Transforming Growth Factor- b 1 Enhances Ha- ras -induced Expression of Cyclooxygenase-2 in Intestinal Epithelial Cells via Stabilization of mRNA*

Oncogenic ras induces the expression of cyclooxygen-ase-2 (COX-2) in a variety of cells. Here we investigated the role of transforming growth factor- b (TGF- b ) in the Ras-mediated induction of COX-2 in intestinal epithelial cells (RIE-1). RIE-1 cells were transfected with an inducible Ha-Ras Val12 cDNA and are referred as RIE-iRas cells. the addition of 5 m M isopropyl-1-thio- b - D -galacto-pyranoside (IPTG) induced the expression of Ha-Ras- Val12 , closely followed by an increase in the expression of COX-2. Neutralizing anti-TGF- b antibody partially blocked the Ras-induced increase in COX-2. Combined treatment with IPTG and TGF- b 1 resulted in a 20–50-fold increase in the levels of COX-2 mRNA. The t 1 ⁄ 2 of COX-2 mRNA was increased from 13 to 24 min by Ha-Ras induction alone. The addition of TGF- b 1 further stabilized the COX-2 mRNA ( t 1 ⁄ 2 > 50 min). Stable transfection of a luciferase reporter construct containing the COX-2 3 * -untranslated region (3 * -UTR) revealed that TGF- b 1 treatment and Ras induction each stabilized the COX-2 3 * -UTR. Combined treatment with IPTG and TGF- b 24 h, then the transcription was stopped by the addition of 100 m M DRB (5,6-dichlorobenzimidazole riboside; Sig-ma). The RNA samples were isolated at 0, 10, 20, 30, 40, and 50 min following the DRB treatment and analyzed for mRNA levels by Northern blotting. Immunoblot Analysis— Immunoblot analysis was performed as described previously (38). Briefly, the cells were lysed for 30 min in radio immunoprecipitation assay buffer (1 3 phosphate-buffered saline, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, 10 mg/ml phenyl- methylsulfonyl fluoride, 10 m g/ml aprotinin, 1 m M sodium orthovana-date), then clarified cell lysates were denatured and fractionated by SDS-polyacrylamide gel electrophoresis. After electrophoresis, the proteins were transferred to nitrocellulose membrane. The filters were then probed with the indicated antibodies, developed by the enhanced chemiluminescence system (ECL, Amersham Pharmacia Biotech), and exposed to X-AR5 film (Eastman Kodak Co.). Quantitation was by densitometry. The anti-COX-2 antibody was purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Anti-pan Ras antibody was purchased from Calbiochem. Transfection

effect of Ras-mediated transformation in intestinal epithelial cells (RIE-1) (6), fibroblasts (7), mammary epithelial cells (8), and non-small cell lung cancer cells (9). The mechanisms underlying the regulation of COX-2 expression are complex. Previously, it was reported that COX-2 expression was regulated at the transcriptional level by activated Ha-ras in mammary cells (8,10) and by oncogene v-src (11) and growth factors (12) in NIH 3T3 cells. Interleukin-1␣ induced rapid but transient activation of COX-2 transcription and also prolonged the halflife of the COX-2 mRNA (13). Post-transcriptional regulation of cytokine-induced cyclooxygenase-2 transcript isoforms by dexamethasone has also been reported (14). We have found that the induction of COX-2 in conditionally Ha-Ras Val12 -transformed Rat-1 cells occurs via a modest increase (ϳ50%) in COX-2 transcription and a 3-fold increase in the half-life of COX-2 mRNA (7).
The TGF-␤s are 25-kDa homodimeric polypeptides belonging to a superfamily of growth regulatory molecules. TGF-␤ has previously been characterized as a potent growth inhibitor for cultured rat intestinal crypt cells (15)(16)(17)(18). Through the activation of specific receptors, the TGF-␤ ligands activate the Smad signal transduction pathway, which appears to serve an important tumor suppressor function (19 -22). However, there is mounting evidence that TGF-␤ may enhance malignant transformation and tumor progression for several different epithelial tumors under certain circumstances (23)(24)(25)(26)(27)(28)(29). One of the remarkable effects of TGF-␤ on intestinal epithelial cells and other cell types is the induction or augmentation of COX-2 expression (4, 29 -34). Expression of either activated Ras or Src proteins activates transcription of the TGF-␤1 gene (35,36). Furthermore, TGF-␤ collaborates with oncogenic Ras in the transformation of mammary epithelial cells (28). Based upon the observations that COX-2 was overexpressed in 85-90% of human colon cancers (1) and that TGF-␤ was abnormally expressed in over 90% of human colon cancers (37), we hypothesized that TGF-␤ may play a role in the regulation of COX-2 expression during the adenoma to carcinoma sequence of events that are involved in the neoplastic transformation of colonic epithelial cells (4).
This study describes the observation that TGF-␤ synergistically enhances the expression of COX-2 in conditionally Ha-Ras-transformed intestinal epithelial cells. We have also evaluated the mechanisms by which Ras and TGF-␤ mediate the induction of COX-2. TGF-␤-neutralizing antibody partially inhibits the increased expression of COX-2 after induction of Ha-Ras, suggesting an important autocrine effect of Ras-induced TGF-␤1 expression. Both Ha-Ras Val12 and TGF-␤1-mediated induction of COX-2 involve stabilization of COX-2 mRNA, and the combined effects of Ha-Ras Val12 and TGF-␤ * This work was supported by National Institutes of Health Grants DK-52334 and CA-69457 (to R. D. B.), DK-47297 and ES-00267 (to R. N. D.), P01 CA77839 and CA 68485 (Vanderbilt Cancer Center), (to R. D. B. and R. N. D.), and P01 CA73992 and CA42014 (to S. M. P.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

EXPERIMENTAL PROCEDURES
Cell Culture-RIE-iRas cell line with an inducible activated Ha-Ras Val12 cDNA was generated by using LacSwitch eukaryotic expression system (Stratagene, La Jolla, CA) and was maintained in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum, 400 g/ml G418 (Life Technologies, Inc.), and 150 g/ml hygromycin B (Calbiochem). The Ha-Ras Val12 cDNA is under the transcriptional control of the Lac operon in rat intestinal epithelial (RIE-1) cells. Isopropyl-1-thio-␤-D-galactopyranoside (IPTG; Life Technologies, Inc.) at a concentration of 5 mM was used to induce the expression of mutated Ha-Ras. Anti-TGF-␤ antibody (R&D Systems, Inc. Minneapolis MN) was used to block the endogenous TGF-␤ activity.
RNA Extraction and Northern Blot Analysis-The extraction of total cellular RNA was performed as described previously (7). RNA samples (20 g/lane) were separated on formaldehyde-agarose gels and blotted onto nitrocellulose membranes. The blots were hybridized with cDNA probes labeled with [␣-32 P]dCTP by random primer extension (Stratagene). After hybridization and washes, the blots were subjected to autoradiography. 18 S rRNA signals were used as controls to determine integrity of RNA and equality of the loading. For determination of mRNA stability, RIE-iRas cells were treated with IPTG, TGF-␤1, or both IPTG and TGF-␤1 for 24 h, then the transcription was stopped by the addition of 100 M DRB (5,6-dichlorobenzimidazole riboside; Sigma). The RNA samples were isolated at 0, 10, 20, 30, 40, and 50 min following the DRB treatment and analyzed for mRNA levels by Northern blotting.
Immunoblot Analysis-Immunoblot analysis was performed as described previously (38). Briefly, the cells were lysed for 30 min in radio immunoprecipitation assay buffer (1ϫ phosphate-buffered saline, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, 10 mg/ml phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, 1 mM sodium orthovanadate), then clarified cell lysates were denatured and fractionated by SDS-polyacrylamide gel electrophoresis. After electrophoresis, the proteins were transferred to nitrocellulose membrane. The filters were then probed with the indicated antibodies, developed by the enhanced chemiluminescence system (ECL, Amersham Pharmacia Biotech), and exposed to X-AR5 film (Eastman Kodak Co.). Quantitation was by densitometry. The anti-COX-2 antibody was purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Anti-pan Ras antibody was purchased from Calbiochem.
The construction of reporter expression vectors pLucϩ3Ј-UTR, pLucϩ3Ј-UTR⌬ARE, and pLucϩARE was described elsewhere. 2 Briefly, a reporter vector, pLuc, was generated by inserting the luciferase cDNA into pcDNA3/zeo (Invitrogene, Carlsbad, CA). The addition of the COX-2 3Ј-UTR (1451 bp) was accomplished by polymerase chain reaction amplification of the COX-2 3Ј-UTR using XbaI-tailed primers and inserting them adjacent to the luciferase coding region to yield pLucϩ3Ј-UTR. pLucϩ3Ј-UTR⌬ARE was generated by digesting pLucϩ3Ј-UTR with ApaI and ScaI to release a 1036-bp region of the COX-2 3Ј-UTR, and this 3Ј-UTR region was cloned into the ApaI and filled-in XbaI sites of pLuc. For pLucϩARE, pLucϩ3Ј-UTR was digested with PmeI and ScaI to generate a 1818-bp LucϩARE fragment that was inserted into pcDNA3/zeo vector. RIE-iRas cells were transfected with pLuc (RIE-iRas/luc), pLucϩ3Ј-UTR (RIE-iRas/Lucϩ3Ј-UTR), pLucϩ3Ј-UTR⌬ARE (RIE-iRas/Lucϩ3Ј-UTR⌬ARE), or pLucϩARE (RIE-iRas/ LucϩARE). The stable transfected clones were selected with neomycin (600 g/ml), hygromycin (150 g/ml), and zeocin (250 g/ml). For the luciferase activity assay, cells were treated for the indicated hours and lysed with passive lysis buffer (Promega, Madison WI). Twenty l of lysate was used for the firefly luciferase reading by using a luciferase reporter assay system (Promega) and a model TD-20/20 luminometer. Firefly luciferase values were standardized to the protein contents and presented as mean Ϯ S.E. of assays performed in triplicate.

Ha-Ras-mediated Transformation and Induction of COX-2-
We previously reported that COX-2 expression is significantly increased in a rat intestinal epithelial cell line stably transformed by Ha-Ras (6) and conditional expression of Ha-Ras rapidly induced the expression of COX-2 in rat fibroblasts (7). To determine whether COX-2 is an early target of oncogenic Ha-Ras in RIE cells, we have introduced the inducible Ha-Ras Val12 cDNA vectors into the RIE-1 cells (RIE-iRas). Noninduced RIE-iRas cells displayed the same nontransformed morphology as the parental RIE-1 cells. Morphological transformation of the RIE-iRas cells was observed between 24 -48 h after IPTG treatment. During this interval, cell-cell contact inhibition was lost. The cells acquired a spindly appearance and grew in overlapping clusters (Fig. 1A). The morphological transformation of RIE-iRas cells could be completely reversed upon withdrawal of IPTG for 72 h (Ϫ72 h). As shown in Fig. 1B, the addition of IPTG into the culture medium induced activated Ha-Ras protein by 4 h. Thereafter, the level of Ras protein was continuously elevated for the duration of IPTG treatment. COX-2 was expressed at very low levels in RIE-iRas cells before IPTG treatment. The elevation of COX-2 protein was detected by 8 h after the addition of IPTG. The induction of COX-2 temporally coincided with the induction of Ha-Ras protein.
We next investigated whether Ras-induced expression of COX-2 involves signaling through TGF-␤. Northern analysis revealed that induction of Ras significantly increased the levels

FIG. 1. Ras-mediated transformation and induction of COX-2.
A, morphological transformation of RIE-iRas cells. RIE-iRas cells were grown on 60-mm tissue culture dishes. The cells were treated with 5 mM IPTG for 0, 24, or 72 h, then IPTG was removed for 72 h (Ϫ72 h). The pictures were taken by using an inverted microscope (original magnification, ϫ100). B, induction of Ha-Ras and COX-2. RIE-iRas cells were treated with IPTG and lysed in radio immunoprecipitation assay buffer (1ϫ phosphate-buffered saline, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, 10 mg/ml phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, 1 mM sodium orthovanadate) at the indicated time points. 50 g of each cell lysates were fractionated by SDS-polyacrylamide gel electrophoresis. After electrophoresis, the proteins were transferred to nitrocellulose membranes. The filters were blotted with the indicated antibodies and developed by the ECL chemiluminescence system. of TGF-␤1 mRNA ( Fig. 2A). In the absence of Ras induction, the addition of exogenous TGF-␤1 transiently increased the levels of COX-2 in RIE-iRas cells, with a peak between 12-24 h following TGF-␤ treatment (Fig. 2B). To determine whether endogenous TGF-␤ mediates a component of the Ras-induced COX-2 expression, TGF-␤-neutralizing antibody was added before the IPTG treatment. Indeed, induction of COX-2 by activated Ha-Ras was partially blocked by TGF-␤-neutralizing antibody (Fig. 2C), implying a functional autocrine role for TGF-␤ in Ras-mediated induction of COX-2.
TGF-␤ Enhances Ras-mediated Induction of COX-2-Although TGF-␤ expression is normally restricted to the lumenal one-third of the intestinal epithelium (16,40), the interstitial cells may produce large amounts of paracrine TGF-␤. It was of interest to determine whether there was a cooperative effect of exogenous TGF-␤1 and oncogenic Ras on the expression of COX-2. The RIE-iRas cells were treated with 3 ng/ml TGF-␤1 without IPTG, and the levels of COX-2 mRNA were analyzed by Northern blotting. As shown in Fig. 3A, in the absence of Ras induction, treatment with TGF-␤1 increased the level of COX-2 mRNA, which reached a peak (3.5-fold) by 8 h after the treatment. Induction of Ras by IPTG treatment increased the levels of COX-2 mRNA by 8 h, and the levels remained at an elevated plateau between 12-72 h. Interestingly, combined treatment with TGF-␤1 and IPTG resulted in a marked induction of COX-2. The levels of COX-2 mRNA were synergistically elevated by the combination of TGF-␤1 and Ras by 24 h after the combined treatment. A 20 -50-fold increase in the levels of COX-2 mRNA was observed between 24 -72 h. Western analysis further confirmed that the increased levels of COX-2 mRNA level also reflected an increase in the level of COX-2 protein (Fig. 3B).
Stabilization of Ras-induced COX-2 mRNA by TGF-␤ Treatment-We have previously reported that induction of Ha-Ras stabilizes COX-2 mRNA in Rat-1 fibroblasts (7). To determine the mechanism of synergistic induction of COX-2 that results from oncogenic Ras and TGF-␤1, we examined the stability of COX-2 mRNA in RIE-iRas cells after TGF-␤1 and IPTG treatment. RIE-iRas cells were treated with IPTG, TGF-␤1, or both IPTG and TGF-␤1 for 24 h, then transcription was stopped by the addition of 100 M DRB. The RNA samples were isolated at 0, 10, 20, 30, 40, and 50 min following the DRB treatment and analyzed for mRNA levels by Northern blotting. As demonstrated in Fig. 4, A and B, COX-2 mRNA was rapidly degraded in noninduced RIE-iRas cells (t1 ⁄2 ϳ 13 min). Individually, either TGF-␤1 or IPTG treatment increased the stability of COX-2 mRNA to a similar extent (t1 ⁄2 ϳ 24 -30 min). The COX-2 mRNA from the RIE-iRas cells after combined TGF-␤1 and IPTG treatment was extremely stable, and the t1 ⁄2 was greater than 50 min.
To further study transcriptional and post-transcriptional regulation of COX-2 in RIE-iRas cells, we transfected the luciferase reporter gene linked with either COX-2 promoter region or 3Ј-UTR into RIE-iRas cells. The 5Ј-flanking region of the human COX-2 gene (nucleotides Ϫ327 to ϩ59) includes the nuclear factor responsible for Interleukin-6 expression (NF-IL6) site and the cyclic AMP response element (CRE) (39). This reporter gene exhibited promoter activity that was modestly increased by induction of oncogenic Ras or by the addition of exogenous epidermal growth factor (EGF) (Fig. 5A). Treatment with TGF-␤1 for 6 and 24 h did not alter the activity of this promoter region.
We next engineered the RIE-iRas cells to express the cytomegalovirus promoter-driven luciferase reporter gene alone (RIE-Ras/luc) or linked with 1.5 kilobases of COX-2 3Ј-UTR (RIE-iRas/lucϩ3Ј-UTR). Stably transfected clones were selected and analyzed. Treatment with IPTG for 24 h increased the luciferase activity by 100% in RIE-iRas/lucϩ3Ј-UTR cells but did not alter the luciferase activity in RIE-iRas/luc control cells (Fig. 5B). These results suggested that induction of Ras stabilized luciferase mRNA via the linked COX-2 3Ј-UTR and not through a transcriptional induction of the luciferase gene alone. Northern blot analysis confirmed that the increased luciferase activity resulted from an increase in the levels of luciferase mRNA (Fig. 5C). The level of luciferase mRNA was clearly elevated by 8 h after the induction of Ha-Ras. A 2-3-fold increase in the level of luciferase mRNA was observed between 24 -48 h after Ha-Ras Val12 was induced by the addition of IPTG (Fig. 5C).
There was no significant induction of luciferase activity in control RIE-iRas/luc cells that were treated with IPTG, TGF-␤1, or both IPTG and TGF-␤1 for 6, 24, and 72 h (Fig. 5D). However, the luciferase activity in RIE-iRas/lucϩ3Ј-UTR cells was induced by either IPTG or TGF-␤1 treatment. The luciferase activity was additively increased by 24 h after combined treatment with TGF-␤1 and IPTG. A synergistic increase in luciferase activity was observed by 72 h after the RIE-iRas/ lucϩ3Ј-UTR cells were treated with both TGF-␤1 and IPTG. These results are consistent with the findings of Northern analysis described in Fig. 3A.
The 3Ј-untranslated region of the COX-2 transcript is extremely AU-rich and contains 14 copies of the Shaw-Kamens sequence (AUUUA), otherwise known as AU-rich elements (AREs) (41). To determine whether the stabilization of COX-2 mRNA by oncogenic Ras and TGF-␤1 treatment was dependent upon highly conserved AU-rich elements located in the proximal COX-2 3Ј-UTR (14), we constructed two additional reporter vectors. The pLucϩ3Ј-UTR⌬ARE construct was derived from pLucϩ3Ј-UTR by removal of a fragment of 415 bp of nucleotide including 8 conserved AU-rich elements from the proximal COX-2 3Ј-UTR. For the pLucϩARE, the 415 bp of conserved AU-rich fragment was linked to the downstream of luciferase cDNA in pLuc vector (see "Experimental Procedures"). RIE-iRas cells were transfected with either pLucϩ3Ј-UTR⌬ARE or pLucϩARE. Stably transfected clones were selected by the treatment with zeocin, pooled, and referred to as RIE-iRas/ Lucϩ3Ј-UTR⌬ARE or RIE-iRas/LucϩARE. IPTG treatment (Ha-Ras induction) of the RIE-iRas/Lucϩ3Ј-UTR⌬ARE cells resulted in a 44% increase in luciferase activity after 24 h (Fig. 6A) and a 3-fold increase after 72 h (Fig. 6B). Treatment of the RIE-iRas/Lucϩ3Ј-UTR⌬ARE cells with TGF-␤1 alone altered luciferase activity by no more than 27%, and the combination of TGF-␤1 and IPTG did not increase the levels over that observed with IPTG alone (Fig. 6, A and B). In contrast, in the RIE-iRas/LucϩARE cells, IPTG treatment increased luciferase activity 5.4-fold by 24 h (Fig. 6A) and 7.2-fold by 72 h (Fig. 6B). TGF-␤1 treatment of the RIE-iRas/LucϩARE cells increased luciferase activity 4.7-fold by 24 h (Fig. 6A) and 2.9-fold by 72 h (Fig. 6B) after treatment. Treatment of RIE-iRas/LucϩARE cells with the combination of IPTG and TGF-␤1 increased the luciferase activity 15.7-fold by 24 h (Fig. 6A) and 23-fold by 72 h (Fig. 6B). Thus, the conserved 415-bp ARE region of the COX-2 3Ј-UTR appears to be necessary for the synergistic increase in expression caused by the combination of activated Ras and TGF-␤1. DISCUSSION Numerous studies have suggested that cyclooxygenase activity and prostaglandin synthesis may be involved in intestinal carcinogenesis. COX-2 expression is increased in human colorectal adenocarcinomas when compared with normal adjacent colonic mucosa (1)(2)(3). Furthermore there is mounting evidence that COX-2 expression in colorectal cancer cells provides a growth and survival advantage (42), increases tumor cell invasiveness (43), and enhances tumor angiogenesis (44). The importance of cyclooxygenase function in colorectal tumorigenesis is further supported by the observations that chronic ingestion of nonsteroidal anti-inflammatory drugs is associated with a reduced incidence of colorectal cancer in humans (45). Nonsteroidal anti-inflammatory drugs can also significantly inhibit colorectal tumorigenesis in animal models (46,47). Selective inhibition of COX-2 activity is particularly effective in this preventive effect (48,49). The present study demonstrates that We had previously observed that TGF-␤1 transiently induced the expression of COX-2 in nontransformed RIE cells (34). TGF-␤ expression is increased in a wide variety of cancers (including colon cancer) relative to adjacent normal tissues (50). We previously observed that chronic TGF-␤1 treatment resulted in both morphologic transformation and constitutive overexpression of COX-2 in cultured rat intestinal epithelial cells (29). We have also observed coordinate high level expression of both COX-2 and TGF-␤1 in colorectal adenomas and carcinomas (4). Transformation of cells with dominant oncogenes (e.g. Ha-ras or v-src) results in transcriptional activation of the TGF-␤1 gene (35,36). Activation of the Ki-ras oncogene occurs frequently in colorectal cancers (51). Increased expression of COX-2 after Ras transformation has also been reported in both epithelial cells (8) and in fibroblasts (7). Our observa-tion that TGF-␤-neutralizing antibody partially inhibited the increase in COX-2 induced by activated Ha-Ras suggests that TGF-␤ contributes to the increase in COX-2 in an autocrine manner. In cooperation with oncogenic Ras, the addition of exogenous TGF-␤1 caused a persistent synergistic induction of COX-2 at both the mRNA and protein levels. Thus, autocrine and paracrine TGF-␤ may contribute to the overexpression of COX-2 that occurs in transformed intestinal epithelial cells.
Thus far, most studies of the COX-2 regulation have focused on the transcriptional mechanism of induction (8, 10 -12, 52-54). An exception was the observation of post-transcriptional destabilization of cytokine-induced cyclooxygenase-2 transcript isoforms by dexamethasone as reported by Ristimaki et al. (13,14). Our previous studies demonstrated that the induction of COX-2 in Rat-1 cells conditionally transformed by Ha-Ras Val12 occurred due to the combined effects of a modest (ϳ50%) increase in transcription and a more significant (3-fold) increase FIG. 5. Luciferase activity assay. A, transfection of reporter vector under the transcriptional control of COX-2 promoter (Ϫ327/ϩ59). RIE-iRas cells were co-transfected with phPES2(Ϫ327/ϩ59) containing the 5Ј-flanking region of the human COX-2 gene (nucleotides Ϫ327 to ϩ59) and pcDNA3/zeo. Pooled stable transfectants were selected by the addition of 300 g/ml zeocin. Cells were plated in 24-well plates and treated with or without 5 mM IPTG for 24 h. Cells were washed twice with phosphate-buffered saline and lysed with passive lysis buffer. 20 l of lysate was used for the firefly luciferase readings. Plotted is the mean Ϯ S.E. B, stable transfection of reporter constructs linked with COX-2 3Ј-UTR. RIE-iRas cells were transfected with pcDNA3/Luc or pcDNA3/Lucϩ3Ј-UTR, and stable transfectants were selected by addition of 300 g/ml zeocin in cultural medium. For the luciferase assay, established clones (15 RIE-iRas/luc and 31 RIE-iRas/Lucϩ3Ј-UTR clones) were plated in 24-well plates and treated with or without 5 mM IPTG for 24 h. Cells were washed twice with phosphate-buffered saline and lysed with passive lysis buffer. 20 l of lysate was used for the firefly luciferase readings. Plotted is the mean Ϯ S.E. C, levels of luciferase mRNA in RIE-iRas/Lucϩ3Ј-UTR cells. Cells were treated with IPTG, and total RNA was collected at the indicated time points. The Northern blot was probed with [ 32 P]dCTP-labeled luciferase cDNA. The levels of 18 S rRNA are shown as loading controls. D, stabilization of COX-2 3Ј-UTR by Ras and TGF-␤1 treatment. Representative clones of RIE-iRas/luc or RIE-iRas/Lucϩ3Ј-UTR cells were treated with IPTG, TGF-␤1, or both IPTG and TGF-␤1 for the indicated hours. Luciferase activity was measured and plotted as mean Ϯ S.E.
in the stability of COX-2 mRNA (7). In the present study, we confirmed that induction of Ha-Ras modestly increased the promoter activity of the COX-2 5Ј-flanking region (Ϫ327/ϩ59), whereas TGF-␤ did not activate the COX-2 promoter. In contrast, either Ha-Ras Val12 induction or TGF-␤1 treatment individually increased the half-life of COX-2 mRNA, whereas the combination of Ras induction and TGF-␤1 treatment markedly increased the half-life of COX-2 mRNA. Both the induction of the Ras and TGF-␤ treatment increased the expression of luciferase, an effect that was mediated by the COX-2 3Ј-UTR contained in the chimeric Luc-COX-2 3Ј-UTR vector. Combined Ras and TGF-␤1 effects resulted in synergistic increase in the expression of Luc-COX-2 3Ј-UTR. We conclude that this inducible increase in expression of Luc-COX-2 3Ј-UTR was due to an increase in the stability of the mRNA. Neither Ras induction nor TGF-␤1 treatment induced luciferase activity in control cells that were transfected with the cytomegalovirus-driven luciferase controls containing no COX-2 3Ј-UTR. Our finding of increased steady-state levels of luciferase mRNA in the absence of increased transcription leads us to conclude that the increased luciferase activity reflects an increase in mRNA stability that has been conferred by the COX-2 3Ј-UTR. While we have confirmed that a significant portion of the increase in COX-2 expression after Ras activation and exposure to TGF-␤ is due to increased mRNA stability, we have not excluded the possibility that increased translational efficiency may also contribute the increased expression of COX-2.
The 3Ј-untranslated region of the COX-2 transcript is extremely AU-rich, and contains 14 copies of the Shaw-Kamens sequence (AUUUA) otherwise known as AU rich elements (AREs) (41,55,56). This motif is present in many immediate early genes and is thought to be involved in the regulation of mRNA degradation (57,58). In our study, removal of 415 bp of nucleotide including first 8 AREs from proximal COX-2 3Ј-UTR significantly reduced and delayed the Ras induced stabilization of COX-2 3Ј-UTR. TGF-␤1-induced stabilization of COX-2 3Ј-UTR was almost completely abolished by removal of this AUrich region. The synergistic increase in Luc-COX-2 3Ј-UTR expression clearly required the presence of this AU-rich region. In addition, linkage of the eight ARE clusters to a luciferase reporter gene resulted in an enhanced stabilization of luciferase by TGF-␤1 and Ras. Synergistic induction of luciferase activity by Ras and TGF-␤1 indicates the important role of this AU-rich region in stabilization of COX-2 mRNA. Chen et al. (58) recently report that interleukin-2 mRNA contains at least two cis elements that mediate its stabilization in response to different signals. A cis element encompassing the 5Ј-UTR and the beginning of the coding region mediates the stabilization that resulted from activation of c-Jun amino-terminal kinase (JNK), whereas other signals may modulate the stability of interleukin-2 mRNA through the 3Ј-UTR. These findings suggest that multiple elements within interleukin-2 mRNA modulate its stability in a combinatorial manner. The present studies do not exclude the possibility of cis elements in addition to those in the 3Ј-UTR that may contribute to COX-2 mRNA stability.
There are several lines of evidence that neoplastic transformation results in abrogation of TGF-␤-mediated growth inhibitory and tumor suppressive effects. For example, the loss of growth inhibitory responsiveness after Ras transformation is associated with a decrease in the levels of type II TGF-␤ receptor (17,59) and inhibitory phosphorylation of the TGF-␤ signaling proteins Smad2 and Smad3 (60). On the other hand, TGF-␤ may actually promote malignant transformation and tumor progression by several different mechanisms. TGF-␤ may suppress tumor immunosurveillance (23). TGF-␤ exhibits growth inhibitory effects on moderate to well differentiated primary colon carcinomas but stimulates the proliferation and invasion of poorly differentiated and metastatic colonic carcinomas (61, 62). TGF-␤1 can induce estrogen-independent tu-FIG. 6. The role of AREs in stability of COX-2 mRNA. A, stable transfection with pLucϩ3Ј-UTR⌬ARE and pLucϩARE. A conserved 415-bp AU-rich region including 8 AUUUA elements was removed from the proximal COX-2 3Ј-UTR. The mutated COX-2 3Ј-UTR was inserted into pcDNA3/Luc reporter vector (pLucϩ3Ј-UTR⌬AR). RIE-iRas cells were stably transfected with pLucϩ3Ј-UTR⌬ARE (RIE-iRas/Lucϩ3Ј-UTR⌬ARE) and were treated with IPTG or TGF-␤1 or both IPTG and TGF-␤1 for 24 h (⌬ARE). Luciferase activity was measured and plotted as mean Ϯ S.E. The conserved 415-bp AU-rich region including 8 AUUUA elements was linked to the downstream luciferase gene in pLuc vector (pLucϩARE). RIE-iRas cells were transfected with pLucϩARE. Stable transfected cells were treated with IPTG or TGF-␤1 or both IPTG and TGF-␤1 for 24 h (ϩARE). Luciferase activity was measured and plotted as mean Ϯ S.E. B, RIE-iRas/Lucϩ3Ј-UTR⌬ARE cells (⌬ARE) and RIE-iRas/LucϩARE (ϩARE) cells were treated with IPTG or TGF-␤1 or both IPTG and TGF-␤1 for 72 h. Luciferase activity was measured and plotted as mean Ϯ S.E. morigenicity of human breast cancer cells (63). TGF-␤ treatment can also substitute for wounding as a promoter of fibrosarcomas in chickens infected with the Rous sarcoma virus (24,25). There is mounting evidence that autocrine TGF-␤ expression by tumor cells plays a major role in the epithelialto-fibroblastoid conversion in mammary cells (28) and in keratinocytes (27,64) that accompanies malignant transformation. TGF-␤1 overexpression in transgenic mouse keratinocytes tends to inhibit the appearance of carcinogen-induced benign skin tumors, but for more advanced lesions, TGF-␤1 overexpression enhanced progression toward the malignant spindlecell phenotype (27). The TGF-␤-induced epithelial-to-mesenchymal transition in Ha-Ras-transformed mammary epithelial cells involves disrupted cell-cell adhesion and the loss of epithelial cell polarity in addition to causing the cells to become more spindle-shaped and invasive (28). Furthermore, chronic exposure to high concentrations of TGF-␤1 promotes the spontaneous neoplastic transformation of cultured rat liver epithelial cells (26).
We have recently reported that chronic exposure of rat intestinal epithelial (RIE-1) cells resulted in neoplastic transformation as defined by loss of growth inhibitory response to TGF-␤, loss of contact inhibition, acquisition of a fibroblast-like spindle cell morphology, anchorage-independent growth, and tumorigenicity in athymic mice (29). These TGF-␤-resistant RIE-1 cells (called RIE-TR) also exhibited markedly increased expression of COX-2 and TGF-␤1. Taken together, these observations in experimental tumor models suggest that tumor cells have not only lost the growth-inhibitory response to TGF-␤, but that TGF-␤ then promotes the malignant phenotype in transformed cells. TGF-␤ induction of COX-2 in cancer cells may contribute to this aggressive phenotype. TGF-␤ expression in human cancers may also have detrimental significance. In human colorectal cancers, high level expression of TGF-␤1 in the primary tumor is an independent predictor of risk (18-fold) for recurrence (65) and advanced stage with reduced survival in colorectal cancer patients (66).
Our present findings demonstrate that TGF-␤1 synergistically increases Ras-induced expression of COX-2, largely via an increase in the stability of COX-2 mRNA. The synergistic increase in COX-2 mRNA stability requires the presence of the conserved ARE region found within the proximal 500 bp of the 3Ј-UTR. These results point to a novel mechanism by which TGF-␤ and activated Ras collaborate to increase the expression of an important effector of tumor progression. Further studies to determine the mechanism of this induced stabilization of the COX-2 mRNA in transformed cells may yield important insights into the process of tumor progression.