the expression of PEA3.
Furthermore, NO treatment also stimulated the DNA-binding activity of PEA3. Consistent with the work by Howe et al. (8), we confirmed that ␤-catenin increased the expression of PEA3 in RKO cells. Corroborating these findings, we showed that IMCE cells had higher levels of both PEA3 and ␤-catenin than YAMC cells. From the sum of these findings, we concluded that NO stimulated PEA3 expression and DNA binding, which was mediated by ␤-catenin signaling. Consistent with this role of ␤-catenin, pea3 proximal promoter contains two putative TCF-1 CCTTTG binding sites, localized at Ϫ294/Ϫ289 and Ϫ287/Ϫ282. Taken in total, PEA3 appears to be a downstream gene of the ␤-catenin pathway (8).
There are binding sites for various transcription factors in the COX-2 promoter, and some of these transcription factors may induce COX-2 activity in different tissues or cells (8, 16 -20). In this study, in contrast to the weak effect on COX-2 activity by most transcription factors, the stimulation of PEA3 was potent. NO induced ␤-catenin signaling, but it did not induce COX-2 promoter activity on it own (Fig. 4). This is consistent with the idea that COX-2 probably is not a direct target of ␤-catenin signaling (8), even though TCF binding elements have been recently identified in COX-2 promoter (8,17).
A variety of genes are regulated by closely spaced PEA3-Ets and AP-1 sequences, and the expression of these genes is associated with tumor progression (40,41). There is a canonical AP-1 site (Ϫ59/Ϫ53) and several ETS sites in the COX-2 promoter. The ETS site (Ϫ75/Ϫ72) close to the AP-1 binding site has been identified as a functional one, and the juxtaposed Ets/AP-1 sites are necessary for Her-2/neu-mediated induction of COX-2 in breast cancer cells (20). PEA3 also synergized with ␤-catenin and c-Jun to transactivate the matrilysin promoter (30). In line with this idea, our results indicate that the synergy between PEA3 and AP-1 is important for COX-2 induction and perhaps also for the progression of some colorectal tumors. Additionally, p300 increased induction by both AP-1 and PEA3 (Fig. 3). It has been suggested that the amount of p300/CBP in cells may be rate-limiting and that different transcription factors may compete for rate-limiting amounts of these coactivators and thus provide mechanisms for cross-talk in the regulation of gene expression (42,43). We also found that p300 greatly enhanced AP-1/PEA3 induction of COX-2 promoter activity (data not shown), suggesting that p300 may serve as a bridge between AP-1, PEA3, and basic transcription machinery to modulate COX-2 transcription.
We were surprised that, of all the transcription factors tested singly or in combination, only the PEA3/p300 combination was affected by nitric oxide. Strikingly, the COX-2 promoter activity induced by PEA3/p300 was increased nearly 2-fold by cotreatment with the NO donor, SIN-1. Although others have FIG. 7. Proposed mechanism by which NO stimulates the expression and activity of COX-2. NO treatment stimulates the activity of matrix metalloproteinases (MMPs), which in turn causes the degradation of E-cadherin and the subsequent dissociation of ␤-catenin. ␤-Catenin accumulates in the cytoplasm, relocates to the nucleus, forms the complexes with TCF/LEF, and activates the ␤-catenin⅐TCF/LEF signaling pathway. ␤-Catenin⅐TCF/LEF activates PEA3 transcription factor, which stimulates the COX-2 activity. Moreover, NO, generated from iNOS transfection or from outside NO donors, also stimulates the interaction between PEA3 and transcription coactivator CBP/p300, which facilitates the induction of COX-2.
suggested that a number of transcription factors are activated by either NO-mediated cyclic GMP signaling or by peroxynitrite, few studies showing activation of promoter activity have been reported. Certainly, in our studies with the COX-2 promoter, NO is not a promiscuous activator. Instead, its activation is restricted to the combination of PEA3/p300. Importantly, not only is this activation by NO seen with exogenous NO donors, but also with endogenously produced NO from transfected iNOS, and the effect on promoter activity was abolished by a specific inhibitor of iNOS catalytic activity. Thus, NO stimulates COX-2 expression at two sites. Not only does NO increase PEA3 expression through ␤-catenin signaling, but also it directly augments the COX-2 promoter activity of the PEA3/p300 combination. Consistent with published data (8), we found that one ETS site (ETS-1, Ϫ75/Ϫ72) and the NF-IL6 site (Ϫ132/Ϫ124) are both critical for induction of COX-2 promoter activity by PEA3 in this cell model. When induction of PEA3 was abolished, the synergism between PEA3 and p300 was abolished, and NO no longer stimulated PEA/p300-induced COX-2 activity. Thus both sites are responsible for NO stimulation of PEA3/p300induced COX-2 activity. When mapping the COX-2 promoter, we first used truncation mutations and found that the range between Ϫ327 and Ϫ125 was critical for the induction by NO and PEA3 (Fig. 5A). But subsequently we found the ETS-1 site (Ϫ75/Ϫ72), which is not located within the aforementioned sequence range (Fig. 5B) was also important. Upon reviewing the data and the constructs, we considered that the difference between the backbones of the constructs, Ϫ124/ϩ59, and ETSM-1 could account for the apparent inconsistency. Although the former was truncated to contain only Ϫ124/ϩ59, the latter had a backbone (Ϫ327/ϩ59) that contained an additional 203 bp. When Ϫ124/59 was used, the ETS-1 site was shown to be non-contributory. By contrast, when Ϫ327/ϩ59 was used, the same ETS-1 site was critical. This indicated to us that the induction of COX-2 by PEA3 through the ETS-1 site needs the sequence between Ϫ327 and Ϫ125, i.e. this induction requires collaboration with one or more other transcription factors whose binding sites are located between Ϫ327 and Ϫ125. Furthermore, this collaboration may require cofactors, such as the interaction with p300 reported here. Thus, synergism or cross-talk between transcription factors plays a critical role in COX-2 induction. Indeed, it is also important for COX-2 induction by NO.
Finally, the functional interaction of p300 with members of the Ets family has been supported by studies showing binding between p300 and Ets-1 and Ets-2 (21). In the current study, we showed that PEA3 co-immunoprecipitates with p300. To our knowledge, this is the first demonstration of a direct interaction between p300 and PEA3 and indicates that PEA3 requires p300 to transactivate the COX-2 promoter. Additionally, the marked effect of NO on the functions of this interaction suggest that NO may enhance the recruitment of p300 by PEA3. Initial attempts to show that NO increases the interaction between PEA3 and p300 in co-immunoprecipitation studies have not been successful. Whether NO modulates the interaction at the protein-protein level or at the level of the interacted protein binding to DNA is of great interest and is currently under intense investigation in our laboratory.
In conclusion, we proposed a pathway that mediates COX-2 induction by NO, in which ␤-catenin-regulated PEA3 plays a critical role. It involves the activation of matrix metalloproteinases, the degradation of matrix and E-cadherin, the activation of various transcription factors, and the synergy between transcription factors/coactivator. As this pathway connects the proinflammatory effect of NO to its role in carcinogenesis, invasion, and metastasis in colorectal cancer cells, it is of clinical importance. Targeting this pathway to block the induction of COX-2 may eventually inhibit colorectal carcinogenesis and progression.