Vinexin β Interacts with the Non-phosphorylated AF-1 Domain of Retinoid Receptor γ (RARγ) and Represses RARγ-mediated Transcription*

Nuclear retinoic acid receptors (RARs) are ligand-dependent transcription factors that regulate the expression of retinoic acid target genes. Although the importance of RAR phosphorylation in their N-terminal domain is clearly established, the underlying mechanism for the phosphorylation-dependent transcriptional activity of the receptors had not been elucidated yet. Here, using a yeast two-hybrid system, we report the isolation of vinexin β as a new cofactor that interacts with the N-terminal A/B domain of the RARγ isotype. Vinexin β is a multiple SH3 motif-containing protein associated with the cytoskeleton and also present in the nucleus. We demonstrate that vinexin β colocalizes with RARγ in the nucleus and interacts with the non-phosphorylated form of the AF-1 domain of RARγ. We also show that this interaction is prevented upon phosphorylation of the AF-1 domain. Using F9 cells stably overexpressing vinexin β or vinexin knockdown by RNA interference, we demonstrate that vinexin β is an inhibitor of RARγ-mediated transcription. We propose a model in which phosphorylation of the AF-1 domain controls RARγ-mediated transcription through triggering the dissociation of vinexin β.

Retinoic acid (RA), 1 the most potent biologically active metabolite of vitamin A, influences the proliferation, differentiation, and apoptosis of a variety of cell types through modifications of expression of subsets of RA target genes (1)(2)(3). The effects of RA are mediated by two classes of nuclear receptors, the retinoic acid receptors (RAR␣, RAR␤, and RAR␥) and the retinoid X receptors (RXR␣, RXR␤, and RXR␥), which function as ligand-dependent heterodimeric RAR/RXR transcription activators (4 -6). RARs and RXRs exhibit a conserved modular structure (see Fig. 1A) with a central DNA-binding domain and two activation domains (AF-1 and AF-2) that synergize for the activation of RA target genes.
Ligand-induced conformational changes in the AF-2 domain of RARs bound at cognate response elements (RA response elements) located in the promoter of target genes cause the dynamic, coordinated, and combinatorial recruitment of coactivators and large complexes with chromatin-modifying and chromatin-remodeling activity, which will decompact repressive chromatin to allow positioning of the transcription machinery at the promoter (2,7). Other proteins are also recruited and serve as connections with the transcription machinery. In line with this, RARs interact with the general transcription factor TFIIH (8,9). This results in the phosphorylation of one residue located in their N-terminal AF-1 domain (Ser 77 in RAR␣1, Ser 79 in RAR␥1, and Ser 68 in RAR␥2) (see Fig. 1A) by the Cdk7 subunit of TFIIH, which has cyclin H-dependent kinase activity. This phosphorylation process, which has been extensively studied especially in the case of RAR␣ (10), plays a critical role in the response to RA.
However, in the particular case of the RAR␥ isotype, phosphorylation by TFIIH, although necessary, is not sufficient. Indeed, to be transcriptionally active, RAR␥ needs to be also phosphorylated at an additional nearby residue (Ser 77 in RAR␥1 and Ser 66 in RAR␥2) (see Fig. 1A) by p38 MAPK subsequent to its activation by RA (11,12). As phosphorylation by both TFIIH and p38 MAPK is required for the activation of RAR␥-controlled genes (2), we hypothesized that phosphorylation of the N-terminal AF-1 domain might regulate the dissociation and/or association of proteins involved in blocking or stimulating RAR␥ activity. With this aim, we performed yeast two-hybrid screening experiments to characterize new cofactors interacting with the phosphorylated or non-phosphorylated forms of the AF-1 domain of RAR␥ and therefore regulating RAR␥ activity. By this approach, we isolated vinexin as a partner for the non-phosphorylated AF-1 domain of RAR␥. Vinexin is a recently identified cytoskeletal protein that exists as two isoforms, vinexin ␣ and vinexin ␤ (13). Both proteins are devoid of any enzymatic activity, but regulate cell adhesion/ cytoskeleton organization as well as signal transduction pathways (13)(14)(15). They share a common C-terminal sequence containing three SH3 domains that bind proline-rich sequences (16). Vinexin ␣ has an additional N-terminal sequence containing a sorbin homology (SoHo) domain that mediates the translocation of the protein to lipid rafts (17).
In this study, we report that vinexin ␤ colocalizes with RAR␥ in the nucleus, interacts with the non-phosphorylated N-terminal AF-1 domain of RAR␥, and represses RAR␥-mediated transcription. We also demonstrate that phosphorylation of the AF-1 domain prevents the interaction of vinexin ␤ with RAR␥. Based on these results, we propose that the underlying mechanism for the phosphorylation-dependent transcriptional activity of RAR␥ involves, at least in part, the dissociation of vinexin ␤.
The AF-1 domain of hRAR␥1 with Ser 77 and Ser 79 substituted with alanines or glutamic acids was amplified by PCR from the corresponding pSG5 vectors and inserted into the XhoI/BamHI-digested pBTM116mod plasmid, which directs synthesis of LexA-DNA-binding domain fusion proteins in yeast. The AF-1 domain of either wild-type (WT) mRAR␣1 or mRAR␣1(S77A) was also inserted into the pBTM116mod plasmid following the same protocol.
The cDNA of vinexin ␤ was amplified by PCR and cloned into the pCX vector driven by the cytomegalovirus immediate-early enhancer and already containing the hemagglutinin, FLAG, and yellow fluorescent protein tags (a gift from T. Lerouge). All constructs were generated using standard cloning procedures and were verified by restriction enzyme analysis and automated DNA sequencing. 2 The DR5-tk-CAT and DR1-tk-CAT reporter constructs were described previously (18). The plasmids encoding FLAG-tagged vinexins ␣ and ␤ were provided by Dr. N. Kioka (13). All-trans-RA was from Sigma. The synthetic RAR␥ (BMS961) and pan-RXR (BMS649) agonists were gifts from Bristol-Myers Squibb Co.
Antibodies-Rabbit polyclonal antibodies raised against the F domain of RAR␥ (antibody RP␥(F)) and mouse monoclonal antibodies raised against the same F domain (mAb4␥(F)) or the N-terminal A domain (mAb441␥(A)) were as described (8,22). Rabbit polyclonal antibodies specific to RAR␥ phosphorylated at Ser 77 or Ser 79 were described previously (12). Anti-FLAG monoclonal antibody M2 (immobilized or not on agarose) was obtained from Sigma, and goat anti-␤-actin polyclonal antibody C-11 was from Santa Cruz Biotechnology Inc. Rabbit polyclonal antibodies against LexA were as described (23). Mouse monoclonal antibodies against vinexin were generated using a synthetic peptide corresponding to amino acids 211-223 of mouse vinexin ␤ according to standard procedures (22). Cy3-conjugated goat antimouse antibodies were from Amersham Biosciences, and Alexa Fluor 488-conjugated goat anti-rabbit antibodies from Molecular Probes, Inc.
Protein-Protein Interactions Using the Yeast Two-hybrid System-The yeast strain L40 was cotransformed with the plasmids encoding the LexA-RAR␥1(A/B) fusion protein and the VP16 acidic activation domain fused to the isolated interacting protein. The cells were grown overnight in selective liquid medium containing histidine, and a quantitative ␤-galactosidase assay was performed as described (24).
GST Pull-down Assays-The GST, GST-RAR␣1, GST-RAR␥1, GST-RXR␣1, and GST-RAR␥1(A/B) proteins were produced in Escherichia coli strain NB42 and purified on glutathione-Sepharose 4B beads (Amersham Biosciences) as described (25). Equimolar amounts of the GST fusion proteins bound to the beads were incubated in GST buffer (50 mM Tris-HCl (pH 8), 0.05% Nonidet P-40, 0.3 mM dithiothreitol, 10 mM MgCl 2 , and 5% glycerol) containing a protease inhibitor mixture and 150 -500 mM NaCl with COS-1 cell extracts expressing the FLAGvinexin protein. After three washes in GST buffer, bound proteins were recovered in SDS loading buffer, subjected to SDS-10% PAGE, and analyzed by immunoblotting. Cells, Transfections, Immunoprecipitations, and Chloramphenicol Acetyltransferase (CAT) Assays-COS-1 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 5% fetal calf serum and transiently transfected using the DMRIE-C reagent (Invitrogen) as described (11). After a 20-h incubation with DNA, the cells were washed and maintained for the indicated times in medium with or without ligand. Cells were harvested, and whole cell extracts were prepared in lysis buffer (25 mM Tris-HCl (pH 8), 1 mM EDTA, 1 mM EGTA, 150 mM NaCl, 1% Triton X-100, and protease inhibitor mixture). Where mentioned, immunoprecipitation was performed by incubation of the extracts with the indicated monoclonal antibodies in 50 mM Tris-HCl (pH 8.0) containing 100 mM NaCl, 10 mM MgCl 2 , 0.3 mM dithiothreitol, 5% glycerol, 0.05% Nonidet P-40, 0.5 mg/ml bovine serum albumin, and protease inhibitor mixture with protein G-Sepharose beads. Proteins with or without prior immunoprecipitation were resolved by 10% SDS-PAGE, electrotransferred to nitrocellulose membranes, immunoprobed, and detected by chemiluminescence according to the protocol of Amersham Biosciences. CAT assays were performed using the enzyme-linked immunosorbent assay method (Roche Diagnostics). All results were normalized to equal ␤-galactosidase activity and to the activity of each receptor in the absence of vinexin ␤ and without ligand.
F9 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum as described (26). F9 cells ablated for RAR␥ (RAR␥ Ϫ/Ϫ ) and re-expressing either WT RAR␥ or RAR␥ mutated at its phosphorylation sites were described previously (26 -28). To establish stable lines overexpressing vinexin ␤, F9 cells were electroporated with the pCX construct along with a plasmid conferring neomycin resistance. After 24 -36 h, the cells were selected with neomycin for 10 days as described (26,27) and analyzed for the expression of the transgene by quantitative reverse transcription (RT)-PCR and immunoblotting. Several clones were isolated and one, V␤(5), was selected and used in these experiments. Cytosolic and nuclear extracts were prepared as described (12).
Immunofluorescence-COS-1 cells cotransfected with the FLAG-vinexin (␣ or ␤) and RAR␥ expression vectors were seeded onto glass coverslips coated with 0.1% gelatin. After 16 h, the cells were fixed with 2% paraformaldehyde in phosphate-buffered saline, permeabilized with 0.1% Triton X-100, and saturated with 5% bovine serum albumin in phosphate-buffered saline. The cells were then incubated with rabbit polyclonal antibodies against RAR␥ (antibody RP␥(F)) and mouse monoclonal anti-FLAG antibodies, followed by Alexa Fluor 488-conjugated goat anti-rabbit and/or Cy3-conjugated goat anti-mouse secondary antibodies. Nuclei were counterstained with 4Ј,6-diamidino-2-phenylindole (Sigma), and the coverslips were mounted on glass slides. The cells were analyzed by fluorescence microscopy using an epifluorescence microscope or a confocal laser scanning microscope.
Small Interfering RNA (siRNA)-The 19-nucleotide RNA oligonucleotides corresponding to vinexin with a 3Ј-dTdT overhang (UGAG- 2 The oligonucleotide sequences are available upon request. GACGAGCUGGAACUUdTdT and dTdTACUCCUGCUCGACCUUGAA) were designed, synthesized, and annealed (to generate duplex siRNA) by Eurogentec S. A. The siRNA corresponding to vinexin was transfected into F9 cells at a final concentration of 50 nM using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol. At 48-h post-transfection (with an intermediate re-transfection at 24 h), the cells were treated with vehicle or RA. At the indicated times, the cells were harvested and subjected to RNA and protein analysis.

Isolation of Vinexin as a Protein That Binds to the Nonphosphorylated AF-1 Domain of RAR␥ in Yeast Two-hybrid
Screening-The yeast two-hybrid system was used to isolate cDNAs encoding mouse proteins interacting with the N-terminal domain of RAR␥. The N-terminal A/B domain of hRAR␥1 in which the two phosphorylatable serine residues had been substituted with alanines, RAR␥(A/B)(S77A/S79A) (Fig. 1B), was used as bait in a screening of a mouse embryo cDNA library. We isolated a clone growing on selective medium and express-ing ␤-galactosidase (Fig. 1C) that contained a 1.5-kb cDNA insert homologous to mouse vinexin (GenBank TM /EBI accession number AF064806) (13). This cDNA fragment (designated RAR␥(A/B)-BP) contained the 3Ј-terminal half of the open reading frame encoding mouse vinexin ␣ (amino acids 360 -733) and 221 nucleotides corresponding to the 3Ј-untranslated region of vinexin ␣ ( Fig. 2A). The encoded protein contained the three SH3 domains of vinexin ␣, which are also shared by the vinexin ␤ isoform (Fig. 2, A and B) (13).
Most interestingly, this cDNA clone was not isolated in a yeast two-hybrid screening with the A/B domain of RAR␥ in which the two serines were substituted with glutamic acids, RAR␥(A/B)(S77E/S79E) (Fig. 1B), which mimic phosphorylated residues. This suggests that the cloned C-terminal half of vinexin would interact specifically with the non- We also tested whether the cloned C-terminal half of vinexin was able to interact with the N-terminal domain of another RAR isotype, RAR␣. With this aim, the corresponding LexA-RAR␣(A/B) protein (either WT or S77A) (Fig. 1B) was expressed in the L40 yeast strain either alone or in combination with the cloned VP16-RAR␥(A/B)-BP hybrid protein. In both cases, ␤-galactosidase was only slightly expressed (Fig. 1C), suggesting that the cloned protein interacts preferentially with RAR␥.
Nuclear Colocalization of Vinexin ␤ with RAR␥-Because our cloned RAR␥(A/B)-binding protein contains the three SH3 domains of vinexin ␣, which are also shared by vinexin ␤, one can speculate that both vinexin isoforms might be candidates for interacting with RAR␥. A number of previous studies (13)(14)(15) indicated that the vinexin ␣ and ␤ proteins are focal adhesion and intermediate junction proteins that play a role predominantly in cytoskeleton organization, cell spreading, and intracellular signaling. However, only vinexin ␤ (not vinexin ␣) could be detected also in the nucleus (13).
Thus, we investigated whether vinexin ␤ is able to colocalize with RAR␥ in the nucleus by performing epifluorescence and confocal imaging with COS-1 cells overexpressing RAR␥ together with FLAG-vinexin (␣ or ␤). WT RAR␥ (Fig. 3, panels 2, 6, 10, and 14) and RAR␥(S77A/S79A) and RAR␥(S77E/S79E) (data not shown) were found exclusively in the nucleus. However, the subcellular localization of vinexin was different depending on the isoform. Indeed, FLAG-vinexin ␣ was found exclusively as dots in the cytosol of the transfected cells (Fig. 3, panels 1 and 9), whereas FLAG-vinexin ␤ was present in both the nucleus and the cytoplasm (Fig. 3, panels 5 and 13). Finally, the distribution of vinexin ␤ in the nucleus appeared to be similar to that of RAR␥ (Fig. 3, panels 8 and 16). In contrast, vinexin ␣ and RAR␥ were distributed in different compartments (Fig. 3, panels 4 and 12). Collectively, these results indicate that vinexin ␤ colocalizes with RAR␥ in the nucleus and thus would be the vinexin isoform that interacts with the N-terminal domain of RAR␥ in vivo.
Vinexin ␤ Co-immunoprecipitates with the Non-phosphorylated AF-1 Domain of RAR␥-To study further the data obtained by yeast two-hybrid screening and fluorescence imaging, co-immunoprecipitation experiments were performed with re- combinant proteins overexpressed in COS-1 cells. FLAG-vinexin ␤ was coexpressed with full-length RAR␥(S77A/S79A), which is not phosphorylatable. After immunoprecipitation of the extracts with monoclonal antibodies raised against the F domain of RAR␥ (mAb4␥(F)), immunoblotting with anti-FLAG antibodies showed that vinexin ␤ interacted with RAR␥(S77A/ S79A) (Fig. 4, A, lane 3; B, lane 4). No interaction could be seen upon deletion of the N-terminal A/B domain (Fig. 4A, lane 7), in agreement with the fact that vinexin ␤ had been isolated in the two-hybrid system as a protein that interacts with this region of RAR␥. Similar results were obtained in the absence and presence of RA (Fig. 4A, compare lanes 3 and 4 and lanes 7 and  8), confirming that the interaction of vinexin ␤ with RAR␥ does not involve surfaces within the AF-2 domain that are reorganized upon ligand binding (4). Finally, no interaction was seen with RAR␥(S77E/S79E) (Fig. 4B, lane 5), which mimics a phosphorylated receptor, in line with the observed interaction of vinexin ␤ with the non-phosphorylated A/B domain of RAR␥.
These results were confirmed in co-immunoprecipitation experiments performed with COS-1 cells overexpressing FLAGvinexin ␤ in combination with a fusion protein in which the N-terminal A/B domain of RAR␥ was fused to the DNA-binding domain of the ER. After immunoprecipitation of the extracts with monoclonal antibodies raised against the A domain of RAR␥ (antibody mAb41␥(A)), immunoblotting with anti-FLAG antibodies showed that only a small amount of vinexin ␤ coimmunoprecipitated with the WT RAR␥ A/B domain (Fig. 4C,  lane 3), in agreement with our previous demonstration that a fraction of overexpressed RAR␥ is phosphorylated in COS-1 cells (8). However, vinexin ␤ did co-immunoprecipitate with the N-terminal domain of RAR␥ in which Ser 77 and Ser 79 had substituted with alanines (Fig. 4C, lane 6). Collectively, these results confirm that vinexin ␤ interacts preferentially with the non-phosphorylated N-terminal domain of RAR␥.
Vinexin ␤ Interacts Specifically with RAR␥ in Vitro, but Not with RAR␣ or RXR␣-The interaction of vinexin ␤ with RAR␥ was further investigated in in vitro protein-protein interaction assays using recombinant GST-WT RAR␥ expressed in E. coli and bound to glutathione-Sepharose beads. When expressed in E. coli, WT RAR␥ is not phosphorylated at Ser 77 and Ser 79 within the A/B domain (8) and thus is more suitable than the corresponding alanine mutant to demonstrate that the interaction with vinexin ␤ occurs with the non-phosphorylated receptor. After incubation of the beads with extracts from COS-1 cells overexpressing FLAG-vinexin ␤, we found that vinexin ␤ interacted with RAR␥ even in the presence of high salt concentrations (Fig. 5A, lanes 3, 5, and 7). Vinexin ␤ also interacted with the isolated N-terminal A/B domain of RAR␥ fused to GST and expressed in E. coli (Fig. 5B, lane 3), confirming that this interaction concerns this non-phosphorylated domain.
Overexpression of Vinexin ␤ Inhibits RAR␥ Transcriptional Activity-Because vinexin ␤ can bind RAR␥, we sought to assess whether it could also regulate the transcriptional activity of this receptor. With this aim, increasing amounts of vinexin ␤ were coexpressed in COS-1 cells along with RAR␥ (WT, S77A/ S79A, or S77E/S79E), the heterodimeric partner RXR␣, and a CAT reporter gene under the control of a DR5 response element (Fig. 6A).
RAR␥(S77E/S79E), which behaves as the phosphorylated receptor, was almost as efficient as WT RAR␥ in inducing the expression of the CAT reporter gene in the presence of ligand (Fig. 6A, compare bars 4 and 16). Overexpression of vinexin ␤ did not affect the transcriptional activity of this mutant in both the absence and presence of the RAR␥/pan-RXR agonist combination (Fig. 6A, bars 13-18), in line with the absence of any interaction between the phosphorylated form of RAR␥ and vinexin ␤.
In contrast, RAR␥(S77A/S79A), which is devoid of any phosphorylation-dependent regulation, was markedly less efficient than the WT receptor in transactivating the CAT reporter gene in response to the ligand (Fig. 6A, compare bars 4 and 10), in agreement with our previous report (8). Overexpressed vinexin ␤ further inhibited the transcriptional activity of RAR␥(S77A/S79A) (Fig. 6A, bars 7-12) in both the absence and presence of ligand. Consistent with the ability of vinexin ␤ to interact with this mutant, vinexin ␤ may be repressor of RAR␥-mediated transcription.
Overexpression of vinexin ␤ also inhibited significantly, in a dose-dependent manner, basal transcription mediated by WT RAR␥ in the absence of ligand (Fig. 6A, bars 1-3). This is in line with the fact that, in the absence of ligand, WT RAR␥ bound at an RA response element is not phosphorylated due to the absence of interaction of the receptor with TFIIH associated with the transcription machinery and to the absence of p38 MAPK activation (2). In response to the ligand, RAR␥ becomes fully active subsequent to the phosphorylation of Ser 79 and Ser 77 by Cdk7 within TFIIH and by p38 MAPK, respectively (11,12), as assessed by immunoblotting with antibodies specifically recognizing RAR␥ phosphorylated at these residues (Fig.  6B, lane 2) (data not shown). Overexpression of vinexin ␤ did not prevent the increase in RAR␥ phosphorylation that occurred in response to the ligand (Fig. 6B, lane 3). However, overexpression of vinexin ␤ inhibited WT RAR␥-mediated transcription in the presence of ligand (Fig. 6A, bars 4 -6). Collectively, these results suggest the hypothesis that vinexin ␤ would be a repressor of RAR␥ through its interaction with the non-liganded and non-phosphorylated form of the receptor. They also emphasize the possibility that an excess of vinexin ␤ 2, 4, and 6) or GST-RAR␥ proteins (lanes 3, 5, and 7) immobilized on glutathione-Sepharose beads. Washing was performed in GST buffer containing 150 -500 mM NaCl. Bound proteins were analyzed by immunoblotting with anti-FLAG antibodies. Lane 1 corresponds to 5% of the amount of loaded vinexin ␤. B, the conditions were the same as described for A with GST and GST-RAR␥(A/B) in buffer containing 150 mM NaCl. C, increasing amounts of extracts from COS-1 cells transfected with the FLAGvinexin ␤ expression vector were incubated with GST (lanes 2, 5, and 8), GST-RAR␣ (lanes 3, 6, and 9), or GST-RAR␥ (lanes 4, 7, and 10) and processed as described for B. D, the conditions were the same as described for B with GST, GST-RXR␣, and GST-RAR␥. would drive RAR␥ to a transcriptionally inactive state even in the presence of ligand.

FIG. 5. Vinexin ␤ interacts in vitro with the non-phosphorylated form of RAR␥, but not with RAR␣ and RXR␣. A, extracts (300 g) from COS-1 cells transfected with the FLAG-vinexin ␤ expression vector were incubated with GST (lanes
Note that no effect of vinexin ␤ overexpression could be detected when the DR5 response element was mutated and unable to bind the RAR␥/RXR␣ heterodimers (data not shown). Moreover, vinexin ␤ did not affect the transcriptional activity of RXR␣ homodimers on a DR1-tk-CAT reporter gene (Fig. 6C), in accordance with the absence of any significant interaction of vinexin ␤ with this receptor (Fig. 5D). This also indicates that the observed inhibition of RAR␥ activity does not reflect a general inhibition of transcription.

In F9 Cells, Expression of RA Target Genes Requires Phosphorylation of the N-terminal Domain of RAR␥, and Vinexin
␤ Colocalizes with RAR␥ in the Nucleus-Having demonstrated the importance of vinexin ␤ in the transcriptional activity of RAR␥ overexpressed in COS-1 cells, we asked whether the same conclusions could be made in vivo in mouse embryo carcinoma cells (F9 cell line), which constitute a well established cell autonomous model system for investigating RA signaling (29). In F9 cells, RA induced the expression of several target genes within 24 h as assessed by quantitative RT-PCR (Fig. 7A). The RA-induced expression of these genes involves the activation of RAR␥/RXR␣ heterodimers, as it was reduced in F9 RAR␥ Ϫ/Ϫ cells (Fig. 7B) (data not shown). Most interestingly, the activation of the RA target genes was less efficiently restored upon re-expression of RAR␥ mutated at the phosphorylation sites than upon re-expression of WT RAR␥ (Fig. 7B) (data not shown), indicating that phosphoryl-ation of the AF-1 domain of RAR␥ plays a crucial role in RAR␥-mediated transcription.
We also investigated whether vinexin ␤ is present in the nucleus of F9 cells, in addition to RAR␥. With this aim, we generated mouse monoclonal antibodies against vinexin. As shown in Fig. 7C, these antibodies efficiently recognized recombinant FLAG-vinexin (␣ and ␤) overexpressed in COS-1 cells. Nuclear and cytoplasmic extracts were prepared from F9 cells and immunoblotted with these anti-vinexin antibodies. A protein species with an apparent molecular mass corresponding to that of vinexin ␣ (82 kDa) was detected only in the cytosol (Fig.  7D, middle panel, lane 1). In contrast, vinexin ␤ (37 kDa) was detected not only in the cytosolic compartment, but also in the nucleus of F9 cells (Fig. 7D, lower panel). Thus, in F9 cells, vinexin ␤ appeared to colocalize with RAR␥ (Fig. 7D, upper  panel). Collectively, these results indicate that F9 cells would constitute a good model to determine whether vinexin ␤ modulates the expression of RAR␥ target genes.
Vinexin ␤ Is an Inhibitor of Several RAR␥ Target Genes in F9 Cells-To determine the role of vinexin ␤ in RAR␥-mediated transcription, we analyzed the consequences of siRNA-mediated knockdown of vinexin in F9 cells and investigated whether reduction in the level of vinexin would influence the expression of the RA target genes. Transfection into F9 cells of the siRNA targeting vinexin reduced the expression levels of both vinexins ␤ and ␣ at both the mRNA and protein levels as shown by quantitative RT-PCR and immunoblotting, respectively (Fig. 8,  A, bars 1-3; and B). In contrast, the expression of RAR␥ was not affected (Fig. 8, A, bars 4 -6; B, lower panel). When vinexin was reduced by RNA interference, the RA-induced expression of all tested RA target genes such as CYP26, Hoxb-1, CRABPII, Stra4, Hoxa-1, HNF3␣, and HNF1␤ ( Fig. 8C) (data not shown), which depend on RAR␥ phosphorylation (see above), was significantly up-regulated. The expression of RAR␥ was not affected (Fig. 8C), indicating that our results do not reflect an increase in RAR␥ levels. Thus, decreasing vinexin levels appears to improve the efficiency of transcription of RAR␥ target genes, consistent with an inhibitory role of vinexin.
To confirm the inhibitory role of vinexin ␤ in the expression of RAR␥ target genes, a stable cell line overexpressing vinexin ␤, the F9 V␤(5) cell line, was established from F9 WT cells. In this cell line, vinexin mRNA levels were significantly increased as assessed by quantitative RT-PCR (Fig. 9A, bar 2). The corresponding protein was also increased in both the cytosolic and nuclear compartments (Fig. 9B) as assessed by immunoblotting. We investigated whether the expression of the RA target genes was affected in this cell line compared with the WT counterpart. Our results clearly show that, in the F9 V␤(5) cell line, the RA-induced expression of some RA target genes such as CYP26, Hoxb-1, and CRABPII (Fig. 9C) was significantly lower than in the parental F9 WT cell line, confirming that vinexin ␤ is a repressor of RAR␥ transcriptional activity. The expression of RAR␥ was not affected (Fig. 9C), indicating that the observed results do not reflect an inhibition of RAR␥ levels. However, the expression of some other RA target genes such as Stra4, Hoxa-1, HNF3␣, and HNF1␤ (Fig. 9C) (data not shown) was not affected, although it was up-regulated in the siRNA vinexin knockdown cells (Fig. 8C). This suggests that some genes may be differentially regulated by overexpressed vinexin ␤, probably due to different conformations of the transcriptional complexes.

DISCUSSION
In this study, we have isolated vinexin ␤ as a novel cofactor of the RAR␥ isotype. Vinexin ␤ was originally identified as a focal adhesion and intermediate junction protein that plays a role in cytoskeleton organization and cell spreading as well as in intracellular signaling (13)(14)(15). In addition to these membrane-associated functions, we present evidence that vinexin ␤ is present in the nucleus, interacts with the non-phosphorylated form of RAR␥, and negatively controls its transcriptional activity.
Vinexin ␤ Interacts with the Non-phosphorylated AF-1 Domain of RAR␥-During the last decade, a number of co-regulators, including the p160 family of coactivators, have been identified as nuclear receptor-binding proteins that are recruited by the AF-2 domain subsequent to ligand-induced conformational changes (30 -33). In contrast, only a few proteins have been reported to interact with the N-terminal domain of nuclear receptors (34 -38). This study is the first reporting the identification of a protein that interacts with the AF-1 domain of an RAR. It is also the first reporting the interaction of an RAR with an adaptor protein (vinexin ␤) known to regulate cytoskeleton organization and signal transduction.
The main characteristic of the vinexin ␤ protein is the pres- ence of three SH3 domains (13) that are known to interact with proline-rich sequences containing the PXXP core (16). Such proline-rich motifs are relatively rigid and exposed and thus are favorable for interactions with SH3, WW, and many other protein interaction domains (16,39). The interactions with these proline-rich motifs are not as tight as those with globular domains, but they present the advantage of being modulated rapidly upon covalent modifications such as phosphorylation. Indeed, phosphorylation of serine residues flanking proline motifs has the ability to positively or negatively regulate the binding of SH3 domains (16,40).
Interestingly, the AF-1 domain of RARs is characterized by the presence of a conserved proline-rich motif (Fig. 1A). In RAR␥, this proline-rich motif contains two serine residues that can be phosphorylated by two types of kinases, Cdk7 within TFIIH and p38 MAPK (2,11). Here, we have demonstrated that vinexin ␤ interacts with the non-phosphorylated form of the N-terminal AF-1 domain of RAR␥ and that its phosphorylation prevents the interaction with vinexin ␤. These results challenge the hypothesis (see below) that the phosphorylationdependent modulation of the interaction with vinexin ␤ is, at least in part, one of the mechanisms involved in the control of RAR␥ activity.
According to our data, the interaction with vinexin ␤ appears to concern only RAR␥ and not the other RAR isotypes or RXR␣. However, it must be noted that other nuclear receptors have been recently found to interact with proteins of the same vinexin family (41). Indeed, the progesterone receptor interacts with c-Cbl-associated protein/ponsin through a proline-rich motif in its N-terminal domain (42). The ER has also been recently reported to interact with the ␣ isoform of vinexin through its N-terminal domain (43). Based on these observations, it appears that this family of proteins, which can be present not only in the cyto-plasm, but also in the nucleus, would be novel regulators of nuclear receptors.
Vinexin ␤ Is an Inhibitor of RAR␥-mediated Transcription-Although the potential importance of phosphorylation of the N-terminal domain of RAR␥ has been established, the underlying mechanism for the phosphorylation-dependent transcriptional activity of RAR␥ has not been elucidated yet. Here, we propose the hypothesis that non-phosphorylated RAR␥ is transcriptionally inactive due, at least in part, to the interaction of its AF-1 domain with vinexin ␤. Consequently, phosphorylation of the AF-1 domain of RAR␥ would control RAR␥-mediated transcription through triggering the dissociation of vinexin ␤. To address this hypothesis, we used not only transfected COS-1 cells, but also F9 cells, which constitute a well established cell autonomous model system for investigating RA signaling (29). Most interestingly, in these cells, the functional RAR isotype, RAR␥, controls most RA-induced events through its phosphorylation and colocalizes with vinexin ␤ in the nucleus. Vinexin ␤ overexpression and siRNA knockdown experiments led us to propose the following model.
The traditional view is that, in the absence of ligand, RAR␥ resides in the nucleus, binds to response elements of target genes, and interacts with corepressors associated with large complexes with histone deacetylase activity, resulting in chromatin compaction and transcriptional repression (30,44,45). In this context, due to the absence of recruitment of the general transcription factor TFIIH by the promoter and to the absence of p38 MAPK activation, RAR␥ is not phosphorylated (2). Thus, from our data, one can suggest that, at that stage, vinexin ␤, which colocalizes with RAR␥ in the nucleus, interacts with the non-phosphorylated N-terminal AF-1 domain of RAR␥ and participates with corepressors in transcriptional repression. In support of this hypothesis is the recent description of a direct interaction between vinexin ␤ and SAFB2, a novel nuclear receptor corepressor (46). It is also consistent with our observation that, in transfected COS-1 cells, overexpression of vinexin ␤ completely blocked the activity of RAR␥(S77A/S79A) (Fig. 6A), which mimics a non-phosphorylated receptor and is devoid of any phosphorylation-dependent regulation.
Then, upon ligand binding, subsequent to the combinatorial recruitment of chromatin-modifying and chromatin-remodeling complexes, repressive chromatin is decompacted, allowing the positioning of the transcription machinery at the promoter and the interaction of the general transcription factor TFIIH with RAR␥ (2,7,45). This makes the Cdk7 subunit of TFIIH able to phosphorylate one serine residue located in the prolinerich motif of the AF-1 domain of RAR␥ (2,8). In the same time slot, p38 MAPK becomes activated and phosphorylates the second nearby serine residue (2,11). Our data suggest that phosphorylation would induce the dissociation of vinexin ␤, therefore allowing transcription to proceed. Such a model is supported by other studies showing that phosphorylation processes also promote the dissociation of vinexin ␤ from other partners (14,15). Whether vinexin ␤ dissociates subsequent to RAR␥ phosphorylation by TFIIH and/or by p38 MAPK is a current matter of investigation in our laboratory. It also remains to be determined whether vinexin ␤ dissociation would allow the subsequent recruitment by the phosphorylated AF-1 domain of other complexes participating in the initiation of transcription. The characterization of these complexes is presently under investigation. According to this model, vinexin ␤ appears to be a repressor of RAR␥, which is recruited through the non-phosphorylated N-terminal domain of the receptor. This is substantiated by our results showing that siRNA knockdown of vinexin potentiates RA-induced activation of most RAR␥ target genes. It is also corroborated by our vinexin ␤ overexpression experiments in COS-1 and F9 cells showing a significant decrease in the RAinduced expression of several RA target genes. This implies that an increased ratio of vinexin ␤ drives the equilibrium to an inactive state, even though the RA-induced phosphorylation of the RAR␥ AF-1 domain is not compatible with vinexin ␤ interaction. Note that a similar dynamic inhibition of nuclear receptor activation has been recently reported upon corepressor overexpression even in the presence of ligand (47,48). Regardless of the detailed mechanism involved, our results strongly emphasize the possibility that, when present at an increased ratio, vinexin ␤ does not impede RAR␥ phosphorylation, but would function as a scaffolding protein that blocks RAR␥ in an inactive state, unable to recruit, through its phosphorylated AF-1 domain, the co-regulators necessary to achieve transactivation. Note, however, that this mechanism depends on the promoter context, as the expression of some genes was not affected upon vinexin ␤ overexpression. The reason why genes are differentially regulated by vinexin ␤ is not yet known. Nevertheless, one can suggest that, depending on the promoter context, the different RAR␥-interacting protein complexes are endowed with different and specific conformations that dictate the consequences of vinexin ␤ overexpression.
In conclusion, we have identified vinexin ␤ as a novel repressor of RAR␥ that is recruited through the non-phosphorylated N-terminal domain of the receptor. We propose that the dissociation of vinexin ␤ upon phosphorylation of the AF-1 domain of RAR␥ would be a physiologically relevant part of the phosphorylation-dependent transcriptional activation process of RAR␥. However, it must be noted that vinexin ␤ can also be phosphorylated, raising the hypothesis that the association/dissociation of vinexin ␤ and RAR␥ might be controlled by the phosphorylation of both partners. Therefore, we can conclude, along with others, that vinexin ␤ functions as a scaffolding protein that controls RAR␥ function through its association/dissociation in response to phosphorylation processes.