Cell culture and transient transfection

Mouse NIH3T3 cells, human embryonic kidney 293T, colorectal adenocarcinoma HT-29 cells (American Type Culture Collection) were cultured in DMEM with 10% FBS and transfected using lipofectAMINE (Invitrogen). Luciferase activity was determined using the dual-luciferase assay system (Promega) according to the manufacturer's directions, and measured with a Turner Luminometer TD-20/20.

Preparation of RNA transcripts

To determine the β-catenin binding site, the full-length COX-2 3′-UTR served as template and was amplified by PCR using primers containing the T7 polymerase promoter. The primers for the F1 UTR were forward, 5′-AGTTAATACGACTCACTATAGGGAAGTCTAATGATCATATT; reverse, 5′-AGTCTAGATCACAAGTATGACTCCTT. The primers of F2 UTR were forward, 5′-AGGTAATACGACTCACTATAGGGTATTAAGGTGGTGGAGCC; reverse, 5′-GGTCTAGATTGGTTATTGCTTTATGT. The primers of F3 UTR were forward, 5′-AGATAATACGACTCACTATAGGGATGACCTCATAAAATACC; reverse, 5′-GATCTAGAGTCTCTTAGCAAAATGGC. The oligonucleotides corresponding to F1-1, F1-2, F1-2 (MT) and F1-3 indicated in Figure 3A were synthesized by Bioneer (Korea). For in vitro synthesis of RNA transcripts, template DNA was incubated in a standard transcription reaction containing [α-³²P]-UTP. All the labeled RNA transcripts were gel purified and quantified by liquid scintillation counting.

Analysis of mRNA decay

NIH3T3 cells transfected with the pBBB vectors were serum starved for 24 h, followed by stimulation with 20% fetal bovine serum. Total cytoplasmic RNA was extracted at intervals for the time-course experiments (27). An aliquot of 1–2 μg of the DNAse I-treated RNA was reverse transcribed with Superscript II Reverse transcriptase (Stratagene). Rabbit β-globin mRNA levels were measured by RT–PCR or real-time PCR. The primers for pBBB^ΔUTR were forward, 5′-TAGAATTCCTCCTGGGCAACGTGCTG-3′; reverse, 5′-CGTCTAGATCAGTGGTATTTGTGAGC-3′. The luciferase mRNA of pGL3 (Promega) was used in transient transfection experiments as an internal control and standard. Real-time PCR was performed with a LightCycler system (Roche). Reactions were amplified using an LC FastStart reaction mix SYBR Green I kit (Roche), and quantified with the LightCycler analysis software (Roche). Relative levels of β-globin mRNA are expressed as the ratio of Cp to the internal luciferase control. The primers for the degradation of endogenous COX-2 mRNA were forward, 5′-TTCAAATGAGATTGTGGAAAAAT-3′; reverse, 5′-AGATCATCTCTGCCTGAGTATCTT-3′. The in vitro decay assay was performed as described (28).

EMSAs and supershift analyses

Cytoplasmic extracts (10 μg) and radiolabeled RNA transcripts were incubated at room temperature for 15 min in a binding buffer containing 15 mM HEPES (pH 7.5), 10 mM KCl, 5 mM MgCl₂, 0.2 mM DTT and 5% glycerol. Subsequently, unbound RNA was digested for 15 min with 10 U of RNase T1 (Ambion) at 37°C. RNA–protein complexes were resolved on 6% native polyacrylamide gels. For supershift assays, 1 μg of antibody was incubated with each binding reaction for 30 min on ice before the complexes were resolved by electrophoresis. To test whether β-catenin binds directly to the UTR, the labeled RNA was incubated with GST-fusion proteins and assayed by electrophoretic mobility shift assay (EMSA) as described (29).

The proximal region of the 3′-UTR is critical for β-catenin-induced COX-2 mRNA stabilization

To map the regions of the COX-2 3′-UTR responsible for stabilization we utilized luciferase reporters containing the full-length COX-2 3′-UTR or various deletions (Figure 2A). LiCl treatment has no significant effect on luciferase activity in cells transfected with either the luciferase control (ΔUTR) or luciferase lacking the F1 region of the COX-2 3′-UTR (ΔF1). In contrast, it increased the luciferase activities derived from the full-length COX-2 3′-UTR (Full) or the F1 region (F1), suggesting that the F1 region is responsible for the β-catenin-mediated stabilization of COX-2 mRNA (Figure 2B).

Open in a separate window

Figure 2

Requirement of the proximal region of COX-2 3′-UTR for β-catenin-dependent stabilization. (A) The reporter construct containing 1455 nt of the COX-2 3′-UTR (connecting line) together with the luciferase coding region (filled box). AU-rich elements are indicated as vertical lines. (B) Luciferase assay. Cells were transfected with various luciferase reporters. Luciferase activities were measured after treatment with LiCl or KCl for 24 h. Three independent experiments were performed. (C) Schematic representation of the locations of the F1, F2 and F3 regions. Vertical lines represent AREs. (D) In vitro RNA degradation assays to identify the β-catenin-responsive element in the COX-2 3′-UTR. [α-³²P]-labeled RNA substrates were incubated with cytoplasmic extracts from either vector or β-catenin-expressing NIH3T3 cells, and the reactions were stopped by adding stop buffer at the indicated times. Processed RNA was resolved on a 7 M urea/5% acrylamide gel and visualized by autoradiography. (E) In vitro analysis of mRNA decay upon MG132 treatment. Labeled F1 mRNA was incubated with cytoplasmic extracts from either DMSO (Control) or MG132-treated NIH3T3 cells, and decay was analyzed as in (D). (F) Stabilization of the F1 UTR by LiCl treatment. Labeled F1 mRNA was incubated with cytoplasmic extracts from either KCl (Control) or LiCl-treated NIH3T3 cells, and degradation was examined as in (D).

There are multiple copies of the AU-rich sequence element (ARE) in the 1455 nt 3′-UTR of COX-2 mRNA. To identify the cis-acting sequences modulated by β-catenin, we examined a cell-free mRNA decay system that has proven useful in mammalian cells (24,28). We synthesized RNA substrates comprising three distinct regions of the COX-2 3′-UTR (F1, F2 and F3 in Figure 2C). Subsequently, the ³²P-labeled RNA substrates were incubated with cytoplasmic extracts from either vector-transfected or β-catenin-transfected NIH3T3 cell, and their degradation was measured using denaturing polyacrylamide gels. As shown in Figure 2D, the F2 and F3 mRNA substrates, which do not contain the ARE region, were as stable as the GAPDH mRNA, and were not affected by β-catenin. On the other hand, the F1 mRNA which had a relatively short half-life was stabilized by cytoplasmic extracts containing overexpressed β-catenin. Stabilization of the COX-2 3′-UTR by β-catenin was confirmed using MG132- or LiCl-treated cytoplasmic extracts (Figure 2E and ​andF).F). These results show that the proximal region of the 3′-UTR is critical for β-catenin-induced COX-2 mRNA stabilization.

The class I AREs in proximal COX-2 UTR are the β-catenin-responsive element

Since the F1 region of the COX-2 UTR contains multiple copies of ARE, we dissected the F1 sequence. AREs are classified into three groups according to features of their sequence and their RNA decay characteristics (36,37). The F1 region of the COX-2 3′-UTR contains both class I ARE with one to three scattered copies of the AUUUA motif and class II ARE with multiple overlapping copies of that motif. To further define the β-catenin-responsive sequences in the F1 region, we further divided the 145 nt of F1 into 40 nt fragments (Figure 3A) and evaluated their degradation rates. We found that both the F1-1 (class II) and F1-2 (class I) fragments were unstable, whereas the F1-3 (non-ARE) fragment was as stable as the control GAPDH mRNA (Figure 3B). We next tested which RNA substrates responded to β-catenin-induced stabilization. As shown in Figure 3C, strong expression of β-catenin stabilized F1-2 rather than F1-1. Furthermore, mutation of the UUU residues of AUUUA to CGC in F1-2 (F1-2 MT) completely abolished the ARE-induced instability. These in vitro decay data indicate that β-catenin is responsible for stabilization of the class I ARE region of the F1 UTR.

Open in a separate window

Figure 3

Mapping of the β-catenin-dependent stabilizing element in the F1 UTR. (A) Schematic representation of the different F1 UTR probes. ARE sequences are indicated as boxes and mutant sequences are shown by underlining and in italics. (B) RNA substrates were incubated with cytoplasmic extracts from NIH3T3 cells, and the reactions were stopped by adding stop buffer at the indicated times. Processed RNA was resolved on a 7 M urea/5% acrylamide gel and visualized by autoradiography. (C) RNA substrates were incubated with cytoplasmic extracts from either vector (control) or β-catenin-expressing NIH3T3 cells. mRNA degradation was examined as in (B).

β-Catenin binds to the F1 region of COX-2 3′-UTR

To stabilize COX-2 mRNA β-catenin could either directly interact with the ARE of COX-2 or indirectly affect other signaling proteins. To determine whether β-catenin binds to the COX-2 3′-UTR, RNA-EMSAs were performed with recombinant β-catenin and radiolabeled RNA probes. The β-catenin protein was able to complex labeled F1 RNA (Figure 2A), but did not interact with labeled F2 or F3 RNA (data not shown). The specificity of the RNA binding was also confirmed by competition with excess unlabeled F1 or non-specific NC RNA (38). To characterize the binding specificity of β-catenin to COX-2 3′-UTR, other recombinant proteins were incubated with labeled F1 RNA. The armadillo repeats (Arm 1-12) of β-catenin as well as the ARE-binding HuR protein (26) formed RNA–protein complexes, but the other proteins did not (Figure 4B). RNA-EMSAs also showed that the binding affinity (K_d) of Arm 1-12 for F1 RNA was ~25 nM (Figure 4C). As shown in Figure 4D, the specificity of the RNA binding was confirmed by competition with excess unlabeled F1 or a β-catenin-specific RNA aptamer (H. K. Lee et al., manuscript submitted). Moreover, we found that β-catenin bound to the minimal β-catenin-responsive element, F1-2, but not to mutant F1-2 (data not shown). We conclude that β-catenin associates directly with ARE located in the first 150 nt of the COX-2 3′-UTR.

Open in a separate window

Figure 4

Specific association of β-catenin with the F1 region of COX-2 mRNA. (A) Specificity of β-catenin binding to RNA. [α-³²P]-labeled F1 RNA (80 pM) was co-incubated with recombinant β-catenin supplemented with unlabeled F1 RNA (F1) or non-specific NC RNA (10-, 50-, 200-fold each). Complexes are indicated as the arrow. (B) RNA-EMSAs. [α-³²P]-labeled F1 RNA was incubated with C-terminal domain of β-catenin (CTD), HuR, Arm 1-12, T-cell factor (TCF), or GST (200 nM each). (C) The RNA-EMSAs were performed with recombinant Arm 1-12 protein (2.5, 5, 10, 25, 50, 75, 100 and 200 nM), and the complexes are labeled 1–3. The HuR–F1 complex as control is indicated by an open arrowhead. (D) Specificity of Arm 1-12 binding to RNA. Labeled F1 RNA was co-incubated with Arm 1-12 supplemented with unlabeled F1 RNA (F1), aptamer (Apt; 10-, 50-, 200-fold each) or F2 RNA (200-fold). Following incubation, RNA–protein complexes were resolved on 5% native gels and visualized by autoradiography. Input indicates RNA only.

Intracellular interaction of β-catenin with F1 RNA

Since we observed that β-catenin bound to COX-2 3′-UTR with high affinity in vitro, we tested for the interaction in vivo by RNA immunoprecipitation assays. LiCl-treated NIH3T3 cells were transfected with various luciferase reporters (Figure 2A and Antisense of F1, F1AS) and fixed with formaldehyde. Subsequently lysates were immunoprecipitated with either β-catenin or normal IgG, and bound RNA was purified and analyzed by RT–PCR. Only the luciferase mRNA with full-size and F1 UTR were found in the anti-β-catenin immunoprecipitates (Figure 5A). We obtained similar results in β-catenin-overexpressing HCT116 colon cancer cells (data not shown), showing that the F1 region in the proximal COX-2 3′-UTR is sufficient for interaction with cellular β-catenin.

Open in a separate window

Figure 5

In vivo interaction of COX-2 mRNA with β-catenin. (A) RNA immunoprecipitation assay. NIH3T3 cells co-transfected with various reporters (ΔUTR, Full, F1, ΔF1 as in Figure 2A and Antisense of F1, F1 AS). After formaldehyde fixation, immunoprecipitations were performed with normal IgG or anti-β-catenin antibody. Bound RNA was extracted from the immune complexes and analyzed by RT–PCR. (B) Supershift assay. Labeled F1 RNA was incubated with cytoplasmic extracts of NIH3T3 cells containing either empty vector (control) or β-catenin expression in the presence of normal IgG, anti-HuR, or anti-β-catenin antibodies and analyzed by 5% native gels. The supershifted bands of β-catenin (arrow) as well as of HuR (open arrowhead) are indicated. (C) Cytoplasmic (CE) and nuclear (NE) extracts were prepared from either normal (Mock) or β-catenin-overexpressing (β-cat) NIH3T3 cells and analyzed by western blotting.

We also carried out supershift assays with cytoplasmic extracts prepared from vector- or β-catenin-expressing NIH3T3 cells. Figure 5B shows that RNA–protein complexes were readily detected with both of these cytoplasmic extracts. Control IgG had no effect on the migration of RNA–protein complex whereas a prominent slowly migrating band was detected with anti-HuR antibody, and the intensity of this supershifted band was stronger when a β-catenin-overexpressing cytoplasmic extract was used. Moreover, a supershifted band was also observed when we used anti-β-catenin antibody, demonstrating that β-catenin was present in this RNA–protein complex. Since overexpression of β-catenin promoted the interaction of HuR with F1 RNA but did not alter the level of HuR (Figure 1B), we tested whether HuR localization was affected by overexpressing β-catenin and found that it caused HuR to move from the nucleus to the cytoplasm (Figure 5C). These data indicate that overexpressed β-catenin associates with the COX-2 3′-UTR and induces HuR to translocate to the cytoplasm where it acts to stabilize COX-2 mRNA.

The generous gifts of various DNA constructs from Drs Nakamura, Nagamine, Chun, Prescott and Han are greatly appreciated. This study was supported by grants from the Basic Science Research Fund of the Korea Research Foundation (2004-015-C00378), the Korea Science and Engineering Foundation (R01-2006-000-10194-0 and 2005-01131). The Open Access publication charges for this article were waived by Oxford University Press.

Conflict of interest statement. None declared.