Hyperphosphorylation of the Retinoid X Receptor α by Activated c-Jun NH2-terminal Kinases*

The nuclear receptor mouse retinoid X receptor α (mRXRα) was shown to be constitutively phosphorylated in its NH2-terminal A/B region, which contains potential phosphorylation sites for proline-directed Ser/Thr kinases. Mutants for each putative site were generated and overexpressed in transfected COS-1 cells. Constitutively phosphorylated residues identified by tryptic phosphopeptide mapping included serine 22 located in the A1 region that is specific to the RXRα1 isoform. Overexpression and UV activation of the stress-activated kinases, c-Jun NH2-terminal kinases 1 and 2 (JNK1 and JNK2), hyperphosphorylated RXRα, resulting in a marked decrease in its electrophoretic mobility. This inducible hyperphosphorylation involved three residues (serines 61 and 75 and threonine 87) in the B region of RXRα and one residue (serine 265) in the ligand binding domain (E region). Binding assays performed in vitro with purified recombinant proteins demonstrated that JNKs did not interact with RXRα but bound to its heterodimeric partners, retinoic acid receptors α and γ (RARα and RARγ). Hyperphosphorylation by JNKs did not affect the transactivation properties of either RXRα homodimers or RXRα/RARα heterodimers in transfected cultured cells.

are at least two main isoforms, which differ in their NH 2terminal A regions and are generated by differential promoter usage and/or alternative splicing (5,6,8). Several lines of evidence support the conclusion that RAR/RXR heterodimers are the functional units transducing the retinoid signal in vivo 9, and 10 and references therein). However, RXRs are also able to heterodimerize with other members of the nuclear receptor superfamily, such as the thyroid hormone receptors, the vitamin D 3 receptor, and the peroxisome proliferator activated receptors (11)(12)(13)(14)(15).
As other members of the nuclear steroid/thyroid hormone receptor superfamily, RARs and RXRs exhibit a conserved modular structure with six variably conserved functional regions (A to F) ( Fig. 1 and Refs. 5 and 6). The amino-terminal A/B region of RARs and RXRs contains a ligand-independent transactivation function AF-1 (16,17), while the highly conserved C region is included in the DNA-binding domain. The E region is more complex, since, in addition to the ligand-binding domain, it contains a dimerization surface and a ligand-dependent transcriptional activation function AF-2 (6, 16 -19). In the COOH-terminal part of the E region, there is a well conserved amphipathic helix (the AF-2 AD core, helix 12 of the ligand-binding domain), which has been shown to be an essential element of the AF-2 function. Upon ligand binding, there is a major transconformational change of the ligand-binding domain that involves the folding back of helix 12 and the formation of a new surface required for interaction with coactivators that relay the AF-2 activity to the transcriptional machinery and/or to the chromatin template (6, 19 -21).
As most members of the nuclear receptor family (22), RARs have been shown to be phosphoproteins (23)(24)(25). In mouse RAR␣, a phosphorylated serine residue has been identified in region B; it is phosphorylated by Cdk7 associated with the general transcription factor TFIIH, and this phosphorylation has been shown to be crucial for AF-1 activity in transfected COS cells (26). In addition, mRAR␣ is phosphorylated by protein kinase A at a serine residue located in the ligand-binding domain/AF-2 domain (27). These serine residues that are conserved among RARs were also found to be phosphorylated in RAR␥. 2 Most interestingly, phosphorylation of residues in the AF-1 and AF-2 domains of both RAR␣ and RAR␥ has been shown to be indispensable for differentiation of embryonal carcinoma F9 cells upon retinoic acid and cyclic AMP treatment (28).
In the present study, we report that RXR␣ is also a phosphoprotein. As for RAR␣, phosphorylation sites for proline-directed protein kinases are located in the A/B region of mouse RXR␣1 and are constitutively phosphorylated in transfected COS-1 cells. In addition, we demonstrate that under stress conditions * This work was supported by funds from CNRS, INSERM, the Collège de France, the Hôpital Universitaire de Strasbourg, the Association pour la Recherche sur le Cancer, and Bristol-Myers Squibb. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ Supported by the Ministère de la Recherche et de l'Enseignement Supérieur.

EXPERIMENTAL PROCEDURES
Plasmid Constructions-The pSG5-based expression vectors for mRAR␣1 (16), mRXR␣1 (33), and murine RXR␣ deleted for the A/B region (mRXR␣⌬AB) (16) were as described. For the construction of mRXR␣1 mutants, mRXR␣1 was first subcloned into the XhoI/BglII sites of pSG5-Cas (16) after polymerase chain reaction amplification of the A to E regions. The mRXR␣1 S22A, S44A, S48A, S54A, S61A, S75A, T87A, S96A, and S101A expression vectors were constructed by double polymerase chain reaction amplification reactions (27), according to Ho et al. (34), generating a XhoI/EcoRV fragment containing the appropriate mutation. The double mutant RXR␣S75A/T87A was constructed according to the same protocol by introducing the T87A mutation into the RXR␣S75A mutant. Similarly, the RXR␣S61A/S75A/T87A expression vector was constructed by introducing the S61A mutation into the RXR␣S75A/T87A double mutant. RXR␣S265A was also constructed by double polymerase chain reaction amplification reaction, generating an EcoRV/BamHI fragment containing the mutation. The double mutant RXR␣S22A/S265A was prepared by subcloning the EcoRI/EcoRV fragment containing the S22A mutation into the same sites of RXR␣S265A. The same strategy was followed for constructing the RXR␣S61A/S75A/ T87A/S265A mutant. All plasmids were verified by automated DNA sequencing. Additional details of constructions and oligonucleotide sequences are available upon request.
The reporter genes DR1G-tk-CAT and mRAR␤2-CAT have been previously described (16). The expression vectors for dominant active Ras (Ras Val-12 ) and dominant negative Ras (Ras Asn-17 ) were gifts from B. Wasylyk (35) and G. M. Cooper (36,37), respectively. Those for human JNK1 and JNK2 were gifts from M. Karin (38,39), and that of the Cdk7 expression vector was as described (26). Dominant active Cdk1 (A14F15) expression vector was a gift from P. Nurse (40). In lane 2, cells were treated with 10 Ϫ7 M 9cRA for 1 h. Whole cell extracts (WCEs) were immunoprecipitated with mAb 4RX3A2, resolved by SDS-PAGE, and electrotransferred onto nitrocellulose membranes. Phosphorylated RXR␣ was detected by autoradiography ([ 32 P]) and chemiluminescence (WB) after immunoreaction with RPRX␣(D). B, phosphorylated and immunoprecipitated RXR␣ was resolved by SDS-PAGE, electrotransferred onto Immobilon membranes, and subjected to acid hydrolysis. Phosphoamino acids were separated by two-dimensional cellulose thin layer electrophoresis (27,45) and visualized by autoradiography. The dotted lines correspond to the position of the ninhydrin-stained phosphoamino acid standards superimposed on the autoradiograms. C, two-dimensional tryptic phosphopeptide map of RXR␣WT, ⌬AB, and S22A. 32 P-Labeled and immunoprecipitated RXR␣ was electrotransferred onto nitrocellulose filters as in A and digested with trypsin. Phosphopeptides were resolved in two dimensions on cellulose plates (27,45). Purified recombinant RAR␣1WT, RAR␣⌬AB, RAR␥1WT, RXR␣1WT, and RXR␣⌬AB overexpressed in Escherichia coli were gifts from H.
Cells, Transfections, and Chloramphenicol Acetyltransferase (CAT) Assays-COS-1 cells maintained in Eagle's modified Dulbecco's medium supplemented with 5% fetal calf serum were transiently transfected using the calcium phosphate precipitation technique (25). In addition to the expression vectors or reporters described in each figure legend, all transfections contained the ␤-galactosidase expression vector pCH110 (1 g) to correct for variations in transfection efficiency and Bluescript DNA as a carrier. After a 20-h incubation with calcium phosphate-precipitated DNA, the cells were washed, maintained for 8 h in the appropriate medium, and incubated for another 20 h in medium containing 0.5% charcoal-treated serum along with the ligand (10 Ϫ7 M 9cRA). Where mentioned, cells were UV-irradiated (40 J/m 2 ) 4 h before harvesting (42). CAT assays were performed using the enzyme-linked immunosorbent assay method (CAT ELISA, Roche Molecular Biochemicals). Results were expressed as pg of CAT/unit of ␤-galactosidase.
The F9 -1.8 reporter cell line stably transfected with the mRAR␤2 promoter coupled to the lacZ gene was maintained in Dulbecco's modified Eagle's medium containing 7.5% fetal calf serum as described by Maden et al. (43).
Immunoprecipitations and CIP Treatment of the Immunoprecipitates-Whole cell extracts were prepared from transfected COS as de-scribed (26) and incubated with Protein A-Sepharose beads cross-linked with the indicated monoclonal antibodies in IP buffer (50 mM Tris-HCl, pH 7.5, 10% glycerol, 0.1 mM EDTA, 150 mM KCl, 5 mM MgCl 2 , and 0.1% Nonidet P-40) for 2 h at 4°C. The beads were washed in IP buffer; resuspended in 100 l of phosphatase reaction buffer (100 mM Tris-HCl, pH 9.8, 1 mM MgCl 2 , 0.1 mM ZnCl 2 ) containing 20 units of calf intestinal alkaline phosphatase (Roche Molecular Biochemicals) in the absence or presence of sodium orthovanadate (50 M) and incubated at 37°C for 3 h. After washing, the immunoprecipitated proteins were resolved by SDS-10% PAGE, electrotransferred onto nitrocellulose membranes, and detected by immunoblotting and chemiluminescence according to the manufacturer's protocol (Amersham Pharmacia Biotech).
Cytosolic Extracts for Detection and Isolation of Activated JNKs-Cells were washed and lysed as described by Sadowski and Gilman (44) in ice-cold hypotonic buffer (20 mM HEPES, pH 7.9, 20 mM NaF, 1 mM Na 3 VO 4 , 0.125 M okadaic acid, 1 mM EDTA, 1 mM EGTA, 1 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, and a mixture of protease inhibitors) containing 0.2% Nonidet P-40. After centrifugation at 16,000 ϫ g for 20 s, supernatants were supplemented with NaCl to 120 mM and clarified (16,000 ϫ g for 20 min), and glycerol was added to 10%.
In Vitro and in Vivo Phosphorylation-In vitro phosphorylation reactions were performed as described by Rochette-Egly et al. (27) with purified bacterially expressed RXR␣1 (1 g), using either p44 mitogenactivated protein kinase (20 ng) or p34 Cdk1/cyclin B (20 ng) (Upstate Biotechnology, Inc., Lake Placid, NY). In the case of JNKs, activated JNKs were first isolated from cytosols of UV-irradiated COS cells by immunoprecipitation with JNK1 antibodies conjugated to agarose beads (Santa Cruz Biotechnology), and the reaction was initiated by the addition of purified RXR␣1 to the beads. Phosphorylated proteins were resolved by SDS-PAGE, electrotransferred onto nitrocellulose membranes, and visualized by autoradiography and immunoblotting.
Two-dimensional phosphoamino acid analysis and tryptic phosphopeptide mapping were carried out on thin layer cellulose plates using the Hunter thin-layer electrophoresis system as described (27,45).

RXR␣ Overexpressed in COS-1 Cells Is Phosphorylated in Its
Region-To determine whether the nuclear RXR␣ is a phosphoprotein, COS-1 cells were transfected with RXR␣WT (33) expression vector and labeled with [ 32 P]orthophosphate. RXR␣ was phosphorylated irrespective of the addition of 9cRA (10 Ϫ7 M) to the culture medium ( Fig. 2A, lanes 1 and 2). Phosphoamino acid analysis indicated that serine residues were phosphorylated (Fig. 2B). Tryptic phosphopeptide mapping of RXR␣WT yielded two main phosphopeptides named a and aЈ and an array of additional peptides, named x, lying on two parallel diagonals (Fig. 2C, panel 1). It must be stressed that, depending on the experiments, these x phosphopeptides were variably distinct, suggesting that they may be partial digestion products or phosphoisomers (45). Identical phosphopeptide patterns were obtained whether or not COS-1 cells were treated with 9cRA (data not shown).
To characterize the phosphorylated regions, COS-1 cells were transfected with RXR␣⌬AB and subsequently labeled with [ 32 P]orthophosphate. The level of phosphorylation was not significantly affected in RXR␣⌬AB ( Fig. 2A, compare lanes 3  and 4). However, the phosphopeptides observed in RXR␣WT were lacking in RXR␣⌬AB (Fig. 2C, panel 2), and two additional phosphopeptides (b and bЈ), which were not detectable in RXR␣WT, appeared. These results indicate that RXR␣ was phosphorylated mainly in the A/B region and also suggest that the A/B region may prevent the phosphorylation of sites located elsewhere in the protein.
There are nine potential phosphorylation sites for prolinedirected kinases in the RXR␣1 A/B region (see Fig. 1). The serine and threonine residues of these putative sites (serines 22,44,48,54,61,75,96, and 101 and threonine 87) were individually mutated to alanine, and the corresponding mutants were expressed in COS-1 cells. Their level of phosphorylation and phosphopeptide maps were not significantly different from those of RXR␣WT except for RXR␣S22A, which lacked phosphopeptides a and aЈ (Fig. 2C, panel 3, and data not shown). The observation that a single mutation (S22A) abrogated two phosphorylated spots (a and aЈ) lying on a diagonal suggests that they may correspond to interdependent phosphorylation of adjacent serines (at positions [17][18][19], with the slowest migrating peptide toward the anode (peptide a) containing only a single phosphate (45). Note that a third spot situated on the same diagonal was sometimes observed (see Fig. 3B, panel 1). The nature of the phosphoresidues present in peptides x, which could be possibly located outside of the A/B region, remains to be identified.
In Vitro Phosphorylation of RXR␣ by Cyclin-dependent Kinases (CDKs) and Mitogen-activated Protein Kinases (MAPKs)-Serine 22 belongs to a conserved motif for proline-directed Ser/Thr kinases, such as the CDKs (46) and the MAPK family. The latter includes extracellular signal-regulated kinases (ERKs), as well as SAPKs, such as the JNKs (29 -32).
Thus, the phosphorylation pattern of RXR␣1 overexpressed in COS cells is similar to that observed in vitro with p34 Cdk1/cyclin B, while ERKs and JNKs appear to phosphorylate different residues located both in the A/B region and the remaining protein. Overexpression of Activated JNKs Increases the Phosphorylation of RXR␣ in COS-1 Cells, whereas ERKs and CDKs Are Inefficient-The above in vitro results suggest that RXR␣ could be a target for CDKs. However, overexpression of a dominant active Cdk1 (40) or of Cdk7, which was previously shown to increase the phosphorylation of RAR␣ (26), had no effect on the level of RXR␣ phosphorylation and on its phosphopeptide maps (Fig. 4A, compare lanes 4 and 5, and data not shown). Whether CDKs other than those tested here could be implicated in the basal phosphorylation of RXR␣ in COS-1 cells remains to be investigated.
Since RXR␣ was a substrate for p44 MAPK in vitro, we also examined whether stimulation of the MAPK pathways could affect the phosphorylation of RXR␣ in transfected COS-1 cells. The ERK pathway is stimulated in response to growth factors through Ras activation, while JNKs are activated by stress stimuli or UV irradiation. Activation of ERKs and JNKs involves their own phosphorylation by other kinases located further upstream in the specific signaling cascade (for reviews, see Refs. 29 -32, 47, 48, and references therein). The phosphopeptide map of RXR␣ was not affected by epidermal growth factor treatment or by overexpression of either an activated Ras (Ras-Val-12 (35)) or a dominant negative Ras (Ras Asn-17 (36, 37)) ( Fig.  4A, lanes 1-3, and data not shown). Similar results were obtained by overexpressing MAPK kinase or the MAPK kinasespecific phosphatase, CL100 (49) (data not shown). Thus, the Ras-ERK cascade of the growth factor receptor tyrosine kinase signaling pathway does not appear to be involved in RXR␣ phosphorylation in COS cells.
In contrast, overexpression and UV activation (40J/m 2 ) of the stress-activated protein kinase JNK1 increased the level of RXR␣ phosphorylation concomitantly with a marked decrease in its electrophoretic mobility that is characteristic of a hyperphosphorylation (Fig. 4A, compare lanes 6 and 9). This upward shift of RXR␣, which was visible both by immunoblotting and by incorporation of 32 P (Fig. 4A, compare lanes 6 and 9, upper and lower panels), could be detected within 5 min after UV irradiation and persisted for at least 2 h (Fig. 4B, lanes 6 -11). Overexpression of JNK2, another c-Jun NH 2 -terminal kinase, had the same effect (Fig. 4D, lane 9). Overexpression of JNK1 without UV irradiation did not cause this mobility shift (Fig.  4A, lane 8). In contrast, UV irradiation in the absence of cotransfected JNK expression vector induced within 1 h a slight but significant reduction of the electrophoretic mobility of RXR␣ (Fig. 4A, lane 7, and Fig. 4B, lanes 1-5). Treatment of cell extracts with calf intestinal alkaline phosphatase in the absence of vanadate (a phosphatase inhibitor) abrogated the upward shift induced by either JNK1 or JNK2 overexpression and UV treatment (Fig. 4D, compare lanes 7 and 11 with lanes 5  and 9, respectively). Note that, as expected, both overexpressed JNK1 and JNK2 (46 and 54 kDa, respectively) were activated by UV irradiation, as determined by Western blot analysis with anti-ACTIVE TM JNK antibodies that recognize the phosphorylated form of JNKs (Fig. 4C, upper panel), while the JNK protein content was not affected (Fig. 4C, bottom, and data not shown). Furthermore, serum starvation (4 h) that is also known to activate JNKs (50) similarly induced an upward shift in RXR␣ electrophoretic mobility (data not shown). In contrast,  3, 5, 7, 9, 11, 13, 15, 17, 19, and 21) or presence (lanes 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, and 22) of JNK1 vector. Where indicated, cells were UV-irradiated 1 h before harvesting. WCEs were immunoprecipitated with mAb 4RX3A2 and processed for immunoblotting with RPRX␣(A). B, two-dimensional tryptic phosphopeptide map of 32 P-labeled immunoprecipitated RXR␣WT, RXR␣S22A/S265A, RXR␣S61A, and RXR␣S61A/S75A/T87A/S265A, with or without cotransfected JNK1 and UV irradiation, as indicated. no upward shift was seen upon retinoic acid treatment (either tRA or 9cRA at 10 Ϫ7 M) for up to 24 h (data not shown).
Altogether, these results demonstrate that RXR␣WT is inducibly hyperphosphorylated by activated JNKs, whereas under the same conditions, there is no hyperphosphorylation of RAR␣ (Fig. 4E), and its phosphorylation pattern is not affected (data not shown).
Activated JNKs Phosphorylate Serine Residues Located in both the B and E Regions of RXR␣1-Two-dimensional tryptic phosphopeptide mapping was used to determine which RXR␣ residues were phosphorylated in transfected COS cells upon activation of JNKs. Several novel RXR␣ phosphopeptides (y1, y2, and z) were generated (Fig. 5B, panel 4) in addition to those obtained from control COS cells (Fig. 5B, panel 1); they were similar to those derived from RXR␣ phosphorylated in vitro with ERKs and JNKs (see Fig. 3B, panel 2). Note that a third y spot (y3) was often observed and that similar phosphopeptide maps were obtained whether JNKs were activated by UV irradiation or serum deprivation (4 h, data not shown). Interestingly, RXR␣⌬AB did not yield phosphopeptide z while phosphopeptides y1-y3 were still present (Fig. 5B, panel 5), thus suggesting that peptide z was generated from the A/B region, whereas peptides y1-y3 originated from elsewhere in the protein.
Since there is a consensus phosphorylation site for prolinedirected kinases in the NH 2 -terminal end of the RXR␣ E region at position 265 (Fig. 1), we mutated the serine residue at this site into alanine (RXR␣S265A). This mutation resulted in the loss of phosphopeptides y1-y3, indicating that serine 265 is a target for activated JNKs (Fig. 5B, panel 6) and that phosphopeptides y1-y3 may correspond to partial digestion products (containing serine 265) and/or to phosphoisomers (45) resulting from the interdependent phosphorylation of the adjacent serine at position 264. The S265A mutation did not suppress the upward shift of RXR␣ upon hyperphosphorylation induced by activated JNKs (Fig. 5A, lanes 3 and 4), thus indicating that phosphorylation of this residue is not sufficient for that process.
The next set of experiments was aimed at identifying the phosphoresidues contained in spot z. Note that the presence of spot z was associated with the upward shift of RXR␣, since RXR␣⌬AB, which did not yield spot z (Fig. 5B, panel 5), was not upward shifted (Fig. 5A, lanes 5 and 6). The mutation of six sites (serines 22,44,48,54,96, and 101) among the nine potential phosphorylation sites present in the A/B region, either individually or in association with mutation of serine 265, had no apparent effect on the upward shift of RXR␣ (Fig. 6A,  lanes 1-8, and data not shown) and did not affect the presence of phosphopeptide z (Fig. 6B, compare panels 2 and 6, and data not shown). In fact, the upward shift was decreased when serine 61, serine 75, or threonine 87 was individually mutated to alanine (Fig. 6A, lanes 9 -16), whereas it was abrogated upon simultaneous mutation of the three residues (Fig. 6A, lanes 19  and 20), irrespective of mutation of serine 265 (Fig. 6A, lanes 21  and 22). Thus, our results suggest that serine 61, serine 75, and threonine 87 are involved in the electrophoretic upward shift of RXR␣ induced by activated JNKs.
Altogether, our results demonstrate that serine 265 is phos-phorylated by activated JNKs and is contained in spots y. In addition, serine 61 in association with two other residues located in the B region (serine 75 and threonine 87) is involved in the appearance of phosphopeptide z and the upward shift of RXR␣ induced by activated JNKs. JNKs Do Not Bind RXR␣ but Bind RAR␣ and RAR␥ in Vitro-Binding assays between RXR␣ and JNKs were performed in vitro with purified recombinant proteins to investigate whether RXR␣ and JNKs could stably interact with each other. Purified bacterially expressed RXR␣ was mixed with GST-JNK2 (also named GST-SAPK␣ (48)) fusion protein attached to glutathione-Sepharose beads, and bound protein was revealed by immunoblotting. No significant binding was detected between RXR␣ and JNK2, either in the absence or presence of 9cRA (1 M) (Fig. 7A, lane 3, and data not shown). In contrast, the GST-JNK2 beads retained purified bacterially expressed RAR␥ and RAR␣ (Fig. 7A, lane 6; Fig. 7B, lane 2; and data not shown) in a ligand-independent manner. Neither RAR␥1 nor RAR␣1 was detected on control GST beads (Fig. 7A,  lane 5, and Fig. 7B, lane 3). RAR␣1 also interacted with the GST-JNK3 fusion protein (JNK3 is also known as SAPK␤ (48)) (data not shown), and purified bacterially expressed RAR␣⌬AB also interacted with either GST-JNK2 or GST-JNK3 fusion  3 and 6) fusion proteins bound to glutathione-Sepharose beads. Bound RXR␣ and RAR␥ were detected by immunoblotting with RPRX␣(A) (lanes 1-3) and mAb 1␥(A1) (lanes 4 -6), respectively. Lanes 1 and 4 correspond to 50% of the input RXR␣WT and RAR␥ proteins. B, purified bacterially expressed (0.5 g) RAR␣1WT (lanes 2 and 3) or ⌬AB (lanes 5-7) were incubated with control GST (lanes 3 and 5), GST-JNK2 (lanes 2 and 6), or GST-JNK3 (lane 7) fusion proteins bound to glutathione beads. Bound RAR␣ was detected by immunoblotting with RP␣(F). Lanes 1 and 4 correspond to 30 and 20%, respectively, of the input RAR␣WT and ⌬AB proteins. C, purified bacterially expressed RAR␣1WT (0.5 g) was incubated with recombinant JNK1 (0.4 g) (Calbiochem) immunoadsorbed on agarose beads crosslinked with JNK1 antibodies. After SDS-PAGE, bound RAR␣ was detected by immunoblotting with RP␣(F).  1-4) or with RAR␣1 or RXR␣ expression vectors (0.1 g), either individually (lanes 5-8 and 9 -12, respectively) or in combination (lanes 13-16). Cells were treated with proteins, irrespective of the presence of tRA (Fig. 7B, lanes 6 and 7, and data not shown). In addition, purified RAR␣1 was retained by JNK1 immobilized onto agarose beads cross-linked with JNK1 polyclonal antibodies (Fig. 7C, lane 2), thus corroborating the above results.
Phosphorylation by Activated JNKs Does Not Affect Transactivation by RXR␣-The ability of activated JNKs to influence RXR␣-mediated activation of transcription was analyzed. COS-1 cells were cotransfected with a reporter construct containing the CAT gene under the control of a retinoic acidinducible promoter, the natural mRAR␤2 promoter that is preferentially activated by RXR␣/RAR␣ heterodimers or the synthetic DR1G-tk promoter that is preferentially activated by RXR homodimers (16).
The effect of activated JNKs on transactivation by RXRs and RARs was also studied in F9 cells stably transfected with a lacZ reporter gene under the control of the murine RAR␤2 promoter (43). As observed with transiently transfected COS cells, activation of endogenous JNKs by UV irradiation and subsequent hyperphosphorylation of endogenous RXR␣, 3 had no effect either on the increase of ␤-galactosidase activity upon tRA and 9cRA treatment (Fig. 8F, compare lanes 9 -12). DISCUSSION RXR␣ Overexpressed in COS-1 Cells Is "Constitutively" Phosphorylated-We have shown here that the ␣1 isoform of mRXR␣, like other nuclear receptors, is a phosphoprotein when overexpressed in COS-1 cells. RXR␣ is phosphorylated in the absence of 9cRA, and no changes occur upon ligand binding. As other nuclear hormone receptors (22,26), RXR␣ is phosphorylated at several residues located in the A/B region that contains the AF-1 transactivation domain. Interestingly, one of the phosphorylated residues has been identified as serine 22, which is unique to the A1 region of the RXR␣1 isoform ( Fig. 1 and Ref. 8), while the location of the others remains to be identified. This is in contrast to the case of RAR␣1, for which no phosphorylation in the isoform-specific A1 region has been found (26).
Serine 22, which is followed by a proline residue, is in a favorable context for phosphorylation by proline-directed kinases that includes CDKs and MAPKs. Although RXR␣ could be phosphorylated in vitro by either of these protein kinases, only CDKs yielded a pattern of phosphorylated peptides identical to that obtained in transfected COS-1 cells. However, in COS cells, coexpression of Cdk1 or Cdk7 did not affect RXR␣ phosphorylation, while the latter increased the phosphorylation of RAR␣ (26). Whether other cyclin-dependent kinases or another proline-dependent kinase could be involved in this constitutive phosphorylation of RXR␣1 needs further investigation. Furthermore, we found no evidence supporting an in vivo involvement of the tyrosine kinase receptor/Ras/Raf/ERK cascade in the phosphorylation of RXR␣1, although it was also an in vitro target for ERKs that belong to the MAPK family.
In the present study, we have shown that RXR␣ is hyperphosphorylated by either JNK1 or JNK2 upon activation by UV irradiation, resulting in an upward shift in its electrophoretic mobility. Whether JNK3 has a similar effect remains to be seen. In contrast to c-Jun, RXR␣ was unable to interact stably with JNKs, suggesting that a labile and transient interaction between JNKs and RXR␣ is sufficient for its phosphorylation. RXR␣ hyperphosphorylation involves residues that are distinct from those that are involved in constitutive phosphorylation; one of these residues (serine 265) is located at the NH 2 -terminal end of the E region, while the three others (serines 61 and 75 and threonine 87) are located in the B region. It is interesting to note that serine 265 corresponds to a readily accessible phosphoacceptor site, since it is exposed outside the ligandbinding domain ␣-helical sandwich, within the ⍀ loop between ␣ helices H2 and H3 (20). Among the three residues located in the B region, serine 61 was clearly phosphorylated by activated JNKs, since its mutation results in the disappearance of phosphopeptide z (Fig. 6B), whereas the mutation of serine 75 and threonine 87 only decreased the intensity of its 32 P labeling. Interestingly, these residues belong to a conserved element involved in JNK binding (TPTPT) that is present in ATFa and ATF2 proteins (55). Thus, serine 75 and threonine 87 might be instrumental in serine 61 hyperphosphorylation through their involvement in a labile RXR␣-JNK interaction.
What Could Be the Function of Hyperphosphorylated RXR␣?-Phosphorylation is an essential prerequisite for the transcriptional activity of various transcription factors such as c-Jun, ATF2, and RARs (Refs. 26, 27, 31, 52, and references therein). However, hyperphosphorylation by activated JNKs did not increase the transcriptional activity of RXR␣ homodimers. We have shown here that RAR␣ and RAR␥, the heterodimeric partners of RXR␣, are able to bind JNKs, although they are not efficient substrates for JNKs in COS cells, thus suggesting that the presence of RAR␣ or RAR␥ may enhance the phosphorylation of RXR␣ by activated JNKs. However, cotransfection of RAR␣ along with RXR␣ did affected neither the phosphorylation of RXR␣ by activated JNKs 3 nor the transcriptional activity of RXR␣/RAR␣ heterodimers using a reporter gene with a promoter containing a DR5 RARE (the natural mRAR␤2 promoter). The same observations were made with endogenous RAR␣, RXR␣, and JNKs in an F9 reporter cell line containing the RAR␤2 promoter coupled to lacZ. Therefore, JNK-mediated RXR␣ hyperphosphorylation does not seem to be involved in the transcriptional synergy of RXR␣ and RARs (59 -61). However, due to the promoter and cell context specificity of the transcriptional functions of RARs and RXRs, the possibility cannot be excluded that RXR␣ hyperphosphorylation is involved in the transactivation of other genes in other cell types.
What could then be the function of hyperphosphorylated RXR␣? One possibility might be that JNK-mediated phosphorylation stabilizes RXR␣ by protecting it from ubiquitination and subsequent proteolytic degradation as previously reported for c-Jun and p53 (for a review, see Ref. 62). Alternatively, RXR␣ hyperphosphorylation may play a role in apoptosis. UV radiations, as well as other stress agents, in addition to being JNK activators, are known to be DNA-damaging agents and to induce apoptosis (for a review, see Ref. 31). Interestingly, RXR␣ has been shown to be essential for the induction of apoptosis in F9 embryocarcinoma cells in response to retinoids (60). Thus, our data suggest the existence of cross-talks between the stress-activated kinases and the RA signaling pathways, both leading to apoptosis. Studies are in progress to investigate whether RXR␣ hyperphosphorylation is actually involved in apoptosis.