A Unique Cytoplasmic Localization of Retinoic Acid Receptor-γ and Its Regulations*

Recent evidence suggests that extranuclear action of retinoid receptors is involved in mediating the pleiotropic effects of retinoids. However, whether they reside in the cytoplasm remains elusive. Here, we showed that retinoic acid receptor-γ (RARγ) was cytoplasmic in confluent cells, or when cells were released from serum depletion or treated with growth factors. In studying the regulation of RARγ subcellular localization, we observed that ectopically overexpressed RARγ was mainly cytoplasmic irrespective of serum concentration and cell density. The cytoplasmic retention of RARγ was inhibited by ligand retinoic acid (RA). In addition, coexpression of retinoid X receptor-α (RXRα) resulted in nuclear localization of RARγ through their heterodimerization. Mutagenesis studies revealed that a C-terminal fragment of RXRα potently prevents RA-induced RARγ nuclear localization and transcriptional function. Furthermore, our results showed that the cytoplasmic retention of RARγ was due to the presence of its unique N-terminal A/B domain, which was subject to regulation by p38 MAPK-mediated phosphorylation. Deletion or mutation of the N-terminal A/B domain largely impaired its cytoplasmic localization. Together, our data demonstrate that the subcellular localization of RARγ is regulated by complex interactions among ligand binding, receptor phosphorylation, and receptor dimerizations.

The pleiotropic effects of retinoids, natural and synthetic vitamin A derivatives, is mediated by two classes of nuclear receptor family, the retinoic acid receptors (RARs) 3 and the retinoid X receptors (RXRs), which are encoded by three distinct genes, ␣, ␤, and ␥ (1, 2). In addition, different isoforms can be generated from each receptor, which differ in their N-terminal sequences, through differential promoter usage and alternative splicing (1,2). The evolutionary conservation of these receptor subtypes and isoforms and their distinct patterns of expression during development and in the adult organism suggest that each of them has discrete functions (3,4). RAR␥, but not RAR␣, plays a role in predisposing murine keratinocytes to Ras-induced tumorigenesis, in retinoic acid (RA)-induced cell cycle arrest and apoptosis in keratinocytes (5), and in the regulation of the balance between hematopoietic stem cell self-renewal and differentiation (6). At the level of transcriptional regulation, overexpression of RAR␥l inhibits transactivation of RA target genes by other RARs (7), and several specific RAR␥ target genes have recently been identified in F9 cells (8).
RARs activate transcription by binding to cis-acting response elements in the promoter/enhancer region of target genes as homodimer or heterodimer with RXRs (1,2,9). RXRs, besides forming heterodimer with RARs in retinoid signaling, can act as heterodimerization partners for many other nuclear receptors to activate transcription of their target genes in a variety of signaling pathways (1,2). Retinoid receptors, like other nuclear receptors, consist of three main functional domains: the nonconserved N-terminal A/B domain, the central DNA-binding domain containing two zinc finger motifs and nuclear localization signal (NLS), and the multifunctional C-terminal ligandbinding domain (LBD) containing regions for receptor dimerization, ligand binding, and the ligand-dependent transactivation function. Dimerization interfaces that largely mediate heterodimerization of RXR␣ with RARs have been mapped to regions in the C terminus, corresponding to helices 9 and 10 in the canonical nuclear receptor LBD structure (10,11). The N-terminal A/B domain contains conserved serine residues, which belong to consensus phosphorylation sites for prolinedependent protein kinases such as cyclin-dependent kinases (Cdks) and the mitogen-activated protein kinases (MAPKs) (12)(13)(14), and phosphorylation of these sites can regulate RAR transactivation (15). Phosphorylation of the A/B domain of RAR␥ is indispensable for differentiation of F9 cells upon RA and cAMP treatment (16), whereas RAR␣ binds to CDK-acti-vating kinase, resulting in an enhanced CDK-activating kinase activity and cell proliferation (17).
Unlike steroid hormone receptors such as glucocorticoid receptor (GR) and androgen receptor that rapidly translocate from the cytoplasm to the nucleus upon ligand-induced activation, retinoid receptors are considered to mainly reside in the nucleus independent of the presence of ligand (18 -20). However, certain rapid retinoid responses, such as activation of GTPase Rac (21), protein kinase C (22), and ERK2 (23), and phosphorylation of cAMP-response element-binding protein (24,25), cannot be explained by classic transcriptional regulation. Thus, subcellular localization of retinoid receptors changes during development of the testes (26) and during various stages of the menstrual cycle (27). Cytoplasmic localization of retinoid receptors in some cell types has been shown to be associated with important biological processes, including growth (28), apoptosis (29), differentiation (30), and inflammation (31,32). Different RAR␤ isoforms also exhibit distinct subcellular localization, with RAR␤2 being primarily nuclear and RAR␤4 being cytoplasmic (33). Thus, subcellular localization may represent another mechanism by which retinoid receptors mediate the pleiotropic effects of retinoids.
Cytoplasmic localization of nuclear receptors not only regulates transcription by modulating their availability in the nucleus but also plays an active role in the cross-talk with other signal transduction pathways. Nur77 migrates from the nucleus to the cytoplasm where it targets mitochondria by binding to Bcl-2 (34 -36), providing a molecular basis for integration of nuclear receptor signaling to mitochondrial apoptotic machinery. RXR rapidly inhibits Rac activation and intracellular calcium release by binding to G protein G q in human platelets that contain no nucleus (21). RXR migrates from the nucleus to the cytoplasm during nerve growth factor-induced PC12 cell differentiation (30) and targets mitochondria to induce apoptosis (29). Interaction of RAR␥ with cytoplasmic c-Src mediates neuritogenic differentiation (28). Altered retinoid receptor subcellular localization has been shown to be associated with cancer progression (37,38). Thus, understanding how subcellular localization of retinoid receptors is regulated will provide additional important information regarding the mechanism of retinoid receptor action.
Here we report that RAR␥ often resides in the cytoplasm when cells are cultured at high density, released from serum starvation, or treated with growth factors. Using transient transfection assays, we show that the cytoplasmic accumulation of RAR␥ is largely dependent on its N-terminal A/B domain, which is modulated by phosphorylation of its Ser-77 and Ser-79 residues by p38 MAPK. Cytoplasmic retention of RAR␥ is also inhibited by ligand binding and heterodimerization with RXR␣. Thus, our results demonstrate that the subcellular localization of RAR␥ is regulated by complex interplays among ligand binding and receptor phosphorylation and that its N-terminal A/B region plays a critical role in determining its cytoplasmic accumulation.
Western Blotting-Cell lysates were boiled in SDS sample buffer, resolved by SDS-PAGE (8 or 12.5% polyacrylamide), and transferred to nitrocellulose (35,36,41). After transfer, the membranes were blocked in 5% milk in TBST (10 mM Tris-HCl, pH 8.0, 150 mM NaCl, and 0.05% Tween 20) for 30 min and incubated with primary antibody in 5% milk in TBST for 2 h at room temperature. The membranes were washed three times with TBST, and incubated for 1 h at room temperature in TBST containing horseradish peroxidase-linked anti-mouse or rabbit immunoglobulin (Santa Cruz Biotechnology). After washing in TBST for three times, immunoreactive products were detected by chemiluminescence with an enhanced chemiluminescence system (ECL, Amersham Biosciences).
Co-immunoprecipitation Assays-For Co-IP assay (29,36), HEK293T cells grown in 60-mm dishes were transfected with indicated expression vectors for 24 h and treated with ligands for 1 h. After washing three times with cold PBS, cells were lysed in 1 ml of P-RIPA buffer (1% Nonidet P-40, 0.5% deoxycholic acid, 0.1% SDS, 10 mM EDTA in PBS) containing protease inhibitors (Sigma). Lysate was precleared by incubating with normal mouse IgG and protein A/G-Sepharose (Santa Cruz Biotechnology) for 2 h at 4°C. Precleared lysate was then incubated with 1 g of anti-Myc monoclonal antibody (9E10, Santa Cruz Biotechnology) overnight at 4°C. Immunocomplexes were then precipitated with 40 l of protein A/G-Sepharose. After extensive washing with P-RIPA buffer, beads were boiled in 40 l of loading buffer and analyzed by Western blotting.
Immunofluorescence Microscopy-Cells seeded on cover slips in 24-well plates overnight were transfected with appropriate expression vectors for 24 h and treated with ligands for 1 h. The cells were fixed in 4% paraformaldehyde in PBS for 10 min, washed twice with PBS, and then permeabilized with 0.1% Triton X-100 in PBS for 10 min (29,35,36). Fixed cells were preincubated for 30 min in PBS containing 5% bovine serum albumin at room temperature. Cells were stained with primary antibody (anti-Myc monoclonal antibody, 1:400 dilution) for 1 h at room temperature followed by incubation with secondary antibody conjugated with Alexa Fluor568 (Invitrogen) or Cy5 (1:1000 dilution). For staining endogenous RAR␥, cells grown on coverslips were stained with rabbit polyclonal anti-RAR␥ antibody (M-454, Santa Cruz Biotechnology) as described above. 4,6-Diamidino-2-phenylindole (0.1 g/ml) was added to the secondary antibody mixture to visualize nuclei. Fluorescence images were collected and analyzed using an inverted fluorescence microscope or MRC-1024 MP laser-scanning confocal microscope (Bio-Rad).

Subcellular Localization RAR␥ Is Regulated by Growth
Conditions-In studying whether retinoid receptors exhibited differential subcellular localization in response to cellular stimulation, we observed that RAR␥ exhibited strikingly different intracellular localization patterns in H460 lung cancer cells depending on culture conditions. In non-confluent cells, immunostaining with anti-RAR␥ antibody revealed that typically Ͼ95% of cells had RAR␥ nuclear staining. In contrast, RAR␥ was predominantly (Ͼ70%) cytoplasmic in confluent cells (Fig. 1A). The subcellular localization of RAR␥ was also regulated by serum concentration. When subconfluent H460 cells were cultured in serum-free medium, RAR␥ was predominantly nuclear. However, cytoplasmic localization of RAR␥ was observed when cells were released from 24 h of serum starvation (Fig. 1B). Furthermore, treatment of H460 cells cultured in serum-free medium with various growth factors, such as epidermal growth factor or platelet-derived growth factor, also resulted in increased RAR␥ cytoplasmic staining (Fig. 1B). Such effects of cell density and growth factors were not observed when RAR␣ was studied (data not shown). Thus, RAR␥ frequently resides in the cytoplasm, depending on cell density, serum concentration, and the presence of growth factors.
Transfected RAR␥ Predominantly Resides in the Cytoplasm-To study the regulation of subcellular localization of RAR␥, we transfected Myc-RAR␥ (RAR␥1 tagged with the Myc epitope) expression vector into HeLa cells. Immunostaining of transfected cells with anti-Myc antibody showed that ectopically expressed Myc-RAR␥ was mainly cytoplasmic ( Fig. 2A). Cytoplasmic localization of Myc-RAR␥ was independent on cell density (data not shown), which was in contrast to the endogenous RAR␥ whose subcellular localization depended on growth conditions (Fig. 1). To study whether cell density-independent cytoplasmic localization of Myc-RAR␥ was due to Myc epitope tagged, RAR␥ fused with green fluorescent protein (GFP) was transfected into HeLa cells. Fluorescence microscopy analysis showed that GFP-RAR␥ also resided in the cytoplasm. Furthermore, untagged RAR␥ displayed similar cytoplasmic localiza-tion when examined by anti-RAR␥ antibody ( Fig. 2A). Cytoplasmic localization of Myc-RAR␥, GFP-RAR␥, and RAR␥ was potently inhibited by ligand binding as treatment of cells with RA resulted in their exclusive nuclear localization. To determine whether the cytoplasmic localization of ectopically expressed RAR␥ could be observed in other cell types, we transfected Myc-RAR␥ into several cell lines, including ZR-75-1 breast cancer, SW480 colon cancer, H460 lung cancer, and HEK293T embryonic kidney cells (Fig. 2B). Similar to that observed in HeLa cells, Myc-RAR␥ was mainly cytoplasmic in these cells, demonstrating that cytoplasmic localization of overexpressed RAR␥ is not cell type-specific. As nuclear receptors are known to shuttle between the cytoplasm and the nucleus and their subcellular localization is often determined by a balance between activities of nuclear export sequence (NES) and NLS (43), we determined whether cytoplasmic retention of RAR␥ might depend on its sustained nuclear export. Myc-RAR␥-transfected cells were treated with leptomycin B (LMB), which is known to inhibit classic Crm-1-dependent nuclear export (44). Fig. 2C shows that treatment of cells with LMB had no apparent effect on cytoplasmic accumulation of Myc-RAR␥, suggesting that retention of transfected RAR␥ in the cytoplasm was not dependent on its continuous nuclear export.

FIGURE 1. Regulation of intracellular localization of RAR␥ by cell density, serum concentration, and growth factors.
A, effect of cell density on RAR␥ subcellular localization. H460 cells were seeded at low density (0.2 ϫ 10 6 cells per 100-mm plate) or high density (7.0 ϫ 10 6 cells per 100-mm plate) and 12 h later, cells were immunostained using anti-RAR␥ antibody (Santa Cruz Biotechnology, M-454). Nuclei were counterstained with 4,6-diamidino-2-phenylindole (DAPI). More than 95% of low density cells showed RAR␥ nuclear staining, whereas ϳ70% of confluent cells displayed RAR␥ cytoplasmic staining. B, effect of serum and growth factors on RAR␥ subcellular localization. H460 cells were seeded at 1.0 ϫ 10 6 cells per 100-mm plate in complete medium overnight and then changed to serum-free medium for 24 h. Cells were then treated with either serum (10%), epidermal growth factor (100 ng/ml), or platelet-derived growth factor (10 ng/ml) for 6 h and immunostained using anti-RAR␥ antibody. About 90% subconfluent serum-deprived cells showed RAR␥ nuclear staining, whereas treatment with serum, epidermal growth factor, and platelet-derived growth factor resulted in 70%, 60%, and 50% of cells displaying RAR␥ cytoplasmic staining, respectively.

Regulation of RAR␥ Subcellular Localization by RXR␣-Our
observations that the cytoplasmic localization of endogenous RAR␥ depended on culture condition while transfected RAR␥ was constitutively cytoplasmic suggested the presence of endogenous RAR␥-binding protein(s) that acted to keep RAR␥ in the nucleus and were limiting. Because RXR␣ was known to modulate subcellular distribution of its heterodimerization partners (29,43,45,46), we examined whether RXR␣ represented such a limiting factor. As shown in Fig. 3A, when RXR␣ was cotransfected with GFP-RAR␥, both receptor proteins were found exclusively in the nucleus. Such a relocalization of RAR␥ from the cytoplasm to the nucleus by RXR␣ occurred in the absence of RA treatment and was specific, because the cytoplasmic localization of GR was not affected by RXR␣ overexpression. In addition, cotransfection of RAR␣ had no effect on the cytoplasmic localization of GFP-RAR␥. Thus, RXR␣ is an important regulator of the subcellular localization of RAR␥, similar to the effect of RAR␥ ligand.
LBD of RXR␣ Specifically Blocks Nuclear Import of RAR␥-To study how RXR␣ regulated RAR␥ subcellular localization, several RXR␣ mutants were constructed. An excess amount of expression vector encoding each of the RXR␣ mutants was transfected with Myc-RAR␥ to determine their effect on the cytoplasmic localization of Myc-RAR␥. Deletion of the A/B domain from RXR␣ had no effect on its ability to induce RAR␥ nuclear import (not shown). Although cotransfection of RXR␣ completely retained Myc-RAR␥ in a ligand-independent manner, removal of the very C-terminal region (amino acids 386 -462) from RXR␣ completely impaired its effect, as cotransfection of RXR␣-(1-385) did not result in relocalization of RAR␥ into the nucleus in the absence of RA (Fig. 3B). Co-IP assay showed that RXR␣-(1-385) could not heterodimerize with RAR␥ ( Fig.  3C), suggesting that heterodimerization of RAR␥ and RXR␣ was required for their nuclear localization. Consistently, RXR␣-(1-385) did not interfere with the effect of RA on inducing Myc-RAR␥ nuclear import. We also evaluated the effect of the RXR␣ LBD, RXR␣-(222-462), which alone is cytoplasmic (29). The mutant bound strongly with RAR␥ in Co-IP assay (Fig. 3C). When RXR␣-(222-462) and RAR␥ were cotransfected, both receptor proteins were almost exclusively localized in the cytoplasm. Surprisingly, RA failed to induce nuclear localization of RAR␥, suggesting that RXR␣-(222-462) blocked the effect of RA on inducing RAR␥ nuclear import. The nuclear presence of RXR␣-(222-462) in the presence of RA likely reflected the formation of RXR␣-(222-462) homodimer due to excess amount of the receptor mutant in the presence of RA. Because RXR␣-(222-462) lacks the putative RXR␣ NLS (between amino acids160 -165) (43), these results suggest that activation of the RXR␣ NLS is crucial for the nuclear localization of RXR␣/RAR␥ heterodimer. To further determine the effect of RXR␣-(222-462), two smaller fragments of the RXR␣ LBD were generated. RXR␣(324 -462), which bound strongly with RAR␥ (Fig. 3C), was able to inhibit the ability of RA in inducing RAR␥ nuclear import, whereas RXR␣-(324 -402), which bound weakly to RAR␥, had no effect. Thus, induction of RAR␥ nuclear localization by RXR␣ requires its NLS and the C-terminal region of the RXR␣ can inhibit the ability of RA in inducing RAR␥ nuclear import.
LBD of RXR␣ Inhibits RAR␥ Homodimerization and Heterodimerization-The observation that the LBD of RXR␣ could potently antagonize the effect of RA on regulating the subcellular localization of RAR␥ was interesting, because there have been several reports describing the generation of truncated RXR␣ fragment by proteolytic cleavage in cancer cells (47)(48)(49). To gain insight into the mechanism of RXR␣-(222-462) action, we determined its effect on RAR␥ homodimerization and heterodimerization by Co-IP experiments. GFP-RAR␥ and Myc-RAR␥ were expressed together with or without an excess amount of GFP-RXR␣-(222-462), and Myc-RAR␥ was immunoprecipitated using antibody against the c-Myc epitope. In the absence of GFP-RXR␣-(222-462), GFP-RAR␥ was co-immunoprecipitated with Myc-RAR␥ in a RA-independent manner (Fig. 4A). When GFP-RXR␣-(222-462) was cotransfected, dimerization of GFP-RAR␥ with Myc-RAR␥ was largely abolished, likely reflecting higher affinity of RAR␥/RXR␣-(222-462) heterodimerization than RAR␥ homodimerization (42). We also examined the effect of RXR␣-(222-462) on RAR␥/ Myc-RAR␥ (100 ng/well in 24-well plates) was transfected into the indicated cell line, and its localization was examined by immunostaining using anti-Myc antibody. More than 50% of cells showed RAR␥ cytoplasmic staining. C, cytoplasmic localization of RAR␥ does not depend on its nuclear export. HeLa cells were transfected with Myc-RAR␥ (100 ng/well in 24-well plates) and treated with or without LMB (2.5 ng/ml, Sigma). Subcellular localization of Myc-RAR␥ was examined by immunostaining.
LBD of RXR␣ Inhibits RAR␥ Transactivation-The fact that RXR␣-(222-462) strongly antagonizes the effect of RA on inducing RAR␥ nuclear import prompted us to examine whether it could interfere with the effect of RA on inducing RAR␥ transactivation. Reporter assays using TREpal-tk-CAT reporter known to be activated by RARs (42) showed that RA strongly induced RAR␥ transcriptional activity. However, when RXR␣-(222-462) was cotransfected, RAinduced RAR␥ transactivation was completely inhibited (Fig. 4E). Interestingly, activation of the TREpal-tk-CAT reporter by RAR␣ was only slightly inhibited by RXR␣-(222-462) cotransfection, demonstrating a selective inhibitory effect of RXR␣-(222-462). The The indicated receptor expression vectors (GFP-RAR␥, 100 ng; GFP-GR, 100 ng); RXR␣, 300 ng; RAR␣, 300 ng) were transfected into HeLa cells cultured in 24-well plates, and their subcellular localization was analyzed by immunostaining using anti-RXR␣ (for staining RXR␣) or anti-RAR␣ (for staining RAR␣) antibody. More than 95% of cotransfected cells showed nuclear localization of RXR␣/RAR␥ or RXR␣. B, C-terminal fragments of RXR␣ block ligandinduced nuclear translocation of RAR␥. HeLa cells cultured in 24-well plates were cotransfected with expression vectors for Myc-RAR␥ (100 ng/well) and indicated RXR␣ fragments fused with GFP protein (300 ng/well), and cells were then treated with or without RA (0.1 M). Cells were immunostained using anti-Myc antibody and examined by fluorescence microscopy. About 80% of cotransfected cells showed the images presented. C, interaction of RAR␥ with RXR␣ and RXR␣ mutants. The indicated GFP-tagged RXR expression vector (1 g/dish) were cotransfected with Myc-RAR␥ (1 g/dish) into HEK293T cells cultured in 60-mm dishes, and their interactions were analyzed by Co-IP assays using anti-Myc antibody. Blots were probed with either anti-GFP or anti-Myc antibody to determine the efficacy and specificity of interaction. Input represents 5% of lysates used for Co-IP experiments.

N-terminal A/B Domain of RAR␥ Is Crucial for Its Cytoplasm
Localization-Although ectopically expressed RAR␥ was localized in the cytoplasm, transfected RAR␣ and RAR␤ were nuclear under the same conditions (Fig. 5A), suggesting that RAR␥-specific sequences were responsible for its cytoplasmic retention. Among three RAR subtypes, their N-terminal A/B domains display a significant variance in sequences, whereas the sequences in their DNA-binding domain and LBD are very similar (1,2). Therefore, we examined the role of the N-terminal A/B domain of RAR␥ by constructing several RAR␥ mutants (Fig. 5B). Deletion of the A/B domain from RAR␥ completely impaired its ability to reside in the cytoplasm, as the mutant (⌬N-RAR␥) was exclusively nuclear (Fig. 5C). Cotransfection of ⌬N-RAR␥ with GFP-RAR␥ resulted in their nuclear localization (Fig. 5D), likely due to their homodimerization (Fig. 4A). When the N-terminal A/B domain of RAR␥ was replaced by the corresponding domain of RAR␤, the resulting chimeric protein (␤AB-RAR␥) (Fig. 5B) was mainly found in the nucleus. To directly test the role of the A/B domain, an expression vector encoding the A/B domain of RAR␥ (RAR␥/AB) was transfected into HeLa cells. Interestingly, immunostaining showed that RAR␥/AB was predominantly cytoplasmic. The RAR␥/AB does not contain classic leucine-rich NES sequences, consistent with the inability of LMB to block RAR␥ cytoplasmic localization (Fig.  2C). Taken together, these results demonstrate that the N-terminal A/B domain of RAR␥ is responsible for its unique cytoplasmic localization.
Regulation of RAR␥ Cytoplasmic Localization by Phosphorylation-When RAR␥ was transfected into cells we noticed that it migrated as double bands in SDS-PAGE, and treatment of RAR␥-containing lysates with alkaline phosphatase resulted in disappearance of the slow-migrating band (Fig. 6A). Thus, RAR␥ was phosphorylated in HeLa cells. The serine residues (Ser-77 and Ser-79 in human RAR␥1) in the A/B domain of RAR␥ can be phosphorylated by cdk7 and p38 MAPK (12,13). To determine whether they were responsible for RAR␥ phosphorylation, we constructed a RAR␥ mutant, in which Ser-77 and Ser-79 were mutated to alanine. The resulting mutant, RAR␥(S77/79A), when transfected into HeLa cells, migrated as a single band (Fig. 6A), suggesting that Ser-77 and Ser-79 were mainly responsible for RAR␥ phosphorylation in the cells. Treatment of RAR␥-transfected cells with SB203580, an inhibitor of the p38 MAPK known to phosphorylate Ser-77 and Ser-79 (13), resulted in disappearance of the slow migrating RAR␥ band (Fig. 6A), suggesting that RAR␥ was mainly phosphorylated by the p38 MAPK in cells, consistent with previous results (13). We also examined the effect of ligand binding and RXR␣ heterodimerization on RAR␥ phosphorylation. Again, ectopically expressed RAR␥ exhibited double bands on SDS-PAGE. Interestingly, the fast migrating band was reduced upon treatment of cells with RA or cotransfected with RXR␣ expression vector (Fig. 6B). Such an effect of RA and RXR␣ expression was not observed when RAR␥(S77/79A) was used, suggesting that RA treatment and RXR␣ expression induce similar conformational change of RAR␥, which renders the N-terminal phosphorylation sites of RAR␥ more accessible to responsible kinases.
To determine the role of RAR␥ phosphorylation in its subcellular localization, we examined the effect of SB203580 on subcellular localization of RAR␥ and RAR␥/AB. Cells transfected with RAR␥ or RAR␥/AB were treated with SB203580, and their subcellular localization was examined. In the absence of SB203580 treatment, RAR␥ and RAR␥/AB were mainly cytoplasmic. In contrast, they were nuclear when cells were treated with the inhibitor (Fig. 6, C and D). When Ser-77 and Ser-79 in the A/B domain of RAR␥ were mutated to alanine, the resulting mutant, RAR␥/〈〉(S77/79A), was predominantly nuclear (Fig.  6E). The observation that the A/B domain became nuclear in response to the p38 MAPK inhibitor or mutations of its phosphorylation sites was surprising because the domain does not contain a classic nuclear import signal. Thus, it is likely that nuclear localization of the A/B domain was due to its nuclear retention by binding to certain nuclear proteins under these conditions. Indeed, a previous study showed that phosphorylation of RAR␥ inhibited its interaction with nuclear vinexin ␤ (50). We also examined whether SB203580 could affect the cytoplasmic localization of endogenous RAR␥, and our result showed that treatment of serum-stimulated H460 lung cancer cells with SB203580 also induced nuclear localization of the endogenous RAR␥ (Fig. 6F). Together, these results demonstrated that phosphorylation of Ser-77 and Ser-79 is required for efficient RAR␥ cytoplasmic localization and that the p38 MAPK is likely involved in the phosphorylation.

Interplay between RXR␣ LBD and RAR␥ N-terminal A/B Domain-
The above data demonstrated that cytoplasmic localization of RAR␥ depended on the presence of its N-terminal A/B domain and was inhibited by ligand binding and RXR␣ heterodimerization. The final localization of RAR␥ is likely determined by balance among these regulations. To address the contribution of each regulatory parameter, we first determined the subcellular localization of RAR␥(S77/79A) and its regulation by RXR␣ and ligand binding. Unlike RAR␥, RAR␥(S77/ 79A) was diffusely distributed in both the cytoplasm and nucleus of cells, with predominant nuclear accumulation (Fig.  7A), likely due to lack of RAR␥ phosphorylation. Like RAR␥, the mutant was exclusively nuclear when cells were treated with RA (Fig. 7A). Cotransfection of GFP-RXR␣ also led to exclusive nuclear localization of RAR␥(S77/79A) (Fig. 7B). In contrast, coexpression of RXR␣-(222-462) resulted in exclusive cytoplasmic localization of the RAR␥ mutant even in the presence of RA (Fig. 7B).
The role of the A/B domain was further illustrated by our examination of the regulation of the subcellular localization of ␤AB-RAR␥. Although ␤AB-RAR␥ alone was predominantly nuclear (Fig. 7E), a majority of ␤AB-RAR␥ became cytoplasmic when coexpressed with RXR␣-(222-462). The effect of RXR␣-(222-462) however was largely abolished by RA as ␤AB-RAR␥ was nuclear when cells were treated with RA (Fig. 7E). These results suggested that A/B domain of RAR␤ could compromise the effect of RAR␥ A/B domain in retaining RAR␥/RXR␣-(222-462) heterodimer in the cytoplasm in the absence of RA. Together, our results demonstrate that an appropriate balance among the phosphorylation of the N-terminal A/B domain of RAR␥, its ligand binding, and RXR heterodimerization determines the final localization of RAR␥ in cells.

DISCUSSION
Regulated protein movement between the nucleus and the cytoplasm provides a simple, reversible, and rapid means to regulate nuclear and cytoplasmic events and to coordinate interaction of signal transduction pathways. Recent studies have demonstrated the importance of rapid non-genomic action of nuclear receptors, including steroid hormone receptors and retinoid receptors. We report here that the subcellular localization of RAR␥ is unique among three RAR subtypes in that it often resides in the cytoplasm depending on cellular environment and growth conditions. We further demonstrate that the N-terminal A/B domain of RAR␥ is a major determinant of its cytoplasmic retention, which acts in coordination with phosphorylation, ligand binding, and RXR␣ heterodimerization to determine the final destination of RAR␥ protein in cells. Lysates prepared from RAR␥-transfected cells were also treated with alkaline phosphatase (AP). Cell lysates were subject to Western blot analysis using anti-RAR␥ antibody. B, regulation of RAR␥ phosphorylation by RXR␣ and RA. HEK293T cells cultured in 6-well plates were transfected with RAR␥ or RAR␥ (S77/79A) expression vector (0.5 g/well) with or without RXR␣ (1.5 g/well) and treated with RA for 1 h. Cell lysates were subject to Western blot analysis using anti-RAR␥ antibody. C and D, inhibition of RAR␥ cytoplasmic accumulation by SB203580. HeLa cells cultured in 24-well plates were transfected with Myc-RAR␥ (C) or Myc-RAR␥/AB (D) (100 ng/well), and then treated with or without SB203580 (10 M) for 1 h. Cells were stained with anti-Myc antibody and examined by fluorescence microscopy. About 80% of transfected cells showed RAR␥ or RAR␥/AB nuclear staining when treated with SB203580. E, mutation of phosphorylation sites in RAR␥/AB impairs its cytoplasmic localization. HeLa cells cultured in 24-well plates were transfected with Myc-RAR␥/AB(S77/79A) (100 ng/well). Cells were stained with anti-Myc antibody and examined by fluorescence microscopy. F, SB203580 inhibits serum-induced RAR␥ cytoplasmic localization. H460 cells cultured in serum-free medium for 24 h were stimulated with serum (10%) in the presence of SB203580 (10 M) for 6 h. Cells were immunostained using anti-RAR␥ antibody.
Cytoplasmic Localization of RAR␥-Endogenous RAR␥ is nuclear in cells grown under normal conditions. However, when cells were cultured at high density, RAR␥ was predominantly cytoplasmic (Fig. 1A). Cell density-dependent intracellular localization of proteins is not unprecedented. For example, subcellular localization of von Hippel-Lindau tumor suppressor (51), the ERM family of proteins (52), aryl hydrocarbon receptor (53), and adenomatous polyposis coli (54) is controlled by cell density. RAR␥ was also cytoplasmic when cells cultured in serum-free medium were released from serum starvation or treated with growth factors (Fig. 1B). Thus, modulation of signal transduction pathways during growth and differentiation can alter the subcellular localization of RAR␥.
Consistently, a previous study showed that the subcellular localization RAR␥ is altered during cell growth and differentiation in the endometrium (27). In addition to being a mechanism regulating receptor transactivation, cytoplasmic localization of RAR␥ may be important for its non-genomic actions involved in the regulation of cell growth, apoptosis, and differentiation. For instance, RA treatment of NIH-3T3 cells resulted in accumulation of RAR␥ in the plasma membrane and its interaction with PI3K, an event that is required for RA-induced cell differentiation (55).
While cytoplasmic localization of many nuclear receptors have been shown to be dependent on the classic leucine-rich NES, which is CRM-1-dependent, a number of nuclear receptors can be localized in the cytoplasm through CRM-1-independent mechanism (56 -59). Thus, protein-protein interaction with factors other than nuclear receptors has been suggested as an important regulatory mechanism for subcellular localization of nuclear receptors. Calreticulin mediates cytoplasmic localization of GR independently of classic leucine-rich NES (56), whereas RXR␣ is capable of binding to cytoplasmic G protein (21). The cytoplasmic accumulation of RAR␥ is unlikely due to increased nuclear export of the receptor protein because blocking Crm-1-mediated nuclear export with LMB had no effect on its cytoplasmic localization (Fig. 2C). Such an observation suggests that cytoplasmic accumulation of RAR␥ may be due to its retention in the cytoplasm through interaction with cytoplasmic proteins.
Inhibition of Cytoplasmic Localization of RAR␥ by Ligand Binding and RXR␣ Heterodimerization-Our studies demonstrate that ligand binding and RXR␣ heterodimerization represent two important activities that retain RAR␥ in the nucleus. Unlike other retinoid receptors, transfected RAR␥ is cytoplasmic but resides in the nucleus in response to RA ( Fig. 2A). This is in analogues to steroid hormone receptors, such as GR and androgen receptor that translocate from the cytoplasm into nucleus upon ligand binding (18). In the context of GR, ligand binding induces a receptor conformation that dissociates GR from Hsp90, leading to its activation of NLS and nuclear accu- mulation. It is likely that RA binding induces a RAR␥ conformation that disfavors its binding to cytoplasmic proteins and/or activates its NLS. Ligand-induced RAR␥ nuclear translocation may lead to cotranslocation of cytoplasmic RAR␥binding proteins into the nucleus, thereby inhibiting their activities. It is noteworthy however that binding of RAR␥ to c-Src was RA-dependent (28) and that RA treatment is required for accumulation of RAR␥ in the plasma membrane in NIH-3T3 cells (55). Thus, ligand binding may act to induce RAR␥ cytoplasmic localization in some cell types and under certain conditions.
Our results show that RXR␣ induces RAR␥ nuclear localization through their heterodimerization (Fig. 3). Thus, the ratio of RXR␣ and RAR␥ proteins is another critical determinant of the subcellular localization of RAR␥. RXR␣ heterodimerization likely inhibits the cytoplasmic accumulation of RAR␥ through its inhibition of RAR␥ binding to cytoplasmic proteins and/or activation of their NLS. On the other hand, our observation that RXR␣-(222-462) suppressed RA-induced nuclear localization of RAR␥ (Fig. 3B) suggests that the RXR␣ NLS is required for its ability to retain RXR␣/RAR␥ heterodimer in the nucleus. RAR␥ is not the first one whose subcellular localization is regulated by RXR␣. Previous studies have shown that RXR␣ plays a dominant role in the nuclear localization of RXR/VDR heterodimer (46). However, RXR␣ is required for nuclear export of Nur77 through their unique heterodimerization in response to certain apoptotic stimuli (29).
Cytoplasmic localization of transfected RAR␥ was potently inhibited by ligand binding or RXR␣ heterodimerization, suggesting that interaction of RAR␥ with cytoplasmic proteins was not sufficient to antagonize the effect of ligand-binding or RXR␣ heterodimerization on inducing its nuclear translocation. However, under some conditions, such as superconfluent culture condition, endogenous RAR␥ was cytoplasmic (Fig. 1), presumably as an RAR␥/RXR␣ heterodimer. Thus, interaction of RAR␥ with cytoplasmic proteins could be enhanced by certain cellular stimuli to overcome the effect of RXR␣ heterodimerization. Interestingly, recent studies have shown that RXR can also be cytoplasmic in response to stimuli that induce apoptosis (29), inflammation (31,32), and differentiation (30). It remains to be seen whether and how cytoplasmic RXR␣ regulates RAR␥ activities under these conditions.
Our analysis of regulation of RAR␥ subcellular localization by RXR␣ revealed an unexpected function of RXR␣ LBD fragment, RXR␣-(222-462), in the regulation of the localization and dimerization capacity of RAR␥. The mutant prevented RA-induced RAR␥ nuclear localization (Fig. 3B), presumably through its inhibition of RAR␥ homodimerization and consequently the RAR␥ NLS activity (Fig. 4A). The mutant also inhibited RAR␥/RXR␣ heterodimerization (Fig.  4, B and C). Of interest is that RXR␣ can actually be cleaved in cancer cells by cathepsin L-type protease to produce fragments of similar sizes to RXR␣-(222-462) mutant (47)(48)(49). Whether proteolytic cleavage of RXR␣ functions as a regulatory mechanism of RAR␥ activities remains interesting to study.
N-terminal A/B Domain of RAR␥ Is Essential for Its Cytoplasmic Localization-Our results demonstrate that the unique N-terminal A/B domain of RAR␥ is the major determinant of RAR␥ cytoplasmic accumulation. This was illustrated by our finding that deletion of the A/B domain from RAR␥ abolished its cytoplasmic accumulation (Fig. 5). In addition, replacement of the A/B domain of RAR␥ with the corresponding domain of RAR␤ impaired the cytoplasmic retention of RAR␥. Furthermore, the A/B domain of RAR␥ alone was exclusively cytoplasmic (Fig. 5). The role of the A/B domain can also be shown by our observation that RXR␣-(222-462) retained RAR␥ (Fig. 3B), but not ⌬N-RAR␥ (Fig. 7), in the cytoplasm. A recent study demonstrates that the A/B domain of retinoid-related orphan receptors (RORs) also mediates their cytoplasmic localization through an undefined mechanism (60). How the A/B domain acts to retain RAR␥ in the cytoplasm remains to be determined. As discussed above, the cytoplasmic localization of RAR␥ was not inhibited by LMB (Fig. 2C). The fact that the A/B domain does not contain classic leucine-rich nuclear export signal again argues against a CRM1-mediated nuclear export mechanism. Thus, the A/B domain of RAR␥ contains proline-rich sequences capable of binding to a number of signaling or adaptor proteins (61,62), suggesting that the A/B domain may reside in the cytoplasm through its interaction with cytoplasmic retention factors. Support of this notion comes from a recent report that the N-terminal proline-rich sequences of RAR␥ are required for RAR␥ interaction with c-Src (28).
Binding of proline-rich sequences to signaling or adaptor proteins is subjected to regulation by phosphorylation of Ser/ Thr residues within or adjacent to the proline-rich sequences (61,62). Thus, it is interesting to note that Ser-77 and Ser-79 are located in the proline-rich sequences in the A/B domain. Consistently, phosphorylation-defective RAR␥/AB (Fig. 6E) and RAR␥ mutant (Fig. 7A) were predominantly nuclear, and inhibition of the p38 MAPK known to phosphorylate RAR␥ resulted in nuclear localization of RAR␥ and RAR␥/AB mutant ( Fig. 6, C, D, and F).
Although the N-terminal A/B domain of RAR␥ and its phosphorylation are involved in RAR␥ cytoplasmic localization, their activities are regulated by RAR␥ ligand binding and RXR␣ heterodimerization. RAR␥ was phosphorylated and yet nuclear upon RA treatment or RXR␣ heterodimerization (Fig. 6B). Because ligand binding and receptor heterodimerization are known to activate NLS, such effects of RA treatment and RXR␣ heterodimerization on RAR␥ nuclear localization and phosphorylation would suggest that their activation of NLS predominates over their effect on phosphorylation. Thus, the steady-state distribution of RAR␥ is subjected to multiple levels of regulations, including ligand binding, receptor heterodimerization, and receptor phosphorylation, which need to act in a coordinated manner to determine its final destination in cells.
The existence of three RAR subtypes and their distinct distributions during development and in the adult organisms suggest that they have distinct modes of action and are involved in the control of different biological activities (1,2). For instance, a recent study demonstrated a non-redundant role of RAR␥ in mediating the hematopoietic stem cell self-renewing effects induced by RA treatment (6). However, three subtypes show extensive sequence homolog in their DNA-binding domain and LBD, suggesting that the difference in their transcriptional regulation cannot satisfactorily explain their distinct action in vivo. Thus, our illustration of the unique property of the RAR␥ A/B domain and the possible underlying mechanism of action suggest that different RAR subtype may have distinct modes of action, dictated by their N-terminal A/B domains.
In summary, our results demonstrate that RAR␥ often resides in the cytoplasm due to its unique N-terminal A/B domain and that the subcellular localization of RAR␥ is subject to multiple levels of regulations that act in concert to determine the final destination of RAR␥ protein in cells. Ligand binding and RXR␣ heterodimerization act to retain RAR␥ in the nucleus, whereas the N-terminal A/B domain confers its cytoplasmic localization, which is also subject to regulation by phosphorylation. Loss of balance among these regulatory events will result in a shift of RAR␥ subcellular localization. It can be envisioned that activation of certain signal transduction pathways may result in activation and/or induction of cytoplasmic RAR␥-binding protein, which, together with modulation of RAR␥ phosphorylation will shift RAR␥ protein from the nucleus to the cytoplasm. Subcellular localization of RAR␥ likely serves as a biological switch to mediate the cross-talk between retinoid signaling and signal transduction pathways in response to various cellular stimulations. Phosphorylation-dependent subcellular retention of RAR␥ will provide a simple, reversible, and rapid means to control cellular responses to different environmental signals and physiological/pathophysiological conditions.