The RING finger protein, RNF8, interacts with retinoid X receptor alpha and enhances its transcription-stimulating activity.

Retinoid X receptor alpha (RXR alpha) is a member of the steroid hormone receptor superfamily. Using yeast two-hybrid screening, beta-galactosidase assays, and pull-down assays, we show that RNF8, a RING finger protein recently isolated as a protein binding to a ubiquitin-conjugating enzyme, binds to RXR alpha through the N-terminal regions of both proteins. In COS7 cells, overexpressed RNF8 colocalized and interacted with RXR alpha in the nucleus, as shown by fluorescence resonance energy transfer. A point mutation of RNF8, Cys-403 to Ser (C403S), which disrupts the RING finger structure, or deletion of the N-terminal region (Delta N) of RNF8 prevented localization of RNF8 to the nucleus without affecting nuclear localization of RXR alpha. Although transient overexpression of RNF8 had little effect on RXR alpha ubiquitination, RNF8 dose-dependently enhanced RXR alpha-mediated transactivation of the RXR-responsive element (RXRE)-bearing gene promoter without the addition of its ligand, 9-cis-retinoic acid (RA), and up-regulated the expression of the genes downstream of RXRE as well as an RA-response element. This transactivation-enhancing activity was not seen with either the C403S point mutant or the Delta N deletion mutant of RNF8. These results suggest a novel function of RNF8 as a regulator of RXR alpha-mediated transcriptional activity through interaction between their respective N-terminal regions.

Retinoic acid (RA), 1 an active metabolite of vitamin A, has profound effects on the proliferation and differentiation of var-ious cell types (1), mainly through two members of the nuclear receptor superfamily, retinoic acid receptor (RAR) and retinoid X receptor (RXR) (2,3), which are ligand-dependent transcriptional regulators. Although all-trans-RA and 9-cis-RA (9cRA) both bind to the RAR, only 9cRA binds to and activates RXR (2).
Like other members of the nuclear receptor superfamily, RXR has a conserved structure made up of five domains, designated N terminus to C terminus as A to E. These domains have distinct functions and can act independently. The Nterminal region (domains A and B) contains an autonomously functioning region called activation function 1 (AF-1), which is involved in ligand-independent transcriptional transactivation and is not highly conserved between receptors. Domain C, which contains two zinc binding motifs (zinc finger-like), corresponds to the core of the DNA binding domain responsible for recognition of cognate response elements. Domain D (hinge region) is involved in ligand-induced functional changes and in the binding of receptors to co-repressors. Domain E, which is moderately conserved, is thought to be involved in ligand binding, contains ligand-dependent activation function 2 (AF-2), and provides a surface for dimerization (4).
Both the RAR and RXR have three subtypes, ␣, ␤, and ␥, characterized by differences in the AF-1 domain (5). They usually form an RAR/RXR heterodimer and exert transcriptionregulating activity by binding to a specific DNA sequence known as the RAR-responsive element in the promoter regions of target genes (6). RXR also forms heterodimers with several other members of the nuclear receptor superfamily and transactivates or inhibits target gene promoters by binding to a specific DNA sequence known as the hormone-responsive element (7). RXR can also form homodimers, which transduce signals by binding to a specific DNA sequence, known as the RXR-responsive element (RXRE), in the promoter regions of target genes (7)(8)(9). The transcriptional regulatory activity of these hetero-or homodimers is modulated by a variety of transcriptional co-activators and co-repressors, depending on the cell and tissue type involved (7). The ligand-dependent transcriptional activity is switched on by the binding of ligand to these retinoid receptors, which induces a conformational change, leading to either the dissociation of co-repressors from or the binding of co-activators to RAR/RXR or RXR/RXR dimers. The co-activators are components of multiprotein complexes with histone acetyltransferase activity that acetylate the N-terminal region of histones, leading to nucleosomal repulsion and chromatin decondensation, which is thought to be indispensable for transcriptional activation by the retinoid receptors (6). In contrast, little is known about ligand-independent transcriptional activation.
The retinoid receptors play a pivotal role in preventing carcinogenic processes in various types of cancers (10 -15). We have shown that normal functioning of RXR␣ is necessary to suppress aberrant growth and induce apoptosis in hepatocellular carcinoma cells (16 -19). Phosphorylation of RXR␣ reduces RXR␣-dependent transactivation and is closely related to hepatocarcinogenesis (20), and restoration of RXR␣ activity, therefore, provides a key target for the prevention of liver cancer (21).
We screened for proteins that bind to RXR␣ and found a really interesting new gene (RING) finger protein, RING finger protein 8 (RNF8), recently isolated as a protein of unknown physiological function, which binds to the ubiquitin (Ub)-conjugating enzyme, E2 (UBE2E2) (22). RNF8 contains the forkhead-associated domain, a phosphopeptide binding motif, in its N-terminal region and a RING finger domain in the C-terminal region (we searched the Conserved Domain Data base for conserved domain RNF8 at www.ncbi.nlm.nih.gov:80/Structure/ cdd/cdd.shtml). RING finger proteins are involved in the ubiquitination of short-lived proteins (23)(24)(25) and in the modulation of transcriptional activation by certain nuclear receptors (26 -28). Although RNF8 binds to UBE2E2, which promotes Ub/ proteasome-mediated degradation of RXR␣ (29), here we have found that RNF8 did not participate in Ub/proteasome-mediated degradation of RXR␣ but, rather, as an enhancer of transactivation activity via RXR␣.
N-terminal polyhistidine His 6 Xpress-tagged WT RNF8 and its deletion mutants and C403S point mutant (Cys-403 mutated to Ser), presented in Fig. 1B, were constructed as described previously (22). The deletion and point mutations were confirmed by sequencing.
Expression vectors encoding fusion proteins consisting of the maltose-binding protein (MBP) and WT RXR␣ or its ⌬A/B mutant were constructed by subcloning cDNA encoding WT RXR␣ or the ⌬A/B mutant, amplified by PCR using, respectively, pAS2-WT or ⌬A/B RXR␣ as template, into the pMAL-p2X vector (New England Biolabs, Beverly, MA).
Enhanced green fluorescent protein (EGFP)-RXR␣ was constructed by subcloning PCR-amplified RXR␣ cDNA with BglII sites attached at both ends into the BglII/BamH1 sites of the pEGFP-C1 vector (Clontech). Blue fluorescent protein (BFP)-RNF8 was generated by cloning the BglII-BglII fragment from pACT2-RNF8 into the BamHI-BamHI site of the pQBI50-fC1 vector (Qbiogene, Carlsbad, CA). The BFP-40-AA-EGFP plasmid encoding EGFP coupled to BFP by a 40-amino acid linker was prepared by ligation of EGFP cDNA into the BFP vector, placing the EGFP-coding sequence in the BFP-reading frame.
Yeast Two-hybrid Assay-A yeast two-hybrid library was screened to isolate cDNAs encoding RXR␣-binding proteins. A bait plasmid, constructed by subcloning the open reading frame of RXR␣ into the pAS2-1 vector as a GAL4 DNA binding domain fusion (pAS2-RXR␣), as described above, was used to identify binding proteins from a human liver cDNA activation domain fusion library in the pACT2 vector. The yeast strain PJ69-2A was transformed with both plasmids. Positive clones were first selected on synthetic dropout (SD) medium in the absence of two nutrients (Leu and Trp). The colonies obtained were resuspended in water at equal densities, spotted onto SD medium in the absence of four nutrients (adenine, His, Leu, and Trp), and scored for differential growth after 3-5 days. The results were confirmed by a quantitative assay testing for ␤-galactosidase activity in the yeast strain Y-187. The two-hybrid system was also used to assess the binding of WT RNF8 to various RXR␣ deletion mutants (⌬A/B, ⌬E terminus, A/B/C, A/B, 1-28, and ⌬A/B/C) and of WT RXR␣ to various RNF8 deletion (1-440, 1-403, ⌬N, and 403-485) or point (C403S) mutants.
Quantitative ␤-Galactosidase Assay-␤-Galactosidase activity in the yeast strain, Y-187, was determined by liquid culture as a quantitative assay. This assay was carried out using o-nitrophenyl-␤-D-galactopyranoside (Sigma) as the substrate. Briefly, 2 ml of overnight culture grown in liquid SD selection medium was used to inoculate 8 ml of yeast complete medium. After 3-5 h of growth (A 600 ϭ 0.5-0.8) in the absence or presence of 1 M 9cRA, 1.5-ml aliquots of culture were placed in microcentrifuge tubes, and cells were collected by brief centrifugation. Each cell pellet was washed with 1.5 ml of Z buffer (60 mM Na 2 HPO 4 , 40 mM NaH 2 PO 4 , 10 mM KCl, 1 mM MgSO 4 , pH 7) and resuspended in 150 l of Z buffer. 100 l of resuspended pellets was transferred in a new 1.5-ml microcentrifuge tube. Cells were then frozen in liquid nitrogen and thawed at 37°C in a water bath 3 times to break open the yeast cells. Then 0.7 ml of Z buffer containing 0.27% (v/v) ␤-mercaptoethanol was added to the tube and mixed well followed by 160 l of o-nitrophenyl-␤-D-galactopyranoside in Z buffer (4 mg/ml, pH 7). The tube was incubated at 30°C until a yellow color developed; 0.4 ml of Na 2 CO 3 was then added to stop the reaction, and the reaction time was recorded. The tubes were centrifuged for 10 min, and the absorbance of the supernatant at 420 nm was measured. The ␤-galactosidase activity was calculated by the equation, ␤-galactosidase units (u) ϭ 1000 ϫ A 420 /(t ϫ V ϫ A 600 ), where t is elapsed time (in min) of incubation, V is 0.1 ml ϫ concentration factor (10 in this experiment), and A 600 is absorbance at 600 nm of 1 ml of culture.
In Vitro Binding Assay-The MBP-WT RXR␣ and MBP-⌬A/B RXR␣ fusion proteins were expressed in Escherichia coli and purified on amylose resin (New England BioLabs, Beverly, MA). COS7 cells overexpressing His-and Xpress sequence-tagged WT RNF8 or its N-terminal deletion mutant (⌬N) were harvested and lysed in radioimmune precipitation assay buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.1% Nonidet P-40, 0.1% SDS, 0.5% sodium deoxycholate, 1 mM phenylmethylsulfonyl fluoride, and 1 mM ZnCl 2 ). Purified MBP-RXR␣s (500 ng) were added to the lysates, and the mixture was incubated for 2 h at 4°C in the absence or presence of 1 M 9cRA (Sigma), then amylose resin was added, and incubation was continued for 30 min at 4°C. The resin was then washed with radioimmune precipitation assay buffer and boiled in SDS sample buffer, and the eluted proteins were separated by 8% SDS-PAGE and analyzed by Western blotting using anti-Xpress antibody (Invitrogen), which recognizes the DLYDDDDK Xpress sequence adjacent to His 6 .
Cell Culture and Transfection-COS7 cells were maintained in Dulbecco's minimum essential medium (DMEM) containing penicillin (100 units/ml), streptomycin (100 g/ml), and 10% fetal calf serum (FCS). The human colorectal carcinoma-derived cells, Caco-2 (Japanese Cancer Research Resources Bank, Tokyo, Japan), were maintained in DMEM/F-12 (Invitrogen) containing penicillin (100 units/ml), streptomycin (100 g/ml), and 10% FCS. Transfections were performed using LipofectAMINE Plus (Invitrogen) according to the manufacturer's protocol. Briefly, before transfection Caco-2 cells were cultured for 24 h in phenol red-and FCS-free medium supplemented with ITS Supplement (Invitrogen) to give final concentrations of 10 mg/liter insulin, 5.5 mg/l transferrin, and 6.7 g/liter sodium selenite. The medium was changed to phenol red-free DMEM containing 10% charcoal-stripped FCS for 6 h, then the cells were transfected and harvested 48 h later. In some experiments, 9cRA (1 M) or vehicle (ethanol, final concentration Ͻ0.05%) was added to the medium 24 h after transfection.
Immunostaining-COS7 cells, plated on coverslips in 35-mm dishes, were transfected with plasmids encoding WT Myc-His-RXR␣ (1 g/dish) and WT or mutant (C403S or ⌬N) His-Xpress-RNF8 (2 g/dish, respectively) as described above, then the cells were washed in phosphatebuffered saline (PBS), pH 7.4, fixed for 10 min in 4% paraformaldehyde in PBS, washed twice in PBS, and permeabilized for 3 min with 0.2% Triton X-100 in PBS. After blocking with 0.1% BSA in PBS, the coverslips were incubated for 2 h with mouse monoclonal anti-Xpress antibody (Invitrogen) and rabbit polyclonal anti-myc antibodies (Santa Cruz Biotechnology, Santa Cruz, CA), washed with PBS, and then incubated for 2 h with Alexa 488-conjugated goat anti-mouse antibody and Alexa 546-conjugated goat anti-rabbit antibodies (Molecular Probes, Eugene, OR), respectively. All procedures were at room temperature. Images were acquired using an Olympus IX70 inverted microscope equipped with differential interference contrast (DIC) optics. Both the fluorescence and DIC images were captured as described previously (22) using a C5985 CCD camera with on-chip integration (Hamamatsu Photonics, Hamamatsu, Japan) and a Macintosh 8500 computer running C5985 imaging software. The fluorescence light source was a 75-W xenon arc lamp shuttered using a custom-made shutter unit so that cells were only illuminated during image acquisition.
Fluorescence Resonance Energy Transfer (FRET)-COS7 cells, grown on 35-mm glass-bottom culture dishes (Matsunami, Osaka, Japan), were transfected with plasmids encoding BFP, BFP-RNF8, EGFP-RXR␣, BFP-EGFP tandem, or a combination of EGFP-RXR␣ and either BFP-RNF8 or BFP. To obtain images of living cells, the culture dishes were observed using an IX70 inverted microscope; the imaging system used was the same as that used for immunostaining experiments. The filter sets for discriminating between BFP and EGFP fluorescence were obtained from Omega Optical Inc. (Brattleboro, VT). For BFP detection, cells were viewed using a filter set with a 365/50 nm excitation filter, a 400-nm dichroic beam splitter, and a 450/65-nm emission filter (BFP channel). EGFP was detected using a filter set with a 475/40-nm excitation filter, a 505-nm dichroic beam splitter, and a 535/45-nm emission filter (EGFP channel). The filters used for FRET were a 365/50-nm excitation filter, a 400-nm dichroic beam splitter, and a 535/45-nm emission filter (FRET channel). The emission filters for the BFP channel and the FRET channel were switched using a filter changer (Lambda 10 -2; Sutter Instrument, Novato, CA) coupled between the output port of the microscope and the C5985 CCD camera. To measure FRET, three images were acquired sequentially through the BFP, FRET, and EGFP channels. FRET quantification between the two fluorophores was performed using the normalized FRET value (N FRET ), as described previously by Xia and Liu (30). Because the amounts of BFP donor and EGFP acceptor vary from cell to cell in transient transfections and the intensity of energy transfer is related to the number of EGFP acceptor/BFP donor pairs, it was necessary to normalize the FRET to the expression levels of BFP and EGFP. The N FRET value was calculated on a pixel-by-pixel basis for the entire image using the equation, where I FRET , I EGFP , and I BFP correspond to background-subtracted images of cells co-expressing EGFP and BFP acquired through the FRET, EGFP, and BFP channels, respectively. Constants a and b represent the fraction of bleed-through of BFP and EGFP fluorescence, respectively, through the FRET channel and were determined using cells expressing either BFP-RNF8 or EGFP-RXR␣. In our system ϳ24% of the BFP signal and 2% of the EGFP signal was detected in the FRET channel. N FRET images are presented in pseudocolor mode, according to a temperature-based lookup table, with blue (cold) indicating low values and red (hot) indicating high values. For quantitative comparison, the N FRET for each pixel in nuclear areas was calculated, and the number of pixels with a given N FRET value was counted and plotted as a distribution histogram. All calculations were performed using the public domain program, Image J Version 1.29 (developed at NIH), with plug-ins for Calculator Plus and SixteenBit Histogram.
In Vivo Ubiquitination Assays-Ubiquitination of proteins requires the covalent attachment of 8.6-kDa Ub to multiple lysine residues, forming poly-Ub chains bound to target proteins, and can be seen as a ladder of high molecular mass species on SDS-polyacrylamide gels (29). COS7 cells were transfected with various combinations of plasmids encoding RXR␣ (1 g), HA-tagged Ub (kindly provided by Dr. D. Bohmann, European Molecular Biology Laboratory) (3 g), or His-RNF8 (WT or C403S) (3 g). MG132 (5 M; Calbiochem), a potent proteasomal inhibitor, was added 24 h after transfection, and the cells were harvested 27-30 h after transfection. Cells were sonicated for 30 s in lysis buffer (100 mM sodium phosphate, pH 8.0, containing 6 M guanidine-HCl and 10 mM imidazole), then RXR␣ was immunoprecipitated using anti-RXR␣ antibodies (DN197; Santa Cruz Biotechnology) and subjected to SDS-PAGE, and ubiquitinated RXR␣ was detected by Western blotting using anti-HA antibodies (12CA5; Roche Applied Science) as described previously (29).
Luciferase Reporter Assays-A reporter plasmid, tk-CRBPII-Luc, which contains multiple copies of the DR1 type RXRE sequences within the rat cellular retinol-binding protein type II (CRBPII) promoter upstream of luciferase cDNA, was kindly provided by the late Dr. K. Umesono (Kyoto University, Kyoto, Japan). COS7 cells, seeded in 12well plates, were transfected with WT RXR␣-expressing plasmid (100 ng/well), WT, or mutant (C403S or ⌬N) RNF8-expressing plasmid (10 or 100 ng/well), and tk-CRBPII-Luc reporter (350 ng/well) was transfected with pRL-CMV-Luc (Renilla luciferase, 50 ng/well, Promega, Madison, WI) as an internal standard to normalize transfection efficiency. After exposure of the cells for 24 h to the transfection mixture, the medium was replaced with DMEM containing 10% charcoal-stripped FCS in the presence or absence of 1 M 9cRA. Forty-eight hours after transfection cell lysates were prepared, and the luciferase activity of each cell lysate was measured using a LumiCount microplate luminometer (Packard Instrument Co.). Changes in firefly luciferase activity were calculated and plotted after normalization to changes in Renilla luciferase activity in the same sample.
Reverse Transcription-PCR-Total RNA was extracted from Caco-2 cells using ISOGEN (WAKO Pure Chemical, Osaka, Japan) and isopropanol precipitation, then 1 g was reverse-transcribed using a High Fidelity RNA PCR Kit (TAKARA Biomedicals, Tokyo, Japan) and an oligo(dT) primer according to the manufacturer's protocol. mRNAs for RNF8, RXR␣, CRBPII, RAR␤, and glyceraldehyde 3-phosphate dehydrogenase (G3PDH) were detected using the primer pairs derived from the human sequences shown in Table I. For RNF8, RXR␣, RAR␤, and G3PDH, the PCR conditions were 1 cycle of 94°C for 5 min, 25 (for RAR␤ and G3PDH) or 30 (for RNF8 and RXR␣) cycles of denaturation at 94°C for 30 s, annealing at 56 -60°C for 30 s, and extension at 72°C for 1 min, and 1 cycle of 72°C for 10 min; for CRBPII, conditions were 1 cycle of 94°C for 5 min, 22 cycles of denaturation at 94°C for 30 s, annealing at 56°C for 30 s, and extension at 72°C for 30 s, and 1 cycle of 72°C for 10 min. The PCR products were run on 2% agarose gels and visualized by ethidium bromide staining.

RESULTS
Discovery of RNF8 as an RXR␣-binding Protein by Screening a Human Liver cDNA Library-Using the yeast two-hybrid screening system and a human liver library, we picked out a cDNA encoding an RXR␣-binding protein and identified it as the RING finger protein, RNF8, by comparing the sequence with the data base in the DNA Data Bank of Japan. RNF8 was originally isolated as a partner of the ubiquitin-conjugating enzyme, UBE2E2 (22). By testing the binding of various mu- Enhanced Transactivation via RXR␣-RNF8 Interaction tants of both RNF8 and RXR␣ (shown in Fig. 1), we determined the interacting sites in these molecules in the presence or absence of 9cRA. Both ␤-galactosidase and growth assays demonstrated that all RXR␣ constructs containing residues 1-131, except for the ⌬E terminus, bound well to RNF8, showing that this region, which includes the AF-1 region, was important for binding (Fig. 1A). A/B/C was more effective than ⌬E terminus (compare columns 3 and 4). It was to be noted that of the constructs containing the AF-1 region, the ⌬E terminus was the only one having the D domain, also called hinge region, and this domain was responsible for the folded structure of RXR. Accordingly, it was reasonable to consider that lack of the D domain led to exposure of RNF8 binding site (AF-1). Construct 1-28, containing only the N-terminal 28 amino acids, still showed some binding, but the greater binding seen with construct A/B (residues 1-131) (compare column 5 and 6) showed that important binding residues are also present in region 29 -131. The greater binding seen with construct ⌬A/B/C (com-pare columns 6 and 7), which in addition to residues 1-28 contains all residues from 198 to the C terminus, suggests that residues in the C-terminal half of the molecule also contribute to the binding of RNF8. Similarly, as shown in Fig. 1B, the N-terminal 209 residues of RNF8 (column 11) were required for binding to RXR␣, whereas deletion (column 10) or disruption (column 13) of the RING finger domain or deletion of the C-terminal 45 amino acids (column 9) had little effect. The fact that the binding of RXR␣ to RNF8 was mediated by the N-terminal region of each of the molecules was corroborated by pull-down experiments (Fig. 2). Purified WT or ⌬A/B mutant RXR␣-MBP fusion proteins or MBP from overexpressing bacteria were incubated with cell lysates prepared from His-Xpress-RNF8-overexpressing COS7 cells in the absence ( Fig. 2A, odd lanes) or presence (even lanes) of 9cRA, and then proteins bound to MBP or MBP fusion protein were pulled down using amylose resins and analyzed by Western blots using anti-Xpress antibody ( Fig. 2A, upper panel). His-Xpress-

Enhanced Transactivation via RXR␣-RNF8 Interaction
RNF8, which did not bind to MBP alone (lanes 1 and 2), was pulled down by MBP-WT RXR␣ (lanes 3 and 4) but not to any significant degree by MBP-⌬A/B RXR␣ (lanes 5 and 6), irrespective of the presence of 9cRA. The amounts of WT RXR␣ and ⌬A/B RXR␣ added were shown to be similar ( Fig. 2A, lower  panel, lanes 3-6). In the converse experiment (Fig. 2B, upper  panel), the MBP-WT RXR␣ fusion protein, which did not bind to the His-Xpress tag alone (lanes 7 and 8), bound to His-tagged WT RNF8 (lanes 9 and 10) but not to His-tagged ⌬N RNF8 (lanes 11 and 12). 9cRA did not affect the interaction between WT RNF8 and RXR␣. The levels of WT or ⌬N RNF8 in the cell lysate assessed by Western blotting are shown in the lower panel of Fig. 2B (lanes 9-12). Subcellular Localization of RNF8 and Its Mutants in COS7 Cells-We then analyzed the subcellular localization of WT RNF8 and its mutants as well as RXR␣ in COS7 cells by immunostaining (Fig. 3). We first confirmed that the protein expression levels of WT (lane 1), C403S (lane 2), and ⌬N (lane 3) His-Xpress-RNF8 were the same by Western blotting (Fig.  3A). DIC and fluorescence microscopy showed that, as reported previously (22), WT RNF8 was expressed predominantly in the nucleus (Fig. 3B, panels a and c), whereas C403S RNF8 and ⌬N RNF8 were expressed predominantly in the cytoplasm (panels d and f, and panels g and i, respectively). RXR␣ is localized mainly in the nucleus as reported previously (31), and neither type (WT, C403S, or ⌬N) of RNF8 affected the localization of RXR␣ (Fig. 3B, panels b, e, and h). These results suggested that RNF8 and RXR␣ are co-localized in the nucleus, potentially as a complex.
Visualization of RNF8-RXR␣ Binding in Living Cells Using FRET-FRET microscopy was used to visualize the binding between EGFP-fused RXR␣ (acceptor) and BFP-fused RNF8 (donor) overexpressed in living COS7 cells (Fig. 4). The EGFP and BFP combination has excitation and emission properties favorable for FRET, since the emission wavelength of BFP partially overlaps with the excitation wavelength of EGFP (32,33), and binding of the two fluorescence-tagged proteins can be detected by measuring the emission intensity of the acceptor fluorophore after exciting the donor fluorophore. Fig. 4A shows FRET image analysis of cells transiently transfected with both BFP-RNF8 and EGFP-RXR␣ and of positive and negative control cells. COS7 cells transfected with EGFP-WT RXR␣ both with and without BFP-WT RNF8 showed nuclear localization of fluorescence (Fig. 4A, panels a and d). Simultaneous expression of BFP-WT RNF8 resulted in localization of RNF8 predominantly to the nucleus (Fig. 4A, panel b). Thus, the presence of the BFP and EGFP moieties in these chimeric proteins Enhanced Transactivation via RXR␣-RNF8 Interaction did not affect their nuclear localization. As clearly indicated in the N FRET image (Fig. 4A, panel c), a strong FRET signal was observed in the nucleus when BFP-WT RNF8 and EGFP-WT RXR␣ were co-expressed, showing direct binding between the two proteins in the living cell. In negative control cells expressing EGFP-WT RXR␣ and BFP (Fig. 4A, panels d and e), the N FRET image did not show a significant FRET signal (panel f). As a positive control, we analyzed cells expressing the BFP-EGFP tandem, in which BFP and EGFP are linked by a 40amino acid peptide. The BFP-EGFP tandem was uniformly distributed (Fig. 4A, panels g and h), and a high N FRET value, indicating intramolecular FRET, was observed throughout the cell (panel i), showing that the method was reliable. Again, the binding between RNF8 and RXR␣ did not require addition of 9cRA.
Quantitative analysis of the N FRET values in the nucleus was performed on several cells, and the results are presented as pixel distribution histograms (Fig. 4B). Most pixels from cells cotransfected with BFP-WT RNF8 and EGFP-WT RXR␣ (panel j) showed N FRET values much higher than those in negative control cells (panel k), and the N FRET value distribution in the BFP-RNF8/EGFP-RXR␣-cotransfected cells overlapped with that for the positive control cells (panel l). These results showed efficient transfer of energy in the nucleus when both proteins were present, supporting direct binding between RNF8 and RXR␣ in the living cell nucleus.
RNF8 Does Not Affect RXR␣ Ubiquitination-We previously reported that UBE2E2 acts as a Ub-conjugating enzyme during RXR␣ ubiquitination in hepatoma cells (29) and that RNF8 binds to UBE2E2 (22). We, therefore, examined the effects of RNF8 on RXR␣ ubiquitination in COS7 cells transfected with plasmids expressing RXR␣-and HA-tagged Ub in the presence or absence of plasmids expressing His-Xpress-tagged WT or C403S RNF8 or control empty vector by immunoprecipitation of the lysate with anti-RXR␣ antibodies and analysis of the precipitate by SDS-PAGE and Western blotting using anti-HA antibody. As shown in Fig. 5, cells expressing RXR␣ and HAtagged Ub showed multiple Ub conjugation to RXR␣, seen as a ladder of high molecular mass species of ubiquitinated RXR␣ (Fig. 5, compare lanes 1 and 2). Unexpectedly, RXR␣ ubiquitination was neither enhanced by simultaneous overexpression of WT RNF8 (compare lanes 3 and 4) nor reduced by simultaneous overexpression of C403S RNF8 (compare lanes 3 and 5), showing that RNF8 was not involved in RXR␣ ubiquitination by endogenous UBE2E2.
RNF8 Enhances Promoter Transactivation via RXR␣ and Expression of Its Downstream Gene(s)-Because some RING finger proteins have been reported to modulate transactivation by certain nuclear receptors (26 -28), we examined the effects of RNF8 on the RXR␣-driven transactivation of the luciferase reporter gene, CRBPII-Luc, the promoter of which contains RXRE sequences. COS7 cells were cotransfected with WT RXR␣, CRBPII-Luc, and increasing amounts of WT RNF8 expression vectors together with pRL-CMV-Luc as an internal control (Fig. 6). Transactivation of this reporter was dependent on both endogenous and exogenous RXR and 9cRA (Fig. 6A,  compare column 1 and columns 4, 7, and 10). Interestingly, WT-RNF8 enhanced the transcription in a dose-dependent manner in the absence of 9cRA (Fig. 6A, columns 1-3), and the effect was more obvious when WT RXR␣ was overexpressed (columns 4 -6). However, the enhancement of transactivation by 1 M 9cRA in the presence of RNF8 was limited to ϳ2.5-fold compared with the level without the ligand, both in the absence (compare columns 3 and 9) and presence (compare columns 10 and 12) of exogenous RXR␣, whereas in the absence of RNF8 9cRA up-regulated the CRBPII promoter activity by 4 -5-fold both in the absence (compare columns 1 and 7) and presence (compare columns 4 and 10) of exogenous RXR␣. C403S RNF8 or ⌬N RNF8 did not increase RXR␣-dependent transcription either in the absence (Fig. 6B columns 1-10) or presence (columns [11][12][13][14][15][16][17][18][19][20] of 9cRA, irrespective of cotransfection with WT RXR␣ (compare columns 1-5 and 6 -10, and columns 11-15 and 16 -20). These observations corroborate the result that C403S and ⌬N RNF8 did not enter the nucleus (Fig. 3B).
Finally, we examined whether enhanced RNF8-mediated transactivation of the CRBPII promoter via the RXRE has biological relevance using colon carcinoma-derived Caco-2 cells that constitutively express CRBPII (34). Caco-2 cells also constitutively express endogenous RXR␣ and RNF8 (data not shown). Overexpression of WT RNF8 resulted in a significant 2.1-fold increase in CRBPII mRNA levels (Fig. 7A, lane 2), whereas overexpression of C403S or ⌬N RNF8 had no effect (lanes 3 and 4, respectively). Moreover, because RXR forms heterodimers with RAR and is essential for the regulation of RXR/RAR-mediated gene expression (1-3), we also examined the effect of RNF8 on RXR/RAR-regulated expression of RAR␤ (Fig. 7B), which is endogenously expressed in Caco-2 cells. WT RNF8 up-regulated RAR␤, which has typical RAR-responsive element in its promoter region (Fig. 7B, lane 6). Neither the C403S nor ⌬N RNF8 affected the expression of RAR␤ (Fig. 7B,   lanes 7 and 8, respectively). Similar results were obtained when the effects of WT and mutant RNF8 on the expression of CRBPII and RAR␤ were examined in the presence of 9cRA (data not shown). These results suggest that RNF8 may be an important factor promoting RXR-mediated as well as RXR/ RAR-mediated gene expression by physical interaction. DISCUSSION We demonstrated here that a novel RING finger protein, RNF8, colocalized with RXR␣ in the nucleus (Fig. 3B) and bound to RXR␣ (Fig. 4) through the N-terminal domain of each molecule (Figs. 1 and 2), leading to ligand-independent transactivation of the target genes (Figs. 6 and 7). A ligand for RXR␣, 9cRA, did not affect the interaction between RNF8 and RXR␣ ( Figs. 1 and 2), and localization of RNF8 in the nucleus required both a normal RING finger structure and the Nterminal region (Fig. 3B). Previous studies have demonstrated that the RING finger structure plays a crucial role in the nuclear localization of certain transcriptional regulators. For instance, both the tumor suppresser, BRCA1, and its associated protein, BARD1, contain N-terminal RING structures that are essential for their nuclear transport (35) and, thus, for their regulation of gene transcription (36). In addition, the N-terminal region of RNF8 contains the forkhead-associated domain that was originally identified as a phosphopeptide binding domain in a group of forkhead transcription factors but is now known to be present in a wide variety of proteins (37). Many forkhead-associated domain-containing proteins are found in the nucleus and are involved in DNA repair, cell cycle arrest, and pre-mRNA processing (37).
Because RNF8 is reported to bind to the Ub-conjugating FIG. 6. RNF8 increases transactivation of CRBPII promoter via RXR␣. COS7 cells were cotransfected with the tk-CRBPII-Luc reporter with or without the WT RXR␣-expressing plasmid plus either increasing amounts (0, 10, and 100 ng) of the plasmid encoding WT RNF8 (A) or mutant (C403S or ⌬N) RNF8s (B) and pRL-CMV-Luc (Renilla luciferase) as an internal standard. A, the amount of cDNA used was maintained constant by the addition of empty His-Xpress vector (). Twenty-four hours later, the medium was replaced with DMEM containing charcoalstripped FCS in the absence (A, columns 1-6, and B, columns [1][2][3][4][5][6][7][8][9][10] or presence (A, columns 7-12, and B, columns 11-20) of 1 M 9cRA, and incubation was continued for another 24 h, then the luciferase activity in each cell lysate was measured and plotted as the fold induction compared with that in cells transfected with empty vector instead of the RNF8-expressing plasmid (column 1 in panels A and B) after normalization to the Renilla luciferase activity. The results shown are the mean Ϯ S.D. for a single experiment performed in triplicate and are representative of those for three independent experiments. A, *, **, †, † †, p Ͻ 0.05, significant difference compared with columns 1, 4, 7, and 10, respectively enzyme, UBE2E2 (22), which is involved in the ubiquitination of RXR␣ (29), we speculated that RNF8 might enhance UBE2E2dependent RXR␣ ubiquitination by connecting the two molecules. However, this was found not to be the case, since no difference in the degree of RXR␣ ubiquitination by endogenous UBE2E2 was seen in control COS7 cells and RNF8-overexpressing cells (Fig. 5).
We found that RNF8 promoted transactivation via RXRE of the CRBP II promoter (Fig. 6) and an increase in CRBP II mRNA levels (Fig. 7A) via the RXR␣ without the addition of the RXR␣ ligand, 9cRA. That CRBPII is a key protein in the absorption of retinoids by intestinal cells suggests possible involvement of RNF8 in retinoid absorption. Moreover, RNF8 also regulated the RXR/RAR heterodimer-mediated gene, RAR␤ (Fig. 7B), which is related to cell differentiation and growth suppression (38). Similar transactivation-promoting activity of RING finger proteins has been reported for ARA54/ SNURF and TIF1␤, which activate transcription of the androgen receptor and the glucocorticoid receptor, respectively (26 -28).
The RXR contains two transcription activation domains, the ligand-independent AF-1 site, located in the N-terminal A/B domains, and the ligand-dependent AF-2 site, located in the C-terminal ligand binding domain (4). These domains participate in the transactivation of target genes, leading to the modulation of various physiological processes. Binding of 9cRA to RXR␣ causes an interaction between AF-1 and AF-2, leading to exposure of the AF-1 region due to a conformational change of RXR␣ (39). Such a conformational change is thought to be necessary for the recruitment of co-activators, including p300 and CBP (40 -42). However, the effect of RNF8 in increasing RXRE-mediated transcriptional activation was relatively weak (Fig. 6A) compared with that of the co-activators. In addition, the increment induced by RNF8 appeared similar between in the absence and presence of ligand. These results suggest different mechanisms of action between a ligand, 9cRA, and RNF8. RNF8 might cause some conformational change of RXR␣ that enhances transactivation independent of 9cRA. We are now examining whether p300 and CBP play a role in the RNF8-mediated enhancement of RXRE-dependent transactivation and testing the possibility that RNF8 might link RXR␣ to these co-activators. In addition, it should be noted that we used cells expressing endogenous RNF8, and it is possible that cells lacking endogenous RNF8 expression might show greater induction of RXRE-dependent transactivation on transfection with RNF8, a possibility we are now exploring.
In summary, we have described a novel function of RNF8, regulating RXR␣-mediated gene expression, as a result of the direct binding of RNF8 to RXR␣. Because both RNF8 and RXR␣ are reported to be expressed in a variety of tissues (43)(44)(45), we are now testing whether this interaction yields a similar biological result across these tissues.